Law, Economics, and Morality

Law, Economics, and Morality This page intentionally left blank Law, Economics, and Morality eyal zamir • barak m...

Author: Eyal Zamir | Barak Medina

241 downloads 3116 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Law, Economics, and Morality

This page intentionally left blank

Law, Economics, and Morality eyal zamir

•

barak medina

1

1 Oxford University Press, Inc., publishes works that further Oxford University’s objective of excellence in research, scholarship, and education. Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam

Copyright © 2010 by Oxford University Press, Inc. Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 Oxford is a registered trademark of Oxford University Press Oxford University Press is a registered trademark of Oxford University Press, Inc. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press, Inc. _____________________________________________ Library of Congress Cataloging-in-Publication Data Zamir, Eyal. Law, economics, and morality / Eyal Zamir, Barak Medina. p. cm. Includes bibliographical references and index. ISBN 978-0-19-537216-8 ((hardback) : alk. paper) 1. Law and economics. 2. Law–Philosophy. 3. Law–Economic aspects. 4. Economics–Moral and ethical aspects. 5. Law–Moral and ethical aspects. I. Medina, Barak. II. Title. K487.E3Z36 2010 340′.11—dc22 2009042476 _____________________________________________ 1 2 3 4 5 6 7 8 9 Printed in the United States of America on acid-free paper Note to Readers This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is based upon sources believed to be accurate and reliable and is intended to be current as of the time it was written. It is sold with the understanding that the publisher is not engaged in rendering legal, accounting, or other professional services. If legal advice or other expert assistance is required, the services of a competent professional person should be sought. Also, to confirm that the information has not been affected or changed by recent developments, traditional legal research techniques should be used, including checking primary sources where appropriate. (Based on the Declaration of Principles jointly adopted by a Committee of the American Bar Association and a Committee of Publishers and Associations.)

You may order this or any other Oxford University Press publication by visiting the Oxford University Press website at www.oup.com

Contents

Acknowledgments xi Introduction 1 part one

Theory 9

1. The Consequentialist Nature of Economic Analysis 11 A. General 11 B. Normative Economics 11 C. Consequentialism and Its Critique 18 D. Responses to the Lack-of-Constraints Critique 21 1. Long-Term and Indirect Effects 22 2. Rule-Consequentialism 24 3. “Preferences for Constraints” 27 4. Feelings of Virtue and Remorse 29 5. An Improved Theory of the Good: Ideal Preferences 30 6. Summary 32 E. Responses to the Demandingness Objection 33 F. Conclusion 40 2. Threshold Deontology and Its Critique 41 A. Deontology 41 B. Critique of Deontology in General 49 C. Critique of Threshold Deontology 51 D. Concluding Remarks 56 3. Private and Public Morality 57 A. General 57 B. The Private/Public Distinction 58 C. Doing and Allowing 60 v

vi

contents

D. Intending and Foreseeing 63 E. Acting and Enacting 70 F. Concluding Remarks 76 4. Constructing Threshold Functions 79 A. Introductory Remarks 79 B. General Structure of a Threshold Function 84 C. Relevant Types of Benefits and Costs 86 D. Shape and Size of the Threshold 93 E. Other Concerns 96 F. Threshold Options 98 G. Concluding Remarks 103 5. Addressing Possible Objections 105 A. Undermining the Normative Neutrality of Economic Analysis 105 B. Quantification and Monetization Difficulties 108 1. General 108 2. Anti-Commodification 111 3. Incomparability 112 4. Incommensurability 113 C. Setting Constraints Too Low 116 D. Incompatibility with the Expressive Role of Law 117 E. Conclusion 122 part two

Applications 125

6. The Fight Against Terrorism 127 A. Introduction 127 B. Economic Analysis of the Fight Against Terrorism 128 C. The Constraint Against Harming Persons and the Fight Against Terrorism 135 1. General Considerations 135 2. Harming Aggressors as a Constraint Infringement 136 D. Constrained Economic Analysis of Intended Harm 140 1. General 140 2. Goals of Anti-Terrorist Measures: Preemption, Retribution, Deterrence, and Pressure 140 3. Basic Elements of the Threshold Function 145

contents

4. The Net Benefit 147 (a) The Relevant Variables 147 (b) Marginal Net Benefit and Alternative Courses of Action 149 5. The Threshold 150 (a) General 150 (b) Probability of the Terrorist Attack 151 (c) The Aggressor’s Culpability 160 (d) Summary 160 6. Torture 161 E. Constrained Economic Analysis of Unintended Harm 166 1. General 166 2. Constructing the Threshold Function 169 3. Killing Persons Who are Doomed 170 4. Victims’ Moral Responsibility and Nationality 171 F. Measures Involving Both Intended and Unintended Harm 174 G. Conclusion 175 7. Freedom of Speech 177 A. General 177 B. Doctrinal Background 178 C. Cost-Benefit Analysis of Free Speech and Its Critique 184 D. The Constraint Against Suppressing Free Speech 189 E. Constrained Cost-Benefit Analysis 195 1. The Regulation’s Net Benefit 195 (a) General 195 (b) Chronologically Remote Harms 196 (c) Low-Probability Harms 199 (d) Small Harms 201 (e) Harms Brought About Through Rational Persuasion 202 (f) Offensiveness 206 (g) Combining Excluders 210 2. The Threshold 211 (a) General 211 (b) The Threshold’s Shape 211 (c) Different Thresholds for Different Bases of Regulation 213

vii

viii

contents

(d) Different Thresholds for Different Categories of Speech 218 (e) Summary 221 3. Choosing Among Permissible Courses of Action 222 F. Conclusion 224 8. Antidiscrimination Law 225 A. Introduction 225 B. Current Legal Norms 228 C. Motivations for Discrimination 231 D. Standard Normative Economic Analysis 234 E. Integrating Deontological Constraints with Economic Analysis 240 1. The Constraint Against Discrimination and Its Incidence 240 2. Moral Constraints and Redistributive Goals 246 3. Integrating Threshold Constraints with Economic Analysis 251 F. Conclusion 255 9. Contract Law 257 A. Introduction 257 B. Economic Analysis of Contract Law 258 C. Deontological Constraints and Contract Law 260 1. The Pertinent Constraints: An Overview 260 2. The Economic Response and Its Critique 262 D. Constrained Economic Analysis of Mistake and Misrepresentation 267 1. A Brief Doctrinal Background 267 2. Standard Economic Analysis 269 3. Deontology and Deception 274 4. Constrained Economic Analysis 277 (a) Integrating Constraints 277 (b) Integrating Options 289 E. Remedies for Breach of Contract 292 1. A Brief Doctrinal Background 292 2. Standard Economic Analysis 294 3. Deontology: Promises, Harms, and Contractual Obligations 298

contents

4. Deontological Features of Contract Remedy Rules 301 5. Challenges Facing Constrained Economic Analysis of Remedy Rules 305 F. Conclusion 310 10. Legal Paternalism 313 A. General 313 B. Paternalism: Classifications and Prevalence 315 C. Normative Economic Analysis of Paternalism 318 1. The Compatibility of Efficiency and Paternalism 318 2. Possible Objections 327 3. A Simple Model 332 D. Incorporating Deontological Constraints 335 1. Deontological Perspectives on Paternalism 335 2. Size of the Threshold 339 3. Relevant Types of Benefits and Costs 342 4. Marginal Net Benefit and Alternative Measures 346 E. Conclusion 347 Conclusion 349 Index 353

ix


Acknowledgments

we are most grateful to Shmuel Becher, Samantha Brennan, Miriam Gur-

Arye, Hanoch Dagan, Meir Dan-Cohen, Guy Davidov, Ronit Donyets-Kedar, David Enoch, Ruth Gavison, Kent Greenawalt, Alon Harel, Roy Kreitner, Mordechai Kremnitzer, David Kretzmer, Itzhak Kugler, Daphna LewinsohnZamir, Ariel Porat, Eric Posner, Yuval Procaccia, Ram Rivlin, Re’em Segev, Yuval Shany, Doron Teichman, David Weisburd, and Richard Zerbe for their invaluable suggestions and comments on earlier versions of parts of this book. We also wish to thank the participants in seminars and workshops held at the Georgetown University Law Center, Haifa University, the Hebrew University of Jerusalem, Tel Aviv University, and the 2007 Annual Conference of the European Association of Law and Economics for their helpful comments. Fruitful correspondence with Shelly Kagan and numerous enlightening discussions with David Enoch are particularly appreciated. We thank Yoel Ben-Or and Netta Corren for outstanding research assistance and Efrat Hakak for her excellent editing. This book incorporates, in revised form, ideas that have previously been published in our joint article Law, Morality, and Economics: Integrating Moral Constraints with Economic Analysis of Law, 96 California Law Review 323 (2008), and in Eyal Zamir’s article The Efficiency of Paternalism, 84 Virginia Law Review 229 (1998). The book was written in part during Eyal Zamir’s stay at the NYU School of Law (2005/06) and Georgetown University Law Center (2008, 2009), and during Barak Medina’s stay at Columbia Law School (2006/07); we thank these institutions for their hospitality. Finally, we gratefully acknowledge the generous financial support we received from the Milton and Miriam Handler Foundation.

xi


Introduction

economic analysis has transformed legal theory. Even its detractors can

hardly deny the enormous contribution made by economic methodology to legal thinking in practically every field of law. Cost-benefit analysis (CBA) forces one to consider the interrelations between goals, means, incentives, and outcomes in a systematic and sophisticated way. It assists analysts and policy-makers in identifying false intuitions and cognitive biases, thus rationalizing decision-making.1 The very act of economic modeling compels one to determine the crucial variables pertinent to any issue. At the same time, even avid supporters of economic analysis can hardly deny the fundamental normative flaws that exist in standard economic analysis.2 In particular, the criteria of economic efficiency tend to ignore fundamental ethical norms such as the inherent immorality of deliberately harming other people. The consequentialist nature of economic analysis, namely its denial of the intrinsic value of any factor other than the goodness of outcomes, makes it normatively unacceptable for many philosophers and lawyers. Deontological moral theories maintain that although the goodness of outcomes counts, it is not the only morally relevant factor.3 The pursuit of good consequences is subject to constraints. Certain acts are inherently wrong and therefore impermissible, even as a means to furthering the overall good. The central constraint is against harming other people.4 Additional

1. Cass R. Sunstein, Cognition and Cost-Benefit Analysis, 29 J. Legal Stud. 1059 (2000). 2. See, e.g., Lewis A. Kornhauser, On Justifying Cost-Benefit Analysis, 29 J. Legal Stud. 1037, 1037 (2000); Richard A. Posner, Cost-Benefit Analysis: Definition, Justification, and Comment on Conference Papers, 29 J. Legal Stud. 1153, 1154 (2000). 3. See infra pp. 41–48. 4. This book mostly discusses deontological constraints rather than (moral or legal) rights, thus avoiding the questions of what are rights, and what is the exact relationship between rights and constraints. On this complex issue, see generally Alon Harel, Theories of Rights, in Blackwell’s Guide to the Philosophy of Law and Legal Theory 191, esp. at 197–201 (Martin P. Golding & William A. Edmundson eds., 2005); Shelly Kagan, Normative

1

2

law, economics, and morality

constraints prohibit such conduct as lying and breaking promises. Currently prevailing deontological theories are moderate rather than absolutist. They admit that constraints have thresholds. A constraint may be overridden for the sake of furthering good outcomes or avoiding bad ones if enough good (or bad) is at stake.5 Thus, while consequentialism at least presumably approves of the deliberate killing of one innocent person to save the lives of two, moderate deontology may justify such killing only for the sake of saving many more people, perhaps hundreds or thousands. Similarly, while consequentialism supports the breaking of a promise whenever it would produce slightly more net benefit than keeping one’s word, moderate deontology would justify breaking a promise only to avoid very considerable losses (an absolutist would object to killing or breaking a promise under any circumstances). Deontology differs from consequentialism not only in holding that there are constraints to attaining the best outcomes but also in recognizing options. People are sometimes allowed to refrain from maximizing the good, even if no constraint infringement is involved. At least under some circumstances, people may legitimately prefer their own welfare, or the welfare of their family, friends, or community, over the overall good.6 Moderate deontology conforms to prevailing moral intuitions (“commonsense morality”).7 At the same time, it arguably lacks the methodological rigor and determinacy characteristic of economic analysis.8 Therefore, the argument goes, policy-makers and legal academics should better ignore nonefficiency considerations or, at most, consider them separately, outside of the economic model. For instance, along with the possibilities of considering deontological considerations separately from CBA or by a different governmental branch, Matthew Adler and Eric Posner mention the possibility of a “superprocedure” through which both deontological and welfarist

Ethics 170–77. For an account of moral rights that is closely related to deontological constraints, see Judith Jarvis Thomson, The Realm of Rights (1990). 5. Kagan, supra note 4, at 78–80. Following Thomson’s terminology, we denote morally impermissible infringements of a constraint as “violations,” describe permissible infringements as those that “override a constraint” or “meet the threshold,” and use the term “infringement” to cover both. See Judith Jarvis Thomson, Some Ruminations on Rights, in Rights, Restitution, and Risk: Essays in Moral Theory 49, 51–52 (William Parent ed., 1986). 6. See infra pp. 41, 46 and 98–103. 7. See infra pp. 48, 78. 8. See, e.g., Jody S. Kraus, Reconciling Autonomy and Efficiency in Contract Law: The Vertical Integration Strategy, 11 Phil. Issues, Supp. to Nous 420, 431–34 (2001).

introduction

considerations would be considered together. Regarding this possibility, they write: “We suppose that that is a theoretical possibility—but we have absolutely no idea what the superprocedure would consist in.”9 Can the normative flaws of standard economic analysis be rectified without relinquishing its methodological advantages? Can deontological moral constraints and options be formalized so as to make their analysis more rigorous? This book examines the possibility of combining economic methodology and deontological morality through explicit and direct incorporation of moral constraints (and options) into economic CBA. It argues that such incorporation would improve economic analysis of law and economic analysis in general, not only as a normative theory but also as a descriptive and predictive tool, without considerably compromising its methodological rigor. At the same time, it maintains that deontologists and jurists who oppose consequentialism have been too hasty in disqualifying economic analysis as a fruitful analytical methodology.10 This book thus develops a detailed framework for incorporating threshold constraints (and options) into CBA. It addresses the challenges facing the formulation of threshold functions and illustrates the construction and use of threshold functions to analyze several prominent legal issues. Deontologically constrained CBA is more complex than standard CBA. Yet we maintain that it is superior to its alternatives. It rectifies some of the normative flaws of conventional CBA without significantly compromising its methodological rigor. It also improves deontology by making the analysis of threshold constraints more precise and its policy implications potentially more determinate. For the deontologist, direct and explicit incorporation of deontological constraints into economic models is vital to make the analysis normatively acceptable. Less obviously, most of the consequentialist responses to the deontological critique, such as the move from act- to rule-consequentialism,

9. Matthew D. Adler & Eric A. Posner, New Foundations of Cost-Benefit Analysis 154–58 (2006). Clearly, when talking about scholarly analysis of law, an institutional separation between economic and deontological analysis is implausible. See also Posner, supra note 2, at 1157 (arguing that attempts to improve “the normative flavor” of CBA “by modifying or even rejecting the Kaldor-Hicks assumption gain less in normative plausibility than they lose in complication and uncertainty”); infra pp. 105–08. 10. Cf. Kraus, supra note 8. Kraus seeks to reconcile autonomy-based theories of contract law, which lack in determinacy and operationality, with economic analysis of law, whose normative foundations are deficient, through “vertical integration.” While Kraus’s proposal leaves the fine-grained analysis of contract doctrine to standard economic analysis, our proposal operationalizes deontology through its combination with economic methodology.

3

4


actually recognize or at least imply that sound CBA should incorporate constraints for practical or instrumental reasons. A consequentialist who embraces one of these responses may thus welcome our proposal without converting to deontology. Furthermore, since people’s behavior is commonly influenced by social norms and prevailing moral intuitions, any theory seeking to explain and predict people’s behavior should take threshold constraints and options into consideration.11 The same is true when explaining existing legal doctrines. Many legal norms fall in line with moderate deontology. For instance, under current constitutional law, statutes infringing upon “fundamental” rights are invalid unless the infringement is necessary to promote a compelling governmental interest.12 Thus, threshold constraints are essential to understanding and explaining existing legal doctrines. Our project is both ambitious and modest. It is ambitious in the sense that it proposes a general framework for analyzing and resolving a great variety of legal and policy issues. It also ambitiously strives to bridge the increasing gulf between economic analysis and other approaches to law and legal policy, and between economic analysis and deontological morality.13 The project is modest, however, for we do not claim that by using the proposed analytical framework, one can avoid difficult normative judgments. The proposed framework can enable one to more adequately grasp the pertinent issues and their interrelations, which may in turn facilitate sounder solutions. The book is divided into two parts. Part One, the first five chapters, lays the theoretical groundwork. It establishes the need for integrating deontological

11. Cf. Amitai Etzioni, The Moral Dimension: Toward a New Economics (1988); John Broome, Deontology and Economics, 8 Econ. & Phil. 269 (1992); Moral Markets: The Critical Role of Values in the Economy (Paul J. Zak ed., 2008); Walter J. Schultz, The Moral Conditions of Economic Efficiency (2001) (arguing that moral normative constraints are essential to facilitating a competitive market); Michael B. Dorff & Kimberly Kessler Ferzan, Is There a Method to the Madness? Why Creative and Counterintuitive Proposals Are Counterproductive, in Theoretical Foundations of Law and Economics 21 (Mark D. White ed., 2009). 12. See, e.g., infra pp. 76–77, 181, and 213–14. 13. On the fragmentation of current legal scholarship, see, e.g., Sanford Levinson & J.M. Balkin, Law, Music, and other Performing Arts, 139 U. Pa. L. Rev. 1597, 1653 (1991); Reinhard Zimmermann, Law Reviews: A Foray through a Strange World, 47 Emory L.J. 659, 688 (1998); Jerry Mashaw, Deconstructing Debate, Reconstructing Law, 87 Cornell L. Rev. 682, 682 (2002).

introduction

constraints (and options) with CBA, explains how such integration may be accomplished, and responds to plausible critiques of such integration. Chapter 1 presents welfare economics and its consequentialist nature. It first discusses the main features of positive and normative economic analysis and the meaning of “consequentialism.” It then analyzes the deontological critique of consequentialism and specifically consequentialism’s lack of constraints on attaining the best outcomes. It critically examines various attempts at defending consequentialism in general, and welfare economics in particular, against this critique. It concludes that all of the attempts to downplay, deny, or circumvent the deontological critique are doomed to failure. The responses that come closest to actually addressing the critique do so by endorsing deontological constraints (and options) on the factoral level. They imply that agents and policy-makers should only strive to attain the overall best outcomes subject to constraints and that agents sometimes have options not to attain the best outcomes. Chapter 2 discusses moderate (or threshold) deontology, its critique, and possible responses. Deontological theories prioritize values such as autonomy, human dignity, and keeping one’s promises over the promotion of good outcomes. In prohibiting the infliction of harm on other people, they resort to distinctions such as that between actively doing harm and merely allowing it, and between intending to do harm and merely foreseeing it. Moderate deontology holds that constraints (and options) have thresholds. A constraint may be overridden for the sake of furthering good outcomes or avoiding bad ones if enough good (or bad) is at stake; and an option not to promote the good may be overridden for the sake of attaining enough good or avoiding enough bad. This chapter analyzes the main critiques leveled against deontology in general and moderate deontology in particular. While recognizing the challenges faced by threshold deontology, we conclude that threshold constraints (and options) are an indispensable part of any acceptable factoral moral theory. At the same time, as it is conventionally portrayed, threshold deontology suffers from a lack of methodological rigor and precision. Chapter 3 addresses the argument that even if moderate deontology is the correct moral theory for individuals, consequentialism is the appropriate moral theory for legal policy-makers such as legislators, judges, and regulators (and for academic policy-analysts). It claims that this argument confuses, among other things, between constraints and options, and between the actor’s perspective and the perspective of an external reviewer. It ultimately rejects the alleged dichotomy between personal and public morality.

5

6


Chapter 4 discusses various substantive and methodological choices involved in formalizing deontological constraints and options. It proposes to determine the permissibility of any act or rule infringing a deontological constraint by means of a threshold function. It outlines the scope of this proposal by describing the role of threshold functions within a broader context that may involve conflicting constraints and require a choice between several deontologically permissible acts. It then delineates the general structure of threshold functions. To capture the essence of deontological constraints, threshold functions set the magnitude and shape of the threshold, as well as the types of costs and benefits that are taken into account in determining whether the act’s (or rule’s) net benefit meets the threshold. This chapter also briefly discusses the construction of threshold functions for deontological options. Chapter 5 tackles a number of methodological and principled objections to the incorporation of deontological constraints into economic analysis. It discusses the claims that such incorporation would adversely affect the normative neutrality of economic analysis; that monetizing deontological constraints faces insurmountable obstacles; that it would lead to setting too low thresholds for constraints; and that it is incompatible with the expressivist role of the law. We conclude that most of these objections are unpersuasive, and none is conclusive. Part Two of the book illustrates the implementation of deontologically constrained CBA in five legal contexts. Each chapter in this part critically discusses standard normative economic analysis of a socially important legal field or part of it. It then demonstrates how a constrained CBA of the pertinent issues might look, paying heed to extant legal norms and pertinent moral considerations. Like standard economic analysis of law, some of the illustrations use mathematical functions to present the pertinent variables and constants, while others suffice with verbal presentation. Chapter 6 presents a constrained CBA of measures taken in the fight against terrorism. It begins by characterizing and criticizing existing normative economic analysis of the fight on terror as reflecting a simplified ad hoc balancing. It then presents the central deontological constraints pertaining to the fight on terror. The bulk of this chapter discusses threshold functions that should be employed to determine the permissibility of such measures as targeted killings and torture. We discuss the factors affecting the evaluation of the act’s relevant net benefit and those determining the amount of net benefit required to justify an infringement. We argue that standard economic analysis fails to take into account critical distinctions. These include the

introduction

distinction between different goals of anti-terrorist measures (including retribution, deterrence, and preemption); the difference between harms the state inflicts through anti-terrorist measures and those resulting from unthwarted terrorist attacks; and the distinction between intended and unintended harm. Deontologically constrained CBA, which incorporates all of these distinctions, is shown to be methodologically workable and normatively superior. Chapter 7 discusses freedom of speech. It briefly describes current constitutional protection of this freedom and surveys its standard economic analysis. It then introduces the deontological constraint against curtailing free speech and analyzes in some detail the normative judgments involved in conducting a constrained CBA of speech regulation. As to calculating the benefit of speech regulation—which is tantamount to calculating the speech’s expected harm—it examines the desirability of excluding, or radically discounting, various types of harms, such as chronologically-remote and low-probability harms, small harms, harms brought about through rational persuasion, and mere offensiveness. Various ways of formalizing such excluders and combining them are examined. The chapter then analyzes the threshold that has to be met to justify speech regulation, including its shape, the setting of different thresholds for content-based and for content-neutral regulation, and different thresholds for different categories of speech. Chapter 8 examines discrimination in the marketplace and, more specifically, the legitimacy and appropriate scope of antidiscrimination legislation. Following a brief survey of current legal norms, it relies on positive economic analysis to explain possible motivations for discrimination. It then examines the efforts made to justify antidiscrimination legislation on standard efficiency grounds. These efforts correspond to the various attempts at defending consequentialism against the deontological critique discussed in chapter 1. It argues that these efforts are unsuccessful. Rather, the appropriate way to adequately capture the issues pertaining to market discrimination is to directly take into account a deontological constraint against harming people by discriminating against them. We analyze the deontological constraint against discrimination, examine the relationships between this constraint and distributive bases for antidiscrimination legislation, and demonstrate how current legal norms are best understood as resting on moderate deontology and embodying threshold constraints. Chapter 9 discusses contract law. Ordinarily, market transactions do not involve infringements of deontological constraints. For this reason (and since they usually involve money or easily monetized goods), standard CBA

7

8


is particularly apt for analyzing contract law. Nevertheless, we maintain that certain deontological constraints apply to contracting behavior and that combining deontological constraints with economic analysis of contract law may be fruitful. The chapter briefly surveys the deontological constraints pertinent to contract law and critically examines the standard economic response to them. It then demonstrates how deontological constraints may be integrated with economic analysis of the contracting stage, focusing on the doctrines of mistake and misrepresentation. Last, it highlights the differences between economic and deontological analyses of contract performance and breach and discusses the difficulties facing integration of deontological constraints with the economic analysis of contract remedies, given the current state of the pertinent theories. Finally, chapter 10 analyzes legal paternalism in its various manifestations. It first argues that contrary to prevailing notions, normative economics does not entail principled antipaternalism. In fact, the consequentialist nature of standard welfare economics—namely the absence of constraints on promoting good outcomes—opens the door to limiting people’s freedom with a view to promoting their own good. Economists ordinarily object to paternalism, but rather than pointing to the intrinsic value of freedom, they base their antipaternalistic stance on various secondary considerations, thus missing the real dilemmas inherent in paternalism. Adding deontological constraints to the analysis better captures the pertinent issues and provides more accurate yardsticks with which to evaluate paternalistic legal norms and explain existing ones. We construct formal models to evaluate the desirability of paternalistic legal norms from both economic and moderate deontological perspectives.

•

part one

Theory


•

one

The Consequentialist Nature of Economic Analysis

•

A. General

this chapter sets out the motivation for incorporating deontological con-

straints into economic analysis of law. It first introduces, in section B, the basic tenets of welfare economics, the normative branch of economic analysis. These include the assumption of rationality, preference satisfaction as the underlying theory of human welfare, a limited regard for distributive issues, and the consequentialist nature of welfare economics. Focusing on the last feature, section C describes consequentialism and its main critiques. These include the lack of constraints on maximizing good outcomes and the lack of options not to maximize the good. Section D discusses, in some detail, the attempts to defend consequentialism against the claim that it allows too much (lacks constraints), and section E surveys the responses to the objection that consequentialism demands too much (lacks options). We conclude that none of the responses to the deontological critique of consequentialism are satisfactory, and that in fact, most of them imply that adding constraints (and sometimes options) to economic analysis is warranted.

•

B. Normative Economics

Economics is conventionally divided into positive and normative fields. Positive economic analysis explains and predicts human behavior—and social outcomes—on the basis of rational choice theory, which assumes that people act “rationally.” This assumption includes both cognitive elements and motivational ones. Cognitive rationality (also known as thin rationality) entails that each person’s set of preferences conforms to formal requirements, such as transitivity and completeness, and that people make their decisions based upon all available relevant information, the exclusion of all irrelevant information, the correct use of the rules of probability, and 11

12


so forth.1 Motivational rationality (or thick rationality) further assumes that each person aims to maximize her own well-being. It excludes both true altruism (actions aimed solely at furthering the well-being of others) and idealism or commitment (actions undertaken out of a sense of duty, even when they conflict with a person’s self-interest and sympathetic preferences).2 Normative (or welfare) economics—which is the focus of our study—is a consequentialist theory, as it evaluates the desirability of acts, rules, policies, projects, etc., solely according to their outcomes. More specifically, normative economics is a welfarist theory. It contends that the only factor which ultimately determines the desirability of anything is its effect on individuals’ welfare. The theory of the good underlying normative economic analysis is preference satisfaction, according to which people’s well-being is enhanced to the extent that their desires are fulfilled. Like utilitarianism, it attributes equal weight to the welfare of every person. It focuses on incentives for future behavior.3 A central debate within normative economics has revolved around measures of welfare. The so-called “old-style” welfare economics—associated with Pigou’s 1920 book Economics of Welfare 4—was based on a rather vague notion of welfare. It did not take sides in the debate among utilitarian thinkers regarding the proper notion of utility that should be aggregated. The “new” welfare economics, in contrast, searches for welfare propositions that do not rest on direct, interpersonal comparisons of utility, happiness, or wellbeing. One can distinguish between three main schools of thought within this approach, all of which are based on a preference-satisfaction theory of welfare.

1. Definitions of thin rationality vary with regard to the elements they include in the list. See, e.g., Jon Elster, Sour Grapes: Studies in the Subversion of Rationality 1–15 (1983); Donald P. Green & Ian Shapiro, Pathologies of Rational Choice Theory: A Critique of Applications in Political Science 14–17, 19 (1994). Some economic models, especially those relating to the behavior of firms, assume unlimited capacity to gather and process information. See, e.g., Green & Shapiro, id. 2. Amartya K. Sen, Rational Fools: A Critique of the Behavioral Foundations of Economic Theory, 6 Phil. & Pub. Aff. 317 (1977); Daniel M. Hausman & Michael S. McPherson, Economic Analysis and Moral Philosophy 51–65 (1996). 3. On the normative foundations and main characteristics of welfare economics, see generally Hausman & McPherson, supra note 2. 4. Arthur Pigou, The Economics of Welfare (1920; revised 1924, 1929, 1932).

the consequentialist nature of economic analysis

The first school, identified with the Pareto Principle, avoids interpersonal comparisons altogether. According to this approach, an individual i ’s preferences are described by a utility function ui, such that ui(A) > ui(B) if and only if the individual i prefers A to B. Thus, the utility function is not a direct measure of well-being, happiness, or welfare, but only a description of the order in which the individual ranks different alternatives. According to the Pareto principle, state A is socially preferable (or Pareto superior) to state B if at least one person prefers A to B, and all others are either indifferent between the two states or prefer A to B. State A is a Pareto optimum if there is no other possible state that is socially preferable to A in the above sense. This principle is the basis of the two fundamental theorems of welfare economics. The first theorem states that under certain conditions, any competitive equilibrium satisfies the conditions for a Pareto optimum. The second theorem states that under other specific conditions, any Pareto optimum can be obtained as a competitive equilibrium after the agents’ initial endowments have been modified by suitable lump-sum transfers.5 Applying only the Pareto criterion and the two theorems of welfare economics, economists are handicapped in providing policy recommendations. In practically every state, there are some people who are worse off compared to another state, and thus no policy is Pareto superior to any other.6 This weakness spawned two different schools of thought within the “new” welfare economics. The more traditional approach uses a social welfare function (SWF) (also known as a Bergson-Samuelson welfare function).7 The SWF is written as follows: W = f (z1, z2 , ..., zn), where zi’s and f represent society’s ethical values, such that W is a numerical representation of the social welfare in a given state of the world. Although in principle, any variable related to a society’s well-being might be included in the SWF, economists have focused on SWF’s in which the arguments in the welfare function are utility indexes of each individual, i.e., W = f (u1, u2, ..., un). The SWF thus assigns a value to each possible distribution of individual utilities in society. Depending on its

5. See, e.g., Allan M. Feldman & Roberto Serrano, Welfare Economics and Social Choice Theory 51–75 (2006). 6. Guido Calabresi, The Pointlessness of Pareto: Carrying Coase Further, 100 Yale L.J. 1211 (1991); Michael B. Dorff, Why Welfare Depends on Fairness: A Reply to Kaplow and Shavell, 75 S. Cal. L. Rev. 847, 858–59 (2002). 7. Abram Bergson, A Reformulation of Certain Aspects of Welfare Economics, 52 Q. J. Econ. 314 (1938); Paul A. Samuelson, Foundations of Economic Analysis (1947).

13

14


form, the social welfare function will embody different normative judgments about distribution.8 The aim was that cardinal, interpersonally comparable utility functions would not be needed for SWF. However, it follows from Arrow’s Impossibility Theorem that SWF must be based on cardinal rather than ordinal utility functions, and interpersonal comparability is required.9 The only way to get a single outcome from a SWF whose arguments are ordinal utility indicators is to define it lexically, that is, to state that society prefers any increase in a certain person’s utility, however small, to any increase in another person’s utility, however large, and to have this hold true independently of the initial utility levels. From a normative perspective, however, this solution is utterly unacceptable.10 The alternative to this approach is the one identified with the Compensation Principle (also known as Kaldor-Hicks or Potential Pareto). It is an attempt to go beyond the Pareto Principle, while stopping short of utilitarianism, by measuring welfare in monetary terms rather than by happiness or well-being. This principle asserts that a state A is socially preferable to state B if those who prefer A to B gain, in monetary terms, from being in A rather than B, more than those who prefer B to A lose. Thus, a social change that does not meet the Pareto criterion should still be carried out if it is possible for the gainers from the change to compensate the losers and remain better off.11 In accordance with the assumption that people’s preferences are

n

1

8. For instance, consider the function W = (∑ ui 1− p )1− p where p is an inequality index. SWF in i =1

which p = 0 represents a utilitarian aggregation. As p approaches infinity, we get a Rawlsian “maximin” function, W = min(u1, u2 , ..., un). When p=1, the function is a sum of the n

logarithms of the utility indexes, W = ∑ lnui . i =1

9. According to Arrow’s Theorem, when there are three or more discrete options to choose from, it is impossible to formulate a social preference ordering that satisfies a certain set of reasonable criteria such as transitivity, independence of irrelevant alternatives, and nondictatorship. See Murray C. Kemp & Yew-Kwang Ng, On the Existence of Social Welfare Functions: Social Orderings and Social Decision Functions, 43 Economica 59 (1976), which base their proof on Kenneth J. Arrow, Social Choice and Individual Values (1951, rev. ed. 1963). See also Dennis C. Mueller, Public Choice III 563–68 (2003); Paul A. Samuelson, Reaffirming the Existence of “Reasonable” Bergson-Samuelson Social Welfare Functions, 44 Economica 81 (1977). 10. See, e.g., Douglas H. Blair & Robert A. Pollak, Collective Rationality and Dictatorship: The Scope of the Arrow Theorem, 21 J. Econ. Theory 186 (1979). 11. See, e.g., Richard A. Posner, Frontiers of Legal Theory 95–141 (2001).


complete (for any two alternative combinations of goods, services, or anything else (A and B), they prefer A to B, B to A, or are indifferent), it is assumed that every person can compare any entitlement to a sum of money. Preferences are therefore measured by people’s willingness to pay (WTP) for their satisfaction.12 In more technical terms, each person’s gain or loss from a shift from state A to state B is measured by a compensating variation (CV): the subtraction (in case of a gain) or the addition (in case of a loss) that is required to each person’s budget in the original state A, to ensure that he will be indifferent between the two states. The social decision should be based on an aggregation of all persons’ CV’s. This view is the basis of the procedure known as Cost-Benefit Analysis (CBA), which assumes that each person’s CV (or her WTP) is an adequate representation of the difference in the person’s utility as between the status quo and a given alternative state.13 Like utilitarianism, the Kaldor-Hicks criterion and CBA thus ordinarily assess the desirability of any act, rule, policy, or project according to its effect on the total welfare of all people. Each of these features of economic analysis is subject to criticism. To begin with, the very reliance on people’s preference satisfaction as the yardstick for human welfare is problematic. People often err as to what is good for them. They make choices on the basis of partial information, psychological biases, and faulty reasoning. A myriad of empirical and experimental studies have demonstrated that people’s preferences and choices significantly deviate from the standard assumptions of rational choice theory. While there is a considerable variance in the ways people perceive facts, process information, frame their decision tasks, and make choices, the deviations from the standard assumptions of cognitive and motivational

12. Cass R. Sunstein, Lives, Life-Years, and Willingness to Pay, 104 Colum. L. Rev. 205 (2004); Elizabeth Hoffman & Matthew L. Spitzer, Willingness to Pay vs. Willingness to Accept: Legal and Economic Implications, 71 Wash. U. L. Q. 59 (1993). 13. See Anthony E. Boardman et al., Cost-Benefit Analysis: Concepts and Practice (1996). The term “cost-benefit analysis” has various meanings on different levels of generality (Richard A. Posner, Cost-Benefit Analysis: Definition, Justification, and Comment on Conference Papers, 29 J. Legal Stud. 1153, 1153–56 (2000)). It may refer to a particular decision procedure used by regulatory agencies (see generally Cass R. Sunstein, The Cost-Benefit State (2002); Matthew D. Adler & Eric A. Posner, New Foundations of Cost-Benefit Analysis (2006)), or more generally to the normative criterion of Kaldor-Hicks efficiency. In this book, we use CBA in the latter meaning. For a closer look at the differences between well-being maximization, the Kaldor-Hicks criterion, and CBA, see Adler & Posner, id. at 9–24.

15

16


rationality are rather systematic.14 Moreover, people’s preferences are sometimes objectionable, reflecting jealousy, sadism, racism, and so forth. Finally, people whose economic or health conditions are very poor may have limited aspirations—so limited that even their fulfillment will not necessarily make their lives much better.15 In recent years, the mounting evidence of people’s bounded rationality has led some economists to endorse rational preference not merely as a proxy for actual ones but as a superior measure of well-being.16 This endorsement usually refers only to cognitive rationality, i.e., taking into account preferences that satisfy such conditions as transitivity, completeness, and dominance, even if actual preferences do not. Other analysts go one step further and also examine the motivational rationality of people. They are willing to discount or disregard not only choices based on misinformation or cognitive biases but also choices based on, for instance, sadistic and prejudiced preferences.17 Many economists insist, however, that the assumption of economic rationality approximates human behavior well enough, and that it is preferable to adhere to this assumption rather than encumber the analysis by making more realistic assumptions.18 More fundamentally, the theory of the good underlying economic analysis may be criticized for denying the intrinsic value of anything but human

14. See generally Robin Hogarth, Judgement and Choice (2d ed. 1987); Choices, Values, and Frames (Daniel Kahneman & Amos Tversky eds., 2000). See also infra pp. 319–20, 326. 15. Amartya Sen, Well-Being, Agency and Freedom: Dewey Lectures 1984, 82 J. Phil. 169, 191 (1985). 16. See, e.g., Louis Kaplow & Steven Shavell, Fairness Versus Welfare 410–13 (2002) (maintaining that it is only the satisfaction of rational preferences that enhances wellbeing). In fact, the standard assumption of mainstream economic models—that people behave rationally—sometimes leads to conclusions based on rational (rather than actual) preferences. See Eyal Zamir, The Efficiency of Paternalism, 84 Va. L. Rev. 229, 246–51 (1998); infra pp. 323–25. 17. See, e.g., John C. Harsanyi, Problems with Act-Utilitarianism and with Malevolent Preferences, in Hare and Critics: Essays on Moral Thinking 89, 96–98 (Douglas Seanor & N. Fotion eds., 1988); Adler & Posner, supra note 13, at 129–30; Howard F. Chang, A Liberal Theory of Social Welfare: Fairness, Utility, and the Pareto Principle, 110 Yale L.J. 173, 179–96 (2000). In fact, agencies engaging in CBA already screen preferences in this way. See Adler & Posner, supra note 13, at 129–30. Other analysts oppose this idea. See, e.g., Kaplow & Shavell, supra note 16, at 418–31. 18. See, e.g., John D. Hey & Chris Orme, Investigating Generalizations of Expected Utility Theory Using Experimental Data, 62 Econometrica 1291 (1994); Richard Posner, Rational Choice, Behavioral Economics, and the Law, 50 Stan. L. R. 1551 (1998).


welfare, including the natural environment and the well-being of animals.19 Its disregard for such notions as desert and fairness is likewise objectionable.20 Economic analysis may consider such nonwelfare values and notions instrumentally through their effect on human welfare, that is, on the satisfaction of human preferences—but arguably, such treatment misses their intrinsic importance. A distinctive advantage of economic analysis’s use of WTP as a measure of human well-being is the facilitation of mathematical economic models and formalization of normative issues. However, measuring welfare in monetary terms raises several concerns. To begin with, the assumption that everything a person might desire is commensurable with money is controversial.21 Even if the principled objection of incommensurability is rejected, the WTP criterion has been criticized for systematically favoring the rich. This is because the sum of money one is willing to pay for any entitlement depends on one’s wealth.22 This problem may be mitigated by shifting from WTP to WTA (Willingness to Accept), the minimum amount of money that one would accept to forgo any entitlement.23 This response is, however, incomplete. A person who desperately needs money is likely to be willing to forgo an entitlement for a lower sum of money than a wealthy person, notwithstanding the greater happiness or satisfaction she would derive from the entitlement. WTA is also much more susceptible to manipulations.24 The regressive effect of monetization of preferences through WTP is connected to a much more fundamental critique of the Kaldor-Hicks criterion, namely its disregard for distributive concerns.25 In its basic form, KaldorHicks efficiency only measures total welfare, attributing no intrinsic value to

19. See, e.g., David Degrazia, Taking Animals Seriously 36–74 (1996). 20. See generally Shelly Kagan, Normative Ethics 54–59, 309 (1998). 21. See generally infra pp. 108–10, 113–16. 22. See, e.g., Ronald W. Dworkin, Is Wealth a Value?, 9 J. Legal Stud. 191 (1980); Donald Hubin, The Moral Justification of Benefit/Cost Analysis, 10 Econ. & Phil. 169 (1994); Thomas F. Cotter, Legal Pragmatism and the Law and Economics Movement, 84 Geo. L.J. 2071, 2127 (1996). 23. Hoffman & Spitzer, supra note 12, at 85–87. 24. Daniel Kahneman, Jack Knetsch & Richard Thaler, Experimental Tests of the Endowment Effect and the Coase Theorem, 98 J. Pol. Econ. 1325, 1336 (1990) (suggesting that individuals habitually misstate WTA as greater than WTP because in many contexts they are rewarded for this misstatement). 25. See, e.g., Amartya Sen, The Discipline of Cost-Benefit Analysis, 29 J. Legal Stud. 931, 945–48 (2000).

17

18


its distribution among people. The Kaldor-Hicks criterion may favor redistribution of resources as a means to maximize aggregate social welfare due to the rule of decreasing marginal utility.26 But this is a mere means to maximize total welfare, and it does not refer to distribution of welfare as such. Many economic analyses deviate from this characteristic of CBA by taking into account distributive concerns and incorporating them into predictive and normative economic models.27 Each and every feature of standard economic analysis merits detailed discussion. This book, however, concentrates on the consequentialist aspect of welfare economics. Hence, the reminder of this chapter focuses on consequentialism, its critique, and possible responses to the critique. Other features of economic analysis, including the assumption of rationality, monetization, and distribution will be addressed inasmuch as they relate to the main discussion.28

•

C. Consequentialism and Its Critique

“Consequentialism” has a variety of meanings. It often refers to a normative theory which asserts that the only factor that ultimately determines the morality of an act or a rule (or anything else) is its consequences. In this book, we will use the term in a narrower sense to denote those theories that take into account the well-being of every person.29 This definition excludes, for example, ethical egoism—the view that an act is right if and only if it leads to the best outcomes for the actor. It is not, however, committed to any specific

26. R. Layard & A.A. Walters, Income Distribution, in Cost-Benefit Analysis 179, 192–97 (Richard Layard & Stephen Glaister eds., 2d ed. 1994). 27. See, e.g., Arnold C. Harberger, On the Use of Distributional Weights in Social Cost-Benefit Analysis, 86 J. Political Econ. s87 (1978); Alberto F. Alesina & Paola Giuliano, Preferences for Redistribution IZA Discussion Paper No. 4056. Available at http://ssrn.com/ abstract=1369802. 28. See infra pp. 313–47, 108–16, and 246–51, respectively. Additional critique is leveled against the manner in which CBA has been used in recent decades by regulatory agencies in the United States. See, e.g., Richard L. Revesz & Michael Livermore, Retaking Rationality: How Cost-Benefit Analysis Can Better Protect the Environment and Our Health 13–14 (2008). 29. On this conventional meaning of the term and on other definitions, see generally Samuel Scheffler, Introduction, in Consequentialism and Its Critics 1, 9 (Samuel Scheffler ed., 1988); Kagan, supra note 20, at 59–64; Philip Pettit, Consequentialism, in A Companion to Ethics 230 (Peter Singer ed., 1991).


theory concerning the goodness of outcomes in general or human well-being in particular. Utilitarianism and normative economics are the most famous consequentialist theories. While sharing these common features, consequentialist theories vary in many respects. First, they differ with regard to the underlying theory of the good and particularly of human well-being. Some versions of consequentialism, such as utilitarianism, posit that human well-being consists of enjoying positive mental states and avoiding negative ones; other theories, including welfare economics, consider the satisfaction of people’s actual or ideal preferences as decisive; and others contend that well-being consists of attaining certain objectively defined elements (such as good health, meaningful social relations, and knowledge).30 They also vary in the importance they attribute (if at all) to the distribution of well-being among members of society and to the well-being of future generations. The underlying theory of the good may or may not incorporate notions of equality, culpability, and desert (normative economics does not, in principle, incorporate any of these notions).31 Consequentialist theories also differ regarding the appropriate focal point of analysis (actions, rules, motivations, virtues, etc.).32 This book focuses on one consequentialist theory, namely normative economics, and more precisely, on the consequentialist character of normative economics. Consequentialism in general, and normative economics in particular, have been the subject of two major critiques aimed at the absence of restrictions on pursuing the overall good and at the requirement to prefer the overall good over one’s own interests. The first critique claims that consequentialism allows too much. Consequentialism imposes no restrictions on attaining the best outcomes, thus legitimizing and even requiring harming people, lying, and breaking promises to achieve desirable results.33

30. On different theories of the good, see generally James Griffin, Well-Being: Its Meaning, Measurement, and Moral Importance (1986); Kagan, supra note 20, at 25–69; Derek Parfit, Reasons and Persons 493–502 (1984); T.M. Scanlon, What We Owe to Each Other 108–43 (1998). 31. See generally Kagan, supra note 20, at 48–59, 308–09. On complex theories of the good and consequentialism, see also infra pp. 30–32. 32. See generally Morality, Rules, and Consequences: A Critical Reader (Brad Hooker et al. eds., 2000); Shelly Kagan, The Structure of Normative Ethics, 6 Phil. Perspectives 223, 236–42 (1992). 33. See, e.g., Bernard Williams, A Critique of Utilitarianism, in J.J.C. Smart & Bernard Williams, Utilitarianism—For and Against, 93–107 (1973); Thomas Nagel, The View from Nowhere 175–88 (1986).

19

20


Consequentialism does not recognize the moral rights of people over their body, labor, and talents.34 Thus, consequentialism arguably requires that we harvest the organs of one person to save the lives of two other people, that we torture the baby of a terrorist to force him to reveal information that may save lives, and so forth. According to the deontological critique, some values, such as autonomy and freedom, take precedence over the attainment of best results, and thus promoting the good (e.g., maximizing aggregate social welfare) should be subject to constraints. Deontological constraints usually include restrictions on violating fundamental rights (e.g., the rights to life and bodily integrity, human dignity, and freedom of speech), special obligations created by promises and agreements, and restrictions on lies and betrayal.35 The second critique is that consequentialism demands too much. Consequentialism presumably requires everyone to do what would maximize overall good outcomes, rather than further one’s own personal goals and interests or the interests of her loved ones or her community. It does not allow for agent-relative options. Thus, consequentialism arguably requires the well-off to contribute almost all of their money and dedicate a large portion of their time and energy, to promoting the well-being of the disadvantaged people around the world. It also requires self-sacrifice when the expected benefit to another person (who may be as well-off as the agent) is only slightly larger than the cost to the agent. This requirement of impartiality arguably conflicts with human nature and with the conception of people as separate entities. It also conflicts with one’s obligations toward family, friends, and community.36 In arguing that consequentialism both allows too much (lacks constraints) and demands too much (lacks options), the deontologist calls attention to the fact that consequentialism focuses on outcomes while deontology

34. Larry Alexander, The Jurisdiction of Justice: Two Conceptions of Political Morality, 41 San Diego L. Rev. 949, 952 (2004) (making this observation regarding conceptions of justice characterized by “unrestricted impartialism”). 35. See infra pp. 41–48. 36. See generally Samuel Scheffler, Boundaries and Allegiances: Problems of Justice and Responsibility in Liberal Thought 1–130 (2001) (discussing people’s “associative duties” and special responsibilities to their families and the social groups they belong to, including nations); Nagel, supra note 33, at 164–75; Tim Mulgan, The Demands of Consequentialism (2001); James Griffin, Incommensurability: What’s the Problem?, in Incommensurability, Incomparability, and Practical Reason 35, 40–48 (Ruth Chang ed., 1997).


focuses on the morality of actions. Consequentialism is not directly interested in the way a particular outcome has been brought about.37 In the context of constraints, there is, for example, a prevailing notion that there is a substantial difference between doing harm and merely allowing it; between intending harm and merely foreseeing it (or the related distinction between harming a person as a side effect of aiding other people and using a person as a means to aiding others); between harming for the sake of avoiding comparable harm befalling others and harming for the sake of increasing others’ well-being.38 Generally, the concern that consequentialism justifies terrible deeds is exacerbated when the theory of the good underlying a consequentialist normative theory is actual preference satisfaction and even more so if preferences are measured by people’s willingness to pay for their satisfaction. People sometimes have prejudiced, xenophobic, and even sadistic preferences, and their willingness to pay for satisfying their preferences directly depends on their affluence. At least theoretically (and most probably not only theoretically), these features of any consequentialist theory resting on actual preferences theory of well-being may lead to justifying “efficient” rapes, murders, and even genocide.39 The next section critically analyzes attempts to defend consequentialism against the critique that it allows too much, and the subsequent section will analyze attempts to answer the critique that it demands too much.

•

D. Responses to the Lack-of-Constraints Critique

Some analysts (including some economic analysts of law) are not overly concerned about the counterintuitive or even morally repugnant conclusions of unconstrained consequentialism,40 but most are. To avoid such

37. For a qualification of this statement, see infra pp. 30–32. 38. See infra pp. 41–46. 39. See, e.g., Williams, supra note 33, at 105; David Dolinko, The Perils of Welfare Economics, Book Review of Fairness Versus Welfare, by Louis Kaplow & Steven Shavell, 97 Nw. U. L. Rev. 351, 356–62 (2002); Harsanyi, supra note 17, at 96. 40. See, e.g., Samantha Brennan, Moral Lumps, 9 Ethical Theory & Moral Prac. 249, 259 (2006) (“Counter-intuitive results aren’t so bad if you are a consequentialist; they are your stock in trade”); Michael B. Dorff & Kimberly Kessler Ferzan, Is There a Method to the Madness? Why Creative and Counterintuitive Proposals Are Counterproductive, in

21

22


counterintuitive or abhorrent conclusions, consequentialists resort to various responses. This section briefly discusses five such responses: (1) demonstrating that, due to long-term and indirect effects, seemingly efficient arrangements that violate deontological constraints are not, in fact, efficient; (2) moving from act- to rule-consequentialism; (3) including “preferences” for deontological constraints within the preferences whose fulfillment constitutes people’s welfare; (4) taking into account people’s feelings of virtue when they act according to commonsense morality and feelings of remorse when they do not; and (5) replacing actual preferences with ideal preferences as the underlying theory of the good. We shall argue that none of these responses successfully addresses the deontological critique of consequentialism.

1. Long-Term and Indirect Effects A common strategy of consequentialists is to demonstrate that the counterintuitive conclusions attributed to consequentialism rest on flawed analysis that disregards or underestimates relevant outcomes. A fuller analysis— so the argument goes—reveals that seemingly efficient arrangements that violate deontological constraints are not, in fact, efficient, and are thus unjustified on purely consequentialist grounds. For example, a consequentialist may argue that killing one person in order to use her organs to save the lives of three other individuals is only seemingly desirable. If the victim were to be selected from among hospitalized patients, such a practice would, in the long run, deter people from being hospitalized lest their organs be harvested against their will. Such a fear might have a detrimental effect on the overall health of the population. Should physicians be allowed to choose the victim from the entire population, this may result in arbitrariness and cause general anxiety. Even if one could guarantee that the choice of the victim would be random and fair, such a scheme might dramatically reduce people’s incentive to look after their own health. In fact, assuming that sick people may not be suitable organ donors, such a general scheme may even create an incentive for people to become moderately sick.

Theoretical Foundations of Law and Economics 21, 21–26 (Mark D. White ed., 2009) (discussing legal economists’ suggestions to legalize baby-selling, racial discrimination, and insider trading; and noting that the “startling quality” of these suggestions may be their “primary virtue” from a “careerist perspective”).


Rather than prolonging people’s life span and enhancing their quality of life, this general program may, in fact, cause more deaths. This rough outline of one example suffices to demonstrate how a consequentialist may respond to at least part of the deontological critique.41 Another well-known example is the framing and executing of an innocent person to prevent serious riots in which hundreds of people will be killed.42 A consequentialist may claim that in the majority of cases, a thorough and sophisticated analysis of an act’s total consequences (direct and indirect, certain and probable) would lead to conclusions similar to those of moderate deontology.43 As for the remaining cases, the consequentialist may claim that they are very rare44 and may insist that in these cases, ordinary morality is simply wrong.45 The claim that a sophisticated analysis of the total set of consequences leads to conclusions that are akin to moderate deontology is more convincing in some contexts than in others. As Bernard Williams has noted, the hypotheses about possible effects that consequentialists often invoke in this debate are “so implausible that [they] would scarcely pass if it were not being used to deliver the respectable moral answer.”46 It is not at all clear

41. The example follows Harris and Singer’s exchange on the “survival lottery.” See John Harris, The Survival Lottery, 50 Philosophy 81 (1975); Peter Singer, Utility and the Survival Lottery, 52 Philosophy 218 (1977). 42. See H.J. McCloskey, An Examination of Restricted Utilitarianism, 66 Phil. Rev. 466, 468–69 (1957); H.J. McCloskey, A Non-Utilitarian Approach to Punishment, 8 Inquiry 249, 255–56 (1965); T.L.S. Sprigge, A Utilitarian Reply to Dr. McCloskey, 8 Inquiry 264 (1965); J.J.C. Smart, An Outline of a System of Utilitarian Ethics, in Utilitarianism—For and Against, supra note 33, at 69–71. A less dramatic example is the forced, uncompensated transfer of property from its owner to a person who values it more highly. While at first glance this may appear efficient, an economist may argue that it is not so “when one considers the incentive effects . . . of allowing such transfers and the alternative of forcing the rich person to transact with the poor person.” (Posner, supra note 13, at 1155). 43. For an account of moderate (or threshold) deontology, as opposed to an absolutist one, see infra pp. 46–48. 44. R.M. Hare, Ethical Theory and Utilitarianism, in Utilitarianism and Beyond 23, 27, 30, 31, 33 (Amartya Sen & Bernard Williams eds., 1982); Robert E. Goodin, Utilitarianism as a Public Philosophy 6 (1995). 45. See, e.g., Kagan, supra note 20, at 76–77; Samuel Scheffler, The Rejection of Consequentialism: A Philosophical Investigation of the Considerations Underlying Rival Moral Conceptions 83 (rev. ed. 1994); Robert E. Goodin, Political Theory and Public Policy 8–12 (1982). 46. Williams, supra note 33, at 100. Cf. Duncan Kennedy, Cost-Benefit Analysis of Entitlement Problems: A Critique, 33 Stan. L. Rev. 387, 398–400 (1981) (demonstrating how incorporation of indirect and remote effects (“externalities run wild”) may result in CBA reaching any desirable conclusion).

23

24


that the cases in which consequentialism leads to horrifying conclusions are rare, especially if consequentialism rests on preference satisfaction as its theory of the good.47 And even if these cases are rare, a moral theory that endorses abhorrent deeds even in rare cases is flawed. Finally, even in cases where long-term effects are likely to lead to a conclusion similar to that of threshold deontology, the deontologist would insist that our deeply held moral intuitions are much stronger than what consequentialist analysis indicates.48

2. Rule-Consequentialism The strategy we have just discussed for defending consequentialism does not transcend act-consequentialism. A different strategy is to move from act- to rule-consequentialism. Even if killing one person to save the lives of two may bring about overall good results, it may still be advisable to adopt an absolute, or almost absolute, prohibition against killing people. Such a rule may bring about overall better results even if in some particular cases it may result in suboptimal outcomes.49 By changing the focal point of a consequentialist theory from acts to rules, we may generate a set of rules that is not very different from commonsense morality.50

47. For a critique of the claim that consequentialism (or utilitarianism) is problematic only in very rare cases, see Amartya Sen, Rights and Agency, 11 Phil. & Pub. Aff. 3, 14 (1982); Chang, supra note 17, at 181. 48. See Scheffler, supra note 36, at 111; Judith Jarvis Thomson, The Realm of Rights 142–43, n.11 (1990). See also Michael S. Moore, Torture and the Balance of Evils, 23 Isr. L. Rev. 280, 295–96 (1989). 49. See, e.g., Richard B. Brandt, A Theory of the Good and the Right (1979). As mentioned above (supra note 32 and accompanying text), acts and rules are not the only focal points available to moral theories. Adopting other focal points such as motives may also narrow the gulf between consequentialism and ordinary morality. Consequentialism may also refrain from choosing any single primary focal point and instead have a comprehensive structure taking into account all focal points at the same time. See Amartya Sen, Utilitarianism and Welfarism, 76 J. Phil. 463, 464–67 (1979) (advocating a comprehensive structure); Shelly Kagan, Evaluative Focal Points, in Morality, Rules, and Consequences, supra note 32, at 134. While character traits and motives are clearly less appropriate objects of economic analysis and legal policy-making, this subsection’s observations on rule-consequentialism are at least partially applicable to other versions of consequentialism, including institution-consequentialism. 50. See, e.g., John C. Harsanyi, Morality and the Theory of Rational Behaviour, in Utilitarianism and Beyond, supra note 44, at 39, 56–60; Goodin, supra note 44, at 71.


As stipulated thus far, this argument is hardly persuasive. Assuming universal compliance with the rule, the command should not be, for example, “Thou shalt not kill,” but rather “Thou shalt not kill unless—all things considered—killing would enhance overall human welfare.” Put differently, under the assumption of universal compliance, rule-consequentialism collapses into act-consequentialism. In fact, the only appropriate rule is “Do whatever maximizes the best results.”51 Some of the critiques of rule-consequentialism, particularly its alleged collapse into act-consequentialism, may arguably be answered if rule-consequentialism is not conceived of as an indirect-actconsequentialism but rather as the moral code whose general internalization would produce the best outcomes.52 This version, however, is still subject to some of the traditional critiques of rule-consequentialism and raises difficulties of its own.53 The move to rule-consequentialism is more compelling if one replaces the ideal, unrealistic assumption of universal compliance with more realistic assumptions. A realistic theory strives to formulate the best set of rules given that some people will not understand, accept, or obey the rules (or simply will not have the time and energy necessary to conduct a comprehensive cost-benefit analysis of every action or inaction). A realistic normative theory takes into account people’s cognitive limitations, self-serving biases, etc. Under such assumptions, the set of rules that would maximize human well-being may be tantamount to threshold deontology (and possibly even to absolutist deontology).54 This is not to say, however, that realistic ruleconsequentialism is unproblematic. Inter alia, it faces considerable difficulties whenever the degree of actual compliance with the rules it advocates

51. See J.J.C. Smart, Extreme and Restricted Utilitarianism, in Theories of Ethics 171 (Philippa Foot ed., 1967). 52. See Brad Hooker, Ideal Code, Real World (2000). 53. See Tim Mulgan, Ruling out Rule Consequentialism, in Morality, Rules, and Consequences, supra note 32, at 212. 54. Allan Gibbard, Utilitarianism and Human Rights, 1 Soc. Phil. & Policy 92 (1984). A somewhat different strategy is to distinguish between intuitive morality, applying to daily actions and decisions that must be taken without much deliberation, and critical morality, applying to extraordinary conditions and requiring thorough, in-depth deliberation. While intuitive morality resembles moderate deontology (based on foundational utilitarianism), the critical normative theory should be act-utilitarianism. For this theory, see Hare, supra note 44.

25

26


deviates from the degree of compliance initially assumed, either downward or upward.55 While the move to rule-consequentialism—coupled with realistic assumptions about people’s behavior—reasonably answers the deontological critique in some instances, it is unsatisfactory in others. Often, institutional arrangements and procedural safeguards may dramatically reduce the risks of miscalculation and bias that serve to instrumentally justify constraints.56 In such cases, even a rule-consequentialist would not avoid conclusions that clearly contradict with ordinary morality. In fact, an avid consequentialist would not aspire to set rules that correspond to ordinary morality but would rather aim to free herself from its constraints.57 Generally, since rule consequentialism treats constraints merely as a means to achieving overall best outcomes, there is no guarantee that it would endorse a factoral moral theory that is identical to any form of moderate deontology. In any event, even if rule-consequentialism of some sort were a valid foundational moral theory, it would not imply that CBA should not include deontological constraints. On the contrary—the more sophisticated versions of rule-consequentialism are more acceptable precisely because they strive to explain and justify the role of deontological constraints on the factoral level. The distinction between factoral and foundational theories will be used throughout the book and thus merits some clarification.58 A factoral moral theory defines the factors that determine the morality of an act, their relative weight, and interaction. Such factors may include the costs and benefits of an act, whether it involves harming other people, and the relationships between the actor and the people affected by the act. In and of itself, a factoral theory neither explains nor justifies the relevance of the various factors and their interaction. This is the role of foundational theories. Interestingly, there is no necessary match between the kind of theory one

55. See generally Kagan, supra note 20, at 223–35; Kagan, supra note 49, at 137–44. Cf. Eric A. Posner & Adrian Vermeule, Terror in the Balance: Security, Liberty, and the Courts 40–41 (2007) (arguing that the attempt to base threshold deontology on secondorder consequentialist arguments is unsuccessful, as it rests on “muted institutional and empirical claims that are unpersuasive and amplified”). See also id. at 187–93. 56. See Jeffrey J. Rachlinksi, The Uncertain Psychological Case for Paternalism, 97 Nw. U. L. Rev. 1165, 1214–19 (2003) (describing how organizational settings can balance and correct mistakes that individuals might make). 57. See, e.g., Kaplow & Shavell, supra note 16, at 79–81, 381–402; Brennan, supra note 40, at 259. 58. In this regard, we follow Shelly Kagan’s taxonomy. See, e.g., Kagan, supra note 32, at 224–36.


adopts on the factoral level and the theory one favors on the foundational level. As we have just seen, foundational consequentialism may endorse threshold constraints on the factoral level.59 Inasmuch as rule-consequentialism endorses deontological constraints on the factoral level, CBA should take deontological constraints into account as well.

3. “Preferences for Constraints” A third response to the anticonsequentialist critique is available only to consequentialist theories—such as welfare economics—whose underlying theory of the good is preference satisfaction. The argument is straightforward: welfare economics strives to maximize the overall satisfaction of people’s preferences without passing judgment on their content or distinguishing between different objects of preferences. Thus, if people have preferences for “fairness,” including a preference for prohibitions against harming others, then these preferences should count in CBA.60 For some economists, the reluctant acknowledgment of the relevance of people’s preferences regarding “fairness” or “justice” seems little more than lip service.61 Others, particularly but not exclusively those interested in the economics of environmental protection, take such preferences more seriously and make space for them in their positive and normative analyses.62 Notably, Richard Zerbe has proposed a modified version of the Kaldor-Hicks criterion in which moral sentiments play an important role.63 Zerbe defines moral sentiments as other-regarding preferences that people are willing to

59. Cf. Stephen Darwall, Agent-Centered Restrictions From the Inside Out, in Deontology 112, 135 n.1 (Stephen Darwall ed., 2003) (explaining that, since under the definition he uses, “a theory is consequentialist if, and only if, it determines whether an act is right by whether the act maximizes good consequences,” rule-utilitarianism would count as a deontological theory). Cf. Mark Tunick, Efficiency, Practices, and the Moral Point of View: Limits of Economic Interpretation of Law, in Theoretical Foundations of Law and Economics, supra note 40, at 80–82. 60. Louis Kaplow and Steven Shavell have popularized this argument in the legal literature. See Kaplow & Shavell, supra note 16, at 431–36. 61. In fact, economists mostly ignore these preferences altogether or attempt to explain them away. Dorff & Ferzan, supra note 40, at 29–31. 62. See, e.g., Guido Calabresi & A. Douglas Melamed, Property Rules, Liability Rules, and Inalienability: One View of the Cathedral, 85 Harv. L. Rev. 1089, 1111–12 (1972). 63. Richard O. Zerbe Jr., Economic Efficiency in Law and Economics (2001).

27

28


pay for their satisfaction even when they are not directly affected by the relevant project, act, or rule. Such sentiments include preferences regarding the fairness and efficiency of rules and the ethical value attached to actions.64 To measure the “existence value” or “non-use” value of things such as environmental protection and wildlife conservation, economists have developed various techniques, including contingent valuation methods (CV or CVM) based on public polls.65 Such methods may in principle be used to elicit information about people’s attitudes to human rights and morality in general. As several scholars have pointed out, the “taste for fairness” or “preferences for constraints” argument is methodologically problematic and conceptually unsound. Methodologically, it is difficult to ascertain, quantify, and aggregate people’s disinterested preferences, whether for ecological values or for moral norms. Conceptually, a judgment regarding the morality of a rule (or an act), based on all relevant factors, must not be confused with one such factor, namely the rule’s (or act’s) effect on the well-being of the person making the judgment.66 There is a fundamental difference between preferences and normative judgments. Individuals have the final say on the content of their preferences, while judgments may be right or wrong. The soundness of a judgment depends on the validity of the arguments underlying it, not on the number of its supporters or the intensity of their support.67 Arguably, pooling together self-interested preferences and judgments (as well as other disinterested, other-regarding preferences) also gives rise to the concern of “double counting.”68 Rather than giving equal weight to the

64. Zerbe, id. at 24–25. 65. See generally Robert Sugden & Alan Williams, The Principles of Practical CostBenefit Analysis 148–67 (1978) (discussing techniques of inferring individual’s valuations of unmarketed goods). On CVM, see generally Using Surveys to Value Public Goods: The Contingent Valuation Method (Robert Cameron Mitchell & Richard T. Carson eds., 1989); A Primer on Nonmarket Valuation (Patricia A. Champ, Kevin J. Boyle & Thomas C. Brown eds., 2003). 66. Matthew D. Adler & Eric A. Posner, Implementing Cost-Benefit Analysis When Preferences are Distorted, 29 J. Legal Stud. 1105, 1112–13 (2000). 67. Lewis A. Kornhauser, Preference, Well-Being, and Morality in Social Decisions, 32 J. Legal Stud. 303, 316–22 (2003); Adler & Posner, supra note 66, at 1125–28; Matthew D. Adler & Eric A. Posner, Rethinking Cost-Benefit Analysis, 109 Yale L.J. 165, 243 (1999); Don Herzog, Externalities and Other Parasites (Book Review), 67 U. Chi. L. Rev. 895, 910–14 (2000). Cf. Jeremy Waldron, Locating Distribution, 32 J. Legal Stud. 277, 292 (2003) (describing the notion of taste for fairness as “disreputable maneuver”). 68. Ronald Dworkin, Taking Rights Seriously 234–37, 275–77 (1977); Chang, supra note 17, at 183–94.


well-being of every person, it encompasses the “preferences” people might have regarding the well-being of other individuals and groups. Moreover, if the political process is a mechanism for aggregating preferences and normative judgments, then almost by definition, any existing legal regime in any society is efficient.69 In fact, if enough people object to consequentialism, aggregating their “preferences” may lead to abandoning consequentialism altogether!70 Including “tastes for fairness” or preferences for deontological constraints within CBA is therefore methodologically problematic, conceptually wrong or tautological, and potentially self-defeating.

4. Feelings of Virtue and Remorse The previous argument in defense of consequentialism rests on disinterested “preferences.” A related but different argument rests on people’s self-regarding preferences or feelings. Most people derive pleasure from behaving morally. For example, people who keep a promise typically feel virtuous, while those who break a promise might feel remorse. The preference for this positive feeling is a component of people’s well-being, and, as Steven Shavell notes, “that in turn means that to maximize social welfare, promises should be kept somewhat more often than would be optimal if the measure of social welfare did not reflect this utility that individuals experience from keeping promises.”71 This conclusion sounds similar to adding a (rather weak) constraint to the cost-benefit analysis of keeping promises, but in fact, it is not.72

69. As Zerbe explicitly writes, this is a corollary of incorporating prevailing moral norms into CBA. See, e.g., Zerbe, supra note 63, at 239 (“a rule which adopts uncontentious norms is necessarily efficient”). 70. For the argument that taking external preferences into account is potentially selfdefeating, see Dworkin, supra note 68, at 235. On the match between ordinary morality and threshold deontology, see, e.g., Samantha Brennan, Thresholds for Rights, 33 Southern J. Phil. 143, 145 (1995); Scheffler, supra note 29, at 9; Shelly Kagan, The Limits of Morality 1–5 (1989). On the prevailing aversion to CBA, see, e.g., W. Kip Viscusi, Corporate Risk Analysis: A Reckless Act?, 52 Stan. L. Rev. 547 (2000); W. Kip Viscusi, The Challenge of Punitive Damages Mathematics, 30 J. Legal Stud. 313 (2001); Jonathan Baron & Ilana Ritov, Intuitions about Penalties and Compensation in the Context of Tort Law, 7 J. Risk & Uncertainty 17 (1993). 71. Steven Shavell, Foundations of Economic Analysis of Law 609 (2004). 72. In fairness to Shavell, it is not clear that he intends this argument to serve as a response to the deontological concerns. However, since the argument is made in the context of the morality of welfare economics and appears to present a practical answer to the deontological concern, it is worth discussing.

29

30


As already noted, the validity of a deontological constraint is independent of people’s feelings about it. It is an empirical question whether and to what extent people feel remorse when behaving contrary to ordinary morality. It stands to reason that some people do not experience this feeling. Deontological constraints, in contrast, are determined normatively. Thus, while people’s feelings of virtue and remorse, as well as their “preferences for constraints” (discussed earlier) may and should constitute part of positive analysis aimed at explaining and predicting human behavior,73 they do not address the deontological normative critique. Moreover, according to the remorse argument, the fact that someone feels bad after breaching a contract makes the breach less efficient—that is, less morally justified—than if breaching the contract made her happy. In contrast, from a deontological perspective (as well as ordinary morality), to the extent that the breacher’s feelings matter at all, their effect on the morality of the breach is the reverse. The fact that the breacher derives pleasure from harming the promisee makes the breach less—rather than more— justifiable.74 Finally, the remorse argument is likely to become circular.75 Once we all turn into (act-) consequentialists, we will no longer feel any remorse for breaking a promise or harming another person if such conduct enhances overall well-being.76

5. An Improved Theory of the Good: Ideal Preferences A well-known strategy for defending consequentialism is to adopt a complex conception of the good with a view toward imitating deontological constraints.

73. For a proposal to incorporate deontological moral judgments into economic models of personal decision-making, see Mark D. White, Can homo economicus follow Kant’s categorical imperative?, 33 J. Socio-Economics 89 (2004). 74. Steven Kelman, Cost Benefit Analysis: An Ethical Critique, 5 Regulation 33, 34 (1981). See also Alan Strudler, Incommensurable Goods, Rightful Lies, and the Wrongness of Fraud, 146 U. Pa. L. Rev. 1529, 1536 (1998) (making a similar claim regarding sadistic pleasures). 75. Williams, supra note 33, at 101; Kelman, supra note 74, at 34. 76. A possible reply to this argument is that there should be an “acoustic separation” between ordinary morality for “ordinary individuals” and consequentialism for academics and government decision-makers. See Kaplow & Shavell, supra note 16, at 381–402. But this (partial) reply is problematic in a democratic legal system. Another possible reply may be to change the focal point and move from act-consequentialism to virtue-consequentialism or character traits consequentialism.


As David McNaughton and Piers Rawling vividly described, “whenever an opponent of a particular consequentialist theory asserts that existing consequentialist theories have ignored some value, the consequentialist can meet the challenge by simply sucking the alleged value into what we might call the consequentialist vacuum cleaner.”77 For example, a theory of the good may take into consideration not only end results but also the acts leading up to them.78 Furthermore, as suggested by Amartya Sen, the goodness of outcomes (or possible world histories) to be maximized may include the realization of moral rights as well as considerations of actions, agency, control, and agent-relativity.79 Such a sophisticated theory of the good is unavailable to economic analysis, which rests on a rather simple theory of human well-being: preference satisfaction.80 Thus, the closest one can get to defending normative economics by improving its underlying theory of the good (without dramatically changing its basic features) is by moving from satisfaction of actual preferences—the traditional measure of well-being in economic analysis—to ideal ones. An ideal preferences theory of the good holds that a person’s well-being consists of the satisfaction of those preferences she would have had if only she calmly and rationally considered the issue, paying heed to all relevant information without any external pressure, bias, or prejudice.81 As indicated, the documented phenomena of systematic cognitive biases have led some economists to endorse cognitively rational preferences as a

77. David McNaughton & Piers Rawling, Agent-Relativity and the Doing Happening Distinction, 63 Phil. Stud. 167, 168–69 (1991). See also Martha C. Nussbaum, The Costs of Tragedy: Some Moral Limits of Cost-Benefit Analysis, 29 J. Legal Stud. 1005, 1029–30 (2000) (noting that some forms of consequentialism “are so elastic that they can include in the statement of consequences things that usually seem like fatal omissions in consequentialism”); Douglas W. Portmore, Consequentializing, 4 Phil. Compass 329 (2009). 78. See Joseph Raz, The Morality of Freedom 269–70 (1986); Sen, supra note 15, at 181–82; Daphna Lewinsohn-Zamir, In Defense of Redistribution through Private Law, 91 Minn. L. Rev. 326 (2006). 79. See, e.g., Sen, supra note 47; Sen, supra note 15. Cf. Posner & Vermeule, supra note 55, at 190–91 (mentioning this possibility in the context of anti-terrorist measures). 80. See, e.g., Hausman & McPherson, supra note 2, at 38–83; Allan M. Feldman, Welfare Economics and Social Choice Theory, 9–22 (1980); G. Peter Penz, Consumer Sovereignty and Human Interests 5-23 (1986); Griffin, supra note 30, at 10. 81. See generally, Griffin, supra note 30, at 10–17; Daphna Lewinsohn-Zamir, The Objectivity of Well-Being and the Objectives of Property Law, 78 N.Y.U. L. Rev. 1669, 1677–1700 (2003).

31

32


superior measure of well-being, and some go one step further and are willing to discount antisocial preferences.82 However, neither sophisticated theories of the good, nor the more modest move from actual to ideal preferences, satisfactorily answer the deontological critique.83 The move to ideal preferences excludes antisocial preferences and reduces the likelihood that purely consequentialist economic analysis would endorse truly deplorable conclusions, but it still does not answer many of the deontologist’s concerns. In contrast to deontology and in line with other consequentialist theories, CBA based on ideal preferences would support violating a constraint now to prevent two similar violations in the future (e.g., murdering one person to prevent two future murders). Furthermore, like the more sophisticated theories of the good, the technique of “laundering” morally objectionable preferences actually aims at imposing constraints on welfare maximization.84 It is unclear if an ideal preferences theory of the good can incorporate such distinctions as doing/allowing or intending/foreseeing, but even if it could, the result would be a recognition that threshold constraints are indispensable on the factoral level.

6. Summary Our discussion has demonstrated that attempts to downplay, deny, or circumvent the deontological critique of consequentialism for its lack of constraints, within a consequentialist factoral framework, are doomed to failure. Numerous additional demonstrations of this claim are provided throughout the second part of the book, where we critically discuss consequentialist attempts to overcome the deontological critique in such contexts as market discrimination, freedom of speech, and legal paternalism.85

82. See supra p. 16. 83. On the critique of such “hybrid consequentialism” from the point of view of nonconsequentialist moral theories, see Elizabeth Anderson, Value in Ethics and Economics 79–86 (1993); Kagan, supra note 20, at 216–18; Robert Nozick, Anarchy, State and Utopia 28–35 (1974); F.M. Kamm, Harming Some to Save Others, 57 Phil. Stud. 227, 251–56 (1989); Moore, supra note 48, at 289–90, 293–94. 84. Cf. Lewinsohn-Zamir, supra note 81, at 1699 (arguing that the laundering technique may assume that satisfying antisocial preferences might contribute to a person’s well-being but then disregards them in shaping social policy). 85. See infra pp. 234–40, 188–89, and 327–32, respectively.


The response that comes closest to addressing the deontological critique satisfactorily is rule-consequentialism. But this success comes at a price. To the extent that it answers the critique, it does so by endorsing deontological constraints on the factoral level.86 This response—as well as, to a considerable extent, the preference for constraints and ideal preferences responses— implies that policy analysis should incorporate deontological constraints.87

•

E. Responses to the Demandingness Objection General

Consequentialism requires everyone to maximize overall good outcomes rather than further one’s own interests or the interests of her family, friends, and community. Given that millions of people around the world are starving to death and many more are living in extreme poverty, a requirement to maximize human welfare implies that the affluent (and even the not-so-affluent) should stop spending their money on such luxuries as watching movies, reading fiction, touring foreign countries, and eating anything but very basic foods. Rather than spending their resources on such goods and services, consequentialism requires people to donate most of their money and devote most of their time to alleviate the suffering of the underprivileged, thereby maximizing good outcomes. More generally, whenever it is possible to produce more good to others than to oneself at the same cost, one should do so. Critics of consequentialism argue that such a requirement is unreasonably demanding. It fails to treat people as separate, autonomous entities who should be given the option of pursuing their own projects and preferring the well-being of their loved ones over the overall good.

86. This is particularly true of theories incorporating agent-relativity into consequentialism, namely theories that allow agents to evaluate the goodness of outcomes depending on their involvement in bringing those outcomes about (or, put differently, redefine the outcome as including such involvement). See Douglas W. Portmore, Can An ActConsequentialist Theory Be Agent Relative?, 38 Am. Phil. Quart. 363 (2001). For a critique of agent-relative consequentialism, see Mark Schroeder, Teleology, Agent-Relative Value, and ‘Good,’ 117 Ethics 265 (2007). We shall refer to such theories in the context of options as well. See infra pp. 36–37. 87. Consequentialists may recognize constraints for additional reasons. For example, at least some elements of well-being (e.g., people’s dignity and privacy) accrue only if they are respected by others as rights and not merely as means to maximize overall well-being. Thus, to maximize overall well-being, a consequentialist should recognize them as rights! See Philip Pettit, The Consequentialist Can Recognise Rights, 38 Phil. Quart. 42 (1988).

33

34


Compared to other consequentialist theories, this objection is less compelling in the context of economic analysis. A fundamental conjecture of normative economics is that under a relatively broad range of circumstances, the best way to maximize overall welfare is for each person to rationally pursue her own interests. Economists rarely, if ever, suggest that people should consciously aim at maximizing overall utility. Rather, they strive to correct market failures or mimic the outcomes of competitive markets. For this reason, this section surveys the responses to this cogent critique rather briefly. Long-Term and Indirect Effects Consequentialists tend to deny that consequentialism actually demands huge sacrifices. For example, they point out that the ability of any individual to considerably improve the position of starving people on the other side of the globe is very limited. In fact, one may not know what would make the lives of such people better or what would be the actual results of making a donation to an aid organization. The probability of effectively promoting the welfare of friends and family is much higher because people derive more benefit from interactions with people they care about, because we know better what the needs of our associates are and how to provide for them, and because it is often easier to promote their welfare.88 Moreover, if the affluent would devote most of their resources, time, and energy to helping the needy, instead of engaging in productive activities, the result might be disastrous— decreasing, rather than increasing, the total human welfare. These replies parallel the long-term and indirect effects defense of consequentialism against the critique that it lacks constraints.89 They are similarly unsuccessful. One need not have intimate acquaintance with starving people to know that clean water and basic food would dramatically enhance their well-being; and at any rate, one can make contributions to reputable aid organizations which possess the pertinent expertise. Even if it is true that a single person’s contribution cannot dramatically alter the fate of millions of poor people, it can surely enhance the well-being of some people to a much greater extent than the corresponding decrease in the

88. See Henry Sidgwick, The Methods of Ethics 430–39 (7th ed. 1907); Frank Jackson, Decision-theoretic Consequentialism and the Nearest and Dearest Objection, 101 Ethics 461 (1991). 89. See supra pp. 22–24.


donor’s welfare (or the welfare of her friends and family) due to the donation. Finally, it may be true that diverting the human and material resources that are currently employed in productive activities to charity would, at some point, decrease aggregate human welfare. The weight of this concern is, however, limited. The level at which any additional diversion of resources to charity would decrease global human well-being greatly exceeds both the current level of charitable contributions and the contributions mandated according to commonsense morality.90 Extremism At this point, consequentialists may concede that their theory is extremely demanding yet insist that it is not unreasonably or unacceptably demanding. The fact that most people fail to live up to the demands of a moral theory does not necessarily mean that something is wrong with the theory. What is possibly wrong is the behavior of people and their self-serving moral intuitions.91 Philosophers who subscribe to this view often argue that no principled departure from the basic requirement to promote the total good withstands theoretical scrutiny, and thus deontological options are not acceptable.92 It is doubtful, however, whether the starting point of these arguments, namely that there is a general requirement to promote the overall good regardless of the proximity between the agent and the needy people, withstands equally stringent scrutiny.93 At any rate, even if an extremely demanding moral theory is acceptable, an extremely demanding legal system—where

90. See Mulgan, supra note 36, at 31–37 (critically discussing consequentialists’ “strategy of denial” in response to the demandingness objection). 91. See, e.g., Peter Singer, Famine, Affluence, and Morality, 1 Phil. & Pub. Aff. 229 (1972); Peter Unger, Living High and Letting Die: Our Illusion of Innocence (1996); Kagan, supra note 70 (labeling the defender of the present position “extremist”). This argument is akin to the claim that in those cases where consequentialism leads to counterintuitive conclusions due to its nonrecognition of constraints, the problem does not lie with consequentialism but rather with the prevailing intuitions. See supra pp. 23, 26. 92. See, e.g., Unger, supra note 91, at 133–57; Kagan, supra note 70 (arguing that recognizing options necessitates the recognition of constraints, which in turn presupposes the doing/allowing, intending/foreseeing, or some such distinctions; and since none of these distinctions is defensible, nor are constraints or options). 93. See Garrett Cullity, International Aid and the Scope of Kindness, 105 Ethics 99, 104–05 (1994); Mulgan, supra note 36, at 29–31.

35

36


norms are centrally, and sometimes violently, enforced by the state—is much less acceptable. The present response is therefore considerably less available to defenders of CBA as a tool for legal policy-making. “Preferences for Options” Another attempt to accommodate options within a consequentialist theory is to take into account people’s desire to have options. Arguably, if people commonly prefer a normative system that includes options, then the maximization of preference satisfaction requires the recognition of options.94 This proposal is, however, methodologically problematic, conceptually unsound, and potentially self-defeating for the same reasons that the notion of preferences for constraints is problematic.95 Modified Theories of the Good More promising suggestions involve fundamental modifications in the theory of the good underlying consequentialism. Thus, it has been suggested that the theory of the good underlying consequentialism should depend on the position of the actor in relation to the resulting state of affairs.96 Accordingly, an evaluation of the goodness of state of affairs should not be impersonal but rather evaluator-relative. An agent should indeed always strive to promote the best state of affairs. However, states of affairs include the position of the agent vis-à-vis the outcome and her involvement in bringing it about.97 From the agent’s position, saving one’s own son rather than two other children may therefore produce more good. Even if the fact that some act would promote one’s own welfare does not constitute a moral reason to perform it; it may constitute a non-moral reason, and a consequentialist theory may admit of an option to always choose between what the balance of moral reasons support doing and what the balance of all reasons,

94. This argument is similar to the notion of “preferences for constraints” discussed supra pp. 27–29. 95. See id. 96. Cf. supra pp. 30–32 (discussing modifications of the theory of the good as a way to overcome the deontological critique of consequentialism). 97. See Amartya Sen, Evaluator Relativity and Consequential Evaluation, 12 Phil. & Pub. Aff. 113 (1983); Douglas W. Portmore, Position-Relative Consequentialism, Agent-Centered Options, and Supererogation, 113 Ethics 303 (2003).


moral and non-moral, supports doing.98 Such theories drastically narrow the gap between consequentialism and deontology.99 This “consequentialist vacuum cleaner” strategy redefines (universal) consequentialism by decoupling it from agent neutrality. It bridges between consequentialism and commonsense morality by adopting an agent-relative assessment of outcomes, which is commonly thought of as a distinctive characteristic of nonconsequentialist theories.100 Contrary to welfare economics, it also rejects welfarism as an appropriate theory of the good, instead taking into account the actions affecting human welfare as part of the outcome. So, while in some sense, agent-relative consequentialism is still a consequentialist theory, at least on the foundational level, it actually endorses deontological options for agents on the factoral level and mandates their consideration by legal policy-makers. Agent-relative act-consequentialism may thus provide a foundational justification for integrating options into CBA while retaining the methodological advantages of the latter as a decision-process mechanism. Collective, Rule-Consequentialism Other responses to the demandingness objection restructure one or more features of act-consequentialism in a similar fashion to the modifications proposed to overcome the critique that consequentialism lacks constraints. Most notably, it has been suggested that rule-consequentialism leads to more acceptable and intuitive conclusions than act-consequentialism with regards to options as well.101 In the present context, in addition to changing the focal point from acts to rules, it is suggested to shift from individual consequentialism to a collective one. Instead of asking how an individual should act, collective consequentialism asks how an individual should act assuming that everybody else will act similarly. Such a theory views welfare maximization as a collective endeavor. Thus, anyone who is able to promote the good by aiding others is required to do so but only to the extent that would have been optimal had everybody else acted in the same way.

98. Portmore, supra note 96. 99. Although, depending on its details, there may still be some differences remaining. See, e.g., Sen’s comparison between his theory of “broad consequentialism” and moderate deontology: Sen, supra note 97 at 128–32. 100. See, e.g., Kagan, supra note 70, at xi–xii, 8; Scheffler, supra note 45, at 1–3. 101. Cf. supra pp. 24–27.

37

38


Rule-consequentialism of this sort is considerably less demanding than act-consequentialism because in reality, almost nobody obeys even this less demanding rule, and thus much greater sacrifice is required of any individual to maximize the good. It is also fairer because one’s moral duty is not increased due to other people’s failure to perform their duties.102 In the context of constraints, we noted that rule-consequentialism is more persuasive as a realistic moral theory than as an ideal one. Assuming universal compliance, rule-consequentialism arguably collapses into actconsequentialism; under realistic assumptions, it may be superior to actconsequentialism. In the context of options, under the assumption of universal compliance, rule-consequentialism may similarly collapse into act-consequentialism. However, it does not fare much better as a realistic theory because it does not escape counterintuitive conclusions. Either the adopted rule would fail to maximize the overall good (in which case, adhering to it seems like rule worship), or it would fail to answer the demandingness objection (because, given the prevailing noncompliance, one should contribute most of her money and devote most of her time and energy to aiding the poor). Furthermore, even the more modest contribution that would be required according to a collective, rule-consequentialism (say, donating 10 percent of one’s income, rather than almost all of it) is far greater than what is expected of people under commonsense morality. The required contribution critically depends on the number of needy people and the extent of their destitution. It also implies that the less effective aid organizations are, the larger one’s sacrifice should be. Most important for our purpose, to the extent that ruleconsequentialism succeeds in bridging the gap between act-consequentialism and commonsense morality, it does so by providing consequentialist justifications for options on the factoral level. Other Modifications Rule-consequentialism is not the only possible modification of simple, actconsequentialism that has been offered in response to the demandingness objection. Another suggestion is to replace maximizing with satisficing, that is, instead of requiring people to bring about the best outcomes, requiring

102. For a critical discussion of rule-consequentialism’s response to the demandingness objection, see Mulgan, supra note 36, at 53–103. A different collective theory is proposed by Murphy. See Liam B. Murphy, Moral Demands in Nonideal Theory (2000). For a critical discussion of Murphy’s theory, see Mulgan, supra note 36, at 104–23.


them to bring about only “good enough” outcomes.103 Alternatively, one may construct a hybrid moral theory incorporating deontological options not to promote the good yet rejecting deontological constraints on promoting it.104 A more complex possibility would differentiate between needs and goals as different components of human well-being, apply act-consequentialism to the satisfaction of the former, rule-consequentialism to the satisfaction of the latter, and a hybrid theory to balance between the demands of these two moral realms.105 All of these theories have been the target of potent critiques, but discussing them here would lead us too far astray. Conclusion None of the consequentialist responses to the demandingness objection are satisfactory. To the extent that some of them get closer to answering the deontological critique, they do so by explicitly or implicitly endorsing deontological options on the factoral level. Even if such options ultimately rest on consequentialist justifications (on the foundational level), these concessions imply that agents may and should make room for options in their normative deliberation. Given the harshness of legal remedies and sanctions, options are all the more so warranted when it comes to the formulation of legal norms by public policy-makers. As indicated, unlike other consequentialist theories, normative economics is less susceptible to the critique of over-demandingness, because across a broad range of circumstances it assumes that the best way to maximize overall welfare is for each person to rationally pursue her own interests. For this reason, our discussion will focus on the integration of constraints, rather than options, into CBA. As will become clear, however, there are some

103. Michael Slote, Commonsense Morality and Consequentialism (1985). For a critique of Slote’s satisficing consequentialism, see, e.g., Philip Pettit, Slote on Consequentialism, 36 Phil. Quart. 399 (1986); Mulgan, supra note 36, at 127–44. 104. Scheffler, supra note 45. For critiques of Scheffler’s theory, see, e.g., Shelly Kagan, Does Consequentialism Demand too Much? 13 Phil. & Pub. Aff. 239 (1984); Stephen Darwall, The Rejection of Consequentialism (Book Review), 81 J. Phil. 220 (1984); Larry A. Alexander, Scheffler on the Independence of Agent-Central Prerogatives from AgentCentered Restrictions, 84 J. Phil. 277 (1987); Mulgan, supra note 36, at 145–66. 105. Mulgan, supra note 36, at 169–294. For critical reviews of Mulgan’s combined consequentialism theory, see Timothy Chappell, The Demands of Consequentialism, 111 Mind 891 (2002); Brad Hooker, The Demands of Consequentialism, by Tim Mulgan (Book Review), 78 Philosophy 289 (2003).

39

40


contexts in which fruitful analysis of legal issues requires us to pay heed to options as well.106

•

F. Conclusion

This chapter provided a brief, general account of economic analysis. Of the various normative and methodological tenets of welfare economics, it highlighted its consequentialist characteristic. We argued that consequentialism is not an acceptable moral theory due to its lack of constraints and lack of options. To the extent that the attempts to defend consequentialism in general, and economic analysis in particular, are successful, they concede that constraints and options should actually be part of normative and policy deliberation, including CBA. Before analyzing the plausibility of integrating constraints (and options) into CBA, one must make sure that there are no other aspects of deontology that make it unacceptable or unattractive. To balance the picture, the next chapter thus describes deontological morality in general and moderate deontology in particular, and presents the major objections to them.

106. See infra pp. 289–91.

•

two

Threshold Deontology and Its Critique

this chapter describes the basic tenets of deontological morality in general

and moderate deontology in particular. It then examines the major critiques leveled against them. The chapter aims to demonstrate that deontology is superior to consequentialism, at least on the factoral level or as a decision procedure.

•

A. Deontology

Deontological moral theories are typically more complex than consequentialist theories and more diverse. It is therefore difficult to state summarily even their basic features. They do, however, share a common denominator. All deontological theories view the goodness of outcomes as a morally relevant factor, but, unlike consequentialism, they do not consider it the only intrinsically important factor.1 Deontological theories prioritize such values as autonomy, human dignity, basic liberties, truth telling, fidelity, fair play, and keeping one’s promises over the promotion of good outcomes. They include constraints on attaining the best outcomes. At the same time, deontology allows people to (sometimes) prefer their own interests over those of others. People may, and sometimes should, pursue the interests of their family, friends, and community, even if such a pursuit conflicts with attaining the overall good. Deontological theories thus recognize agent-relative constraints (on promoting the good) and agent-relative options (not to promote the good).

1. See, e.g., John Rawls, A Theory of Justice 26 (rev. ed. 1999) (“deontological theories are defined as non-teleological ones, not as views that characterize the rightness of institutions and acts independently from their consequences. All ethical doctrines worth our attention take consequences into account in judging rightness.”); Shelly Kagan, Normative Ethics 60, 64, 70–78 (1998); Frances M. Kamm, Morality, Mortality, Vol. I: Death and Whom to Save From It 76 (1993).

41

42


Whereas consequentialism judges the morality of an action (or anything else) according to its outcomes, deontology focuses on the morality of the action itself and on relations between people.2 For instance, a consequentialist would maintain that, all else being equal, the fact that a person was tortured is a bad thing; and if two people were tortured, it is certainly worse. In contrast, a deontologist would hold that it is immoral for an agent to torture another person or to be in a torturer-victim relationship. For the deontologist, the fact that she tortures one person may well be worse than the fact that somewhere in the world, two people are being tortured by someone else. As ordinarily conceived, consequentialism hence adopts an impartial, agentneutral perspective, whereas deontology is distinctively agent-relative. The central deontological constraint is against harming other people. This constraint does not apply to each and every interest an individual may have. (Had it been so inclusive, we would be left with the Pareto principle, which practically blocks any move from the legal status quo.)3 The exact list of deontological constraints is debatable, but it usually includes restrictions on violating fundamental rights, such as the rights to life and bodily integrity, human dignity, and freedoms of religion and speech. It also includes special obligations created by promises and agreements and restrictions on lying and betrayal.4 There is additionally a “deontological requirement of fairness, of evenhandedness or equality in one’s treatment of people.”5 The question of how to establish a defensible list of constraints is part of the broader question of how a normative theory is to be established and defended, which lies beyond the scope of our discussion.6

2. Stephen Darwall, Introduction, in Deontology 1 (Stephen Darwall ed., 2003). 3. Guido Calabresi, The Pointlessness of Pareto: Carrying Coase Further, 100 Yale L.J. 1211 (1991); Michael B. Dorff, Why Welfare Depends on Fairness: A Reply to Kaplow and Shavell, 75 S. Cal. L. Rev. 847, 858–59 (2002). 4. See Kagan, supra note 1, at 84–94 (discussing the scope of the constraint against harming people), 106–52 (discussing lying, breaking promises, failing to meet special obligations, conventions, and duties to oneself); Christopher McMahon, The Paradox of Deontology, 20 Phil. & Pub. Aff. 350, 354–68 (1991) (defining deontological constraints through the notion of unfair treatment of others); Darwall, supra note 2 (providing a long list of relations which are intrinsically important for deontology). 5. Thomas Nagel, The View from Nowhere 176 (1986). On deontological notions of fairness, see also Kamm, supra note 1; Iwao Hirose, Aggregation and Numbers, 16 Utilitas 62 (2004). 6. On this question, see Norman Daniels, Justice and Justification (1996); Brad Hooker, Ideal Code, Real World 4–23 (2000); Kagan, supra note 1, at 4-6, 11–17; Rawls, supra note 1, at 15–19 (introducing the notion of “reflective equilibrium”).

threshold deontology and its critique

The notion of agent-relativity crucially implies that there must be a difference between a person’s duty to refrain from violating a constraint and her duty not to bring about or prevent other violations, even where such violations are the expected outcome of avoiding the current one. Otherwise, the prohibition to kill one person in order to save two would prohibit both killing the person and not killing her (thereby allowing the death of the two). Deontology therefore must resort to a distinction between actively violating a constraint and not preventing constraint violations by others, or to some such distinction.7 In the context of the constraint against harming people, deontology thus distinguishes between actively harming a person and not aiding her (often labeled the doing/allowing distinction).8 While doing harm is at least presumably immoral, allowing harm is not ordinarily so. This is not to say that a duty to aid others never exists. Deontology is less demanding than consequentialism, which requires one to promote overall good outcomes at all times and thus to assist other people whenever the costs to oneself is smaller than the benefits to the other.9 Deontology does require agents to aid others, at least when the loss or suffering experienced by those others is large enough, and the risks and costs to the agent do not exceed a certain threshold.10 What the doing/allowing distinction suggests is that the moral responsibility for actively harming people is greater than for failing to prevent harm, and thus the former is worse than the latter.11 It also implies that the duty to prevent harm befalling other people, in circumstances where it exists, is subject to the constraint against actively harming people.12

7. David Enoch, Intending, Foreseeing, and the State, 13 Legal Theory 69, 97–99 (2007). 8. A useful collection of studies of the doing/allowing distinction is Killing and Letting Die (Bonnie Steinbock & Alastair Norcross eds., 2d ed. 1994). For psychological studies substantiating the prevalence of this intuition, see, e.g., Ilana Ritov & Jonathan Baron, Reluctance to Vaccinate: Omission Bias and Ambiguity, in Behavioral Law and Economics 168 (Cass R. Sunstein ed., 2000). The doing/allowing distinction is sometimes reformulated as a distinction between interfering and not interfering with a person’s welfare. For a critical discussion of this suggestion, see Shelly Kagan, The Limits of Morality 92–101 (1989). 9. See supra pp. 19–21, 33–40. 10. See, e.g., Kagan, supra note 1, at 95–97. See also infra pp. 98–103. 11. Samuel Scheffler, Doing and Allowing, 114 Ethics 215 (2004). 12. Cf. Eric Blumenson, Killing in Good Conscience: What’s Wrong with Sunstein and Vermeule’s Lesser Evil Argument for Capital Punishment and Other Human Rights Violations?, 10 New Crim. L. Rev. 210, 222–23 (2007); infra pp. 60–62.

43

44


Another distinction deontologists often draw is between intending harm and merely foreseeing it. Intending harm is immoral even if the harm is merely allowed, while foreseeing harm is not necessarily immoral.13 In causing an intended harm, one aims at evil. The constraint against intending harm forbids not only harming a person as an end but also as a means to attaining another goal. Killing a person to inherit her money is an intended harming, even if the killer would have preferred that there were other ways to obtain the money. Using a person as a means violates the requirement to respect people as ends. A related distinction is thus drawn between harming a person as a side effect of aiding or saving other people and using a person as a means to aiding or saving others. Allegedly, from the perspective of the harmed person, suffering a certain loss (including loss of life) as a result of being used by another person, or as a mere side effect of another person’s otherwise praiseworthy act is equally damaging. But this is not necessarily so. The harmed person has an interest not only in her fate but also in her status, in what permissibly may be done to her.14 The means/side-effect distinction is often discussed in reference to the “trolley problem.”15 Suppose that an uncontrolled trolley is hurtling down a track. Directly in its path stand five people who cannot escape and will be killed by the runaway trolley. An agent can flip a switch, diverting the trolley to another track, where it will kill a single individual. Should the agent flip the switch? Alternatively, suppose that the only way the agent can save the five people is by pushing another individual onto the track, blocking the trolley, and killing that individual. Should the agent push the other individual? Most people find diverting the trolley morally permitted, perhaps even required, while pushing the individual morally forbidden.16 Deontologists ground the difference in the distinction between killing as a mere side effect (in the diverting scenario) and killing as a means (in the pushing scenario).

13. On this distinction, see Kagan, supra note 8, at 128–82; Nagel, supra note 5, at 179–80; Jonathan Bennett, The Act Itself 194–225 (1995). 14. Adil Ahmed Haque, Torture, Terror, and the Inversion of Moral Principle, 10 New Crim. L. Rev. 613, 634 (2007). 15. See Philippa Foot, The Problem of Abortion and the Doctrine of the Double Effect, in Virtues and vices and other essays in moral philosophy 19, 23 (1978); Judith Jarvis Thomson, The Trolley Problem, 94 Yale L.J. 1395 (1985); F.M. Kamm, Harming Some to Save Others, 57 Phil. Stud. 227 (1989); Alison McIntyre, Doing Away with Double Effect, 111 Ethics 219 (2001). 16. See John Mikhail, Elements of Moral Cognition (forthcoming 2010).


The death of the single person on the other track is a mere side effect because it does not provide the actor with a reason for her behavior or an explanation for it.17 The intending/foreseeing and means/side-effect distinctions may rest on causality tests or on the agent’s motivation. According to the former, one should determine whether inflicting the harm is, objectively speaking, a means to attaining the agent’s goal. Thus, in the first trolley scenario, killing the single person on the other track is not part of the causal chain leading to saving the five, whereas pushing the individual in the second scenario is. According to the latter criterion, the decisive factor is the agent’s mental attitude to the harm. Suppose that the agent knows that the single person on the other track is someone she wishes were dead, and that is her motivation for flipping the switch. According to the motivation test, flipping the switch—which would have otherwise been permissible or even obligatory— is morally wrong and impermissible. Those who favor the causality test may plausibly argue that the question of the morality of the act can be separated from the righteousness of the actor.18 An agent may conceivably fail to do the morally right thing blamelessly or do the morally right thing and still be considered blameworthy. The doing/allowing and the intending/foreseeing distinctions usually overlap; and the former is sometimes rationalized as a proxy for the latter. In most cases of merely allowing harm, the harm is an unintended outcome of people’s inaction, and thus the two distinctions coincide.19 The two distinctions do, however, yield different conclusions when one intentionally allows harm (thus violating only the constraint against intending harm), and when one actively harm another as a mere side effect of attaining another goal (thus violating only the constraint against doing harm). An example of the first scenario would be not treating a terminally ill, suffering patient, with

17. See Robert Cryer & A.P. Simester, Iraq and the Use of Force: Do the Side Effects Justify the Means?, 7 Theoretical Inq. L. 9, 32–33 (2006). 18. For different positions in this debate and for further analyses of the intricacies of the intending/foreseeing distinction (and references to many more discussions), see Judith Jarvis Thomson, Physician Assisted Suicides: Two Moral Arguments, 109 Ethics 497, 514–16 (1999); Enoch, supra note 7, at 79–81; William J. Fitzpatrick, Acts, Intentions, and Moral Permissibility: In Defence of the doctrine of Double Effect, 63 Analysis 317 (2003); Joseph Shaw, Intentions and Trolleys, 56 Phil. Q. 63 (2006). 19. See also Frances Howard-Snyder, Doing vs. Allowing Harm, Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/entries/doing-allowing/#9 (last revised Sep. 21, 2007).

45

46


the intention of letting nature take its course. Diverting the trolley is an example of the second scenario. As will be demonstrated, sometimes both criteria bear on the permissibility of a certain behavior. These distinctions, as well as others to which deontologists resort,20 play a very minor role, if at all, in consequentialist analysis. Deontological moral theories are either absolutist or moderate. While absolutist deontology maintains that constraints must not be violated for any amount of good consequences, moderate deontology holds that constraints have thresholds. A constraint may be overridden for the sake of furthering good outcomes or avoiding bad ones if enough good or bad is at stake. For instance, even the constraint against actively/intentionally killing an innocent person may be justifiably infringed if such killing is the only way to save the lives of thousands of people.21 The thresholds that have to be met to justify the infringement of other constraints, such as those against lying or breaking one’s promise, are possibly much lower. Correspondingly, deontological options need not be absolute: when enough good or bad outcomes are involved, there is no longer an option not to further the good or avoid the bad. Moderate deontological theories may thus be described as pluralistic. In determining the amount of good/bad outcomes that may justify infringement of a constraint, a moderate deontologist may reasonably take into account both the doing/allowing and the intending/foreseeing distinctions. Thus, the threshold that has to be met to justify actively harming another person when the harm is intended is plausibly much higher than when it is a mere side effect. Referring once more to the trolley problem, it stands to reason that diverting the uncontrolled trolley to another track (that is, killing as a side effect) may be justified in order to attain a net saving of a small number of lives, whereas pushing a person onto the track to block the trolley (killing as a means), is only justified for the sake of saving very many people. Suppose now that a person who is stuck on the track is expected to block the runaway trolley, thereby saving the lives of other people. Arguably, not aiding this person to get off the track, thus allowing her to be killed as a means to saving the others, is permitted for the sake of saving less people than in the pushing scenario (involving an active killing).

20. For a discussion of seven different distinctions drawn by deontologists and their correspondence with three aspects of agents’ culpability, see Michael S. Moore, Torture and the Balance of Evils, 23 Isr. L. Rev. 280, 299–314 (1989). See also infra pp. 86–93. 21. See, e.g., Judith Jarvis Thomson, Some Ruminations on Rights, in Rights, Restitution, and risk 49 (William Parent ed., 1986); Samantha Brennan, Thresholds for Rights, 33 Southern J. Phil. 143 (1995); Kagan, supra note 1, at 78–80.


The recognition of thresholds helps illuminate the relationships between deontological constraints and options. Contrary to first appearances, it is sometimes permissible to infringe a constraint as a means to promoting good outcomes, even when promoting such outcomes is not a moral duty but a mere option. An agent may not be under a duty to rescue imperiled people because such a rescue operation would involve considerable costs to her. Nevertheless, if the agent chooses to incur those costs and rescue the people, the expected net benefit of such a rescue operation may be large enough to justify her breaking a promise or even inflicting some physical pain on another person. At the same time, if the agent chooses not to rescue the imperiled people (which, ex hypothesis, is perfectly legitimate), breaking the promise may be impermissible even if keeping it entails larger costs to the agent than those involved in the rescue operation.22 In considering fairness issues, such as whom to save from death when only some imperiled people can be rescued, and in calculating the net benefit of infringing behavior (as threshold deontology requires), deontology typically diverges from consequentialism in yet another respect. Consequentialism usually aggregates all benefits and costs indiscriminately. Deontology, in contrast, often excludes various costs and benefits from the moral calculus. A deontologist may plausibly believe that in choosing between course of action A, in which two people will be saved from death and a third person will be relieved from a mild headache, and course of action B, where two other people will be saved from death and a car will be salvaged, the prospects of relieving the headache and of salvaging the car are simply irrelevant. Similarly, deontological morality may judge that no amount of pecuniary loss may justify an active/intended killing, hence regarding such loss as morally irrelevant in this context. In the same vein, a deontologist may wish to distinguish between harming a person in order to prevent considerable harms befalling other people and harming a person in order to further improve other people’s well-being. Contrary to standard CBA, where forgone benefits are but one type of costs and forgone costs are simply benefits, the deontologist might wish to exclude or drastically discount the enhancement of well-being as justifying infringements of constraints. A variety of such excluders will be discussed later.23

22. Cf. Frances Myrna Kamm, Supererogation and Obligation, 82 J. Phil. 118 (1985). 23. See infra pp. 86–93, 147–49, 195–211, and 342–46.

47

48


More so than its rivals, factoral moderate deontology conforms to prevailing moral intuitions (commonsense morality).24 It may thus be directly defended as an intuitionist theory.25 Factoral deontology may, however, be grounded in other foundational theories. Prominent deontological foundational theories are Immanuel Kant’s universalizability and various brands of contractarianism. According to the former, the key to determining the valid moral rules is that they may be applied universally. Universal binding is not only a necessary condition for sound moral rules but also the ground for their validity.26 According to contractarian theories, since morality first and foremost governs interactions between people, the right moral system is the one to which everyone would reasonably agree. Such a moral system is binding by virtue of the fact that it could be reasonably agreed upon.27 A deontological factoral theory may conceivably rest on a consequentialist foundational theory as well.28 Possibly, the best results overall would be attained if people would always reason and behave as deontologists. This section provided a bird’s-eye view of deontological morality; the following sections examine the key critiques leveled against deontology as a factoral morality. Since our proposed analytical framework rests on moderate deontology, after discussing some criticism of deontology in general in section B, section C will focus on the objections to moderate deontology. This discussion, as well as the analysis in chapters 3 through 5 (dealing with deontology as a public morality, the construction of threshold functions, and possible objections to such construction, respectively), and the application of our proposed analytical framework to various legal issues in the ensuing chapters, will provide a fuller explication of deontology in general and moderate deontology in particular.

24. See, e.g., Brennan, supra note 21, at 145; Samuel Scheffler, Introduction, in Consequentialism and Its Critics 1, 9 (Samuel Scheffler ed., 1988); Kagan, supra note 8, at 1–5. 25. Darwall, supra note 2, at 2; Nancy (Ann) Davis, Contemporary Deontology, in A Companion to Ethics 205, 211–12 (Peter Singer ed., 1991). 26. Immanuel Kant, Groundwork for the Metaphysics of Morals (Mary Gregor trans. & ed., Cambridge Univ. Press, 1998) (1785) (introducing the categorical imperative). 27. Among contemporary contractarian theories, the most influential is Rawls, supra note 1. Rawls is primarily concerned with political philosophy, but his ideas are relevant to normative ethics as well. See David Richards, A Theory of Reasons for Action (1971). Another influential theory is offered by T.M. Scanlon, What We Owe to Each Other (1998). On contractarianism as a foundational moral theory, see generally Kagan, supra note 1, at 240–56. 28. Kagan, supra note 1, at 223–39.

threshold deontology and its critique •

B. Critique of Deontology in General

This section discusses three major critiques of deontological morality.29 One critique often leveled against deontological constraints—moderate and absolutist alike—is that they are irrational. Assuming that there is something intrinsically bad in harming a person or using her as a means to some end, it seems irrational to oppose such harming or such use when the outcome of not harming or not using the person is a greater amount of equally severe harming or using of other people. This critique focuses on the victims of harm, but it also applies to justifications of constraints that focus instead on the agent, the relations between the agent and the victim, and so forth.30 One may respond to the so-called “paradox of deontology” critique by contending that constraints are necessary to cope with people’s biases, shortsightedness, and cognitive limitations.31 The belief that by torturing a suspect, the authorities will successfully prevent a terrorist attack is often misguided, and it is therefore safer to prohibit such torture altogether (or almost altogether). This reply corresponds to the rule-consequentialist response to the deontological critique, namely combining constraints on the factoral level with rule-consequentialism on the foundational level.32 It thus calls for a parallel rejoinder, i.e., the difficulty to justify obeying the deontological constraint when it is certain that such obeisance would actually increase the overall violations of the constraint. Alternatively, one may reject the implicit consequentialist assumption that the desirability of actions is ultimately determined by their outcomes, or, in other words, to divorce the right from the good. Thus, a deontologist may claim that duties arising from special relationships (e.g., between parents and their children) as well as prohibitions against being in certain relationships (e.g., between a torturer and her victim) take precedence over

29. For additional challenges facing nonconsequentialist morality, see Larry Alexander, The Jurisdiction of Justice: Two Conceptions of Political Morality, 41 San Diego L. Rev. 949, 962–66 (2004). 30. Kagan, supra note 8, at 24–32; Samuel Scheffler, The Rejection of Consequentialism: A Philosophical Investigation of the Considerations Underlying Rival Moral Conceptions 80–114 (rev. ed. 1994). For the radical claim that the act/omission distinction is nothing but a psychological error, see Jonathan Baron, Morality and Rational Choice 99–120 (1993). 31. Kagan, supra note 8, at 32–39. 32. See supra pp. 24–27.

49

50


the duty to maximize the good.33 While consequentialism judges acts according to whether they bring about the best state of affairs, deontology judges acts according to whether the actors conduct themselves in ways that maintain their moral integrity.34 Another line of attack on moderate as well as absolutist deontology points to the great difficulties in systematically and generally characterizing the content of the constraint against harming people. Shelly Kagan, among others, has powerfully demonstrated that none of the conventional deontological distinctions—between doing harm and merely allowing it, between intending harm and merely foreseeing it, or between interference and noninterference with others’ welfare—provides a coherent and intuitively acceptable criterion for setting the constraint in all circumstances.35 While some philosophers believe it is nevertheless possible to come up with such a systematic criterion,36 others concede that such a criterion may not exist, yet maintain that the criterion’s absence is not fatal to deontology. It may be that different factors play different roles in different contexts.37 The last critique of deontology to be mentioned here is that any nonconsequentialist concern is incompatible with the (weak) Pareto principle and thus unacceptable. As Louis Kaplow and Steven Shavell have argued, while a consequentialist normative theory, determining the goodness of outcomes solely on the basis of their effect on people’s well-being, will always favor a

33. W.D. Ross, The Right and the Good 16–47 (Philip Straton-Lake ed., 2002) (1930); David McNaughton & Piers Rawling, On Defending Deontology, 11 Ratio 37, 41–44 (1998). 34. Stephen Darwall, Agent-Centered Restrictions From the Inside Out, in Deontology, supra note 2, at 112, 127–35. 35. Kagan, supra note 8, at 83–182. See also Bennett, supra note 13 (criticizing the very distinction between doing and allowing, as well as the normative significance of this distinction and the distinction between intending and foreseeing harm); Killing and Letting Die, supra note 8. 36. See, e.g., Kamm, supra note 15; Frances M. Kamm, Shelly Kagan’s The Limits of Morality, 51 Phil. & Phenomenological Res. 903 (1991) (reviewing Kagan, supra note 8). 37. See, e.g., James Griffin, The Limits of Morality by Shelly Kagan, 99 Mind 128, 129 (1990) (book review) (arguing that the moderate may do without “a simple, fully explanatory moral theory,” instead admitting that “morality is a messy, unsystematic affair, that justifications for moral norms are varied, incomplete, and not articulable in a way that will suit all cases”); F.M. Kamm, Morality, Mortality, Vol. II: Rights, Duties, and Status 49–60 (1996) (indicating that the same factor may have different moral significance in different contexts, possibly because the factors interact differently in each context); Samuel Freeman, Utilitarianism, Deontology, and the Priority of Right, 23 Phil. & Pub. Aff. 313, 349 (1994) (concluding that deontology relinquishes complete rational systematization for the sake of plurality of values and principles).


rule (or anything else) that makes everybody better off, a deontological theory may sometimes favor a rule that makes everybody worse off.38 There are several persuasive responses to this claim. To begin with, deontological theories do not necessarily conflict with the Pareto principle. A deontological theory may plausibly qualify any constraint or option such that it would not apply whenever everybody would be better off without this constraint or option.39 More fundamentally, the conflict between deontology and the Pareto principle is tautological: it merely restates the conflict between deontology and consequentialism. It may therefore be read as establishing the weakness of the Pareto principle.40 Finally, in the context of deontological constraints, Kaplow and Shavell’s argument holds only behind an imaginary contractarian veil of ignorance, for in real life, it is the removal of a constraint against harming people that necessarily violates the Pareto principle.41 Removing the constraint renders the harming of people obligatory whenever such harming produces overall good outcomes, despite the fact that the harmed people are worse off.

•

C. Critique of Threshold Deontology

Above and beyond the general critique leveled at deontological morality, moderate deontology is subject to additional criticism. This section addresses the allegations that moderate deontology is incoherent, disrespectful of human dignity, arbitrary, inappropriately aggregates harms to people, and leads to strange puzzles.

38. Louis Kaplow & Steven Shavell, Fairness Versus Welfare 52–58 (2002). 39. Howard F. Chang, A Liberal Theory of Social Welfare: Fairness, Utility, and the Pareto Principle, 110 Yale L.J. 173 (2000); Richard Craswell, Kaplow and Shavell on the Substance of Fairness, 32 J. Legal Stud. 245, 246–57 (2003); Lawrence B. Solum, Public Legal Reason, 92 Va. L. Rev. 1449, 1496–99 (2006). 40. Amartya Sen, The Impossibility of a Paretian Liberal, 78 J. Pol. Econ. 152 (1970) (referring to a special case of Kaplow & Shavell’s general argument); Mark D. White, Pareto, Consent, and Respect for Dignity: A Kantian Perspective, 67 Rev. Soc. Econ. 49 (2009). See also Jules L. Coleman, The Grounds of Welfare, 112 Yale L.J. 1523–24 (2003) (reviewing Kaplow & Shavell, supra note 38); David Dolinko, The Perils of Welfare Economics, Book Review of Fairness Versus Welfare, by Louis Kaplow & Steven Shavell, 97 Nw. U. L. Rev. 351, 363–64 (2002) (pointing to “blatant circularity” in Kaplow & Shavell’s argument); Dorff, supra note 3, at 860–61. 41. Cf. Jeremy Waldron, Locating Distribution, 32 J. Legal Stud. 277, 285 (2003) (explaining that Kaplow & Shavell’s argument is “refutation by a single case, and a hypothetical case at that”).

51

52


Incoherence Setting thresholds is arguably incoherent: either deontological constraints have primacy over the good, in which case they should never be violated, or they can be overridden by the goodness of results, in which case one becomes a consequentialist.42 The response to this critique is that it is quite coherent to maintain that (contrary to consequentialism), the goodness of outcomes is not the only factor, and that (contrary to absolutist deontology), constraints may be outweighed by enough good outcomes. Recognizing that there is more than one morally relevant factor inevitably implies that under different circumstances, some factors outweigh others and that all factors should be taken into account. Disrespectfulness Relatedly, some argue that the very formulation of rules that determine when it is permissible to kill or torture people is disrespectful of human dignity and thus incompatible with Kantian morality. Extreme emergencies may indeed compel one to do horrible things to prevent catastrophic outcomes. However, respect for people requires that such acts be “unprincipled, context-generated.” They “ought to be performed strictly as acts of necessity, not as acts governed by principles.”43 This argument is unpersuasive. For one thing, many infringements of deontological constraints, including lying, promise breaking, or causing mild physical pain, are deemed permissible under daily circumstances, not merely under extreme ones. Moreover, even if one focuses on such extreme measures like killing and torture, if no moral principles govern the behavior of agents in extreme cases, how would agents decide whether the circumstances they face are truly extreme? How can one judge, in retrospect, whether the infringement was morally justified?

42. See, e.g., Davis, supra note 25, at 215–16. See also Russell L. Christopher, Deterring Retributivism: The Injustice of “Just” Punishment, 96 Nw. U. L. Rev. 843, 877–80 (2002) (arguing that in permitting an innocent person to be punished once the threshold is met, threshold deontology is no better than consequentialism). 43. Alon Harel & Assaf Sharon, “Necessity Knows No Law”: On Extreme Cases and Uncodifiable Necessities, 60 U. Toronto L.J. (forthcoming). See also Charles Fried, Right and Wrong 10 (1978) (arguing that “the concept of the catastrophic is a distinct concept just because it identifies the extreme situations in which the usual categories of judgment (including the category of right and wrong) no longer apply”). A related argument is that, even if moderate deontology per se does not disrespect human dignity, monetizing threshold constraints and integrating them with CBA does. On this argument see infra pp. 108–16.


There may be institutional and “expressivist” reasons to refrain from explicitly authorizing the killing or torturing of innocent people by legislation;44 and there may be instrumental advantages to regarding some acts as morally taboo.45 However, once it is accepted that constraints may justifiably be infringed under certain circumstances, there is no escape from delineating these circumstances. Arbitrariness Moderate deontology is impugned for being arbitrary; it is impossible, some claim, to set the deontological threshold at any particular point in a nonarbitrary manner.46 This is arguably due to the incomparability of consequentialism and deontology: the former is goal-oriented, committed to maximizing good results, while the latter holds that some acts are intrinsically wrong. Similar allegations are made with regard to thresholds for deontological options.47 In reply, one can grant that while any cutoff point is indeed “arbitrary” in the sense that the threshold could have been fixed higher or lower, this is an inevitable feature of any pluralist normative theory incorporating more than one morally relevant factor.48 Seemingly arbitrary rules are often set by the legal system when the benefits of having a bright-line rule offset its obvious costs. At other times, institutional arrangements are established, giving policy-makers or individuals discretion to set the threshold within a rather broad range.

44. See Christopher Kutz, Torture, Necessity and Existential Politics, 95 Cal. L. Rev. 235, 256 (2007) (criticizing moderate deontology on institutional grounds, namely that “institutions and institutional actors tend to abuse the limits of their discretion”); See also infra pp. 117–22. 45. Cf. Moore, supra note 20, at 329–31 (conceding the psychological danger involved in allowing thresholds for constraints but denying that this is a compelling argument against moderate deontology). 46. Anthony Ellis, Deontology, Incommensurability and the Arbitrary, 52 Phil. & Phenomenological Res. 855 (1992); Larry Alexander, Deontology at the Threshold, 37 San Diego L. Rev. 893, 905–10 (2000); Russell L. Christopher, The Prosecutor’s Dilemma: Bargains and Punishments, 72 Fordham L. Rev. 93, 154 (2003). See also Eric A. Posner & Adrian Vermeule, Terror in the Balance: Security, Liberty, and the Courts 188 (2007). 47. Larry Alexander, Affirmative Duties and the Limits of Self-Sacrifice, 15 Law & Phil. 65 (1996). 48. Kagan, supra note 1, at 80–81. See also Moore, supra note 20, at 332. On incomparability, see infra pp. 112–13.

53

54


Inappropriate Aggregation Another difficulty facing threshold deontology is that it arguably requires a morally inappropriate aggregation of harms to people. Judging that it is morally permissible to kill one innocent person to save the lives of a large number of people assumes that human lives may be added up. However, since we are not concerned with loss of objects, but with loss to people, and since for every person the loss of life, liberty, or anything comparable is equally painful, there is arguably no reason to prefer the lives of the many to the life of the one. As John Taurek has famously argued, one should decide whether to let the one or the many die by tossing a coin, thereby giving all people an equal chance of survival.49 Most philosophers reject this counterintuitive argument.50 Some argue that equal respect for human beings means that tossing a coin is appropriate when there is the same number of people in each group, as each person’s claim to be rescued is balanced against the conflicting claim of a person in the other group. However, when one group is larger than the other, equal respect for the marginal or additional people in the larger group (whose claims are not balanced or “neutralized” by conflicting ones) requires saving the larger group.51 Others claim that equal respect for people justifies neither a decision mechanism by which all groups stand a similar chance to be rescued regardless of their size nor a blanket preference for rescuing the largest group. Rather, only a weighted lottery, whereby each group’s chances are proportionate to its size, truly respects each and every person in all the groups, even if her presence neither creates a tie nor breaks it.52 Yet others deny that deontology is incompatible with the aggregation of human lives. Even if losing one’s life is a loss to that person, it is also a loss of a person; since

49. John M. Taurek, Should the Numbers Count?, 6 Phil. & Pub. Aff. 293 (1977). 50. See, e.g., Derek Parfit, Innumerate Ethics, 7 Phil. & Pub. Aff. 285 (1978); Gregory S. Kavka, The Numbers Should Count, 36 Phil. Stud. 285 (1979); James A. Montmarquet, On Doing Good: The Right and the Wrong Way, 79 J. Phil. 439 (1982). 51. See, e.g., Kamm, supra note 1, at 75–143; Scanlon, supra note 27, at 229–41; Rahul Kumar, Contractualism on Saving the Many, 61 Analysis 165 (2001). For a critique of this position, see, e.g., David Wasserman & Alan Strudler, Can a Nonconsequentialist Count Lives?, 31 Phil. & Pub. Aff. 71 (2003); Michael Otsuka, Saving Lives, Moral Theory, and the Claims of Individuals, 34 Phil. & Pub. Aff. 109 (2006). 52. See, e.g., John Broome, Selecting People Randomly, 95 Ethics 38 (1984); Jens Timmermann, The Individualist Lottery: How People Count, But Not Their Numbers, 64 Analysis 106 (2004). For a critique of this view, see, e.g., Iwao Hirose, Weighted Lotteries in Life and Death Cases, 20 Ratio 45 (2007); Wasserman & Strudler, supra note 51, at 85.


people are valuable, one should save more people rather than less (at least sometimes).53 Taurek’s counterintuitive argument is therefore theoretically problematic, at any rate under some plausible deontological theories. It is worth noting, however, that even accepting the “innumerability” argument does not deal moderate deontology a fatal blow. First, this argument does not rule out threshold deontology tout court. Accepting this argument would at most rule out killing one person to save the lives of a large number of other people (and possibly any other infringement of a constraint aimed at protecting comparable interests), but it would not rule out, for example, breaking a promise or causing mild pain to one person in order to save the life of another.54 Second, the counterintuitive implications of Taurek’s argument are troubling not only for moderate deontologists, but also—and primarily— for absolutists. The argument implies that when facing a choice between saving one person and saving a group of five, an agent must not opt for the latter but rather toss a coin. Moderate deontology can actually mitigate this counterintuitive conclusion by conceding that tossing a coin (or conducting a weighted lottery) is the right procedure as long as the difference between the group sizes is not too large, yet insisting that when the difference exceeds a certain threshold (for instance, when facing a choice between saving one person and saving a whole city), it is permissible to save the larger group.55 Puzzles Finally, the discontinuity brought about by setting thresholds arguably leads to strange moral puzzles. Suppose, for example, that it is morally correct to torture a terrorist’s mother (thereby putting pressure on the terrorist to reveal where he planted a bomb) to save the lives of x people but not to save the lives of x–1 people. Must the torture stop the moment the terrorist (or anybody else) reduces the number of threatened people from x to x–1? And if only x–1 people are at risk, could the police justifiably lure one more person into danger, thus legitimizing the torture and saving the lives of all people?56

53. See, e.g., Hirose, supra note 5; John T. Sanders, Why the Numbers Should Sometimes Count, 17 Phil. & Pub. Aff. 3 (1988). 54. See also Judith Jarvis Thomson, The Realm of Rights 166–67 (1990). 55. See Hirose, supra note 5; Sanders, supra note 53, at 14. 56. On these and other puzzles, see Alexander, supra note 46, at 900–05.

55

56


These (somewhat gimmicky) examples do not seem dispositive within the legal context, where a myriad of normative, pragmatic, and institutional considerations either justify somewhat troubling, discontinuous outcomes, or lead to the setting of more or less vague standards to be applied on a case-bycase basis. The problem of transforming gradated, quantitative differences into discontinuous, qualitative outcomes would pervade the legal system even if it were purely consequentialist.

•

D. Concluding Remarks

While we recognize the challenges facing deontology in general and moderate deontology in particular, we conclude that moderate deontology is more attractive than consequentialism, at least on the factoral level. Threshold constraints and options are an indispensable part of any acceptable factoral moral theory. While in the final analysis, they do not necessarily preclude such things as silencing harmful expressions, paternalistically limiting people’s freedom, or breaking a promise, they force one to consider the truly relevant moral factors, including such intrinsically important values as autonomy, human dignity, and loyalty. The compatibility of moderate deontology with commonsense morality makes it particularly apposite to legal analysis and policy-making.57 Before proceeding to show how such constraints and options may be combined with economic methodology, the next chapter examines whether there is a difference in this regard between personal and public morality.

57. See infra pp. 77–78.

•

three

Private and Public Morality

•

A. General

it has been argued that even if moderate deontology is the right moral

theory for individuals, consequentialism is the appropriate moral theory for legal policy-makers, such as legislators, judges, and regulators, and for academic policy analysts. While the impersonal nature of consequentialism may be irreconcilable with the notion of people as separate, autonomous agents, impartiality and impersonality are desirable virtues for the state and for legal policy-makers, whose role is to advance the general good.1 Relatedly, it has been argued that the distinctions central to deontology, such as between actively doing harm and passively allowing it, and between intending harm and merely foreseeing it, are inapplicable in the public sphere. Whereas an individual may not be morally responsible for not helping people she does not know (or even people she knows but with whom she has no special relations), the state bears responsibility for the well-being of all people. For example, if by imposing overly lenient penalties, the state fails to deter offenders and protect their victims, it cannot evade responsibility for the suffering of the latter on the ground that it merely allowed this suffering and did not actively inflict it.2 As for the intending/foreseeing distinction— which deontologists often rely on to justify the infliction of harm that is a mere side effect of preventing greater harms from befalling other people— it is claimed that such a distinction cannot meaningfully be applied to the state. The mental states of decision-makers within public institutions, such as parliament members or members of a school board, are very often unclear, unobservable, and unverifiable; and even if a clear mental state

1. Robert E. Goodin, Utilitarianism as a Public Philosophy 60–77 (1995). 2. Cass R. Sunstein & Adrian Vermeule, Is Capital Punishment Morally Required? Acts, Omissions, and Life-Life Tradeoffs, 58 Stan. L. Rev. 703, 716–28 (2005).

57

58


may be verified and attributed to institutions, it lacks intrinsic normative significance.3 Though attractive, these arguments are ultimately unsuccessful. Some of them conflate and even confuse the distinctions between deontological constraints and deontological options, between morality and law, between individual and collective decision-making, and between the state’s role as an actor and its role as the provider of legal norms. Straightening out these confusions, this chapter demonstrates that while some features of public decision-making indeed render deontology less appealing in the public sphere, other characteristics actually make it more compelling in this sphere.

•

B. The Private/Public Distinction

The claim that consequentialism is the right moral theory for public decisionmaking, even if moderate deontology is appropriate for individuals, presupposes that the two spheres are distinguishable. Indeed, many decisions and actions are distinctively personal, while others are clearly public. A private person’s decision whether to keep her promise to join someone for lunch is personal, while Congress’s enactment of a law is obviously public. However, as has been repeatedly pointed out, the private/public distinction is vague and easily manipulated.4 In fact, there is no sharp dividing line but rather a continuum between private and public entities.5 This vagueness should make one wary of hinging the applicability of deontological constraints and options on this distinction. While this concern may be of limited importance for abstract moral analysis, it is more troubling when pragmatic, legal policymaking is concerned.

3. David Enoch, Intending, Foreseeing, and the State, 13 Legal Theory 69 (2007). 4. See, e.g., Duncan Kennedy, The Stages of the Decline of the Public/Private Distinction, 130 U. Pa. L. Rev. 1349 (1982); Ruth Gavison, Feminism and the Public/Private Distinction, 45 Stan. L. Rev. 1 (1992). 5. See, e.g., Brentwood Acad. v. Tenn. Secondary Sch. Athletic Ass’n, 531 U.S. 288, 302 (2001). In light of the growing role of private entities in performing governmental functions (see, e.g., Gillian E. Metzger, Privatization as Delegation, 103 Colum. L. Rev. 1367, 1369 (2003)), the validity and clarity of the private/public distinction is put into question. See, e.g., Jody Freeman, The Private Role in Public Governance, 75 N.Y.U. L. Rev. 543, 548 (2000) (arguing that there is “no purely private realm and no purely public one . . .[only] the set of negotiated relationships between the public and the private”).

private and public morality

Specifically, if the crucial distinction is between autonomous individuals who do not bear responsibility for the well-being of all people and public authorities that presumably shoulder such responsibility, then state governments, and even more so local authorities, raise the same agent-relative concerns as do individuals.6 It is quite clear that a state is not responsible for the well-being of the citizens and residents of other states to the same extent that it is responsible for the welfare of its own people. At the same time, corporate officers are expected to take into account the well-being of all of the corporation’s shareholders and employees, teachers are expected to pay heed to the success of all their students, and parents are expected to care for all their children. Similarly, if the crucial distinction is between decisions made and actions taken by a single person, on the one hand, and collective decisions and actions on the other, then once again there is a whole spectrum between the two extremes, as often private organizations consist of many people, and public decisions are made by a single person.7 The elusiveness of the private/public distinction renders the “conversion” from deontology in the private sphere to consequentialism in the public sphere rather problematic for legal policy-making; yet it does not rule out this conversion altogether. For one thing, many actions and decisions are clearly private or public. For another, the gradual move from the private to the public sphere may possibly be accompanied by a gradual move from moderate deontology to consequentialism, possibly by a gradual decrease of the magnitude of the thresholds constraints and threshold options (and a growing inclusiveness of benefits and costs taken into account in determining the permissibility of the action).8 The following discussion thus sets aside the imprecision of the private/public distinction and assumes that it is viable.

6. Larry Alexander, The Jurisdiction of Justice: Two Conceptions of Political Morality, 41 San Diego L. Rev. 949, 953 (2004). 7. Apparently, another difference between individuals and the state is that unlike the former, the latter can conduct a thorough CBA of every action or inaction. Thus, the objection that consequentialism is impracticable because it is impossible to know all the consequences of one’s actions and inactions is less powerful when it comes to the state. This argument is, however, misguided. First, states do not have unlimited resources to conduct CBA in each and every case. Second and more fundamentally, the justification for deontological constraints (and options) does not rest primarily on the impracticability of universal consequentialism but rather on the intrinsic immorality of actively/intentionally harming other people (and since deontologists do take consequences into account, they are not immune to this objection anyway). 8. On the size of deontological thresholds and the relevant types of benefits and costs, see infra pp. 93–96 and 86–93, respectively.

59

60

law, economics, and morality •

C. Doing and Allowing

As explained above, deontology focuses on the morality of actions and maintains that it is (at least sometimes) forbidden to violate a constraint even if the violation would have reduced the number of future constraint violations. To differentiate between a person’s duty to refrain from the current violation and her duty not to bring about the other violations (due to avoiding the current one), deontology must resort to a distinction between actively violating a constraint and merely allowing it, between intending harm and merely foreseeing it, or something of the sort.9 Challenging the relevance of these distinctions in the public sphere is therefore an effective way to challenge the appropriateness of deontology as a public morality. In this vein, Cass Sunstein and Adrian Vermeule have argued that the very distinction between doing and allowing is not intelligible when it comes to governments and that even if it is intelligible, it is morally irrelevant.10 Much of their argument focuses on the role of the state as the provider of legal norms and rests on the difficulty of attributing a mental state to the state— two issues we will soon address. The remaining points they make relate to autonomy-based and consequentialist justifications for the doing/allowing distinction, which they believe do not apply in the public sphere. According to the autonomy argument, people have a presumptive right to live their lives without being required to assist others. Such a requirement would greatly interfere with people’s liberty and should therefore be avoided. According to the consequentialist justification, “an all-things-considered assessment of consequences justifies the act/omission distinction, at least in most domains.”11 None of these justifications—so the argument goes—holds true with regard to the government. The first argument confuses deontological options and deontological constraints. Even if it is true that the government cannot shirk its duty to promote the well-being of all citizens (and hence, in this sense the doing/ allowing distinction is inapplicable or less appropriate), it does not follow that the government is not subject to deontological constraints (in which context the doing/allowing distinction is relevant). It is logically possible and normatively plausible that governments are morally required to actively

9. Enoch, supra note 3, at 97–99; supra pp. 43–46. 10. Sunstein & Vermeule, supra note 2, at 720–28, 737–39. 11. Id. at 725.


promote the well-being of all of their people and at the same time are subject to the moral duty to refrain from actively harming them.12 The basic justification for deontological options is the autonomy and separateness of agents; and this justification does not apply to the government, whose raison d’être is to safeguard and promote people’s welfare. At the same time, a basic justification for deontological constraints lies in the autonomy and human dignity of individuals who may be affected by an act. From that angle, people deserve to be treated as ends, and not as mere means, both by other individuals and by the state. The active infliction of an injury is disrespectful of the injured person’s dignity and autonomy in a way that the mere allowing of such injury is not.13 An analogy may be drawn in this respect between the government and a parent. Parents are responsible for promoting the welfare of their children even when it entails the taking of active measures. In that sense, parents may be said to have no options (or very limited ones) vis-à-vis their children. However, this does not mean that parents are not subject to constraints vis-à-vis their children. A parent’s duty to maximize the welfare of all of her children does not imply that she may kill one of them and harvest that child’s organs to save the lives of the others. In the same vein, the state’s duty to prevent harm from befalling its citizens does not imply that its active infliction of harm and its failure to prevent harm are morally equivalent.14 While the special relations between children and parents, between local

12. As regards the logical possibility, one may object that constraints necessarily entail options. If there are constraints but no options, then sometimes the government is required not to maximize the good (constraints), but there are no cases in which it is permissible not to maximize the good (no options). This looks impossible as a matter of the logic of “it is required that” and “it is permissible that.” The answer to this objection is that the denial of options for the government does not rule out the existence of one type of options: the option not to maximize the good when such maximization involves the violation of a deontological constraint. 13. The same is true regarding the intending/foreseeing distinction discussed below: It makes a difference if one is harmed inadvertently, as a regrettable side effect of attaining a legitimate goal, or purposefully, when the harm is intended as a goal in itself or as a means to attaining another goal. Even if the physical harm (including death) is identical in both cases, only the latter disrespects the injured person as an autonomous human being. See Adil Ahmed Haque, Torture, Terror, and the Inversion of Moral Principle, 10 New Crim. L. Rev. 613, 632–34 (2007) (“Victims have an interest not only in what happens to them but also in what permissibly may be done to them, not only in their fate but also in their status [as independent and inviolable persons]”). 14. Eric Blumenson, Killing in Good Conscience: What’s Wrong with Sunstein and Vermeule’s Lesser Evil Argument for Capital Punishment and Other Human Rights Violations?, 10 New Crim. L. Rev. 210, 222–26 (2007); Haque, id. at 631–32.

61

62


residents and the municipality, and between citizens and the state greatly reduce the availability of options to parents, municipalities, and states (vis-àvis their children, residents, and citizens, respectively), these relationships do not necessarily detract from the constraints against actively harming people. These constraints hold irrespective of such relationships. We then turn to Sunstein and Vermeule’s second argument, namely that the consequentialist, “all-things-considered” justification for the doing/allowing distinction applies only in the private sphere. Besides the confusion between options and constraints, this argument falls short for two additional reasons. First, it only applies to those theories that seek to ground deontology on the factoral level on foundational consequentialist theories, such as ruleconsequentialism.15 Our proposed integration of deontological constraints with economic analysis is compatible with such theories; thus to the extent that constraints on the factoral level rest on foundational consequentialism, the present argument is indeed relevant. However, it is safe to assume that most advocates of factoral deontology ground it on nonconsequentialist foundational theories, such as Kant’s universalizability requirement or some sort of contractarianism. The present argument would not appeal to these ethicists. Second and more important, to the extent that one predicates the prohibition on actively harming other people on consequentialist considerations, it seems that this constraint is more, rather than less, vital in the public sphere. Under realistic assumptions about possible errors and biases of government officials, the risks involved in allowing governmental bodies to violate people’s basic liberties, to treat people unevenly, or to lie are much greater than those involved in allowing private people to do the same. For instance, granting governmental agencies carte blanche to torture people suspected of withholding information about future terrorist activities, whenever it is believed that the total expected decrease in the well-being of the potential victims of those activities is greater than the expected total decrease of the well-being of the tortured, seems extremely dangerous even from a (rule-) consequentialist perspective.16 In addition, state violations of basic human rights are arguably more harmful than private violations

15. See Shelly Kagan, Normative Ethics 17–22, 189–94, 299–303 (1989); supra pp. 24–27. 16. See, e.g., Thomas P. Crocker, Torture, with Apologies, 86 Texas L. Rev. 569, 585–601 (2008); Bruce Ackerman, Before the Next Attack: Preserving Civil Liberties in an Age of Terrorism 6 (2006); David Cole, Enemy Aliens: Double Standards and Constitutional Freedoms in the War on Terrorism 228 (2003) (stating that governments overreact in times of crisis). For an opposite position, see Eric Posner & Adrian Vermeule, Terror in the Balance: Security, Liberty, and the Courts 187–93 (2007).


because they not only harm the individual person but also adversely affect the basic tenets of a liberal democracy.17 Both the autonomy-based justifications for deontological constraints (coupled with the basic distinction between deontological options and constraints) and rule-consequentialist considerations thus clearly undermine the argument that the doing/allowing distinction is inappropriate in the public sphere.

•

D. Intending and Foreseeing

Along with the attack on the doing/allowing distinction, scholars who argue that deontology is inappropriate as a public morality also target the relevance of the intending/foreseeing distinction in this sphere. The critique focuses on the difficulty of attributing such mental states as intention and foresight to collective bodies, the difficulty of observing and verifying those mental states, and the alleged absence of intrinsic normative significance to the mental states of organizations, including states.18 To assess these arguments, one should take into account several distinctions: between the actor’s perspective and the perspective of an external (often judicial) reviewer of the act, between collective and single decision-making, and between the provision of legal norms and the performance of particular acts. This section focuses on the first two distinctions, and the next section discusses in greater detail the third (whose significance exceeds the intending/foreseeing distinction). Decisions and actions in the public sphere are varied and multifaceted. While some of them lay down general norms, others—in fact, the overwhelming majority—pertain to concrete cases. While some decisions are made by collective bodies, probably most are made by single decision-makers. One question that should be asked with regard to any public decision (general or particular, collectively or singly made) is whether the normative principles guiding the decision should be deontological, including the intending/ foreseeing distinction. A different question is whether an external observer (in particular, a judge deciding on the legality or constitutionality of an action) can reliably determine the mental state of the decision-maker. Even if one answers both questions affirmatively, it may be asked whether such an

17. Blumenson, supra note 14, at 226–34. 18. Enoch, supra note 3; Cass R. Sunstein & Adrian Vermeule, Deterring Murder: A Reply, 58 Stan. L. Rev. 847, 849–52 (2005).

63

64


observable and verifiable mental state can be attributed to the pertinent governmental body, including the state itself, and whether it should be deemed morally significant. It seems useful to start our inquiry with the decision-maker’s own moral perspective. From this perspective, there seems to be no fundamental difference between decisions made in a personal capacity and decisions made in one’s public role. A person intentionally instructing another to shoot innocent civilians is violating a constraint whether she is a private person or an army officer. In fact, some of the best-known and oft-discussed examples in the consequentialism/deontology debate—such as the framing by the police and executing of an innocent person to prevent riots in which many people will be killed19 and the “survival lottery” scheme for random selection of forced organ donors out of the general population,20 pertain to decisions plausibly made by public officials. Furthermore, from the decision-maker’s perspective, it should make no difference whether she makes the decision on her own or participates in a collective decision process. Deontological morality requires her to shun and to vote against the intentional framing of an innocent person as a means to thwart deadly riots. Likewise, she is required to cast her vote against intentionally killing people to harvest their organs as a means to save others (moderate deontology requires one to oppose these acts at least as long as the deontological threshold is not met). A decision to do these things violates the constraint even if the decision is physically executed by another person who follows the decision-maker’s instructions. Shifting from the decision-maker’s own perspective to that of an observer assessing the morality of the decision may raise epistemological concerns regarding the intentions of the former. Identifying intentions is particularly difficult with regard to decisions made by collective bodies such as parliament.21

19. H.J. McCloskey, An Examination of Restricted Utilitarianism, 66 Phil. Rev. 466, 468–69 (1957); H.J. McCloskey, A Non-Utilitarian Approach to Punishment, 8 Inquiry 249, 255–56 (1965); T.L.S. Sprigge, A Utilitarian Reply to Dr. McCloskey, 8 Inquiry 264 (1965); J.J.C. Smart, An Outline of a System of Utilitarian Ethics, in J.J.C. Smart & Bernard Williams, Utilitarianism—For and Against 69–71 (1973). 20. John Harris, The Survival Lottery, 50 Philosophy 81 (1975); Peter Singer, Utility and the Survival Lottery, 52 Philosophy 218 (1977). 21. On these difficulties and the ways to overcome them, see Andrei Marmor, Interpretation and Legal Theory 122–26 (2d ed. 2005); Elena Kagan, Private Speech, Public Purpose: The Role of Governmental Motive in First Amendment Doctrine, 63 U. Chi. L. Rev. 413, 438–42 (1996); J. Morris Clark, Legislative Motivation and Fundamental Rights in Constitutional Law, 15 San Diego L. Rev. 953, 974–78 (1978).


Issues of observability and verifiability are even more troubling when the external observer is a legal decision-maker, such as a court assessing the legality of an action, who is constrained by various evidentiary rules. However, these difficulties characterize the legal process in general and are not necessarily connected to the private/public distinction. Courts regularly engage in exposing or attributing intentions even when decisions are made collectively, in both private and public spheres. In criminal law, torts, and even contract law, legal liability often hinges on the determination of subtle mental states of individuals, including individuals in their organizational role (such as corporate officials) and on the mental states attributed to the organization itself.22 A multitude of doctrines and techniques have been developed to deduce subjective mental states from objective facts and circumstances.23 In a similar fashion, the law can determine the mental state of decision-makers in the public sphere and attribute them to public bodies. Such determinations are imperfect and incomplete, whether in private or criminal law, or in the public sphere, but the difficulties are not insurmountable.24 Using various direct and indirect techniques for inferring intentions and other mental states (or objectifying them), courts have long invalidated administrative actions based on ulterior motives or aiming at extraneous purposes, governmental decisions disregarding relevant considerations,

22. See, e.g., 1 Kathleen F. Brickey, Corporate Criminal Liability: A Treatise on the Criminal Liability of Corporations, Their Officers and Agents 129–30 (2d ed. 1991) (“[I]t is now generally acknowledged that corporations . . . may be held liable for mens rea crimes, including crimes requiring specific intent”); Amanda Pinto & Martin Evans, Corporate Criminal Liability 39–63 (2d ed. 2008); V.S. Khanna, Is the Notion of Corporate Fault a Faulty Notion?: The Case of Corporate Mens Rea, 79 B.U. L. Rev. 355 (1999). Under American law (28 U.S.C.A. § 2680(h)), the government is liable in tort for claims “arising out of assault, battery, false imprisonment, false arrest, malicious prosecution, abuse of process, libel, slander, misrepresentation, deceit, or interference with contract rights,” (torts that typically require specific intentions), with regard to acts or omissions of “any officer of the United States who is empowered by law to execute searches, to seize evidence, or to make arrests for violations of Federal law.” 23. See, e.g., Glanville Williams, Textbook of Criminal Law 79–82 (2d ed. 1983). 24. Were motives and intentions truly indiscernible when it comes to governmental bodies, it would have posed a serious problem not only to deontologists but also to consequentialists. Whenever one has to determine the meaning of a statute, a governmental policy or a court judgment, such determination should plausibly take into account the motives and purposes underlying the statute, the policy, or the judgment (Cf. Clark, supra note 21, at 974). In this regard, it makes no difference whether the interpreter is a deontologists or a consequentialist.

65

66


and so forth.25 Thus, in the context of limitations on freedom of expression, American courts “have routinely inquired into the motivation underlying executive or administrative decisions in a variety of contexts.”26 Specifically, they have differentiated between restrictions on the time, place, or manner of expression that abridge freedom of expression as a mere side effect (the validity of which is examined according to a balancing test); and the use of such restrictions as a means to limit the dissemination of certain ideas or information (where the judicial scrutiny is much stricter).27 In a similar vein, “the well-settled constitutional [right] not to be disadvantaged by government simply because of one’s race or sex . . . would be largely meaningless if courts were not free to inquire into whether a governmental official or body did in fact discharge a given employee, or arrest a particular individual . . . because of that person’s race or other protected status . . .”28 Courts are increasingly willing to inquire into the motives of legislators,29 and international law takes into consideration the motives of states.30 Furthermore, it may be argued that while inquiring into the motives of private people sometimes infringes their privacy and autonomy (for instance, when it comes to their motives in choosing a spouse),31 governmental officials, acting in their public role, are not entitled to similar privacy privileges.32 True, in the context of judicial review of legislation and governmental actions, a

25. See, e.g., De Smith’s Judicial Review 265–94 (6th ed., Lord Woolf, Jeffrey Jowell & Andrew Le Sueur eds., 2007) (English law); Alfred C. Aman, Jr. & William T. Mayton, Administrative Law 527–29 (2d ed. 2001) (American law). 26. Laurence H. Tribe, American Constitutional Law 814–15 (2d ed. 1988) [ footnotes omitted]. 27. Id. at 789–804, 814–25; infra pp. 180–81, 213–18. 28. Laurence H. Tribe, The Mystery of Motive, Private and Public: Some Notes Inspired by the Problems of Hate Crime and Animal Sacrifice, 1993 Supreme Court Rev. 1, 18. See also Kagan, supra note 21, at 442 (discussing the case of discharging an employee because of union activity). This is not to say that the law cannot or should not proscribe practices causing a ‘disparate impact’ on certain groups, in order to promote equality, for example. See infra pp. 225–56. 29. Caleb Nelson, Judicial Review of Legislative Purpose, 83 N.Y.U. L. Rev. 1784 (2007). 30. For instance, the establishment of customary international law requires not only a consistent practice by states over time but also opinio juris: following the practice from a sense of a legal obligation. See, e.g., 1 Oppenheim’s International Law 1 27–28 (9th ed., Robert Jennings & Arthur Watts eds., 1996); Restatement of the Law, 3rd, on Foreign Relations Law of the United States § 102(2) & comment C (1987). 31. See infra pp. 244–46. 32. Tribe, supra note 28, at 17.


countervailing consideration is the institutional requirement of respect between governmental branches, which militates against the courts openly inquiring into the subjective intentions of legislators and high-ranking officials. However, this hurdle may be overcome. A powerful case has been made for the proposition that the best way to understand First Amendment law in the United States is as “a kind of motive-hunting.”33 Rather than directly discussing motives, the courts have developed objective tests that serve as proxies for motives. Arguably, the very structure of First Amendment law and the various distinctions it draws are best understood as a set of objective tests to determine whether a certain statute would have been enacted, or an administrative action would have been taken, absent an improper motive.34 For this limited purpose, one need not inquire into the intricate thinking processes of government officials. It suffices to assume that governmental bodies, including the legislature, are doing things for reasons, and that these reasons may be comprehended by other people.35 Even if there is no such thing as “the legislator’s intention,” members of the legislative body may certainly have shared intentions—the denial of which “would render the phenomenon of legislation a rather mysterious achievement.”36 Take, for example, the shooting down of a hijacked civilian aircraft that is expected to be used as a deadly weapon against thousands of people (as in the suicide attacks against American targets on September 11, 2001). In the absence of special evidence to the contrary, one may confidently assume that killing the innocent passengers on board the aircraft is neither the goal of shooting it down nor a means to achieve any goal, but rather a regrettable side effect of saving the lives of the terrorists’ intended victims on the ground.37 To arrive at this conclusion, one need not read the minds of the commanders approving the action nor the pilots executing it. At the same time, absent

33. Kagan, supra note 21. 34. Consequentialists may insist that these objective tests are not meant to serve as proxies to the decision-makers’ intentions but to bring about the best outcomes; but this claim sounds a bit strained. 35. Cf. Marmor, supra note 21, at 119–39 (arguing “that laws, at least in certain cases, are enacted with relatively specific intentions, and that this is a matter of fact which is discernible through an ordinary fact-finding procedure” (id. at 120)). 36. Marmor, supra note 21, at 122–26. When a decision process, including legislation, consists of several necessary stages, it may be sufficient to identify the mental state of the person or people in one stage to nullify the validity of the entire process. See Lawrence A. Alexander, Introduction: Motivation and Constitutionality, 15 San Diego L. Rev. 925, 938 (1978). 37. See infra pp. 166–69.

67

68


special circumstances, the mass killing and destruction resulting from the carpet bombing of a city where no military targets exist (nor are believed to exist), is not a side effect but rather the intended outcome of the bombing, possibly serving as a means to breaking civilian morale.38 The possibility that in any of these two examples, what ultimately motivates the decision-maker is, say, the desire to win the next elections (through saving the lives of thousands of people in the first example or winning the war in the second), is basically irrelevant. In this sense, intending and foreseeing are somewhat objectified.39 Moreover, even if the decision-maker’s ultimate goal is to win the elections, killing the passengers on board the aircraft is still a regrettable side effect, and the mass killing resulting from the carpet bombing is still a means to an end. The same analysis applies, mutatis mutandis, to the enactment of a statute permitting a commander to shoot down civilian aircrafts in the above circumstances or authorizing carpet bombing of cities during war. To apply the intending/foreseeing distinction to governmental bodies, one need not only identify or attribute mental states to the individuals playing various roles within governmental bodies but to the governmental bodies themselves. Much has been written about the metaphysical question of whether organizations in general, and the state in particular, have mental states. Without getting into details, it seems plausible that an organization or an institution should neither be treated holistically as if it were a human being nor as a mere aggregation of individuals (so that knowledge, foresight, and intentions can only be attributed to each of them separately).40 The former notion may undesirably hinder the attribution of personal— moral and legal—responsibility to individual members of the organization.

38. Similarly, it is rather obvious that in the case of capital punishment, the death of the convict is the state’s intended outcome (or a means to attain deterrence), whereas the possible (or even certain!) deaths of victims of murders due to insufficient deterrence are at most a foreseen side effect of such insufficient deterrence. See Carol S. Steiker, No, Capital Punishment Is Not Morally Required: Deterrence, Deontology, and the Death Penalty, 58 Stan. L. Rev. 751, 756–64 (2005). 39. See G.E.M. Anscombe, War and Murder, reprinted in Moral Problems: A Collection of Philosophical Essays 393, 403–05 (3d ed., James Rachels ed., 1979); Seana Valentine Shiffrin, Speech, Death, and Double Effect, 78 N.Y.U. L. Rev. 1135, 1155 (2003) (indicating that “objective constructions of intention are not uncommon in discussions of double effect in the philosophical literature and are especially familiar in law,” and arguing that one can evaluate the “justifiability of . . . policies by considering their rationales, even though we lack information about the subjective mental states of the particular people who will later implement these policies . . .”). 40. Celia Wells, Corporations and Criminal Responsibility 151 (2d ed. 2001).


The latter may undesirably preclude the attribution of responsibility to the organization itself (and given the division of roles and functions within large organizations, possibly preclude the attribution of responsibility to anyone else). For different purposes, governmental branches may or may not be taken to have intentions. As the above examples sought to demonstrate and as some philosophers have powerfully argued, it is often sensible to attribute intentions to governmental bodies.41 This is done in a wide range of contexts and may be done in the present context as well. Such attribution facilitates the application of the intending/foreseeing distinction to governmental bodies. The intentions attributed to governmental bodies need not be identical to the motivations of the decision-makers (and thus, the desire to win the elections, mentioned earlier, need not be attributed to the state). This brings us to the last argument; namely, that even if it is metaphysically meaningful and epistemologically feasible to attribute mental states such as knowledge, foresight, and intention to governmental entities, it is normatively implausible that such mental states should have intrinsic moral significance. State intentions—so the argument goes—are highly complex and, in a sense, artificial. They are “determined by the intentions of individuals, by facts about the decision-making mechanism, by matters of institutional design, [and] by internal power struggles.”42 They are, therefore, unlikely to have any intrinsic normative significance. We fail to see why this is the case. The fact that the mental state of organizations, including governmental bodies, is sometimes complex or is the product of intraorganizational struggles does not seem a sufficient reason to deprive it of

41. See, e.g., Philip Pettit, Akrasia, Collective and Individual, in Weakness of Will and Practical Irrationality 68, 68–80 (Sarah Stroud & Christine Tappolet eds., 2003) (discussing the necessary conditions for the existence of an intentional agent and concluding that public bodies, business corporations and private associations may all satisfy these conditions, despite the fact that they operate only by courtesy of individual contributions, with no autonomous means of perception, feelings, or spontaneous inferential dispositions); Christopher Kutz, Complicity: Ethics and Law for Collective Age 107–12, 191–202 (2000) (discussing the conditions under which intentions may be ascribed to collectives); Peter A. French, Collective and Corporate Responsibility (1984) (arguing that corporations are moral agents); John Searle, The Construction of Social Reality 23–26 (1995) (endorsing the notion of collective intentionality); Wells, id. at 78–81 (doubting that corporations can be morally blameworthy). Note that attributing moral agency to organizations and to the state does not ipso facto imply that they should also be treated as moral patients. 42. Enoch, supra note 3, at 86.

69

70


intrinsic normative significance.43 If one accepts that the framing of an innocent person by a private person and the indiscriminate killing of civilians by a terrorist infringe the deontological constraint against intentionally harming other people, then it is unclear why these actions cease to infringe this constraint once they are carried out by the police or by the state’s air force (even if a heated debate between decision-makers preceded the decision to frame the innocent or to kill the civilians). Assuming (as we do in the present discussion), that for some purposes, including judging the moral permissibility and legality of such actions, it matters whether innocent people are harmed as a means to some legitimate end or merely as a side effect, then we do not see why the intentions to frame the innocent or to bomb civilians are normatively insignificant. Governmental bodies do things for reasons, and these reasons are normatively significant. To be sure, the fact that harming the innocent is a side effect of attaining a desirable goal does not necessarily legitimize the action. It may still be impermissible to target a terrorist if it is highly probable that dozens of innocent people (who may be unaware of the terrorist’s presence in their neighborhood) will be killed as well.44 All we argue here is that, to the extent that it makes a difference whether killing the innocent is an intended outcome or merely a foreseeable one, this distinction applies to the state.45 We thus conclude that the intending/foreseeing distinction can and should apply metaphysically, epistemologically, and normatively to decisions and actions made by public bodies, whether by a single official or collectively.

•

E. Acting and Enacting

The arguments we have critically examined thus far aimed to establish that even if deontology is the right personal moral theory, it is inappropriate for the state and other public bodies. A more modest claim might be that

43. It may be noted that even the mental state of a single person is often rather complex and may reflect internal struggles. See, e.g., Thomas Schelling, Egonomics, or the Art of SelfManagement, 68 Am. Econ. Rev. 290 (1978). 44. See infra pp. 166–75. 45. Plausibly, the threshold that would have to be met to render such an action permissible will be higher in the case of intended harm than in the case of a merely anticipated one. See Haque, supra note 12, at 635–36. See also infra pp. 135–75, 213–18.


deontology is an appropriate moral theory both for private actors and for public ones (such as policemen and judges) but not for legislators.46 According to this claim, one should distinguish between the role of the state as an actor and its role as the provider of legal norms. While as an actor, the state’s position is parallel to that of a moral agent and thus subject to deontological constraints, in its legislative role, the state’s position is akin to that of an ethicist who lays down general norms of behavior. Just as the ethicist is not subject to agent-relative constraints when formulating a moral theory, so should the legislator be free of such constraints. According to this claim, while it may be wise to formulate legal rules that put constraints on people’s behavior—including the state’s conduct as an actor—the legislative process itself should strive solely to maximize overall good outcomes.47 Several characteristics of legislation seem to render this claim appealing. To begin with, the direct interpersonal relationships existing among private individuals and between public officers and the persons subject to their power (such as a police investigator and a suspect, a judge and the litigants) are largely missing in the legislative process, which is carried out on much higher levels of abstraction, generality, and impersonality.48 Moreover, whereas particular governmental acts (such as police investigations or court proceedings) often focus on ex post examination of people’s behavior, legislation is distinctively future oriented. Laws, regulations, and general policy guidelines are meant to affect future conduct; and this arguably means that they should concentrate on outcomes.49 Finally, it has been argued that if people who follow the legislator’s norms actively inflict harm on other people, the legislator “will never be guilty of anything more than merely failing to prevent a harm that is actually caused by others.”50 If this is so, then at least according to the doing/allowing distinction, a legislator simply cannot infringe a deontological constraint.

46. For a laconic argument along these lines, see Michael S. Moore, Four Reflections on Law and Morality, 48 Wm. & Mary L. Rev. 1523, 1551–52, esp. n.79 (2007). 47. This argument somewhat resembles the suggestion that moderate deontology on the factoral level should be grounded in rule-consequentialism on the foundational level. See supra pp. 24–27. 48. Cf. Enoch, supra note 3, at 94. 49. Cf. Meir Dan-Cohen, Rights, Persons, and Organizations: A Legal Theory for Bureaucratic Society 136–59 (1986). 50. Richard Craswell, Kaplow and Shavell on the Substance of Fairness, 32 J. Legal Stud. 245, 264 (2003).

71

72


A few preliminary comments are in order before we delve into the merits of these arguments and further discuss the appropriateness of applying deontological morality to the legislature. First, the present argument does not hinge on the collective/single decision distinction. The argument concedes that deontological constraints should apply to particular governmental actions, whether performed by a single public official or by a collective body, and at the same time denies the applicability of constraints to legislation even if the legislator consists of a single person. Second, and for similar reasons, the present argument does not rest on the epistemological difficulties of identifying or attributing mental states to the legislator, difficulties that by now we assume can be overcome. Third, according to at least some of the rationales underlying the present argument, it actually transcends the private/public morality debate and carries much broader implications. Parents, educators, preachers, and moral philosophers regularly engage in laying down and inculcating general norms of conduct for people to follow. According to the argument under discussion, in their “legislative” or educational role, these people are not subject to deontological constraints and should thus reason as consequentialists. While the present claim may seem more compelling than the previous ones, we maintain that it should be rejected as well. One practical difficulty with this claim is that the dividing line between laying down general norms and performing particular actions is not always clear. This is particularly true of the judicial and executive branches. When a court sets a precedent, it both determines the outcome of a specific case and establishes more or less general rules. Even when a judicial or an administrative decision does not formally constitute a precedent, it ordinarily gives rise to expectations—and often to a legal duty—to treat like cases alike in the future, thus having some “legislative” significance. While the vagueness of the acting/enacting distinction is not fatal, it should make one wary of adopting it as a criterion for the applicability of deontological constraints in legal policy-making.51 The first characteristic of legislation noted as supporting the claim that it should rest on consequentialism was the lack of direct interpersonal relationships between the legislature and the law’s addressees. However, such relationships are immaterial in the context of most deontological constraints: one must not intentionally/actively harm others even if they are complete strangers. Moreover, the few constraints that do rest on special relationships,

51. Cf. supra pp. 58–59.


such as the duties to keep one’s promises and to refrain from betrayal, may actually apply to the legislature, at least in liberal democracies where members of legislative bodies are elected on the basis of their declared agendas and trustworthiness. In the same vein, the second characteristic of legislation, namely its distinctively future-looking nature, indeed implies that outcomes should be a primary concern for the legislature. It does not follow, though, that the legislature is not subject to constraints when it strives to maximize good outcomes. We then turn to the argument that by its very nature, the legislator cannot actively infringe a deontological constraint because it is always the people following its rules who cause the harm, rather than the legislator. This claim assumes that if one entity violates a constraint by doing/intending some harm, then no other entity can concurrently infringe a constraint by actively/ intentionally ordering, encouraging, or facilitating that violation. This assumption is unfounded. For instance, a statute authorizing the killing of innocent people infringes a deontological constraint, notwithstanding the fact that the people carrying out the killing infringe a constraint as well. The same is true of a statute that explicitly permits improper discrimination between people on the basis of skin color or gender. In this case, it seems plausible that both the legislator and the people who actually discriminate infringe the constraint against intentionally/actively harming other people’s dignity. As the discrimination example suggests, there are (at least) two different, not mutually exclusive, ways in which a legislator may infringe a deontological constraint. Sometimes, the very enactment of a law constitutes such an infringement, regardless of how it will affect people’s behavior. Thus, a law that overtly legitimizes race discrimination or that implies that men are morally superior to women arguably infringes a constraint even if it does not actually affect anybody’s behavior.52 Similarly, a statute that states the facts underlying its enactment or the purpose it aims to achieve may infringe the constraint against lying if the statements of facts or of the law’s purpose are

52. Possibly, some constraints apply to public bodies even if they do not apply to private individuals. Perhaps a private person who says, “I detest homosexual relationships” infringes no constraint, whereas a public official who, as a public official, says the same thing does infringe a constraint. On expressive theories of legislation, see generally Elizabeth S. Anderson & Richard H. Pildes, Expressive Theories of Law: A General Restatement, 148 U. Pa. L. Rev. 1503 (2000). See also infra pp. 117–22.

73

74


knowingly false.53 The other way in which a law may arguably infringe a deontological constraint is by actively/intentionally bringing about, or increasing the likelihood of, infringements of constraints by the law’s addressees. In the above example, a law that explicitly legalizes race discrimination in all likelihood encourages such discrimination. Likewise, a statute authorizing capital punishment is likely to bring about the execution of convicts by the state. Possibly, even refraining from legislating a certain statute may infringe a deontological constraint if intended to allow constraint violations by citizens. Interestingly, there may be cases in which a legislator (including an executive laying down general guidelines and a court setting a general norm) would face a tension between these two types of constraints. Suppose a legislator is a moderate deontologist who believes that suspects must not be tortured unless the lives of a large number of people could be saved by such torture, the expected death of the victims is imminent, the probability of saving their lives is very high, there is no other way to save them, and the torture is not expected to inflict a lasting harm on the suspect. Suppose further that if the legislator will lay down strict procedural and substantive conditions for the permissibility of torture, it would actually reduce the number of tortures, compared to a legal regime in which torture is never permitted. (While the last assumption is somewhat speculative, it is not implausible as it may be that absent clear guidelines regarding the permissibility and impermissibility of torture, more suspects—rather than less— will actually be tortured by overly motivated or frustrated interrogators.) In this case, it may be argued that even if these rules will actually reduce the number of constraint violations by interrogators, their very enactment by the state violates a deontological constraint. The legislation itself is incompatible with the respect a legislator must demonstrate toward people’s autonomy and human dignity. Putting this difficult question aside,54 we maintain that both types of constraints on legislation not only are conceptually possible but also normatively plausible. This conclusion is straightforward when the very enactment of a

53. While statements of fact are not common in legislation, they may be found, for example, in section 2 of the American with Disabilities Act (ADA), 42 U.S.C. 12101 (2001). 54. See also infra pp. 117–22.


law, regardless of its ensuing effect on people’s conduct, infringes a deontological constraint. The same conclusion holds, at least sometimes, with regard to the second type of constraints (inducing constraint violations), though in that case the picture is more complex. One possibility is that legislation that actively/ intentionally induces the violation of deontological constraints by the law’s addressees may be morally wrong, but it does not infringe any deontological constraint. An opposite position would be that whenever a law actively/ intentionally increases the likelihood that the law’s addressees will violate a deontological constraint, the enactment of this law inevitably violates, or at least infringes, the same constraint. An intermediate position may concede that not just any law that actively/intentionally increases the likelihood that some people will violate a deontological constraint ipso facto infringes the same constraint; yet under certain circumstances, such a law does infringe a deontological constraint by inducing constraint violations. This intermediate position leaves open the possibility that the constraint infringed by the legislator is different from the constraint whose violation the law brings about. Formulating a comprehensive theory of the relationships between constraint violations by the legislator and by the law’s addressees lies beyond the scope of our discussion. As the above examples sought to demonstrate, often enough, active/intentional inducing of active/intentional infliction of harms by a law’s addressees (and perhaps also an active/intentional inducing of other constraint violations) does constitute an infringement of a deontological constraint by the legislator. The powerful measures used to enforce legal norms lend support to the claim that the legislator is indeed subject to (at least some) deontological constraints against inducing constraint violations by the people whose behavior it governs. This notion is also compatible with the special duties the state bears toward its people. The same would apply to other legislative bodies, such as local authorities and regulatory agencies. Substantive moral arguments are necessary to delineate the scope of the deontological constraints applicable to the legislator. For our purposes, we need not rule out the theoretical possibility of cases in which a legislator who actively/intentionally induces or facilitates constraint violations does not thereby infringe a deontological constraint. Even if such cases exist, it is sufficient that typically, when the legislator actively/intentionally induces constraint violations, it infringes a deontological constraint, to justify the incorporation of deontological constraints into CBA.

75

76


F. Concluding Remarks

This chapter sought to refute the claim that even if deontology is the right moral theory for individuals, it is inappropriate for public decisions or for legal review of such decisions. We demonstrated that some of the arguments made to support this claim conflate important distinctions—such as between deontological constraints and deontological options and between the legislation of general norms and the performance of particular acts—and that none of them leads to the alleged conclusion. While some differences between the private and public spheres render consequentialism more appealing in the latter sphere, others actually make it less attractive. While the stakes in the public sphere are often higher than in the private sphere, and thus there are possibly more instances in which the pertinent thresholds are met,55 deontological constraints do apply in the public sphere.56 It is worth noting that the notion that deontology is appropriate for public decision-making—and in particular that public decision-makers, including legislators, are subject to deontological constraints—coincides with prevailing moral intuitions and extant legal norms. For example, an experiment exposing people’s “omission bias” (which correlates with the deontological doing/allowing distinction) found no significant difference between people’s judgments when asked to make a personal decision and their judgments when asked to make a general policy decision on the same issue.57 In the legal sphere, it is commonly held that violating basic civil rights cannot be justified on the ground that the violation produces slightly more overall good.58 Even Sunstein and Vermeule, who adamantly reject the absolute

55. See Stuart Hampshire, Public and Private Morality, in Public and Private Morality 23, 48–52 (Stuart Hampshire ed., 1978). 56. Thomas Nagel, Ruthless in Public Life, in Public and Private Morality, id. at 75, 89. 57. Ilana Ritov & Jonathan Baron, Reluctance to Vaccinate: Omission Bias and Ambiguity, 3 J. Behav. Dec. Making 263, 266–67 (1990) (showing that people’s reluctance to vaccinate even when the expected deaths due to the vaccine are considerably lower than the expected deaths due to failure to vaccinate is similar in the personal decision hypothetical and in the policy decision one). 58. Constitutional protection of basic liberties often includes provisos that are akin to deontological thresholds. One example is the doctrine that only “compelling interests” can justify infringements of fundamental rights. See, e.g., Johnson v. California, 543 U.S. 499, 505 (2005); Grutter v. Bollinger, 539 U.S. 306, 326 (2003). It also applies whenever the government adopts a content-based regulation of speech. See, e.g., Republican Party of Minn. v. White, 536 U.S. 765, 774–75 (2003). As explained by Justice O’Connor, when the government infringes a person’s fundamental right to equal protection on the basis of race, that


opposition to capital punishment, do not propose that the decision whether to use this punishment be based on an unconstrained CBA. Quite the contrary, their claim is that if each execution deters some eighteen murders (as some evidence suggests), then the threshold of the deontological constraint against actively/intentionally killing the convict is met.59 The adoption of consequentialism as public morality, thus relieving public officials and legislators of deontological constraints, would therefore entail far-reaching and mostly undesirable reforms in current legal norms. Inter alia, it would mean that the distinction between content-based and contentneutral restrictions on freedom of speech should be abandoned.60 It would similarly imply that the side effect of criminal punishment (e.g., on the convict’s family) should have the same significance as its intended consequences.61 It would also entail a radical change in the policies of the medical establishment and agencies regulating drugs and medications, which currently attribute considerably greater weight to the avoidance of harm than to the promotion of good outcomes.62 While current legal doctrines and public policies may indeed require reassessment, it does not seem that such reassessment should rest on consequentialist morality. Legal academics and public decision-makers should follow moderate deontology first and foremost because of its merits as both private and public morality (at least on the factoral level, or as a decision procedure). While we do not argue that the law should straightforwardly enforce morality, we do believe that in a liberal democracy, there should be no fundamental divergence between law and morality. General compatibility of legal norms with morality is essential for the moral duty to

person has suffered an injury, and the application of the strict scrutiny standard “determines whether a compelling governmental interest justifies the infliction of that injury.” Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 229–30 (1995). See also infra pp. 180–81, 213–18. 59. Sunstein & Vermeule, supra note 2, at 719, 726–27. For a critique of this proposition, emphasizing that a constraint may be infringed only if the desirable good outcome cannot be attained without such an infringement, see Steiker, supra note 38, at 782–86. See also infra pp. 81–83, 149–50, 222–23, and 346–47. 60. Shiffrin, supra note 39, at 1177–80; Enoch, supra note 3, at 95. See also infra pp. 180–81, 213–18. 61. Enoch, supra note 3, at 96. 62. On the American FDA’s policy of not approving potentially harmful drugs even when their expected net benefit to most users is considerably greater than the expected harm to some users, see Henry I. Miller, To America’s Health: A Proposal to Reform the Food and Drug Administration (2000).

77

78


obey the law.63 Public decision-makers should follow moderate deontology for another reason, namely its conformity with prevailing moral intuitions.64 Even critics of threshold deontology readily admit that it comports with commonsense morality much better than do rival theories.65 Assuming that popular legitimacy is important for public authorities (including the legislature and the courts), and given that moderate deontology is compatible with commonsense morality while consequentialism is not, policy-makers should thus reason as moderate deontologists. Compatibility with commonsense morality may also increase the level of compliance with the legal norms. This is not to say that policy-makers should not avail themselves of the benefits of economic methodology; it only means that they should use a deontologically constrained CBA rather than a standard one.

63. See, e.g., John Rawls, A Theory of Justice 308 (1999); Chaim Gans, Philosophical Anarchism and Political Disobedience (1992); Michael B. Dorff & Kimberly Kessler Ferzan, Is There a Method to the Madness? Why Creative and Counterintuitive Proposals Are Counterproductive, in Theoretical Foundations of Law and Economics 21, 34–39 (Mark D. White ed., 2009); Seana Valentine Shiffrin, The Divergence of Contract and Promise, 120 Harv. L. Rev. 708, 713–19 (2007) (“Especially because there are moral duties to obey the law, legal rules should be sensitive to the demands placed on moral agents so that law-abiding moral agents do not, as a regular matter, face substantial burdens on the development and expression of moral agency.” id. at 715). 64. On this conformity, see, e.g., Samantha Brennan, Thresholds for Rights, 33 Southern J. Phil. 143, 145 (1995); Samuel Scheffler, Introduction, in Consequentialism and Its Critics 1, 9 (Samuel Scheffler ed., 1988). 65. See, e.g., Shelly Kagan, The Limits of Morality 1–5 (1989). Cf. Louis Kaplow & Steven Shavell, Fairness Versus Welfare 62–69 (2002) (providing a similar observation on “notions of fairness,” defined as any notion that is not exclusively consequentialist and restricted to individuals’ welfare).

•

four

Constructing Threshold Functions

•

A. Introductory Remarks

having established the necessity of integrating deontological constraints

with economic methodology, this chapter demonstrates how such integration may actually be accomplished. Given the vast variety of contexts to which deontological constraints are applicable, the considerable diversity within theories of moderate deontology, and the multiplicity of modeling possibilities, we will not try to provide a complete array of plausible mathematical formulae. Rather, we will present the central substantive and methodological choices involved in constructing threshold functions. More specific substantive and methodological issues will be discussed in the second part of this book, which illustrates the implementation of the general framework in several fields. The very use of stylized mathematical representation facilitates a more definite and less ambiguous description of constraints, their scope, and the types of benefits and costs they take into account. A mathematical formulation brings to light the implications of alternative definitions of a constraint and may thus contribute to the normative debate concerning such definitions. A formalized definition of a threshold constraint may then be integrated into an otherwise standard economic analysis, thus redefining its target function. Before getting into the details of the choices involved in the construction of threshold functions, some clarifications regarding the goals and scope of the following discussion are in order. Focus on Methodology While this chapter describes the implications of different variations of deontological constraints for constrained CBA, it does not discuss in any detail the substantive arguments for and against these variations. Rather than directly engaging in the philosophical debate, the goal of this chapter is to demonstrate the methodological plausibility of integrating deontological 79

80


constraints into economic analysis. Sometimes, however, the methodological discussion sheds new light on the substantive issues as well. Object of Analysis In principle, the following analysis applies to both acts and rules, to both moral and legal questions, and to both private and public choices. Of these pairs, our primary focus is on public (rather than private) decisions concerning legal (rather than moral) issues. The analysis of legal issues is often more complex than a comparable analysis of moral issues. In normative ethics, any act (or omission) is morally required, prohibited, or optional (and in the latter case, either praiseworthy or condemnable). Only rarely do ethicists discuss the outcomes of infringements or violations of moral norms. The legal system, in contrast, determines not only the normativity of acts but also provides for detailed sanctions and remedies for violations. Thus, a violation of a legal prohibition may result in criminal, administrative, or civil sanctions, which vary in their severity, intrusiveness, monetary or nonmonetary character, the identity of the enforcer (public or private), and so forth. Moreover, a prohibited act may be void, voidable, unenforceable, or valid—and each of these adjectives bears different meanings in different contexts. Therefore, the law may reflect deontological constraints not only by prohibiting certain activities (e.g., enslavement) but also by restricting their legal effect (e.g., rendering enslavement contracts void). Finally, since legal norms are produced, interpreted, revised, and enforced by governmental authorities, institutional considerations loom large on almost any legal issue. Our analysis applies to rules as well as to actions (such as resolving a particular legal dispute). To avoid repetition, we will mostly refer to “acts” and “actors,” but these terms should be understood as referring to both legal rulemaking and particular decisions. In fact, however, much of the analysis is relevant to moral (rather than legal) questions and to individual (rather than public) choices as well. Standard CBA as the Default The following discussion takes the standard features of conventional CBA, including its underlying theory of the good and economic monetization methods, as the default. We shall not discuss these features unless the adding of threshold constraints to CBA requires modification thereof. Similarly, we shall not discuss the incorporation of distributive concerns into the social utility function underlying CBA. It is not that distributive concerns are not

constructing threshold functions

important; rather, they are important whether or not deontological constraints are considered.1 Conflicting Constraints Legal (and other) policy-making often involves not only conflicts between the promotion of the good and deontological constraints but also conflicts between deontological constraints. For example, the constraint against harming a person’s reputation sometimes conflicts with the constraint against limiting freedom of speech. Similarly, the constraint against harming one’s dignity by discriminating against her in the marketplace conflicts with the constraint against limiting freedom to choose with whom to interact. In such cases, policy-makers need not only weigh the goodness of outcomes against deontological constraints but also (or even primarily) balance or prioritize between the constraints themselves according to some normative theory. We touch upon such conflicts in the second part of this book.2 However, we do not suggest that threshold functions resolve conflicts between constraints and do not explore such conflicts in any detail.3 Choosing Among Prima Facie Permissible Courses of Action The functions described here aim to determine whether acts or rules infringing a deontological constraint produce a sufficiently large net benefit to override the constraint. At times, this is the sole question facing the decisionmaker. This is the case when the infringement under consideration is the only reasonable way to achieve some desirable outcome. Sometimes, however, determining whether one or more of the available courses of action violate a deontological constraint is only part of the inquiry. Consider the choice between the following alternatives: (1) deliberately killing 2 innocent people to save the lives of 280 (out of 500 facing death); (2) deliberately killing 3 innocent people to save the lives of 290 (out of the same 500); and (3) letting all 500 die. If the threshold of the constraint against actively/intentionally killing innocent people is sufficiently high, only the third choice would

1. There may be interesting interactions between introducing deontological constraints and incorporating distributive concerns into CBA, but this issue exceeds the scope of our analysis. On the relationships between deontological constraints and distributive concerns in the context of marketplace discrimination, see also infra pp. 246–51. 2. See infra pp. 244–46. 3. For an interesting proposal to model such conflicts through a “Weight Formula,” see Robert Alexy, On Balancing and Subsumption: A Structural Comparison, 16 Ratio Juris 433 (2003).

81

82


be permissible. If, alternatively, the threshold is met whenever at least 100 people are saved by killing each person, then alternatives (1) and (3), but not (2), are permissible. The choice between (1) and (3) may be made using standard CBA, and since alternative (1) results in a net saving of 278 lives and alternative (3) with the saving of none, alternative (1) should be chosen. Finally, if the threshold is met whenever killing one person saves at least 90 lives, then all three alternatives are permissible. Using standard CBA to choose among them, alternative (2) ranks first (net saving of 287 lives), alternative (1) second (net saving of 278 lives), and alternative (3) last (no lives saved). In situations like these, the decision process consists of two stages: sorting out those courses of action that do not infringe, or that override, the deontological constraint; and then using standard CBA to choose the one that brings about the best outcomes (as long as the chosen alternative both brings about the best outcomes and does not violate any deontological constraint, the order of the two stages is not crucial). The separation between the two stages is necessary because—as demonstrated below—the factors taken into account in each stage and their interrelations may be different. It does not follow, however, that whenever a decision-maker faces more than one deontologically permissible course of action, the choice between them should necessarily rest on standard CBA. Consider the choice between the following possibilities: (1) deliberately killing 1 innocent person to save 101 others; (2) spending $10,000,000 to save the same 101 people; and (3) letting the 101 people die. Presumably, the only alternative infringing a deontological constraint in this example is (1). If alternative (1) overrides the constraint against deliberately killing innocent people, then all three courses of action are permissible. The net outcomes of the alternatives are as follows: (1) net saving of 100 lives; (2) saving of 101 lives at a cost of $10,000,000; and (3) no lives saved and no costs borne. At this point, there seems to be (at least) two ways to proceed. One way is to employ standard CBA, monetize the value of a person’s life, and choose between alternatives (1) and (2) (both of which produce much more good than alternative (3)) based on whether one life is valued at more or less than $10,000,000. Another approach is to maintain that whenever an infringement of a deontological constraint is involved (but not necessarily in other contexts), human life is lexically more important than monetary losses. Thus, one would rank alternative (2) higher than alternative (1), regardless of the monetary value attributed to human life.4

4. Cf. Judith Jarvis Thomson, The Realm of Rights 164 (1990) (arguing that whenever a large increment of good may be achieved by infringing a person’s right, such infringement


The notion of lexical priority may be deemed relevant not only when choosing between a course of action that permissibly infringes a constraint and a course of action that involves no such infringement; but also when choosing between two or more actions, all of which prima facie justifiably infringe some constraints. Assume, for instance, that there are two ways to save a person’s life: one entails lying and one entails bodily injury to a bystander. A deontologist may plausibly argue that the availability of the former course of action (lying) renders the latter (inflicting bodily injury) impermissible, despite the fact that absent the former, the latter would have been considered permissible. The notion of lexical priority may also be deemed relevant when faced with a choice between two prima facie permissible infringements whose expected net benefit is dissimilar. Suppose that to preempt an imminent, deadly terrorist attack, the authorities may either torture a person possessing pertinent information or impose considerable, though temporary, limitations on the freedom of movement of other people. The expected probability of success of the former measure is 85% and the probability of the latter is 75%. If the deontological constraint against torture is lexically more stringent than the constraint against temporarily limiting people’s freedom of movement, then one has to decide whether the increase in the probability of success is great enough to render the torture permissible. Be that as it may, the general discussion that follows focuses on formulating threshold functions to determine whether the good outcomes of a certain act or a rule are sufficiently large to override a deontological constraint (the first stage). The possibility of constructing functions that deal with both stages of the decision process, taking into account the marginal net benefit of infringing the more stringent constraint, will be considered later.5 Threshold Options While incorporating constraints into standard CBA may significantly alter the conclusions of the analysis, incorporating deontological options is unlikely to have such an effect. This is primarily because standard CBA, and economic models in general, ordinarily assume that people are self-interest maximizers rather than maximizers of the total good. If people are not

is impermissible if there exists a way to produce a comparable increment without infringing any right); Carol S. Steiker, No, Capital Punishment Is Not Morally Required: Deterrence, Deontology, and the Death Penalty, 58 Stan. L. Rev. 751, 783–85 (2005) (submitting that the availability of alternative means to prevent future murders renders capital punishment impermissible). 5. See infra pp. 149–50, 155–56, 169–70, 223, and 346–47.

83

84


expected to impartially maximize the total good, then standard CBA already embodies unlimited options (or, more precisely, an invariable priority to one’s self-interest). Incorporation of options into CBA would also be less revealing whenever commonsense morality coincides with the standard economic assumption that people have absolute or almost absolute options. In such contexts, the threshold above which the agent no longer enjoys an option would only come into play in very rare cases, making its formalization rather unhelpful. There are, nevertheless, cases in which, due to market failures, the maximization of aggregate social welfare requires imposing considerable sacrifices on people. Even if such imposition does not constitute an infringement of a deontological constraint by the legislature, it may still restrict otherwise available options. Assuming these options have thresholds, incorporating threshold options into CBA may thus be fruitful. The concluding section of this chapter briefly considers the integration of deontological threshold options with CBA.

•

B. General Structure of a Threshold Function

Incorporating deontological constraints into economic analysis entails the characterization of a threshold function, T, such that an act is permissible only if the product of this function is positive. One possibility is to use an additive function. According to this approach, the infringement of a deontological constraint should be expressed by adding to the act’s harms, C, some factor K (where K>0), such that an infringing act is permissible only if its benefit, B, exceeds the sum of C and K: (1) T = B – (C + K ). Alternatively, using a multiplier function, the infringement can be expressed by multiplying the act’s harms by some factor K,′ such that an infringing act is permitted only if its total benefit, B, exceeds its weighted harms, K ′×C (where K ′>1): (2) T = B – K ′×C. The threshold function may also combine these two forms to include both K (a constant constraint) and K ′ (a multiplier). Note that when K = 0 and K ′ = 1, the threshold function reflects consequentialism and when either of these factors is prohibitively high, the function reflects absolute


deontology. Moderate deontology entails intermediate levels of K and K.′ The threshold function may take other mathematical forms as well. Consider the following function: (3) T = 1n (B – C) – K. This function narrows the spectrum of the possible outcomes and enables one to define K in a different scale than B and C.6 The threshold function determines not only the size of net benefits required to justify overriding the constraint but also the types of benefits and costs that should be considered in this regard. To highlight this aspect, we propose that, as a general matter, a threshold function consists of two elements: B = b( . . . ), the infringing act’s relevant net benefit, and K = k( . . . ), the size of the minimum amount of benefits that is required to justify the infringing act (the threshold). Thus, in its general form, the threshold function is: (4) T = B — K = b(. . .) — k (. . .). To illustrate, assume that the only type of benefit that can justify overriding the constraint against actively/intentionally killing an innocent person is the saving of human lives. The permissibility of such an act may thus be determined according to the following threshold function: (5) T = (x – y) – K ′× y where x is the number of persons who will be saved if the act is committed, y is the number of those who will be killed as a result of the act, and K ′ is a multiplier expressing the magnitude of the constraint against actively/ intentionally killing innocent people. For instance, if K ′ = 100, this function entails that killing one person ( y = 1) is justified only if the act results in saving at least 101 persons, regardless of the act’s other possible costs and benefits. To formalize the notion that a threshold function should sometimes accord different weights to different types of costs and benefits, the thresholds K and K ′ may take the form of vectors of thresholds. Thus in the case of a multiplier threshold, using discount factors for any type of costs and benefits would yield the following general form of the function: (6) T = (d1 x1 + d2 x2 + . . . – h1 y1 – h2 y2– . . .) –(K ′1 h1y1 + K ′2 h2 y2 + . . .)

6. For example, if the costs and benefits are measured in monetary terms, setting K=10 entails that the infringing act is permissible only if its net benefit exceeds approximately $22,000; if K=12, the required net benefit is approximately $163,000; and if K=14, the net benefit should exceed $1,200,000.

85

86


where xi is a certain type of benefit, yi is a certain type of cost, and di and hi are discount factors. To assess more realistic scenarios (particularly when the objects of assessment are rules), one might have to use more sophisticated threshold functions. The following sections describe the main choices required in formulating such functions. We start by discussing what types of benefits and costs are deemed relevant according to the threshold function (section C). We then move to assess what size of net benefit is required to justify an infringement and the relevant considerations in choosing between additive and multiplier functions (section D).

•

C. Relevant Types of Benefits and Costs

Standard CBA monetizes and aggregates all costs and benefits involved in an act. If, for example, an act involves the killing of some people and saving others, body injuries, damage to property, and pecuniary gains and losses, all these costs and benefits are taken into account. While a threshold function may similarly consider all such costs and benefits (thus deviating from standard CBA only in adding a threshold K), such inclusiveness arguably misses significant distinctions between different costs and benefits. Following the philosophical literature, this section examines several limitations on the types of benefits and costs bearing on the permissibility of infringing deontological constraints. The possible “excluders” of costs and benefits described may be endorsed alternatively or cumulatively. Lexical Priority According to commonsense morality, certain values take lexical priority over other values. For instance, human life may be thought of as lexically superior to pecuniary losses. Under this view, one should sort out those types of benefits (and costs) that are not lexically inferior to the harm prohibited by the constraint. This notion is echoed in the discussion of “exclusionary reasons” in constitutional and human rights analysis.7 Accordingly, an infringement of the constraint against torture may be justified only if it is absolutely necessary to generate a sufficient amount of benefits such as saving lives and avoiding comparable tortures, but not for

7. See, e.g., Richard H. Pildes, Against Balancing: The Role of Exclusionary Reasons in Constitutional Law, 45 Hastings L.J. 711 (1994).


generating pecuniary benefits, regardless of their size. Similarly, such characteristics as a person’s life expectancy may be classified as irrelevant in assessing the legitimacy of saving one’s life while infringing a constraint.8 Given the great variety of types and intensity of harms, however, lexically ranking values in the abstract may be misleading.9 One may therefore use constrained CBA without embracing lexical priorities. In any event, such priorities need not apply outside of the realm of deontological constraints. Even if it is impermissible to deliberately kill an innocent person for any amount of money, it may still be permissible—in fact inevitable—to trade safety measures against other uses of money. Small Benefits A more controversial claim is that even benefits similar in nature to the harms prohibited by the deontological constraint (in the sense that they both belong to the same sphere, e.g., the sphere of bodily injuries) should not be considered if they are too small compared to the prohibited harm. An oftdiscussed example is whether it is ever permissible to actively/intentionally kill an innocent person in order to avoid a vast number of minor headaches. A consequentialist would reasonably hold that a world with a vast (but finite) number of temporary, moderate headaches could be worse than a world lacking those headaches and containing one more premature death. There is, in other words, some finite number of headaches such that it is permissible to kill an innocent person to avoid them.10 A deontologist, in contrast, may believe that the number of (at least certain types of) headaches that can be avoided by killing an innocent person is irrelevant in assessing the permissibility of such killing.11 Thus, to be included in the threshold function,

8. On the question of whether it is permissible to actively/intentionally kill people who are doomed to die anyway, in order to save the lives of others, see infra pp. 170–71. 9. Richard Craswell, Incommensurability, Welfare Economics, and the Law, 146 U. Pa. L. Rev. 1419, 1456–57 (1998); Theodor Lenckner, The Principle of Interest Balancing as a General Basis of Justification, 1986 B.Y.U. L. Rev. 645, 645–51 (demonstrating this concern in the context of the criminal defense of justifying necessity). 10. Alastair Norcross, Comparing Harms: Headaches and Human Lives, 26 Phil. & Pub. Aff. 135 (1997); Dan W. Brock, Aggregating Costs and Benefits, 58 Phil. & Phenomenological Res. 963, 967 (1998). 11. For a discussion of this excluder, see Thomson, supra note 4, at 166–69 (labeling it “the distributive constraint”); Samantha Brennan, Thresholds for Rights, 33 S. J. Phil. 143, 149–53, 160–65 (1995) (hereinafter: Brennan, Thresholds) (denoting it “the universal constraint”); Samantha Brennan, Moral Lumps, 9 Ethical Theory & Moral Prac. 249 (2006). Cf. Frances M. Kamm, Morality, Mortality, Vol. I: Death and Whom to Save From

87

88


any single benefit must—according to this view—bear some proportion to the harm whose active/intentional infliction is prohibited by the constraint. If both harms and benefits are monetized, this proportion may straightforwardly be represented in mathematical terms. Some deontologists may want to exclude rather significant benefits, perhaps even benefits whose magnitude is greater than the harm whose active/intentional infliction is prohibited by the constraint. To illustrate, a deontologist may hold that nothing less than the saving of human life can justify the brutal torture of an innocent person. According to this position, anything less than a saved life should be excluded from the threshold function.12 The Existential Constraint13 Whether or not the elimination of a vast number of small harms can ever justify the active/intentional infliction of a major harm, one may hold that to override a constraint, at least one of the benefits produced by the act should bear a certain proportion to the inflicted harm. The one benefit should be, for instance, at least as large as (or nearly as large as)—and possibly of the same nature as—the inflicted harm. Thus, even if there is no number of moderate headaches whose elimination justifies the deliberate killing of an innocent person, such killing is possibly justified for the sake of, say, the saving of one person’s life and eliminating 100,000 moderate headaches.14

It 101–02, 144–64 (1993); T.M. Scanlon, What We Owe to Each Other 235–41 (1998). For a critique of the exclusion of small benefits, see Alastair Norcross, Rights Violations and Distributive Constraints: Three Scenarios, 76 Pac. Phil. Q. 159 (1995); Brock, supra note 10. 12. Brennan, thresholds, supra note 11, at 152. Brennan discusses the possibility that the smallest single benefit which is large enough to be taken into account equals the total benefit required to override the constraint (K in function (4) above). This would be the case, for example, if only preventing a person’s death is significant enough to justify the torture of an innocent person, but no more than one human life is necessary. This is, in fact, Thomson’s position (Thomson, supra note 4, at 166–69). According to this position, it may be permissible to inflict some pain on some people to save others from premature death, but it is impermissible to kill one person to save any number of people, or even to inflict one minor pain to alleviate any number of similar pains. For a critique of this position, see Brennan, thresholds, supra note 11, at 154–65; Richard J. Mooney, The Realm of Rights, by Judith Jarvis Thomson, 90 Mich. L. Rev. 1569, 1972 (1992) (Book Review). See also infra pp. 161–66. 13. The title of this excluder follows Brennan, Thresholds, supra note 11, at 153–54. 14. As Brennan points out, the existential constraint is meaningful only if it is higher than the “universal constraint” (excluding altogether too small benefits) and lower than the total benefit. Otherwise it converges with either of them, and even all three may converge. Brennan, Thresholds, supra note 11, at 152–53, 154.


Chronologically Remote Benefits and Costs Standard CBA employs discount rates to determine the present value of future benefits (and costs).15 Such discount rates may be part of constrained CBA as well. However, ordinary morality and threshold deontology may go one step further and exclude chronologically remote benefits from the threshold function altogether.16 Even if it is permissible to actively and intentionally kill one person to immediately save the lives of 101 others, arguably no number of people who will be saved in twenty years from now justifies such intentional killing. This claim may rest on the typical uncertainty of future events and the fear of errors,17 but it may also manifest people’s myopia.18 The threshold function may reflect any substantive moral judgment regarding the treatment of chronologically remote benefits by employing a suitable discount rate. For instance, in calculating the act’s net benefit, any of the act’s relevant consequences is multiplied by 1/(1+r)t, where r is some per-period (e.g., one year) discount factor, and t is the number of periods. Applying a sufficiently high discount rate will result in marginalizing the weight of future outcomes. Different discount factors may correspond to different types of costs and benefits. Probabilistic Costs and Benefits A threshold function should plausibly take into account the probabilistic nature of harms and benefits.19 A simple approach would be to calculate the

15. On the normative and methodological difficulties involved in CBA’s use of discount rates (especially discount rates for future lives), see generally Lisa Heinzerling, Regulatory Costs of Mythic Proportions, 107 Yale L.J. 1981 (1998); John J. Donohue III, Why We Should Discount the Views of Those Who Discount Discounting, 108 Yale L.J. 1901 (1999); Richard L. Revesz, Environmental Regulation, Cost-Benefit Analysis, and the Discounting of Human Lives, 99 Colum. L. Rev. 941, 943–48, 987–1017 (1999); Symposium: Intergenerational Equity and Discounting, 74 U. Chi. L. Rev. 1–246 (2007). 16. Larry Alexander, Deontology at the Threshold, 37 San Diego L. Rev. 893, 905 (2000). 17. The reluctance to consider remote benefits is particularly strong when their probability is relatively low. See Alexander, id. See also infra pp. 196–201, 210–11. 18. See Daniel A. Farber & Paul A. Hemmersbaugh, The Shadow of the Future: Discount Rates, Later Generations, and the Environment, 46 Vand. L. Rev. 267, 281–84 (1993) (reviewing contradictory studies of people’s discount rates for future lives, ranging from negative to infinite rates). On people’s tendency to attribute too little weight to future benefits and costs, as compared to present ones, see, e.g., Martin Feldstein, The Optimal Level of Social Security Benefits, 100 Q.J. Econ. 303, 307 (1985). 19. See Shelly Kagan, The Limits of Morality 87–91 (1989) (discussing probabilistic harms); Thomson, supra note 4, at 170–74 (concluding that probabilistic benefits should be taken into account even if the benefit has not materialized ex post). Possibly, there

89

90


act’s net benefit according to the expected value of the act’s consequences, i.e., to multiply all relevant harms and benefits by the probability of their occurrence. The probability may reflect the risk that the infringing act will not generate the anticipated benefit or the chance that the benefit will be achieved without infringing the deontological constraint. Arguably, such multiplication does not fully capture prevailing moral intuitions. Ordinary morality and threshold deontology plausibly hold that it is permissible to actively/intentionally cause a serious harm to a person (even as a side effect of saving others and certainly as a means to saving them) only if the probability of saving the others is quite high. Assuming it is permissible to intentionally kill one innocent person to save x lives, it is arguably impermissible to kill one person to save 4x people with a probability of 0.25, or even 2x with a probability of 0.5. This intuition may also be justified on a second-order consideration pertinent to low-probability costs and benefits. The further one is from absolute certainty the more one is prone to misjudge expected outcomes. Various studies have documented people’s systematic errors in dealing with low-probability events and discussed their policy implications.20 This consideration calls for extra caution in assigning to decision-makers the task of calculating and considering such effects, taking into account the decision-maker’s sophistication. Thus, the threshold function must determine how uncertain outcomes (both costs and benefits) are discounted in calculating the act’s net benefit, B. As indicated above, it may reflect “risk-neutrality” by calculating outcomes based on their expected value; but it may well attach different discount factors to uncertain outcomes. Assume that there is some probability p that if a person’s right not to be tortured is not infringed, x persons will be killed. The net benefit of infringing the constraint against torture can be zero for low levels of risks and may be only partially sensitive to changes in the level

is a constraint against actively/intentionally exposing people to a risk of harm, at least when they are aware of this exposure, even if the risk never materializes. Such exposure is certainly a cost that should be taken into account in choosing among deontologically permissible acts. See David McCarthy, Rights, Explanation, and Risks, 107 Ethics 205 (1997). 20. See, e.g., Colin F. Camerer & Howard Kunreuther, Decision Processes for Low Probability Events: Policy Implications, 8 J. Pol’y Analysis & Mgmt. 565 (1989); Paul Slovic, Baruch Fischhoff & Sara Lichtenstein, Regulation of Risk: A Psychological Perspective, in Regulatory Policy and the Social Sciences 241 (Roger G. Noll ed., 1985).


of risk. For instance, the act’s net benefit, B, can be calculated according to the following noncontinuous function: ⎧ 0 ⎪ (7) B = ⎨ 0.5px ⎪ px ⎩

if p < 5% if 5% ≤ p ≤ 50% if p > 50%

According to this function, if the probability of saving lives is less than 5%, no number of expected saved lives would render the torturing of a person permissible, and the expected number of saved lives is discounted by half if the probability of saving lives is between 5% and 50%. Alternatively, low probabilities may be discounted using a continuous function such as: (8) B = p 2x. We shall discuss the normative judgments underlying such functions in specific contexts later.21 Once again, even if low probability outcomes are excluded or discounted in the threshold function, they need not be similarly excluded or discounted when choosing between those acts that do not infringe, or that override, the deontological constraint.22 Last, commonsense morality also possibly distinguishes between saving an unidentified and an identified person from the same harm. Whereas from a consequentialist viewpoint, an act that would harm one unknown person out of a group is identical to an act that would inflict the same harm on a specific person (since the result in both cases is harm to one person), deontology may well distinguish between the two acts and find the latter more objectionable.23 Promoting the Good vs. Eliminating the Bad Contrary to standard economic wisdom, there is a prevailing intuition that normatively, increasing one’s utility is not as important as decreasing

21. See infra pp. 145–46, 147–48, 151–60, 164–65, 169–70, 199–201, 287, 316, 333, and 344–45. 22. Both standard CBA and constrained CBA may of course reflect risk aversion. 23. See, e.g., Sanford H. Kadish, Respect for Life and Regard for Rights in the Criminal Law, 64 Cal. L. Rev. 871, 893–94 (1976) (making this claim in the context of criminal law); T.C. Schelling, The Life You Save May Be Your Own, in Problems in Public Expenditure Analysis 127 (Samuel B. Chase, Jr. ed., 1968) (discussing the difference between individual death and statistical death).

91

92


one’s disutility.24 This moral judgment—often phrased in terms of avoiding pain versus promoting happiness—may be endorsed by consequentialists,25 but is primarily embedded in deontology. Were promotion of the good as compelling as eliminating the bad, the doing/allowing distinction would have collapsed. Whenever an actor abides by the prohibition against actively doing harm (e.g., refraining from killing one person) because the good produced by such harm is not large enough (e.g., saving the lives of two), she simultaneously avoids doing harm (to the one) and avoids doing good (to the two). Therefore, there must be an underlying notion that promoting the good is less morally compelling than eliminating the bad. 26 This judgment may vary from an absolute denial of the moral duty to promote happiness, to a slight preference for the avoidance of unhappiness over the promotion of happiness.27 A plausible version may refer to some reasonably acceptable level of well-being and attribute greater weight to bringing the worse-off to this level than to promoting people beyond it.28 All versions are subject to serious objections,29 yet at least the weaker versions reflect widely held intuitions.30 Thus, for instance, while it may be permissible to actively and intentionally torture one person as a means to preventing the premature death of x people, it seems impermissible to torture one person

24. See, e.g., 1 Karl R. Popper, The Open Society and Its Enemies 284–85, n.2 (5th rev. ed. 1966). This intuition is reflected, for example, in Model Penal Code § 3.02 (setting out that conduct that the actor believes to be necessary “to avoid harm or evil”—but not to produce benefit or good—is justifiable under certain circumstances). 25. James Griffin, Is Unhappiness Morally More Important than Happiness?, 29 Phil. Quart. 47, 47 (1979). In fact, this judgment is often labeled “negative utilitarianism.” See, e.g., R.N. Smart, Negative Utilitarianism, 67 Mind 542 (1958); A.D.M. Walker, Negative Utilitarianism, 83 Mind 424 (1974). 26. On this distinction, see Kagan, supra note 19, at 121–25 (a critique); F.M. Kamm, NonConsequentialism, the Person as an End-in-Itself, and the Significance of Status, 21 Phil. & Pub. Aff. 354, 381–82 (1992) (a reply to the critique). See also N. Ann Davis, The Priority of Avoiding Harm, in Killing and Letting Die 298 (Bonnie Steinbock & Alastair Norcross eds., 2d ed. 1994). 27. See Griffin, supra note 25 (critically discussing different versions of the “negative doctrine”). 28. Id. at 49–51. 29. See, e.g., Smart, supra note 25; R.I. Sikora, Negative Utilitarianism: Not Dead Yet, 85 Mind 587 (1976); Griffin, supra note 25. 30. For psychological evidence of this intuition, see, e.g., Jonathan Baron & Mark Spranca, Protected Values, 70 Org. Behav. & Human Decision Proc. 1, 10–11 (1997) (finding that people are much more supportive of medical intervention aimed at increasing newborns’ IQ from subnormal to normal than from normal to superior).


in order to prolong the lives of x (or more) people beyond their regular life expectancy. In a similar fashion (and more realistically), while preventing a large pecuniary loss may sometimes justify the unauthorized use of another person’s property or the breach of a contractual promise, such use or breach would not be permissible for the sake of making a comparable or even larger gain. To reflect this moral judgment, one may exclude gains from the threshold function altogether. Alternatively, and less drastically, one may attribute different weights to losses and gains: full weight to the disutilities avoided by the act and a reduced weight to the utilities it produces. Another (indirect but often powerful) way to take this distinction into account may rest on the fact that people’s willingness to pay (WTP) for an entitlement is often much smaller than their willingness to accept (WTA) regarding the same entitlement.31 Thus, one can use WTP to measure the gains the infringing act yields and WTA to measure the losses it prevents.

•

D. Shape and Size of the Threshold

Following the proposal to formalize threshold functions as T = B—K and the discussion in the previous section of the net benefit of the infringing act, B, this section focuses on K, the numerical representation of the threshold that has to be met for the act to be permissible. K reflects the strength of the relevant deontological constraint. The threshold can be quantified as some function of (plausibly certain types of) the harm(s) that the act inflicts but may also embody the general importance of the relevant deontological constraint. If active, yet unintended, infliction of harm on a person infringes a constraint and if the threshold that has to be met to justify such harming is lower than the one applying to an active and intended harm, then K should reflect this distinction as well.32 An important question is to what extent the size of the threshold is a function of actual (or probable) harm inflicted by the infringing act. As indicated

31. On this phenomenon (sometimes called “the endowment effect”), see generally Jack L. Knetsch & J.A. Sinden, Willingness to Pay and Compensation Demanded: Experimental Evidence of an Unexpected Disparity in Measures of Value, 99 Q.J. Econ. 507 (1984); David W. Harless, More Laboratory Evidence on the Disparity Between Willingness to Pay and Compensation Demanded, 11 J. Econ. Behav. & Org. 359 (1989). 32. See supra pp. 41–46; infra pp. 135–176, 213–218, 240–44, and 252–53.

93

94


earlier, a threshold function may be additive, multiplier, or a combination of these. To illustrate, the threshold function in the case of the constraint against active/intentional killing of innocent people may be additive and therefore independent of the actual harm, for instance: (9) T = x – y – 100 where x is the number of persons who will be saved if and only if the act is committed, y is the number of those who will be killed as a result of the act, and K (100 in this example) is the constant threshold. Alternatively, the function may be based on a multiplier, and therefore an increasing function of the actual harm: (10) T = (x – y) – 100 y where K ′ (100 in this example) is the multiplier. A third possibility is a combination of these two: (11) T = (x – y) – (50 + 50 y). In many contexts, multipliers or combined functions better capture the intuitions underlying threshold deontology. For instance, all the above three functions yield that it is permissible to kill one person ( y = 1) to save the lives of at least 101 others (x ≥ 101). However, when the actual harm (i.e., the number of people who are killed) is higher, the additive function (9) yields counterintuitive results. It yields that it is permissible to kill 10 to save at least 110 people and to kill as many as 10,000 to save just 10,100 people. In contrast, according to the multiplier function (10), it is permissible to kill 10 to save at least 1,010 and to kill 100 to save 10,100 people. According to the combined function (11), it is justified to kill 10 to save no less than 560, and to kill 197 to save 10,100 people. Indeed, the prevailing view seems to be that the “strength” of the deontological constraint—and thus, the size of the threshold—is an increasing function of the harm that is expected to be inflicted in the specific circumstances.33 A conspicuous shortcoming of a multiplier function is that whenever an infringement of a deontological constraint causes no harm, the threshold will be zero.34 For example, there would practically be no constraint against

33. See, e.g., Alexander, supra note 16, at 898–900; Brennan, Thresholds, supra note 11, at 157. 34. Samantha Brennan, How is the Strength of a Right Determined? Assessing the Harm View, 32 Am. Phil. Q. 383, 388–90 (1995). However, the infringement of the deontological constraint would still require that B ≥ 0 for the act to be permissible. Since B typically includes


breaking promises when it does no harm to the promisee. Furthermore, even if an act (be it breaking a promise or intentionally inflicting physical pain) does cause harm to another person, there would be practically no constraint against such act so long as the actor is willing to fully compensate the injured person (assuming such compensation is possible). While perfectly compatible with consequentialism and economic efficiency,35 this conclusion conflicts with deontological morality and the prevailing conception of moral and legal rights. A defender of multiplier threshold functions may respond that the very infringement of the constraint, the very fact that a lie is told or a promise is broken, constitutes some harm to the right-holder (and thus K is necessarily positive).36 Similarly, she may argue that compensation very rarely puts the injured person in as good a position as she would have occupied but for the breach, and multiplying this remaining harm would thus result in a positive threshold. These arguments carry some weight but do not seem wholly persuasive—which may lead one to the conclusion that the combined function (11) is superior to both the additive and the multiplier ones. Applying a function that sets the level of threshold as an increasing function of harm requires one to determine what types of harms should be counted in this respect. The question is then, which of the concrete elements of the infringement of the constraint are relevant in setting the size of the threshold? As indicated, one plausible consideration is the probability of harm. Arguably, if the probability of harm does not reach a certain minimum (either since there is a chance that the act will not engender harm after all, or that the same harm would come about anyway), the act does not infringe the constraint, and the threshold should indeed be zero (despite the fact that the infliction of the low-probability risk of harm was both active and intentional).37

only some of the act’s costs and benefits, it is at least theoretically possible that the act would be considered efficient under standard CBA, but deontologically impermissible. 35. A well-known example is the “efficient breach” doctrine, according to which remedies for breach of contract should urge the promisor to perform and take precautions to avoid breach so long as performance and such precautions are efficient, and to breach and avoid such precautions if breach is efficient. See, e.g., Richard A. Posner, Economic Analysis of Law 118–30 (7th ed. 2007); Steven Shavell, Foundations of Economic Analysis of Law 304–14, 342–62, 375–80 (2004); Steven Shavell, Damage Measures for Breach of Contract, 11 Bell J. Econ. 466 (1980). See also infra pp. 294–95. 36. Brennan, supra note 34, at 390. 37. For the claim that even imposing low-probability risks infringes moral rights (i.e., constraints), see McCarthy, supra note 19, at 212–15.

95

96


In addition, certain types of harms should plausibly not count in quantifying the threshold. It may be argued that in determining the size of the threshold of the constraint against actively/intentionally killing people, the only relevant factor is the number of people who will be killed, and no other costs engendered by the infringing act should be considered.38 Such additional costs may possibly be relevant in calculating the act’s net benefit (and certainly in choosing between several acts that meet the threshold). Even costs that involve infringements of deontological constraints may seem irrelevant when they are much smaller than the major infringement considered. Suppose that to save the lives of some people, one must not only kill one person but also break a promise or inflict a short, mild pain on another. While promise breaking and the infliction of a mild pain are infringements of deontological constraints, one may hold that they should not affect the size of the threshold determining the minimal number of saved people necessary to meet the constraint against active/intentional killing.

•

E. Other Concerns

Deontological moral theories are varied and complex. Some of the claims made by deontologists are quite compelling whereas others rest on questionable intuitions.39 To illustrate, the causal connection between an infringing act and its expected benefits (or costs) may be indirect. It may be, for example, that one or more of the people saved by an infringing act will then be able, thanks to their unique qualifications, to save the lives of additional people. A deontologist may or may not view the fact that some outcomes are not directly produced by the act as morally significant.40 Similarly, a deontologist

38. Cf. Kamm, supra note 11, at 146. Kamm argues that according to the principle of irrelevant utilities, when considering whether to redirect a threat away from five people in the direction of either Joe or Jim, the fact that directing the threat at one of them would also involve destroying a patch of beautiful flowers which gives pleasure to many people, should not be taken into account in choosing between the two individuals. For a critique of Kamm’s principle of irrelevant utilities, and specifically the claim that it dovetails with commonsense morality, see Brock, supra note 10, at 966. 39. See, e.g., John Broome, Kamm on Fairness, 58 Phil. & Phenomenological Res. 955, 958 (1998) (characterizing some of the arguments in Kamm, supra note 11, as depending entirely on the author’s intuitions, and thus as powerless against people who do not share these intuitions). 40. Kamm, supra note 11, at 104, 107–12.


may or may not distinguish between cases according to the allocation of benefits produced by the act (e.g., between a case in which the infringing act saves ten people from a moderate, temporary pain and the case in which it saves five people from two occasions of such pain each).41 Additional complexities are involved when a single act infringes more than one constraint. It may be that in preventing a person from expressing certain views, I not only infringe her freedom of speech but also break my promise not to act in this way. Plausibly, the accumulative effect of the two constraints requires that a greater amount of good be produced by the infringing act (compared to the case where only one constraint were infringed),42 yet the resulting threshold need not necessarily be a simple sum or a simple multiplication of the two separate thresholds. In principle, almost any distinction between different costs and benefits, any notion of fairness, and any approach to the accumulative effect of different constraints may be incorporated into a threshold function.43 At a certain point, however, the disadvantages of such incorporation (in terms of clarity and manageability of the formal representation) outweigh the advantages. As in economic modeling in general, there is no simple recipe for making such methodological decisions.

41. Brennan, Thresholds, supra note 11, at 151 (concluding that, in the context of excluding small benefits, each benefit should be considered separately, even if a single person gets more than one benefit). For a list of other “candidate factors,” see Brennan, supra note 34, at 390–91. Brennan’s list includes the fact that the beneficiary of the infringing act is the one who performs it; the fact that the beneficiary is also the person whose rights are infringed (see also Samantha Brennan, Paternalism and Rights, 24 Can. J. Phil. 419 (1994); infra p. 342); the number of people participating in infringing the constraint; and the fact that the right-bearer is part of a group whose rights are often infringed. Ritov and Baron found a strong correlation between the belief that the number of actors involved in causing a certain loss counts, and the belief that the number of actions involved in producing the same total loss counts. According to their experiments, however, most people do not share either of these beliefs; and furthermore, no significant correlation was found between holding these beliefs and viewing the infliction of the losses as violating deontological constraints. See Ilana Ritov & Jonathan Baron, Protected Values and Omission Bias, 79 Org. Behav. & Human Decision Processes 79, 89–90 (1999). 42. Brennan, supra note 34, at 390. 43. Some notions of fairness—particularly those opposing the aggregation of harms or those that deny the feasibility of interpersonal comparisons of costs and benefits—may, however, be incompatible with the very use of CBA, whether bounded by deontological constraints or not. On Taurek’s objection to aggregating harms to people, see John M. Taurek, Should the Numbers Count?, 6 Phil. & Pub. Aff. 293 (1977); supra pp. 54–55.

97

98


F. Threshold Options

As mentioned earlier, deontology not only imposes constraints on promoting the good but also recognizes options not to promote the good, at least sometimes. Like constraints, options need not be absolute. Moderate deontology plausibly maintains that options have thresholds. Incorporating threshold options into CBA would be useful when maximizing total welfare requires some sacrifice, which conflicts with the person’s self-interest. At times, a legal imposition of sacrifices would infringe a deontological constraint applying to the state. Compelling a person to sacrifice her life to save the lives of many more people is certainly such a case. The permissibility of such an infringement may be assessed using CBA subject to deontological constraints, along the lines discussed previously. However, not just any sacrifice imposed by the state entails an infringement of a deontological constraint. Under most accounts, collecting taxes to finance the provision of public goods and imposing precontractual disclosure duties to ensure the efficiency of transactions, for instance, do not infringe any constraint. Such legal duties do, however, deprive people of deontological options to use their money as they see fit and to keep information for themselves. Imposing such a duty is thus justified only if the expected good produced by depriving people of their options is large enough to warrant the required sacrifice. In constructing threshold functions for options, one could start from a prima facie (deontological) premise that no person is morally required (and, a fortiori, may not be compelled legally) to maximize the overall good unless the amount of good at stake is large enough to override the threshold option. Alternatively, one could start from the opposite, prima facie (consequentialist) premise: everybody is morally required (and may possibly be legally compelled) to maximize the total good unless the sacrifice involved is large enough to override this duty. We prefer the first formalization because it better conforms with the assumption of economic rationality underlying standard economic analysis and with the recognition of deontological options. Another advantage of this construction is that it captures the deontological notion and prevailing intuition that people have an option to promote other people’s welfare even if the benefit they confer upon those people is smaller than the cost to themselves.44

44. See, e.g., Michael Stocker, Agent and Other: Against Ethical Universalism, 54 Australasian J. Phil. 206 (1976); Shelly Kagan, Normative Ethics 166–67 (1998). By starting from


As a general matter, incorporating deontological options into CBA thus entails the characterization of a threshold function, To, such that an act is required—rather than optional—only if the product of this function is positive. Since threshold functions for options, like threshold functions for constraints, determine not only the size of the net benefits required for justifying the imposition of a duty (with a view to maximizing overall good) but also the types of benefits and costs that should be considered in this regard, we propose that, in its general form, the structure of threshold functions for options would be similar to that of a threshold function for constraint (function (4) above): (12) To = B – K = b(. . .) – k (. . .) where B is the relevant net benefit, and K is the size of the minimum amount of benefits that is required to justify the imposition of a duty to maximize the good (rather than a mere option to do so). Whenever To > 0, a duty may be imposed with a view to maximize the total good, while if To ≤ 0, there is no such duty, but only an option. Constructing particular threshold functions for options entails various normative judgments and methodological choices. Since there is some overlap between these issues and those discussed in the context of constraints, and since threshold functions for options are generally less useful than threshold functions for constraints, we shall suffice with a few comments as to the possible exclusion of certain benefits and costs and to the construction of the threshold. Proximity According to commonsense morality, there is a crucial difference between helping someone who needs immediate assistance and is standing directly in front of you, and helping someone who needs similar help, but who is geographically far away.45 Relatedly, the stronger the family, social, or community ties between the agent and the needy persons, the greater is the sacrifice a

the presumption that people have options unless the amount of total welfare at stake is large enough, one legitimizes options even if the net benefit of the sacrifice is negative (there is certainly no duty to act under such circumstances). In contrast, under the assumption that people should maximize the good unless the cost to themselves is too high, courses of action that yield negative net benefit are never justified (and thus there is no option, but rather sacrificing is immoral). Self-injurious courses of action may call for paternalistic interventions. See infra pp. 313–47. 45. See, e.g., F.M. Kamm, Intricate Ethics: Rights, Responsibilities, and Permissible Harm 345–97 (2007); Violetta Igneski, Distance, Determinacy and the Duty to Aid: A Reply to

99

100


person may be morally required to make to help them.46 There may be various ways to accommodate such judgments within threshold functions for options, ranging from completely denying the relevance of benefits to distant people, to attributing more or less steeply decreasing weight to people’s welfare as the (geographic or other) distance between them and the agent increases. Promoting the Good vs. Eliminating the Bad There is a very strong intuition that people may sometimes be required to make sacrifices in order to eliminate the bad, such as providing bare necessities to the underprivileged, but not in order to promote the good, such as further enhancing the welfare of the privileged (even if such sacrifice would increase aggregate welfare). Even within consequentialist morality, there is a sense that a distinction should be drawn between needs, “the biologically determined necessities of life,” and goals, “the chosen pursuits, projects, and endeavours which give life much of its meaning.”47 We critically discussed this distinction and its possible formalization in the context of threshold constraints.48 The discussion is applicable, mutatis mutandis, in the context of options. Probabilistic Costs and Benefits Arguably, a person may have a duty to bear a certain cost in order to save the life of another person with certainty, yet merely have an option to bear the same cost to save the lives of n people with 1/n probability of success, at least if n is rather large. For instance, even if one is required to donate $250 to save one person’s life, she is not necessarily required (though she has the option) to donate the same amount to save the lives of 100 people when there is a 99% chance that those people will die despite her donation. If one shares this moral judgment, then the benefit from the action may be treated in the same way we proposed to treat probabilistic costs and benefits in the context of constraints.49

Kamm, 20 Law & Phil. 605 (2001) (arguing that the distance merely serves as a proxy to the determinacy of the situation, which is morally significant). 46. See, e.g., Ronald Dworkin, Law’s Empire 195–216 (1986); Samuel Scheffler, Boundaries and Allegiances: Problems of Justice and Responsibility in Liberal Thought 1–130 (2001); Thomas Nagel, The View from Nowhere 164–75 (1986). 47. Tim Mulgan, The Demands of Consequentialism 173 (2001). On Mulgan theory, see generally id. at 169–294. 48. See supra pp. 91–93. 49. See supra pp. 89–91.


The Threshold’s Form: Additive, Multiplier, or Combined In discussing the form of threshold constraints, we concluded that very often, multiplier thresholds better capture the moral judgments of threshold deontology than additive ones.50 In the context of options as well, the level of the threshold should be an increasing function of the sacrifice required of the actor, that is, a multiplier:51 (13) To = (x – y) – K × y where x is the benefit to others, y is the cost to the actor, and K is a multiplier. K determines the inconvenience, y, that the person is required to suffer to bring about the net benefit (x–y). In fact, to the extent that the sacrifice refers to resources with decreasing marginal utility, such as money or time, it stands to reason that the level of the threshold should increase exponentially, as any additional unit of sacrifice in terms of money or time entails an increasing marginal sacrifice by the actor. For instance, the threshold function for options may take the following form: ⎧ (x − y ) − K lny if y > 0 (14) To = ⎨ if y = 0 ⎩ (x − y ) where the exponent is the natural logarithm function of y.52

50. At the same time, since some deontological constraints, such as the constraint against lying, may arguably be infringed even if nobody is harmed, and since some infringements may be accompanied by an offer of compensation, additive, or at least combined—additive and multiplier—thresholds may be warranted to avoid a threshold of zero. See supra pp. 93–95. 51. See, e.g., Samuel Scheffler, The Rejection of Consequentialism: A Philosophical Investigation of the Considerations Underlying Rival Moral Conceptions 20, 169 (rev. ed. 1994); Kagan, supra note 44, at 164. 52. In this function the value of K increases exponentially as y increases. The following table compares the value of the threshold in functions (13) and (14), for K = 10. y

K×y

K lny

1

10

1

2

20

5

5

50

40

10

100

200

100

1,000

40,000

1,000

10,000

8,000,000

101

102


Multiplier threshold functions for options, such as (13) and (14), imply that when a person can promote the total good at zero cost to herself ( y = 0), she enjoys no option. She is under a moral duty to maximize the good.53 In the context of constraints, the implication that lying or promise breaking that involves no (uncompensated) harm does not infringe a constraint, would seem objectionable to some. It is unclear whether the same is true in the context of options. If we judge that a person is under no duty to promote the good, even at no cost to herself, so long as the total good she can produce is lower than a certain minimum, then a multiplier threshold does not reflect this judgment. This is because when the cost to the agent is zero, so will be the threshold above which she is subject to a duty. In this case, one should prefer a combined (multiplier and additive) threshold to a multiplier one. While we have no clear intuition as to the right answer to this question in the moral sphere, it seems to us that such a minimum is indeed warranted in the legal sphere. Given the basic differences between the means used to enforce legal norms and those used to enforce moral norms, it stands to reason that the law should not restrict people’s options unless the amount of good produced by such a restriction exceeds a certain minimum. To express this judgment, combined (multiplier and additive) functions should therefore be employed by legal policy-makers, where the additive factor expresses the minimal net benefit required to give rise to a duty, even when the cost to the actor is zero or trivial. Maximal Sacrifice? Are there sacrifices that are so great that one would never be morally required to make, regardless of the net benefit they would yield? A positive answer to this question may hold that no person should be required to sacrifice her life, her entire career, or her dignity—even for the sake of saving the lives of millions of people.54 One may reflect this judgment by holding that To > 0 is not a sufficient condition to override an option. In addition, the cost to the

53. Even in that case, however, if one excludes the welfare of distant people or benefits whose probability is too low (and so forth, as considered above), then it may still be possible that under a multiplier threshold one would have an option, and not a duty, to act when the cost is zero. This would be the case where all benefits to other people brought about by the sacrifice are excluded from the threshold function. 54. Cf. Larry Alexander, Affirmative Duties and the Limits of Self-Sacrifice, 15 Law & Phil. 65 (1996) (critically discussing the possibility of an “invariant cap” on the sacrifice a person may be expected to make to save people he bears responsibility to).


agent must not surpass a certain upper limit. Different ceilings should be defined for different types of costs, such as monetary and nonmonetary (notwithstanding the fact that all costs might have to be monetized to facilitate the use of the threshold function). Such a ceiling may be set in absolute terms (for instance, no person should be required to donate more than $10,000 to charity), but in some contexts, it would make more sense to set it proportionally (no person should be required to donate more than 10% of her income), or exponentially (the richer a person is, the greater the percentage of her income she may be required to give to charity). Instead of setting such a ceiling, one could retain the agent’s option in all but very extreme cases simply by using an extremely high multiplier.55 Such a construction may, however, be undesirable because it would set a very high threshold even in cases that do not involve extreme sacrifices. Answers to these difficult questions are hard to come by on the moral plane. In the legal sphere, however, the questions themselves take on a different form. Once again, the dissimilarity lies in the different ways in which moral and legal norms are enforced. The question of whether a person is morally required to donate 50% of her income or forgo an extremely significant personal project in order to help the underprivileged, is quite different from the question of whether the legal system may impose such duties. In most, if not all, cases, laying down very demanding legal duties would not only take away people’s options but also constitute an infringement of a deontological constraint by the state. In such cases, the primary question is not whether the agent has an option but whether the state is justified in infringing the constraint against actively/intentionally harming people. Such questions should primarily be approached using deontologically constrained CBA, along the lines described in the previous sections, rather than threshold functions for options.

•

G. Concluding Remarks

This chapter outlined a general framework for implementing our proposal. Normative economic analysis uses formal notation to represent a social welfare function (SWF). Traditionally, the social policy reflected in the SWF is the maximization of aggregate social welfare; hence, the SWF is readily

55. Kagan, supra note 44, at 164–65.

103

104


translated into a simple CBA. Economists have long acknowledged that a SWF may take into account not only aggregate social welfare but also its distribution.56 We suggest that a SWF can and should reflect deontological constraints (and options) as well, using threshold functions. Deontologically constrained CBA does not determine the pertinent constraints on maximizing social welfare based on people’s preferences but rather on a normative judgment. It integrates such values as human dignity and autonomy as constraints on promoting the good, rather than as components of the good to be promoted.57 An infringement of a constraint is not yet another “cost” of the pertinent act or rule, to be considered along with other costs and benefits. Rather, constraints must not be infringed unless sufficiently large good (or bad) outcomes are at stake. Following moderate deontology, threshold functions set the minimal net benefit of the action, policy, or rule that has to be produced to justify an infringement of a constraint (or to remove an option not to promote the good). Typically, the pertinent net benefit includes only certain types of costs and benefits, thus excluding others. Various substantive and methodological objections may be raised against our proposed analytical framework, including its incompatibility with the alleged normative neutrality of standard CBA and the challenges it faces in quantifying and monetizing moral constraints. The following chapter addresses these objections.

56. See supra p. 18. 57. See supra pp. 30–32.

•

five

Addressing Possible Objections

following the discussion of the need to integrate deontological constraints

with CBA and the demonstration of how such integration may be formalized, this chapter considers possible objections to our proposal. These objections are principled, methodological, or both, and may come from economists, deontologists, or both. Section A addresses the claim that adding deontological constraints to standard CBA will undermine its normative neutrality. Section B discusses difficulties of quantifying and monetizing moral constraints. Section C examines the claim that using deontologically constrained CBA would lead to setting the constraints too low. Finally, section D considers the argument that the use of threshold functions by legal policy-makers and the enactment of legal norms that verbally embody such functions are incompatible with the expressive role of legal norms.

A. Undermining the Normative Neutrality of Economic Analysis •

While conceding the normative flaws of standard CBA, economists may still endorse a division of labor between economic analysis and deontological concerns to preserve the alleged value-free and objective nature of economic analysis. However, as many commentators have long pointed out, the idea that standard CBA is a value-free, “scientific” mode of analysis is false.1 The choice between competing notions of distributive justice within the consequentialist framework (such as mere maximization of utility or a Rawlsian maximin) is anything but value-free. The same is true regarding the choice

1. See, e.g., Daniel M. Hausman & Michael S. McPherson, Economic Analysis and Moral Philosophy 109–20 (1996); Michael B. Dorff, Why Welfare Depends on Fairness: A Reply to Kaplow and Shavell, 75 S. Cal. L. Rev. 847, 863–88 (2002); Joseph William Singer, Normative Methods for Lawyers, 56 UCLA L. Rev. 899, 915–21 (2009).

105

106


of the underlying theory of the good.2 Even the Pareto criterion entails nontrivial normative judgments.3 Thus, while adding deontological constraints to economic analysis obviously reflects a normative judgment, disregarding such constraints does so as well. There is nothing more “subjective” in setting objectively defined constraints to maximizing well-being than in not setting such constraints. Rather than claiming that standard economic analysis is free of normative judgments, one may argue that it is neutral in the more limited sense that it rests on uncontroversial normative judgments. Very few would deny that maximizing human well-being is a worthwhile goal. Such a consensus is, however, besides the point. What deontologists deny is not that outcomes count but that outcomes are the only thing that ultimately counts; it is this very controversial claim that lies at the center of our discussion.4 Another possible objection is that even if standard CBA reflects normative judgments, once these judgments have been made, the routine use of CBA requires merely empirical, value-neutral investigations.5 In response, it should first be noted that many of the normative questions underlying CBA are not yet settled. For example, it is unclear whether the theory of the good underlying standard CBA should be the satisfaction of actual preferences or the satisfaction of rational or ideal ones.6 Second, the current state of

2. Cf. David Lyons, The Moral Opacity of Utilitarianism, in Morality, Rules, and Consequences: A Critical Reader 105 (Brad Hooker et al. eds., 2000) (making a similar argument in response to the claim that utilitarianism rests on “objective grounds” and its dictates determined by “empirically determinable facts”); I.M.D. Little, Ethics, Economics, and Politics 23 (2002) (“Some feel that economics becomes unscientific if value judgments are admitted: they are trying to fly without wings”). 3. See, e.g., Amartya Sen, The Impossibility of a Paretian Liberal, 78 J. Pol. Econ. 152 (1970) (demonstrating the potential incompatibility of the Pareto principle with liberal values and concluding that the former is unacceptable as a universal rule); Louis Kaplow & Steven Shavell, Fairness Versus Welfare 52–58 (2002) (demonstrating the potential incompatibility of the Pareto principle with nonconsequentialist concerns and concluding that the latter should be ignored); Howard F. Chang, A Liberal Theory of Social Welfare: Fairness, Utility, and the Pareto Principle, 110 Yale L.J. 173 (2000) (arguing that it is possible to construct a deontological theory subject to the constraint that every constraint would be overridden if following it would infringe the Pareto principle); Mark D. White, Pareto, Consent, and Respect for Dignity: A Kantian Perspective, 67 Rev. Soc. Econ. 49 (2009). 4. Cf. Jeremy Waldron, Criticizing the Economic Analysis of Law, 99 Yale L.J. 1441, 1460 (1990) (reviewing Jules J. Coleman, Markets, Morals, and the Law (1988)). 5. Cf. Partha Dasgupta, What do Economists Analyze and Why: Values or Facts?, 21 Econ. & Phil. 221, 221–22 (2005) (“[T]he ethical foundations of [modern economics] were constructed over five decades ago and are now regarded to be a settled matter.”). 6. See supra pp. 15–16, 31–32; infra pp. 323–25.

addressing possible objections

standard CBA reflects decades of methodological refinement. Constrained CBA may benefit from similar academic refinement over time. More precisely, the emergence of methodological conventions regarding the handling of these questions is likely to make the pertinent normative judgments less conspicuous, as has been the case with standard CBA. A related objection is that even if standard CBA reflects normative judgments and even if some of these judgments are still debated, economists do not have the philosophical expertise necessary to engage in the moral deliberation required for setting deontological constraints. In considering this argument, one should distinguish between academic use of CBA and its use by governmental agencies. As regards academic analysts, and particularly those engaging in economic analysis of law, to the extent that one can distinguish between normative and positive analyses,7 constrained CBA is primarily germane to the former (though it may also contribute to positive analysis of people’s behavior, influenced by ordinary morality, and of the legal system, embodying such constraints). Inasmuch as economic analysis of law aims to enhance the normative discussion of legal issues, consideration of deontological concerns seems both essential and feasible. Just as mainstream legal theory has embraced the economic perspective and as mainstream economic analysis of law is gradually embracing the insights of cognitive psychology (the so-called “behavioral law and economics”), there is no compelling reason why economic analysis of law should not pay heed to moral concerns and incorporate them into CBA. Constrained CBA would contribute to the operationality of moderate deontology and may even enrich the philosophical debate about threshold constraints. As regards regulatory agencies, the requirement to conduct CBA restricts the agencies’ discretion and reduces the risk of error.8 Authorizing the agencies to engage in philosophical inquiries would arguably broaden their discretion and complicate the oversight process. According to this view, deontological constraints should not be incorporated into CBA but rather set by other governmental authorities, along with other constraints imposed

7. In actuality, this distinction is very often blurred. See Hausman & McPherson, supra note 1, at 214–20. 8. Cass R. Sunstein, Cognition and Cost-Benefit Analysis, 29 J. Legal Stud. 1059 (2000); Matthew D. Adler & Eric A. Posner, New Foundations of Cost-Benefit Analysis 101–23 (2006); Richard L. Revesz & Michael Livermore, Retaking Rationality: How Cost-Benefit Analysis Can Better Protect the Environment and Our Health 12–13 (2008).

107

108


on agencies.9 Indeed, constitutional and institutional considerations dictate that politically accountable bodies such as Congress, rather than regulatory agencies, should construct threshold functions, either mathematical or verbal.10 It does not follow, however, that—inasmuch as agencies face decisions involving deontological constraints—these constraints cannot or should not be incorporated into CBA. Once the ethical and policy judgments are made and a threshold function is constructed, the regulatory agencies and other governmental branches may, and should, routinely employ this function. Agencies specializing in measuring and evaluating costs and benefits are well-suited to employ threshold functions requiring assessment of the regulation’s harms and benefits. Rather than broadening their discretion, the explicit application of such functions may actually facilitate greater transparency and accountability of agencies’ activities.

•

B. Quantification and Monetization Difficulties 1. General

Another objection to incorporating deontological constraints into economic models has to do with the difficulty in quantifying and expressing such constraints in monetary terms. Standard economic models assume that people have a complete ordering of preferences regarding anything that may affect their well-being and that they are able to attribute a dollar value to any entitlement (or lack thereof). Thus, at least in principle, economic models may rest on an empirical, factual examination of people’s willingness to pay (WTP) for anything, or the sum for which they would be willing to give away something they already have—their willingness to accept (WTA). Some may argue that deontological constraints cannot be monetized in the same way.

9. Cf. Peter G. Sassone & William A. Schaffer, Cost-Benefit Analysis 160 (1978) (“A project must satisfy a number of diverse constraints. Such constraints may be budgetary, legal, social, political, or institutional . . . .[These constraints] exclude alternative projects that obviously are not feasible . . . . [For example] benefits and costs cannot be divided along racial lines.”). 10. See Matthew D. Adler, Beyond Efficiency and Procedure: A Welfarist Theory of Regulation, 28 Fla. St. U. L. Rev. 241, 313–15 (2000) (tentatively considering the desirability of institutional separation between consequentialist inquiries and setting the deontological constraints).


One possible response to this concern is that quantification and measurement problems are not unique to deontological constraints. Even in distinctively market contexts, the above assumptions of standard economic analysis are problematic. It is often practically impossible to determine, for instance, the market value, or even the value for a certain person, of a specific contractual term. This difficulty is exacerbated when economic analysis addresses nonmarket issues such as the hearsay rule in evidence law and criminal liability. Despite the impossibility of filling in the equations with actual dollar amounts, economic analyses of such issues often yield important insights. As Richard Posner notes about his model of the regulation of free speech: “I offer these formulas as a heuristic, a way of framing and thinking about the regulation of speech, rather than as an algorithm for use by judges.”11 Economic models may highlight the crucial factors and their interrelations, explain and predict the way people would act under different legal regimes, and indicate what rules or standards may bring about the greatest social utility under specified conditions. Rigorous, qualitative economic analysis is not necessarily less revealing than a quantitative one (in fact, it is often more revealing), and the same holds true for constrained economic analysis. Another possible response is that deontological constraints need not be expressed in monetary terms. As demonstrated, threshold functions often disregard some or many of the costs and benefits involved.12 The more costs and benefits are excluded from the function, the less it is necessary to commensurate different costs and benefits. For example, if under a certain threshold function, the only benefit that may ever justify the active/intentional killing of x innocent people is saving the lives of at least K times x people, then no monetization of either human life or of K is necessary. The above-mentioned two responses bypass the objection that deontological constraints are not monetizable by indicating that a meaningful economic analysis subject to deontological constraints may be conducted without such monetization. But is the claim that deontological constraints are not monetizable warranted? Indeed, since deontological constraints rest on normative judgment, the usual technique economists use to monetize goods, legal entitlements, and other things—people’s WTP or WTA—is

11. Richard A. Posner, Frontiers of Legal Theory 62, 68 (2001). See also infra pp. 184–89. 12. See supra pp. 86–93; infra pp. 147–49, 168, 195–211, and 342–46.

109

110


inappropriate in the present context.13 In this respect, even a move from actual to ideal preferences would not overcome the difficulty. If one can sensibly translate deontological constraints into monetary terms, the additional step of imputing the resulting measure to people’s ideal judgments is largely fictitious and essentially superfluous (and if one cannot sensibly make this translation, no reference to ideal preferences would do the work). Hence, we have to address the objection that monetization of deontological constraints is impossible or undesirable because deontological constraints are incommensurable with well-being, or at least with the ordinary measure of well-being used in economic analysis, money. There is indeed a deeply held intuition about the incommensurability of different spheres of values and relations, including the incommensurability of wealth and deontological constraints. Not only such things as human life, trust, and liberty cannot and should not be traded in market transactions; people strongly resent even considering questions such as what is the monetary worth of one’s child or of one’s sexual autonomy.14 Preserving the separation between different spheres of valuation, in particular between the market and nonmarket spheres, is arguably essential for human identity and human flourishing.15 Beyond these general intuitions, there are three more concrete claims that can be made against monetizing deontological constraints: anti-commodification, incomparability, and incommensurability. We shall take these claims in turn.16

13. See supra pp. 27–29. 14. See, e.g., Jonathan Baron & Mark Spranca, Protected Values, 70 Org. Behav. & Human Decision Proc. 1 (1997) (documenting people’s resistance to trade-offs between deontological concerns and economic values); Alan Page Fiske & Philip E. Tetlock, Taboo Trade-offs: Reactions to Transactions That Transgress Spheres of Justice, 18 Pol. Psychol. 255 (1997) (arguing that people find questions regarding the monetary worth of one’s children or one’s loyalty to her country morally offensive). 15. See generally Michael Walzer, Spheres of Justice 95–128 (1983); Elizabeth Anderson, Value in Ethics and Economics (1993); Margaret Jane Radin, Contested Commodities 46–53 (1996) (criticizing the notion of strict separation between the market and nonmarket domains); Wendy Nelson Espeland & Mitchell L. Stevens, Commensuration as a Social Process, 24 Ann. Rev. Soc. 313 (1998). 16. These interrelated concepts and their implications for ethics, law, and economics, have been extensively debated in the literature. Noteworthy contributions include: Joseph Raz, The Morality of Freedom 321–66 (1986); Anderson, supra note 15; Incommensurability, Incomparability, and Practical Reason (Ruth Chang ed., 1997); Rethinking Commodification: Cases and Readings in Law and Culture (Martha M. Ertman & Joan C. Williams eds., 2005); Symposium, Law and Incommensurability, 146 U. Pa. L. Rev. 1169–1731 (1998); Cass R. Sunstein, Incommensurability and


2. Anti-Commodification It has been argued that monetization might bring about commodification. It may lead to expansion of the market domain by turning more and more objects into tradable commodities, or at least to making people view them as commodities, thus ignoring their intrinsic value.17 Such expansion of the market domain is considered detrimental to human flourishing. However, this thesis rests on debated philosophical claims regarding human nature and flourishing, and on questionable empirical assumptions regarding the effect of monetization on people’s perceptions of goods, values, and relations.18 For our purposes, it is sufficient to point out that there is no necessary link between monetization and commodification. When deontological constraints are translated into money for the purpose of incorporation into CBA, this is not done with a view to transforming the values underlying these constraints into tradable goods. On the contrary, it is done to constrain economic analysis that would otherwise treat anything as commensurable with anything else. When monetization is based on normative deliberation rather than on market prices (or people’s preferences more generally), it is perfectly compatible with maintaining the nontradability and even inalienability of the monetized object. Thus, courts routinely award monetary damages for loss of limb or life in tort actions, without thereby transforming them into tradable goods19 or necessarily committing to a commodified conception of compensation.20 In a similar fashion, constrained CBA may quantify deontological constraints in monetary terms without turning them into commodities.

Valuation in Law, 92 Mich. L. Rev. 779 (1994); Donald H. Regan, Authority and Value: Reflections on Raz’s Morality of Freedom, 62 S. Cal. L. Rev. 995, 1056–75 (1989). 17. See Radin, supra note 15, at 95–122 (critically discussing the “domino theory,” according to which one “cannot both know the price of something and know that it is priceless” and therefore commodification of an object precludes coexistent noncommodified understanding of the same object). 18. For a nuanced analysis of the conflicting arguments and assumptions, see Radin, supra note 15. See also Scott Altman, (Com)modifying Experience, 65 S. Cal. L. Rev. 293 (1991); Martha C. Nussbaum, The Costs of Tragedy: Some Moral Limits of Cost-Benefit Analysis, 29 J. Legal Stud. 1005, 1031 (2000). 19. Stephen R. Munzer, An Uneasy Case Against Property Rights in Body Parts, 11 Soc. Phil. & Pol’y 259, 281–84 (1994). 20. See Margaret Jane Radin, Compensation and Commensurability, 43 Duke L.J. 56 (1993).

111

112


3. Incomparability Following Ruth Chang’s suggestion, we shall use the term “incomparability” to denote the inability or undesirability of ranking two options or items and the term “incommensurability” to denote the inability or undesirability of precisely measuring two options or items by some common scale of units of value.21 Two options are thus incomparable if it is neither true that one of them is better than or preferable to the other nor true that they are of equal value.22 Two options may be comparable, that is, ordinally compared, even if a cardinal comparison between them is impossible (i.e., they are incommensurable). Thus, incomparability entails incommensurability, but the reverse is not necessarily true. Note that even incomparabilists do not claim that people do not actually choose among “incomparable” options; their only claim is that such choices cannot be justified on the basis of meaningful ranking of the options.23 While incomparability is intuitively appealing, there are strong arguments against this notion. Even if the compared courses of action or values are different in kind and cannot be evaluated on the same scale, a comparison may still be possible. For example, it seems morally permissible to touch someone’s elbow, or even to push him, to save another person’s home from burning down, and impermissible to torture an innocent person to save another person’s car from being destroyed. These judgments rest on comparisons. Even when it is asserted that no monetary gain whatsoever justifies the deliberate killing or enslavement of a person, it does not imply that human life or liberty is incomparable with money. Rather, it means that life and liberty are considered lexically more valuable than any sum of money.24

21. Ruth Chang, Introduction, in Incommensurability, Incomparability, and Practical Reason, supra note 16, at 1–2. 22. See, e.g., Raz, supra note 16, at 322 (Raz refers to “incommensurability” while we use his formulation to define incomparability). As some commentators have pointed out, there may be an additional relation between two options, namely that of rough equality. See generally Chang, supra note 21, at 4–5, 23–27; James Griffin, Well-Being: Its Meaning, Measurement, and Moral Importance 79–81, 96–98 (1986). While the notion of rough equality may provide a better understanding of apparent incomparabilities, it poses difficulties to money commensurability and thus to CBA. See Jonathan Aldred, Cost-Benefit Analysis, Incommensurability and Rough Equality, 11 Envtl. Values 27 (2002). 23. Raz, supra note 16, at 335–40; Richard Craswell, Incommensurability, Welfare Economics, and the Law, 146 U. Pa. L. Rev. 1419, 1423–24 (1998). 24. Daniel Statman, Moral Dilemmas 55–72 (1995); Chang, supra note 21, at 19–21; James Griffin, Incommensurability: What’s the Problem?, in Incommensurability,


Making a personal choice between a career opportunity and family commitments, or a public choice between different policies having both monetary implications and health and safety effects, may be an agonizing process. But this anguish does not prove that the values involved are incomparable. It simply indicates that the stakes are high and that a cautious weighing of the alternatives is warranted. Had the alternatives been truly incomparable, one could have flipped a coin.25 Finally, even if incomparability is a valid notion, it is neither available to consequentialists nor to moderate deontologists. Global comparability is a characteristic feature of consequentialism, and threshold deontology presupposes comparability between constraints and promoting the good (or avoiding the bad).

4. Incommensurability We thus arrive at the argument that deontological constraints are incommensurable with money. To put this argument in perspective, recall that money commensurability is a standard feature of CBA. Arguments of incomparability and incommensurability are thus often intertwined with critiques of consequentialism and economic analysis, whereas objections to the notion of incommensurability are frequently made by proponents of the economic methodology.26 Drawing the battle lines in this way can be misleading, however. Although economists regularly monetize such things as human lives and body integrity, the method they use to that end— aggregation of WTP or WTA—is inapplicable to deontological constraints, which rest on a normative judgment rather than an aggregation of preferences. An economist may thus hold that there is no acceptable way to

Incomparability, and Practical Reason, supra note 16, at 35; Regan, supra note 16, at 1058–59. 25. Regan, supra note 16, at 1059–64; Larry Alexander, Banishing the Bogey of Incommensurability, 146 U. Pa. L. Rev. 1641 (1998). 26. See Radin, supra note 20, at 64; Jeanne L. Schroeder, The Laconomics of Apples and Oranges: A Speculative Analysis of the Economic Concept of Commensurability, 15 Yale J. L. & Human. 347, 352 (2003). Instances of such correlation include, on the one hand, Frank Ackerman & Lisa Heinzerling, Priceless 39–40 (2004) (criticizing CBA of health and environmental issues), and, on the other hand, Craswell, supra note 23 (arguing that even if theories of incommensurability are correct with regard to individual decisions, as long as individuals do make choices, these theories need not affect policy-making based on people’s revealed preferences).

113

114


monetize deontological constraints. At the same time, the moderate deontologist’s objection to consequentialism need not entail a belief in the incommensurability of constraints with money. To be sure, many deontologists object not only to consequentialism but also to a theory of the good based on preferences satisfaction, and all the more so to quantifying preferences by a monetary scale. Yet as far as we can see, there is no necessary connection between these positions. The following arguments are meant to convince the economist and the deontologist that monetizing deontological constraints is both feasible and worthwhile, at least sometimes. As regards the economist, it is not inconceivable to attribute monetary value to a deontological constraint. First, when it comes to issues such as the monetary value of freedom of speech or truth telling, the supposedly strict division between factual identification and aggregation of preferences, and between normative determinations, is rather illusive. The methodological difficulties of identifying people’s preferences almost inevitably entail the exercise of discretion, and such discretion may well reflect normative choices. In any event, we have already addressed and rejected the idea that economic analysis can ever be value-free. Just as it regularly monetizes such things as human lives, personal injuries, and the existence value of wild ecosystems, economic analysis should also be able determine the monetary value of deontological constraints based on a normative judgment. On a higher level of abstraction, economists should feel more comfortable with monetization of deontological constraints if threshold constraints on the factoral level are conceived of as resting on consequentialism on the foundational level.27 From a different angle, the scope of any threshold constraint in a given legal system may, at least theoretically, be derived from a comparison between the existing rules and the rules that would have been set on a purely consequentialist basis. For example, assume (counterfactually) that standard CBA unequivocally calls for breaching contracts whenever performance is inefficient (the “efficient breach doctrine”).28 The extent to which existing contract law deviates from the efficient breach doctrine by “excessively” deterring breaches may reflect a deontological constraint against promise breaking. If all other variables are monetized, then one should be able, at

27. See Larry Alexander, Deontology at the Threshold, 37 San Diego L. Rev. 893, 910–11 (2000); supra pp. 24–27. 28. See generally infra pp. 294–97.


least in theory, to extract the monetary value of the legally imposed constraint as well.29 As for the moderate deontologist, since she acknowledges that enough good or bad outcomes may override a deontological constraint, she cannot deny the comparability of constraints and the goodness or badness of outcomes. In the legal context, she may deny that it is possible or desirable to set a deontological constraint ex ante in very exact terms and prefer a vague standard. Yet, just as a court must ex post determine the monetary damages for loss of life or limb, so too must it determine ex post the legality, validity, or enforceability of an action according to whether the deontological constraint has been violated or overridden. In general, the notion that such values as basic freedoms and human life are incommensurable with money at least partially stems from the fact that money has no intrinsic value. It is merely a means to achieve other goals. Money often bears a negative symbolic significance, connecting greed and materialism. However, this is a distorted perception. Money can be and is often used to promote the most intrinsically valuable goals, including saving lives, legally protecting human rights, and creating inspiring works of art.30 As unfortunate as it may be, tradeoffs between things such as safety and money (representing any alternative use) are unavoidable. This recognition is one reason to prefer moderate to absolutist deontology. Since it is difficult to draw a boundary between a willingness to monetize things such as physical pain and unwillingness to monetize deontological constraints against actively/intentionally harming people, arguably one should either reject money commensuration tout court, or concede to it entirely, at least under appropriate circumstances. A principled antimonetization implies that CBA and economic methodology should only apply to market issues (and even in this context they would face considerable difficulties).31 Taking this position seriously would not only rule out the

29. Cf. Robert Sugden & Alan Williams, The Principles of Practical Cost-Benefit Analysis 178–97 (1978) (discussing the inference of valuations underlying actual policies). 30. Regan, supra note 16, at 1070; James Griffin, Are There Incommensurable Values?, 7 Phil. & Pub. Aff. 39, 52 (1977). 31. The difficulties stem from the fact that, both descriptively and normatively, even market behavior is subject to various nonconsequentialist moral and social norms. See Hausman & McPherson, supra note 1, at 214–20 (demonstrating how people’s moral dispositions affect, and are affected by, economic outcomes); Radin, supra note 15, at 102–14 (analyzing the pervasive interactions between values of personhood and community, and the market).

115

116


incorporation of threshold constraints into CBA, but also the very use of CBA in such spheres as health and safety regulation and environmental protection (not to mention family law, criminal law, and human rights). To be sure, a world without economic analysis of nonmarket issues is not inconceivable; it existed only several decades ago and still exists in most parts of the globe. But whatever the normative deficiencies of economic analysis, doing away with CBA as a governmental decision procedure and with economic analysis of law seems most undesirable. In sum, while monetization poses a difficulty regarding the incorporation of deontological constraints into economic analysis, such monetization is not necessarily essential to modeling constraints, and monetization based on normative deliberation is not impossible. We submit that it is preferable to address deontological concerns as constraints on CBA rather than to either ignore them altogether or deal with them separately as secondary considerations.

•

C. Setting Constraints Too Low

While moderate deontologists should find economic analysis subject to constraints more acceptable than standard economic analysis, they may nevertheless object on the grounds that it would lead to constraints that are set too low. As a matter of pure logic, deontological thresholds may be set at any point higher than zero (consequentialism) and lower than infinity (absolutist deontology). However, deontological constraints are ordinarily thought of as violable only in unusual cases.32 The deontologist believes that values such as human life and basic liberties take priority over the good and should not be routinely traded off against satisfaction of preferences. The very idea that deontological constraints are commensurable with well-being may push

32. See, e.g., Robert Nozick, Anarchy, State and Utopia 29, n.* (1974) (if constraints may ever be infringed, it would be only “to avoid catastrophic moral horror”); Charles Fried, Right and Wrong 10 (1978) (maintaining that constraints may be infringed to avoid “the catastrophic”); Shelly Kagan, The Limits of Morality 113 (1989) (arguing that moderate deontologists believe the threshold of the constraint against harming is “quite high,” if there is a finite threshold at all); Carol S. Steiker, No, Capital Punishment Is Not Morally Required: Deterrence, Deontology, and the Death Penalty, 58 Stan. L. Rev. 751, 783–84 (2005) (arguing that ordinarily deontological constraints may be overridden in “emergency situations” only).


in the direction of lower thresholds,33 and the incorporation of constraints into economic models is likely to reinforce this tendency. As a practical matter, one would not go to the trouble of calculating costs and benefits against a deontological threshold if it is only in circumstances of colossal catastrophe that the analysis would be relevant. The deontologist’s concern that our proposal would lead to setting lower thresholds is therefore not groundless. In response, we note that nothing in our proposal necessitates lower thresholds. Additionally, one may point out that in certain contexts, such as the constraints against lying or promise breaking, rather low thresholds may actually capture prevailing moral intuitions.34 Moreover, even in the context of the constraint against harming people, facing the questions of when and why extremely high constraints are justified may not be a bad idea. In general, the present concern is considerably less compelling if the justification for threshold constraints on the factoral level rests on foundational consequentialism.35 At the same time, considerations of long-term and indirect effects make even act-consequentialists rather cautious in allowing infringements of commonsensical moral prohibitions.36

•

D. Incompatibility with the Expressive Role of Law

Deontologists (and consequentialists) may object to determining the permissibility of infringements through the construction of explicit threshold functions on the grounds that using such functions is incompatible with the expressive role of law. This objection rests on the notion that legal provisions do not only impose duties and convey rights. They also express attitudes, shape public perceptions, and may thus inflict “expressive” harm. For instance, a statute authorizing the torture of a suspect terrorist or the killing of an innocent person, if and only if such torture or killing is expected to save

33. Michael S. Moore, Torture and the Balance of Evils, 23 Is. L. Rev. 280, 330–31 (1989); cf. Frederick Schauer, Instrumental Commensurability, 146 U. Pa. L. Rev. 1215 (1998) (discussing the likely effects of decision-makers’ dispositions regarding commensurability (or incommensurability) of values and courses of action on their decision procedures and actual decisions); supra pp. 52–53. 34. Itzhak Kugler, Direct and Oblique Intention in the Criminal Law 40–41 (2002). 35. See supra pp. 24–27. 36. See supra pp. 22–24.

117

118


at least 100 lives with sufficiently high probability, can be considered disrespectful of human dignity. This objection partially overlaps with the critique leveled against moderate deontology in general and the opposition to monetizing constraints;37 yet it adds to these critiques an institutional aspect, namely that using threshold functions by policy-makers and enacting legal norms that verbally embody such functions are especially objectionable due to the expressive effect of such schemes. According to this objection, while infringements of deontological constraints may sometimes be justified, it is undesirable to formally authorize such infringements. Agents, including public officials, may act extra-legally when necessary in the face of calamity,38 but such acts should only be evaluated ex post factum.39 This position is implicit, for instance, in the German Constitutional Court’s judgment regarding the validity of a statute authorizing officials to shoot down an aircraft that is being wielded as a deadly weapon and in the judgment of the Israeli Supreme Court regarding the use of force in interrogations.40 In both cases, the courts held that granting ex ante authorization to inflict such harm is unacceptable, but did not rule out the possibility of granting a criminal law defense to officials who resort to such measures if the action is deemed justified ex post.41 At the core of this view lies the concern that formal legitimization of infringements is objectionable from a deontological perspective and undesirable from a consequentialist one. To assess this argument, one should distinguish between different forms of legalizing infringements. The two paradigms are an explicit ex ante authorization to infringe a constraint under specified circumstances and

37. See supra pp. 52–53 and 108–16, respectively. See also supra pp. 73–74. 38. See, e.g., Richard A. Posner, Not a Suicide Pact: The Constitution in a Time of National Emergency 152–58 (2006). 39. See, e.g., Oren Gross, Are Torture Warrants Warranted? Pragmatic Absolutism and Official Disobedience, 88 Minn. L. Rev. 1481, 1526–34 (2004); Alon Harel & Assaf Sharon, “Necessity Knows No Law”: On Extreme Cases and Uncodifiable Necessities, 60 U. Toronto L.J. (forthcoming). 40. Judgment of the Bundesverfassungsgericht [BVerfG] [German Federal Constitutional Court] Feb. 15, 2006, BVerfGE, 1 BvR 375/05, available at http://www.bundesverfassungsgericht. de/entscheidungen/rs20060215_1bvr035705.html; HCJ 5100/94 Public Comm. against Torture in Israel v. State of Israel, 53(4) PD 817. An English translation is available at 38 I.L.M. 1471 (1999) (Isr.). 41. See Judgment of the Bundesverfassungsgericht, supra note 40, at § 128; Public Comm. against Torture in Israel v. State of Israel, supra note 40, at § 38. See also Article 8(2)(b)(xxiii) of the Statute of the International Criminal Court.


an absolute ex ante prohibition coupled with a willingness to justify or excuse infringements ex post. A system of ex ante authorization may vary according to several factors.42 These include the identity of the authorizing body (the legislature, the executive, or the courts);43 the level of specificity of the authorization and the corresponding scope of discretion left to the actors; the public or secret nature of the authorization; and its timeframe (whether it is limited to periods of war or emergency).44 Absolute ex ante prohibition coupled with ex post assessment of infringements may similarly take different forms. The ex post assessment can be formally conducted by a judicial body, or informally carried out through public and political deliberation. The legal effect of finding the infringement permissible may also vary, constituting either a criminal law defense of “excuse” or one of “justification.”45 Taking cognizance of these distinctions, one may object to explicitly authorizing in advance certain infringements because such authorization is likely to bring about unjustified infringements. It may lead to routine use, or at least routine consideration, of measures that should be taken, or even considered, only in extreme and rare circumstances.46 Moreover, granting such powers may be misinterpreted as imposing a duty to infringe a constraint and may thus result in a large number of infringements. In certain

42. For a comparative review, see, e.g., John Ferejohn & Pasquale Pasquino, The Law of the Exception: A Typology of Emergency Powers, 2 Int’l J. Const. L. 210 (2004). 43. For instance, it was suggested that aggressive interrogation techniques should be implemented on the basis of judicial “torture warrants.” See Alan M. Dershowitz, The Torture Warrant: A Response to Professor Strauss, 48 N.Y.L. Sch. L. Rev. 275 (2003); Gross, supra note 39. 44. Cf. Bruce Ackerman, Before the Next Attack: Preserving Civil Liberties in an Age of Terrorism (2006); Oren Gross, Chaos and Rules: Should Responses to Violent Crises Be Constitutional, 112 Yale L.J. 1011 (2003). 45. Miriam Gur-Arye, Can the War Against Terror Justify the Use of Force in Interrogations? Reflections in Light of the Israeli Experience, in Torture: A Collection 183 (Sanford Levinson ed., 2004). See also Posner, supra note 38, at 152–58 (suggesting a judicial doctrine of “national security necessity,” that would extend a form of qualified immunity “to national security officials who violate a constitutional right in good faith in compelling situations of necessity,” as a better and simpler alternative to presidential pardons). 46. Henry Shue, Torture, 7 Phil. & Pub. Aff. 124, 141 (1978) (arguing that while torture may be justified when it is the least harmful means available to secure a supremely important aim, it should nevertheless be strictly prohibited, since “[a]ny practice of torture once set in motion would gain enough momentum to burst any bonds and become a standard operating procedure”); David Luban, Liberalism, Torture, and the Ticking Bomb, 91 Va. L. Rev. 1425, 1446 (2005) (contending that once coercive interrogation is allowed in any circumstances, it will be used casually, and a “culture of torture” will come into being); Harold Hongju Koh, Can the President Be Torturer in Chief ?, 81 Ind. L.J. 1145, 1165 (2006).

119

120


contexts, for instance in the case of authorizing coercive interrogations, it may also induce the establishment of institutions that would train agents to act accordingly, thus making infringements an even more readily available option.47 More generally, it is argued that due to its expressive effects, legitimizing an infringement in certain circumstances may be (wrongly) perceived as implicitly legitimizing it in other circumstances too.48 These “slippery slope” arguments suggest that even though legitimizing a (presumably justified) infringement in one set of circumstances does not logically legitimize it under different circumstances, the former is prone to bring about the latter due to political and psychological reasons.49 While the slippery slope argument cannot be ignored, it does not necessarily preclude predetermined guidelines for the permissibility of infringements. For one thing, the absence of ex ante authorization may result in people (including officials) refraining from infringing constraints when the net benefit of such infringements is large enough to override the constraint. Prior authorization may thus be necessary to encourage risk-averse agents, who are reluctant to “dirty their hands,” to nevertheless promote the overall good when such action involves a justified infringement of a constraint.50 Moreover, such authorization may be accompanied by measures that deter unjustified infringements. Accurately evaluating these and other considerations is difficult, yet at the very least they make the slippery slope argument inconclusive.51

47. See, e.g., Henry Shue, Torture in Dreamland: Disposing of the Ticking Bomb, 37 Case W. Res. J. Int’l L. 231, 238 (2006); Mordechai Kremnitzer, The Landau Commission Report: Was the Security Service Subordinated to the Law or the Law to the ‘Needs’ of the Security Service? 23 Isr. L. Rev. 216, 254–57 (1989). 48. For a discussion of theories of “expressive law and economics,” which accentuate the effect of “what the law says” (rather than of “what the law does”) on preferences and behavior, see, e.g., Richard H. McAdams, An Attitudinal Theory of Expressive Law, 79 Or. L. Rev. 339 (2000); Robert Cooter, Expressive Law and Economics, 27 J. Legal Stud. 585 (1998); Robert D. Cooter, Do Good Laws Make Good Citizens? An Economic Analysis of Internalized Norms, 86 Va. L. Rev. 1577 (2000). 49. Bernard Williams, Which Slopes are Slippery?, in Moral Dilemmas in Modern Medicine 127, 128 (Michael Lockwood ed., 1985); Eugene Volokh, Mechanisms of the Slippery Slope, 116 Harv. L. Rev. 1026 (2003). 50. For the “dirty hands” argument, see, e.g., Michael Walzer, Political Action: The Problem of Dirty Hands, 2 Phil. & Pub. Aff. 160 (1973) (suggesting that moral public officials are the ones willing to get their hands dirty by choosing to violate a constraint to bring about a sufficiently high social good). 51. See, e.g., Eric A. Posner & Adrian Vermeule, Terror in the Balance: Security, Liberty, and the Courts 200–03 (2007). See also generally David Enoch, Once You Start


The current objection may also rest on deontological concerns. According to this view, justifiably infringing a constraint to attain a desirable outcome is different from attaining the same outcome without infringing any constraint. Agents who act in accordance with ex ante authorization to infringe a constraint may lose sight of this fundamental moral distinction. Prior authorization may turn the actor’s decision-making process into a rather technical assessment of whether the conditions set forth by the legislature are met, without giving sufficient attention to the nature of the action as an infringement of a constraint and of the constraint’s underlying rationales.52 Relatedly, a law that expresses, even inadvertently, an improper message should arguably be invalidated even if its content and expected outcomes are desirable.53 The very formulation of rules that determine when it is permissible to kill or torture people is disrespectful of human dignity and thus, so the argument goes, incompatible with Kantian morality. Extreme emergencies may indeed compel one to do horrible things to prevent catastrophic outcomes. However, respect for people requires that such acts be performed, so to speak, by the force of the circumstances and should not be governed by predetermined principles. We have already discussed and rejected this argument.54 Even under extreme circumstances, agents should be guided by some moral principles, and the same is true of ex post judgment of the morality of their actions. Choosing between prior, legislative authorization of infringements and ex post assessment of their permissibility involves additional considerations. Thus, the democratic principle of citizens’ equal participation in delineating the circumstances under which infringements should be deemed permissible lends support to legislative ordering of this issue, following public deliberation.55 Legislative determination of the circumstances in which an

Using Slippery Slope Arguments, You’re on a Very Slippery Slope, 21 Oxford J. Leg. Stud. 629 (2001). 52. Cf. Elizabeth S. Anderson & Richard H. Pildes, Expressive Theories of Law: A General Restatement, 148 U. Pa. L. Rev. 1503, 1512 (2000). 53. Id. at 1513. For a critique, see Steven D. Smith, Symposium: Expressivist Jurisprudence and the Depletion of Meaning, 60 Md. L. Rev. 506, 520 (2001). 54. See supra pp. 52–53. 55. See, e.g., Jeremy Waldron, Law and Disagreement 232–34, 244–49 (1999). See also Samuel Issacharoff & Richard H. Pildes, Emergency Contexts without Emergency Powers: The United States’ Constitutional Approach to Rights During Wartime, 2 Int’l J. Const. L. 296 (2004) (suggesting that where both legislature and executive endorse a particular tradeoff of liberty and security, the courts have accepted that judgment; but where the executive

121

122


infringement is justified also mitigates the concern of the institutional bias of law enforcement and security agencies, whose primary goal is to fight crime and terror.56 The choice between prior statutory authorization and subsequent judicial review further involves complex considerations of comparative institutional capacity.57 Finally, refraining from setting guidelines ex ante is unfair to actors who are exposed to the risk of bearing civil and criminal liability if it turns out that their judgment regarding the permissibility of the infringement is different from that of the ex post reviewer. We need not resolve these difficult questions here. We acknowledge that in choosing between the various alternatives of authorizing or justifying infringements—ex ante vs. ex post; legislative, administrative, or judicial; public vs. secret; explicit vs. implicit; general vs. specific, and so forth—one must carefully consider the expressive effect and the expected outcomes of each alternative. At the same time, we insist that whichever alternative is chosen, deontologically constrained CBA should be employed to determine the permissibility of the infringing act. Hence, even if the expressive concern militates against some of the above schemes in some contexts, it does not militate against our proposed analytical framework.

•

E. Conclusion

Our suggestion to integrate deontological constraints with CBA is subject to much of the same critique leveled against moderate deontology, to many of the objections made against standard CBA, and to additional criticism aimed at the proposed integrative methodology. At the same time, deontologically constrained CBA overcomes at least some of the conspicuous

has flown in the face of legislative policies or without legislative approval, the courts have invalidated executive action, even during wartime, or scrutinized it more closely). 56. See, e.g., Hamdi v. Rumsfeld, 542 U.S. 507, 545 (2004) (Souter, J., concurring). See also Thomas P. Crocker, Torture, with Apologies, 86 Texas L. Rev. 569, 585–93 (2008). 57. For a discussion of this issue in the context of the fight on terror, see, e.g., Posner & Vermeule, supra note 51, at 15; Posner, supra note 38, at 27; David Dyzenhaus, Are Legislatures Good at Morality? Or Better at it than the Courts?, 7 Int’l J. Const. L. 46 (2009). This issue is often discussed in the context of the role of the “proportionality” requirement in judicial review of legislation and administrative decisions. See, e.g., David Beatty, The Ultimate Rule of Law (2004); Mattias Kumm, Political Liberalism and the Structure of Rights: On the Place and Limits of the Proportionality Requirement, in Law, Rights and Discourse: Themes from the Legal Philosophy of Robert Alexy 131 (George Pavlakos ed., 2007).


normative deficiencies of standard CBA (discussed in chapter 1) and of the methodological shortcomings of moderate deontology (discussed in chapter 2). This chapter focused on objections that may be raised against the combination of moral constraints with CBA. We conclude that some of these objections (such as the claim that incorporating constraints into standard CBA would undermine its alleged normative neutrality, or that it would bring about undesirable commodification of moral values) rest on shaky foundations. Other objections are perhaps more powerful, but none of them is conclusive. These objections imply that deontologically constrained CBA is possibly more fruitful in some contexts than in others. Indeed, various considerations should affect the precise manner in which constrained CBA is employed, including the object of the analysis (legal rules, governmental policies, ex post judicial determination of the legality of actions, and so forth), its publicity, and the identity of the analysts. We reiterate that addressing deontological concerns as constraints on CBA is superior to either ignoring them or dealing with them separately as secondary considerations. Since “the proof of the pudding is in the eating,” the second part of this book applies the proposed analytical framework in various legal fields.

123


•

part two

Applications


•

six

The Fight Against Terrorism

•

A. Introduction

the fight against terrorism poses significant moral and legal challenges

to liberal democracies. The attempts of potential terrorists to disguise themselves among innocent, civilian population, as well as the substantial damage created by acts of terror, present policy-makers with tragic choices. For instance: What type of preemptive measures can a state legitimately take against persons before proving their culpability, or even their evil intentions, in a court of law? Is it legitimate to save lives by harming dangerous individuals through extrajudicial methods, such as targeted killings, administrative detentions, or aggressive interrogation techniques? Is it justified to inflict harm on innocent persons in an attempt to save others? The extensive literature evaluating the reaction of the democratic world to the terrible events of September 11, 2001 largely reflects a division between two schools of thought. One camp, which mostly criticizes the United States’ reaction, rests its arguments on “principles.” According to this view, the Constitution “represents a collective commitment to principles . . . [and] the recognition that ‘pragmatic’ cost-benefit decisions will often appear in the short term to favor actions that may turn out in the long term to be contrary to our own best principles.”1 The other camp, to which legal economists usually belong, rejects precisely this commitment to principles. It claims that there is no consensus on the content of these principles, as “text, tradition, precedent, and reason so often tug in different directions.”2 Consequently, “[r]ealism requires recognition that constitutional decision-making . . . is driven in the main by policy judgments,”3 which are based exclusively on

1. David Cole, How to Skip the Constitution, N.Y. Rev. Books, Nov. 16, 2006, at 20, 21. 2. Richard Posner, How to Skip the Constitution: An Exchange, N.Y. Rev. Books, Jan. 11, 2007, at 63. 3. Id.

127

128


estimating the costs and benefits associated with each government practice. This chapter questions the alleged dichotomy between principles on the one hand and cost-benefit analysis on the other, by demonstrating the plausibility and desirability of incorporating deontological constraints (“principles”) with economic analysis (“balancing”).4 We begin in section B by briefly characterizing and criticizing existing normative economic analyses of the fight against terrorism as reflecting a simplified ad-hoc balance of interests (or act-consequentialist) approach. Section C presents the central deontological constraints pertinent to the fight against terrorism. We focus on the constraint against actively/ intentionally inflicting harm. Additionally, we address the debate as to whether an aggressor retains his right to life regardless of his actions or evil intentions, and suggest that preempting an aggressor should be considered as a constraint infringement, at least in cases where the guilt of the harmed person has not been established by fair judicial proceedings. Sections D and E discuss the threshold functions that should apply when determining the permissibility of inflicting intended and unintended harm, respectively. We discuss the factors that should be considered in evaluating the act’s relevant net benefit and those determining the size of net benefit required to justify an infringement. Section F analyzes anti-terrorist measures that involve inflicting both intended and unintended harm, demonstrating an integration of the two types of threshold functions.

B. Economic Analysis of the Fight Against Terrorism •

Terrorism is conventionally defined as the use of violence causing death, great bodily harm, or serious psychological damage to innocent individuals with the intent to cause those harms or with wanton disregard for them, for the purpose of coercing or intimidating a specific group or government or otherwise gaining some perceived political or military benefit.5 Compared to

4. We do not discuss in this chapter the dilemma whether (justified) infringements should be authorized ex ante by legislation or should be subject to an absolute formal prohibition, while government activities that override deontological constraints are to be evaluated only post factum. See, in this respect, supra pp. 117–22. 5. See Christopher L. Blakesley, Ruminations on Terrorism: Expiation and Exposition, 10 New Crim. L. Rev. 554, 580 (2007). Different elements of this definition are debated, but

the fight against terrorism

other criminals, terrorists are often willing to sacrifice their lives for their cause, which significantly diminishes the deterrence effect of legal sanctions on terrorism. Additional distinctive features of terrorism are its typically organized form and the extensive indirect adverse effects of the threat of terror. A substantial portion of the economic analysis of terror is devoted to a positive, rather than normative, inquiry of these distinctive characteristics of terrorism.6 Such positive analyses include theoretical and econometric evaluations of the efficacy of the use of various anti-terror policies,7 politicaleconomy assessments of the characteristics and sources of terrorism,8 and empirical studies of the adverse economic effects of terror.9 These aspects of the economic analysis of terror and the fight against it are important in evaluating existing security policies and in designing new ones. Our focus is, however, on the branch of economic analysis that aims to provide normative evaluation of such policies. Two notable contributions to this branch of inquiry are Richard Posner’s Not a Suicide Pact: The Constitution in a Time of National Emergency, and Eric Posner and Adrian Vermeule’s Terror in the Balance: Security, Liberty, and the Courts.10

nothing in the discussion that follows hinges on the legal classification of a certain act as “terrorism.” 6. For a review of the extensive literature in this area, see, e.g., Fernanda Llussá & José Tavares, Economics and Terrorism: What We Know, What We Should Know, and the Data We Need, in Terrorism, Economic Development, and Political Openness 233 (Phillip Keefer & Norman Loayza eds., 2008). 7. See, e.g., Philip B. Heymann & Juliette N. Kayyem, Protecting Liberty in an Age of Terror (2005); Asaf Zussman & Noam Zussman, Assassinations: Evaluating the Effectiveness of an Israeli Counterterrorism Policy Using Stock Market Data, 20 J. Econ. Persp. 193 (2006); Nuno Garoupa, Jonathan Klick & Francesco Parisi, A Law and Economics Perspective on Terrorism, 128 Pub. Choice 147 (2006). 8. See, e.g., S. Brock Blomberg, Gregory D. Hess & Akila Weerapana, Economic Conditions and Terrorism, 20 Eur. J. Pol. Econ. 463 (2004); Todd Sandler & Walter Enders, An Economic Perspective on Transnational Terrorism, 20 Eur. J. Pol. Econ. 301 (2004); Jeffrey Dunoff & Joel P. Trachtman, The Law and Economics of Humanitarian Law Violations in Internal Conflict, 93 Am. J. Int’l L. 394 (1999). 9. See, e.g., The Economic Analysis of Terrorism (Tilman Brück ed., 2006); Walter Enders, Terrorism: An Empirical Approach, in 2 Handbook of Defense Economics 815 (Todd Sandler & Keith Hartley eds., 2007); Bruno S. Frey, Simon Luechinger & Alois Stutzer, Calculating Tragedy: Assessing the Costs of Terrorism, 21 J. Econ. Surveys 1 (2007); Harry W. Richardson, Peter Gordon & James E. Moore II, The Economic Costs and Consequences of Terrorism (2007); The Economic Consequences of Terror, 20(2) Eur. J. Pol. Econ. 291–515 (2004) (special issue). 10. Richard A. Posner, Not a Suicide Pact: The Constitution in a Time of National Emergency (2006); Eric A. Posner & Adrian Vermeule, Terror in the Balance: Security, Liberty, and the Courts (2007).

129

130


The normative economic analysis of the fight against terrorism advocates an ad-hoc balance of interests approach. It opposes the counterintuitive claims of absolutist deontology11 and holds that the use of a security measure is justified whenever its expected social benefits exceed its costs. As succinctly stated by Richard Posner, “one is not to ask whether liberty is more or less important than safety. One is to ask whether a particular security measure harms liberty more or less than it promotes safety.”12 This approach rejects the notion of civil liberties as constraints on consequentialist balancing. It claims that “security and liberty trade off against one another,”13 and thus “governments should, and do, balance civil liberties and security at all times.”14 The concept of “balancing” is, of course, not unique to the economic approach. Courts frequently employ balancing tests when there are two or more competing interests, construed, for example, as a conflict between an individual right and state need.15 However, the process of balancing advocated by the economic approach is distinctive in several respects. For one thing, CBA weighs competing interests in monetary terms, reflecting people’s preferences.16 Another singular aspect is the use of ad-hoc balancing, rather than a so-called “definitional” one. The balancing endorsed by standard economic analysis is ad-hoc not only in the sense that it prefers a case-by-case inquiry to a global one,17 but also in that it balances the relevant interests, rather than competing constitutional principles or values. For example, when the United States Supreme Court used the terminology of balancing in analyzing the legitimacy of detaining suspected terrorists in the

11. Posner, supra note 10, at 40–42. 12. Id. at 32. 13. Posner & Vermeule, supra note 10, at 24. 14. Id. at 5. 15. For a critical overview, see, e.g., T. Alexander Aleinikoff, Constitutional Law in the Age of Balancing, 96 Yale L.J. 943 (1987). Note that the balancing metaphor takes two distinct forms. One form of balancing is the process of placing the competing interests on a set of scales and ruling the way the scales tip (such that one interest outweighs another). A different form of balancing is used when each interest survives and is given its due consideration. Id. at 946. In the present context, balancing takes on the first meaning. 16. See supra pp. 14–18. 17. This is the meaning that Nimmer gives to the distinction between ad-hoc and definitional balancing. Melville B. Nimmer, Nimmer on Freedom of Speech 2.02–.03 (1984). See also Louis Henkin, Infallibility Under Law: Constitutional Balancing, 78 Colum. L. Rev. 1022, 1027–28 (1978); Aleinikoff, supra note 15, at 948.


case of Hamdi v. Rumsfeld (2004), it referred to “[s]triking the proper constitutional balance” between the security needs and “the values that this country holds dear . . . [such as the] commitment to due process.”18 In contrast, notwithstanding the rhetoric of comparing liberty and security, economic analysts take into account only the concrete individual interests that are at stake, rather than abstract values and principles. In this sense, they are actconsequentialists. Posner and Vermeule concede that in principle, one may apply what they term a “non-welfarist” balancing, in which “the effects of actions on rights are themselves among the consequences to be evaluated.”19 Treating right infringements as a cost does not transcend consequentialism, yet the adoption of such a more sophisticated theory of the good may somewhat narrow the gap between consequentialism and deontology.20 At any rate, they do not adopt such an approach. The economic analysis of the fight against terrorism thus does not consider the mere infringement of a basic liberty or some other fundamental principle as a social cost in the constitutional calculus. What counts is only the policy’s adverse effect on the welfare of individuals. Under this view, the content of a concept such as liberty is nothing but the outcome of the balancing process, as it is determined exclusively by the scope of society’s existing needs: “[it] is the point of balance [that] determines the optimal scope of the right.”21 Liberty is not perceived as a normative concept but rather as a descriptive one. Posner and Vermeule, as well as Richard Posner, oppose setting any deontological constraint that would limit states in taking security measures in the fight against terrorism, for two reasons. First, as far as the constraints are based on second-order institutional and empirical concerns (ruleconsequentialism), they argue that these concerns are highly exaggerated and unsubstantiated.22 In fact, they suggest that the main concern is that deontological constraints might “block government’s attempts to adjust the balance as threats wax and wane.”23 The real danger, so they argue, is that “civil libertarian panic about the specter of authoritarianism” will hinder

18. 542 U.S. 507, 532 (2004) (plurality opinion). 19. Posner & Vermeule, supra note 10, at 190. See also id. at 40. 20. See supra pp. 30–32. 21. Posner, supra note 10, at 31. 22. Posner & Vermeule, supra note 10, at 40–41, 187–93. 23. Id. at 31.

131

132


cost-justified security measures.24 Thus, they not only reject the view of rights as trumps but also reject the position that one should “place a thumb on the scale” to assure that the balance struck in any particular situation properly reflects the central position of the relevant liberty in the constitutional scheme. If anything, they place the thumb on the side of security.25 Second, Posner argues that deontological constraints, including those recognized as constitutional principles, are indeterminate and subjective, as they reflect each decision-maker’s moral and religious values and personal life experience.26 Posner and Vermeule suggest that even though incorporating deontological constraints and setting thresholds can in principle be justified on the basis of first-order moral considerations, this approach is inapplicable, since setting the threshold comes at the price of “reduced theoretical coherence and an arbitrary flavor.”27 Consequently, according to standard economic analysis, to justify a security measure, it is not necessary to show that it would generate sufficiently high social benefits to override a deontological constraint. It suffices that the benefit exceeds the cost of the policy’s direct adverse effects. In accordance with standard act-consequentialism, the economic analysis disregards the doing/allowing distinction. Hence, it equates the active infliction of harm by security measures with not preventing a comparable harm to the potential victims of terrorist activities.28 Standard CBA also rejects the distinction between saving or harming unidentified (“statistical”) people and saving or harming identified ones.29 As a result, human rights and liberty are placed on both sides of the scale, thereby offsetting the status of basic liberties in the process of balancing.

24. Id. at 39. 25. See Thomas P. Crocker, Torture, with Apologies, 86 Tex. L. Rev. 569, 579 (2008). 26. Posner, supra note 10, at 25. See also id. at 26, 41; Louis Kaplow & Steven Shavell, Fairness Versus Welfare 45–47 (2002). 27. Posner & Vermeule, supra note 10, at 40. See also id. at 187–88 (“Standardly, the permissible ratio is one to one: where relevant restrictions are met, a government may kill A to save B, not merely a thousand B’s. . . . [It] is quite mysterious . . . why the sheer catastrophic size of the threatened harm should matter. . . . Oddly, the catastrophe exception builds in an arbitrary threshold . . .”). 28. See, e.g., Posner & Vermeule, supra note 10, at 191; Kaplow & Shavell, supra note 26, at 331–36. Cf. Cass R. Sunstein & Adrian Vermeule, Is Capital Punishment Morally Required? Acts, Omissions, and Life-Life Tradeoffs, 58 Stan. L. Rev. 703 (2005); Karima Bennoune, Terror/Torture, 26 Berkeley J. of Int’l L. 1 (2008); Aleinikoff, supra note 15, at 981–82. On the doing/allowing distinction, see generally supra pp. 41–43, 60–69. 29. For a discussion of this distinction see supra p. 91; infra pp. 168–69.


Legal economists further propose to replace an actual, detailed balance of interests with a hypothetical one, referring to some baseline of normalcy. According to this notion, the state always operates at a “Pareto frontier” with regard to the balance of security and liberty; and hence there is a perfect trade-off between “liberty” and “security.” Any increase in the demand for security, resulting from the threat of terror, entails a decrease in liberty.30 It follows—without even having to conduct a detailed CBA—that infringements of deontological constraints are a necessary and justified corollary of the fight against terrorism. CBA allows legal economists to argue that judges should defer to decisions of the executive branch, and that judicial review of governmental action should be relaxed or even suspended in times of emergency.31 This position rests on the concern that judges have limited institutional capacity and knowledge needed to evaluate security policies.32 Once one assumes that designing anti-terrorist policies entails no principled normative deliberation but merely an aggregation of costs and benefits, governmental agencies seem better qualified than courts to perform the task of evaluating security policies. Not surprisingly, standard normative economic analysis of the fight against terrorism legitimizes a rather broad range of measures. It endorses policies such as indefinite detentions, warrantless domestic wiretapping, and coercive interrogations.33 It justifies restrictive measures against persons whenever the expected benefit of those measures exceeds their costs, even if the probability that these persons would have been involved in terror is marginal. For example, Richard Posner concedes that even in radical Islamist communities, only a small percentage of people are willing to commit, or even abet, terrorist acts. He nevertheless argues that since there are millions of Muslims in the United States, and since a few terrorists may cause catastrophic harms, the government may take measures to curtail the freedom of speech of radical imams.34 Posner and Vermeule even question the desirability of the international laws of war. They contend that “if the laws of

30. Posner & Vermeule, supra note 10, at 26–27. 31. Id. at 15. See also Posner, supra note 10, at 27. 32. See Adrian Vermeule, Judging under Uncertainty: An Institutional Theory of Legal Interpretation 230 (2006). 33. Posner, supra note 10, at 53–71, 80; Sunstein & Vermeule, supra note 28. 34. Posner, supra note 10, at 124.

133

134


war are intrinsically bad, the United States should not only violate those that interfere with the war on terrorism; it should advocate the abolition of all laws of war.”35 For the reasons elaborated in the first part of this book, we find both the consequentialist underpinnings of standard normative economic analysis of the fight against terrorism and much of its implications unacceptable. The claim that a state’s failure to effectively eliminate threats of terrorism is morally equivalent to actively/intentionally harming people disregards the doing/allowing and intending/foreseeing distinctions and fails to distinguish between deontological options and deontological constraints.36 Furthermore, the claim that the risk of terror necessarily justifies losses in liberty is unfounded. The alleged trade-off rests on questionable empirical assumptions. It assumes not only that at the baseline the state operates at a Pareto frontier with regard to the balance of security and liberty, but also that all background factors, such as the resources allocated to tackle the security threat, do not change in the move from normalcy to emergency. These assumptions are implausible.37 More fundamentally, the claim that there is a perfect trade-off between “liberty” and “security” treats the curtailment of liberty as a mere shorthand for reduction in people’s welfare. Once liberty is conceived as entailing a deontological constraint on actively/intentionally taking peoples’ lives or harming their bodily integrity and dignity, then two fundamental modifications of the balancing process are needed. First, to justify the infringement of constraints, the expected good outcomes (or the avoided bad ones) must meet a relatively high threshold. Second, only certain types of benefits and costs may legitimately be taken into account in this context.38 The following discussion describes the central characteristics of such a modified balance between liberty and security.

35. Posner & Vermeule, supra note 10, at 262. In their book, they avoid deciding whether the laws of war are, indeed, bad on the balance, and are willing to assume that “some laws of war” are optimal (id. at 263). See also Eric Posner, A Theory of the Laws of War, 70 U. Chi. L. Rev. 297 (2003); Dan Belz, Is International Humanitarian Law Lapsing into Irrelevance in the War on International Terror?, 7 Theoretical Inq. L. 97 (2006). 36. See supra pp. 41–48, 60–70. 37. Crocker, supra note 25, at 576–77; David Cole, No Reason to Believe: Radical Skepticism, Emergency Power, and Constitutional Constraint, 75 U. Chi. L. Rev. 1329, 1335–42 (2008). 38. Jeremy Waldron, Security and Liberty: The Image of Balance, 11 J. Pol. Phil. 191, 199 (2003).


C. The Constraint Against Harming Persons and the Fight Against Terrorism •

1. General Considerations Anti-terrorist activities may entail infringements of any conceivable deontological constraint, including the constraints against deception and breaking promises. However, such activities primarily infringe the most basic deontological constraints; those against actively/intentionally killing people, injuring them, curtailing their freedom of movement and speech, and harming their human dignity. Identifying the pertinent constraints and setting thresholds for the permissibility of their infringement rest on the ordinary deontological distinctions. These include the distinction between doing harm and merely allowing it. They also include the distinction between intending harm (which may be immoral even if the harm is merely allowed) and foreseeing harm (which is not necessarily immoral), and the related distinction between harming a person as a side effect of attaining a legitimate result and using a person as a means to aiding or saving others. A harm is intended even if the actor has no interest in imposing the harm except as a means to achieve some end. To illustrate, if the government puts pressure on a terrorist by harming innocent persons who are dear to him, the harm to these persons is intended, irrespective of the government’s lack of a direct interest in harming those persons. Harm is unintended only when it is a mere side effect, that is, when the adverse outcome does not provide the actor with a reason for his behavior or an explanation for it.39 Consider the following scenario:40 A state bombs a terrorists’ munitions plant, and as a result, kills civilians living nearby. Is the killing of the civilians intended? To answer this question, one must find out the aim of dropping bombs on the plant. One possibility is that the bombing was aimed at destroying the plant itself, damaging the terrorists’ capacity to build their own weapons in order to prevent future terrorist attacks; another possibility

39. See supra pp. 41–48, 60–70. 40. This example is a variation of the one proposed by Jonathan Bennett, Morality and Consequences, in 2 The Tanner Lectures on human Rights 45, 95 (Sterling M. McMurrin ed., 1980). See also Robert Cryer & A.P. Simester, Iraq and the Use of Force: Do the Side-Effects Justify the Means?, 7 Theoretical Inq. L. 9, 31–34 (2006).

135

136


is that the aim was to terrorize the civilian population supporting the terrorists in order to break their morale and force them to surrender. Under the first possibility, the killing of civilians is a mere side effect. Under the second, the state is intentionally harming civilians as a means to its end. Importantly, from the perspective of moderate deontology, in both cases the state infringes a constraint, yet the threshold that must be met to render the infringement permissible is different. The threshold is much higher in the second case, as intended harm reflects disrespect to human dignity. Victims have an interest not only in their fate but also in their status, in what can permissibly be done to them.41

2. Harming Aggressors as a Constraint Infringement According to most deontological approaches, in evaluating the legitimacy of inflicting harm on a person, it is essential to take into account the person’s relevant moral responsibility. While consequentialism considers one’s fault only instrumentally, as harming the blameworthy is likely to deter future undesirable conduct, deontology plausibly ascribes intrinsic value to one’s fault. Imposing harm on a blameworthy person, at least in a context closely related to his blame, is less morally objectionable than imposing the same harm on an innocent person. The German Constitutional Court referred to this issue in passim, when scrutinizing the constitutionality of a statute authorizing the shooting down of an aircraft intended to be wielded as a lethal weapon.42 The court decided that authorizing the military to shoot down a plane with innocent people on board is unconstitutional, as it infringes their right to dignity (even if these people are doomed to die anyway). However, the court held that if only terrorists are on board the aircraft, it is perfectly legitimate to shoot down the plane. The court reasoned that preempting the attackers does not

41. Adil Ahmed Haque, Torture, Terror, and the Inversion of Moral Principle, 10 New Crim. L. Rev. 613, 634 (2007). On oblique intention, that is, cases in which the actor does not intend harm but is practically certain that her action will bring it about, see generally Itzhak Kugler, Direct and Oblique Intention in the Criminal Law (2002). 42. Judgment of the Bundesverfassungsgericht [BVerfG] [German Federal Constitutional Court] Feb. 15, 2006, BVerfGE, 1 BvR 375/05, available at http://www.bundesverfassungsgericht.de/entscheidungen/rs20060215_1bvr035705.html.


infringe the moral requirement to respect their dignity because they forfeited their rights by their willful conduct.43 A moderate deontologist would agree that the blame of the individual harmed by the state is morally relevant in evaluating the act’s permissibility. However, contrary to the German court’s ruling, it is doubtful that inflicting harm on the guilty does not infringe a deontological constraint. Consider, first, the imposition of criminal sanctions on a person convicted in criminal proceedings. Even if retribution obliges the state to impose the sanction,44 retribution can only justify some types of sanctions—those that are deemed “appropriate” to the severity of the person’s moral blame.45 A person’s guilt is not sufficient to establish the claim that sanctioning him does not infringe a moral constraint, since it is still required to show that the harm is justifiable in terms of the relevant theory of retribution. One’s wrongful actions thus do not negate all of one’s rights and do not necessarily justify infringing the constraint against torturing or killing a convict.46 Moreover, retribution is not the only possible rationale for imposing criminal sanctions. Certainly, if one endorses other rationales for criminal sanctions, such as private and general deterrence, or incapacitation, the fact that imposing the sanctions is justified does not imply that it does not infringe the constraint against actively/intentionally inflicting harm. The difficulties in characterizing the infliction of harm on the guilty as a non-infringement are significantly greater when—as is typically the case in the context of the fight against terrorism—no judicial determination of the person’s guilt precedes the harm. Even if retributive criminal sanctions do

43. See id. §§ 138–39. The Israeli Supreme Court expressed a similar view in addressing the legitimacy of “targeted killings” of terrorists. The court ruled that “[fulfilling the rule] of proportionality is not required regarding harm to a combatant, or to a civilian taking a direct part in the hostilities at such time as the harm is caused. Indeed, a civilian taking part in hostilities is endangering his life, and he might—like a combatant—be the objective of a fatal attack. That killing is permitted. . . . [P]roportionality is required [only] in any case in which an innocent civilian is harmed.” HCJ 769/02 Public Comm. against Torture in Israel v. Government of Israel, § 46 (2005), available at http://elyon1.court.gov.il/files_ eng/02/690/007/A34/02007690.a34.pdf. 44. Immanuel Kant, The Metaphysics of Morals 6:331 (Mary Gregor trans. & ed., Cambridge Univ. Press, 1996) (1797). See also, e.g., Dan Markel, Executing Retributivism, 103 Nw. U. L. Rev. 1063 (2009). 45. See, e.g., Thomas E. Hill, Kant on Wrongdoing, Desert and Punishment, 18 Law & Phil. 407, 409 (1999). 46. See, e.g., David Luban, Unthinking the Ticking Bomb, in Global Basic Rights 181 (Charles R. Beitz & Robert E. Goodin eds., 2009).

137

138


not infringe a constraint, it does not follow that the same is true of measures taken against suspected terrorists outside the criminal justice system. The requirement that retributive sanctions would follow a judicial finding of criminal liability reflects not only the concern of a wrong, and possibly biased, attribution of guilt in the absence of judicial proceedings, but also a normative judgment about fairness. A person can justifiably be subject to harm based on retribution only if his blame was determined in a way that meets the requirements of procedural justice.47 We then come to the difficult case of inflicting harm with the aim of preempting an aggressor. John Locke famously argued that an aggressor forfeits his right to life.48 Following this view, several contemporary scholars endorse the so-called “forfeiture argument”: it is permissible to kill in self-defense, for a person possesses the right to life only so long as he does not pose an unjust, immediate threat to others.49 This position is subject to extensive and, in our view, convincing criticism. The presumption that the aggressor actually chooses to forfeit his right to life is hard to sustain.50 Moreover, this presumption is incompatible with the notion that the right to life is inalienable.51 Thus, even if a would-be suicide-bomber intends to sacrifice his life in order to kill others, it does not follow that when determining what measures are legitimate to preempt the attack, one can assume that the would-be suicide-bomber does not possess a right to life. To the extent that the aggressor’s forfeiture of the right to life rests on his culpability or malice (and not merely on the threat he imposes), the argument fails to justify the right to self-defense against an innocent attacker. It is also subject to the difficulties discussed before, regarding the lack of fair judicial proceedings for establishing one’s guilt.

47. See also infra p. 141. 48. John Locke, Two Treatises of Government 279 (Peter Laslett ed., Cambridge Univ. Press 1988) (1690) (arguing that by his own actions, the aggressor “expose[s] his Life to the other’s Power to be taken away by him”). 49. Judith Jarvis Thomson, Self-Defense, 20 Phil. & Pub. Aff. 283, 302 (1991); Suzanne Uniacke, Permissible Killing: The Self-Defence Justification of Homicide 213 (1994); Fiona Leverick, Killing in Self-Defence 62 (2006). 50. Sanford H. Kadish, Respect for Life and Regard for Rights in the Criminal Law, 64 Cal. L. Rev. 871, 883 (1976). 51. George P. Fletcher, The Right to Life, 13 Ga. L. Rev. 1371, 1382–83 (1979); Boaz Sangero, Self-Defence in Criminal Law 44 (2006).


Most important for our purposes, the forfeiture argument does not quite explain the limitations imposed on the right to self-defense, such as the requirement that the preemption is necessary and that the harm inflicted on the aggressor is proportionate to the harm that the aggressor aims to inflict. If the aggressor has no right to life, why impose these limitations?52 It is also unclear what types of constraints or rights the aggressor forfeits as a result of his aggression. Supporters of the forfeiture argument respond that the aggressor forfeits his right to life only at the point at which killing him is necessary and proportional.53 This response, however, is tantamount to arguing that a person has no right not to be subject to a justified infringement. It acknowledges that the aggression itself is insufficient to render the killing of the aggressor a noninfringement. In sum, it seems that an aggressor retains his right to life regardless of his evil actions or intentions. The German court’s obiter dictum—that shooting down a plane when only the terrorists are on board would not infringe a moral constraint—is therefore unsupported. At least in cases where the guilt of the harmed person has not been established through fair judicial proceedings, any infliction of harm infringes a constraint and can only be justified if enough good (or bad) outcomes are at stake.54 This is not to say that the harmed-person’s guilt is irrelevant. As we argue, a person’s moral responsibility for creating the circumstances that call for infringing the constraint should plausibly affect the threshold function used to determine the permissibility of the infringement.55 Since the measures that democracies employ in their fight against terrorism often infringe the deontological constraint against actively/intentionally harming people, one should assess their permissibility through threshold functions. The following sections discuss the key choices that one must make in constructing such functions.

52. See, e.g., Cheyney C. Ryan, Self-Defense, Pacifism, and the Possibility of Killing, 93 Ethics 508 (1983); Tziporah Kasachkoff, Killing in Self-Defense: A Questionable or Problematic Defense?, 17 Law & Phil. 509, 517 (1998). 53. Leverick, supra note 49, at 66; Uniacke, supra note 49, at 213 (“[The aggressor does] not have a right against [the victim] that [the victim] not use necessary and proportionate defensive force”). 54. For a related discussion regarding the nature of lying to a person who illegitimately threatens to kill an innocent person, see infra pp. 276, 284. 55. See infra p. 160. See also infra pp. 171–74.

139

140


D. Constrained Economic Analysis of Intended Harm •

1. General Evaluating the permissibility of counterterrorist measures that infringe the deontological constraint against actively/intentionally harming persons requires the construction of a threshold function. In what follows, we demonstrate a constrained economic analysis of common activities in the fight against terrorism. We distinguish between the threshold function applying to cases of intended harm, discussed in this section, and the one applying to harms inflicted as a mere side effect, discussed in section E. We begin by distinguishing between different possible goals of anti-terrorist measures and explaining their importance in assessing the action’s permissibility. We argue that preemption is the soundest justification for overriding the constraint against harming people. We then move to present the basic elements of the threshold function. These include the types of benefits and costs that one should take into account in calculating the act’s net benefit and the factors that determine the size of the threshold. Throughout the analysis, we assume, for simplicity’s sake, that the action under consideration would harm only the suspected terrorists, and no adverse side effect is expected (this assumption is relaxed in section F). We use the antiterrorist practice of targeted killing of suspected terrorists as a paradigm and only briefly address other measures. The last subsection then extends the discussion by applying the same considerations to a different practice, namely the use of force in interrogations.

2. Goals of Anti-Terrorist Measures: Preemption, Retribution, Deterrence, and Pressure Anti-terrorist measures have four central goals: preemption, retribution, deterrence, and pressure. Preemptive measures are acts that aim at directly thwarting terror attacks of would-be terrorists, either by capturing them and holding them in detention or by killing or physically disabling them. This goal is the central basis of permissible infringements and is therefore the focus of our analysis. In contrast, it is debated whether the attainment of any of the other goals can justify an active/intentional infliction of harm by the state. We suggest that even if deterrence may sometimes justify anti-terrorist


measures, the net benefit of measures taken for this goal would have to meet a higher threshold to be justified. Retribution As previously discussed, retribution is a legitimate aim of intentionally inflicting harm, provided that the sanction is determined and executed through the criminal justice system, following a judicial finding that the relevant person is criminally liable for his activities. Retributive punishment presupposes guilt and must be proportional to the degree of the wrongdoer’s guilt. The determination of guilt and the apportionment of punishment generally require reliable procedures, such as a fair trial and impartial judges. These constitutive elements are typically lacking in a military or administrative action conducted as part of the fight against terrorism.56 Therefore, the legitimate aims of actions such as targeted killings, torture of suspects, and administrative detentions are commonly perceived as preemptive and forward-looking.57 They are conventionally justified as measures of “selfdefense,” rather than as backward-looking retribution based on desert.58 We shall proceed on the assumption that retribution cannot justify the relevant constraint infringements in the fight against terrorism.59 Deterrence The infliction of harm to achieve deterrence is contested. Inflicting harm to generate general deterrence is incompatible with the Kantian imperative because it uses the targeted person as a means rather than as an end.60 At the

56. This is also true in pursuing war. Jeff McMahan, Aggression and Punishment, in War: Philosophical Perspectives 67, 84 (Larry May ed., 2007). 57. Some argue that even in such cases a state has a due-process obligation to develop fair, rational procedures for its use of targeted killing. John Radsan & Richard W. Murphy, Due Process and Targeted Killing of Terrorists (working paper, 2009), available at http://ssrn. com/abstract=1349357. 58. See, e.g., George P. Fletcher, A Crime of Self-Defense: Bernhard Goetz and the Law on Trial 18–38 (1988); George P. Fletcher, Punishment and Self-Defense, 8 Law & Phil. 201 (1989). 59. One may argue, however, that in exceptional cases, of persons who committed unusually horrible terror attacks, whose blame is certain, and bringing them to justice is impossible, inflicting harm that is aimed at retribution can be justified. 60. Kant, supra note 44, at 6:331. See also Jeff McMahan, Just Cause for War, 19 Ethics & Int’l Aff. 1, 16 (2005); Mark Tunick, Efficiency, Practices, and the Moral Point of View: Limits of Economic Interpretations of Law, in Theoretical Foundations of Law and Economics 77, 77–79, 90–91 (Mark White ed., 2009).

141

142


same time, the fight against terrorism may require dynamic military tactics to affect terrorists’ motivation and willingness to fight. Absent a state reaction, the “successful” execution of one terrorist attack is likely to encourage others to commit similar attacks. Hence, targeting the terrorist does not rest merely on deterrence, as the terrorist is responsible, to some degree, for the increased risk of further wrongful actions by others.61 Legitimizing deterrence also raises second-order concerns. More often than not, it is impossible to scrutinize the claim that an action is efficient in achieving deterrence, as it requires evaluating counterfactual scenarios and controlling for other potentially relevant factors. It is also impractical to evaluate the marginal contribution of each instance of inflicting harm to achieving a certain level of deterrence and thus to determine the necessity of each infringement. Take, for example, the policy of demolishing houses of suicide bombers. The Israeli Supreme Court has forbidden using house demolitions as a means of punishment since it is not executed through normal channels of the criminal justice system and since it inflicts harm on innocents as well.62 However, the court has not voided this measure inasmuch as the government uses it for deterrence purposes.63 This distinction requires the state to show that the policy is effective in discouraging would-be terrorists. However, due to the difficulties in verifying the efficacy of this policy, the result is arguably a judicial endorsement of a policy whose actual, though concealed, motivation is retribution.64 Without trying to resolve this debate, we suggest that even if deterrence can justify a constraint infringement, the threshold of net benefit that has to be met to render such an action permissible should be higher than the one applicable to an infringement aimed at preempting the targeted person. To illustrate, assume that killing a terrorist deters other would-be terrorists and thus saves the lives of, say, eighteen would-be victims. A moderate

61. McMahan, supra note 60, at 16–17. For a broader justification of inflicting harm for the purpose of deterrence, see Thomas Hurka, Liability and Just Cause, 21 Ethics & Int’l Aff. 199 (2007). 62. See, e.g., HCJ 3363/03 Baqer v. IDF Commander in the West Bank (2003), § 10. 63. See, e.g., HCJ 1005/89 Aga v. IDF Commander in Gaza, 44(1) PD 536 (1989), summarized in English in 23 Isr. Y.B. Hum. Rts. 330 (1993); HCJ 124/09 Dawayat v. Minister of Defense (2009). 64. Brian Farrell, Israeli Demolition of Palestinian Houses as a Punitive Measure: Application of International Law to Regulation 119, 28 Brooklyn J. Int’l L. 871 (2002); David Kretzmer, The Occupation of Justice: The Supreme Court of Israel and the Occupied Territories 145–63 (2002).


deontologist may accept that under certain circumstances, an otherwise wrongful killing of a person is permissible to preempt this person from intentionally killing eighteen others. It does not follow, however, that it is ever justified to kill a person who does not impose a threat to the lives of others in order to deter other persons from killing.65 At the very least, a moderate deontologist will plausibly set a much higher threshold for legitimizing the killing of a person (or torturing him), in order to deter other people from committing terrorist activities, than the one that has to be met to legitimize such killing (or torture) to preempt the same person from inflicting harm.66 Moreover, assuming arguendo that deterrence is a legitimate aim, it seems inevitable to base our judgment about the permissibility of an infringement on data regarding the efficacy of the general practice in deterring potential terrorists. It is impractical to measure the deterring effect of a single action. However, to base the decision on the average impact of targeted killings, one must also have reliable information regarding the linearity of the deterrence function. Otherwise, one cannot conclude that each killing deters terrorist attacks which would kill eighteen people. If the statistical mean exceeds the threshold but the median falls below it, each killing will have an expected value that meets the threshold, but most killings will not be justified, as the median falls below the threshold.67 Pressure Inflicting intended harm, exposing a person to the risk of harm, or even threatening to do so, may be an effective method to force terrorists to call off a terror attack, to surrender, or to provide information. However, at least as far as intentional harming of innocent persons is concerned, it is generally

65. The Israeli Supreme Court applied this rationale in determining the authority of an occupying power to assign the place of residence of an individual, by holding that considerations of deterring others are insufficient for making an order of assigned residence. HCJ 7051/02 Ajuri v. IDF Commander in West Bank, 56(6) PD 352 (2002), § 27. An English translation is available at http://elyon1.court.gov.il/files_eng/02/150/070/A15/02070150. a15.htm. 66. Sunstein & Vermeule argue that since (1) threshold deontologists would agree that an otherwise wrongful killing would be permissible to save eighteen lives; and since (2) there are studies indicating that capital punishment prevents eighteen times more killings than are committed through executions; it follows that (3) capital punishment is justified. Sunstein & Vermeule, supra note 28, at 740–41. Assuming claim (2) is correct, the problem is that claim (1) pulls together different types of goals, as explained in the text. 67. See Haque, supra note 41, at 642–43.

143

144


accepted that such tactics are prohibited. These are clearly cases where a person is used as a means for achieving an end. Thus, an army must neither use enemy civilians as “human shields”68 nor take hostages as “bargain chips.”69 In the same vein, the army must not force a local resident to convey a warning message to a wanted person in a place besieged by the army. In both cases, risking the lives of civilians is intended, as it serves as a means to the army’s ends. The Israeli Supreme Court absolutely prohibited even “the solicitation of a local resident’s assistance, for the purpose of relaying an ‘early warning’ . . . when that resident gives his consent and when performance of the role will cause him no damage.”70 Another measure that is sometimes used for pressure purposes is siege. Siege is a legitimate means of warfare if it serves a specific military objective but not if it aims at starving a civilian population in order to force the enemy to surrender.71 A fortiori, it is prohibited to carry out so-called “strategic” or carpet bombing campaigns that result in mass killings of civilians, with the intent of forcing the terrorists to surrender.72 Interestingly, though, in considering the legitimacy of the threat or use of nuclear weapons, the International Court of Justice did not classify this prohibition as absolute.73 Preemption thus seems to be the soundest justification for overriding the constraint against actively/intentionally harming people. However, before

68. See Article 28 of IV Geneva Convention Relative to the Protection of Civilian Persons in Time of War, 1949. 69. Orna Ben-Naftali & Sean S. Gleichgevitch, Missing in Legal Action: Lebanese Hostages in Israel, 41 Harv. Int’l L.J. 185 (2000); CrimFH 7048/97 John Does v. Ministry of Defence, 54(1) PD 721 (2000). English trans. available at http://elyon1.court.gov.il/files_ eng/97/480/070/a09/97070480.a09.htm. 70. HCJ 3799/02 Adalah v. GOC Central Command, IDF 60(3) PD 67, 80 (2005). English trans. available at http://elyon1.court.gov.il/files_eng/02/990/037/A32/02037990.a32.htm. 71. Protocol Additional to the Geneva Conventions of 12 August 1949, and relating to the Protection of Victims of International Armed Conflicts (First Protocol), 8 June 1977, Article 54(1). 72. See, e.g., Eyal Benvenisti, Human Dignity in Combat: The Duty to Spare Enemy Civilians, 39(2) Isr. L. Rev. 81, 102 (2006). For a different, and in our view unfounded view, see Alan Dershowitz, Why Terrorism Works: Understanding the Threat, Responding to the Challenge 177 (2002) (suggesting the destruction of all the houses in the village from which a terrorist originated). 73. Advisory opinion on the Legality of the Threat or Use of Nuclear Weapons, Advisory Opinion, 1996 I.C.J., § 97 (the court declared that it could not “reach a definitive conclusion as to the legality or illegality of the use of nuclear weapons by a State in an extreme circumstance of self-defence, in which its very survival would be at stake.”).


we move on to discuss this aim, one should note that it is often difficult to determine the actual purpose of a specific anti-terrorist measure. We already mentioned the difficulty in ascertaining the decision-maker’s true motive.74 Another difficulty arises when an anti-terrorist measure is taken for more than one purpose. Punitive measures ordinarily serve both backward- and forward-looking aims. When a person who committed a terrorist act is contemplating another attack, targeting him may be retributive, preventive, and deterring, all at the same time.75 Notwithstanding this possible conflation of purposes, in most cases it is possible to determine what the action’s dominant subjective and objective purpose is76 and to set the appropriate threshold function accordingly.

3. Basic Elements of the Threshold Function Recall that T is a threshold function, such that an infringing act is permissible only if the product of this function is positive. In constructing the threshold function, one should define its variables and parameters. The central parameter is the threshold level K, which determines the minimal level of net benefit that is required to justify infringement. The variables are the factors that represent the act’s net benefit. In the context of targeted killing aimed at preemption, the function generally consists of four main variables: (1) The probability that the targeted person(s) will succeed in executing an act of terror unless thwarted ( p) (2) The harm the terror attack is expected to inflict (x) (3) The probability that the preemptive measure will result in killing the targeted person(s), (q) (4) The harm that the preemptive measure is expected to inflict ( y). Accordingly, the action will generate the benefit of preventing the harm of a terror attack x, in probability pq, which is the product of the probability that the targeted person would have executed the terror attack unless

74. See also supra pp. 63–70. 75. McMahan, supra note 56, at 78–82. 76. See, e.g., Seana Valentine Shiffrin, Speech, Death, and Double Effect, 78 N.Y.U. L. Rev. 1135 (2003); Uniacke, supra note 49, at 92–155.

145

146


thwarted p, and the probability that the state action is successful q.77 The action’s cost is denoted as qy. The threshold function evaluates the infringement from either an ex ante perspective, before it is known whether the action will succeed in inflicting harm on the suspect, or ex post, after it is known that harm was inflicted. For simplicity, we assume that the action is certain to succeed (such that q = 1), and focus on the ex ante perspective. The threshold function can take the following form: (1) T = ( px–y) – K For example, if p = 1, x = 2 (it is certain that the terror attack will occur unless thwarted by killing the would-be perpetrator, and will result in killing two persons), and y = 1 (the targeted killing will result in intentionally killing one person, the suspect), the killing is permissible only if the threshold level, K, is lower than 1. If this parameter is higher than 1, as is typically the case, the intended killing is impermissible given these values. As discussed above, the threshold function can also take other forms. The main alternative is to set the threshold level as a multiplier, such that: (2) T = px – yK ′ For reasons discussed in chapter 4,78 we tend to favor combined threshold functions: (3) T = px – yK ′ – K In this function, the amount of (relevant) net benefit necessary to render a constraint infringement permissible should be a function of the actual or probable harm inflicted by the infringing act; yet the threshold is positive even if infringing the constraint causes no harm. The relevant costs that should be multiplied by K ′ may be only part of the act’s actual costs (or harms) y. Moreover, in principle, K ′ may be a vector of multipliers, each multiplied by one type of harm, ( y1, y2 , y3 , . . .).

77. It is assumed that if the action results in killing the targeted person, it is certain that the terror attack will not take place. If this assumption is relaxed, the probability pq should denote only the reduction in the probability of attack as a result of the preemption. 78. See supra pp. 93–96.


We now move to discuss the two elements of the threshold function—first the variables that determine the action’s relevant net benefit and then the parameters K and K ′ that determine the threshold’s shape and magnitude.

4. The Net Benefit (a) The Relevant Variables Standard CBA considers any expected cost of a thwarted terror attack as a benefit of the preventive action. It ordinarily multiplies any such cost by the probability of its occurrence and employs a discount rate to determine the present value of future benefits and costs. Deontologically constrained CBA may treat future, probabilistic, and small outcomes in the same way, but it need not necessarily do so. Excluding or discounting these harms is tantamount to excluding or discounting the benefits generated from the preemptive action. Threshold functions (1) – (3) reflect the view that the infringement is only justifiable if the act directly thwarts a terror attack by physically disabling would-be terrorists, as it calculates the action’s benefit by taking into account only the direct harm of the prevented terror attack. It is unclear whether unintended potential benefits, mainly the indirect effect of reducing terror attacks through general deterrence, should also count in calculating the act’s net benefit.79 We tend to answer this question in the affirmative.80 Focusing on preemptive measures, a moderate deontologist may wish to exclude or discount chronologically remote harms of the relevant would-be terror attack and harms whose probability is low. These excluders may echo

79. For instance, while the arguable aim of the invasion of the United States and its allies to Iraq was related to preemption, it also generated, as a presumably unintended byproduct, the benefit of liberating the people of Iraq from an oppressive regime. See Cryer & Simester, supra note 40 (arguing that only the action’s aim may justify it, not the positive side effects); David Enoch, Ends, Means, Side-Effects, and Beyond: A Comment on the Justification of the Use of Force, 7 Theoretical Inq. L. 43 (2006). 80. Cf. the Israeli Supreme Court decision in Ajuri, supra note 65, at § 27. The court held that the military commander is empowered to issue an order assigning a place of residence only when it can demonstrate that the person is dangerous. The court further stated that once this element is proved, the discretion whether to exercise this power is subject to the proportionality requirement, and in this respect “there is no defect in the military commander taking into account considerations of deterring others in deciding whether to issue the order.”

147

148


the criminal law self-defense doctrine, under which an act is considered preemptive, and thus justifiable, only if it aims at thwarting an imminent harm.81 This requirement can be interpreted as referring to either the temporal element or the probabilistic element of the threat.82 It can be reflected in the analysis in two ways. One way, discussed here, is to exclude such harms from the calculation of the action’s relevant net benefits; another way, discussed in the next subsection, is to set a much higher threshold when the risk is not imminent.83 One may wish to exclude or discount chronologically remote harms of the relevant would-be terror attack based on instrumental and contingent considerations. The more distant the expected harm, the more difficult it is to assess its probability and magnitude, and thus the greater the danger of making an erroneous decision. A moderate deontologist may also exclude chronologically remote harms whenever the state will have an opportunity to prevent the harm without killing the suspect. However, since the availability and effectiveness of such alternative measures vary from one scenario to another, it seems implausible to universally disregard future harms altogether. Similarly, it seems that in setting the minimal probability threshold, one should plausibly take into account the type and magnitude of the expected harm. Thus, one might not wish to exclude the danger of a terror attack that is expected to result in a high number of victims, even if its probability is rather low.84 A moderate deontologist may also exclude harms that the suspect aims to cause, which are lexically inferior to the harm that the preemptive action inflicts. Arguably, the only types of benefit that can justify overriding the constraint against actively/intentionally killing a person is the saving of human lives or preventing serious injuries. If one considers human dignity and bodily integrity as lexically superior to pecuniary losses,85 then an infringement of the constraint against killing is never permissible for the sake of preventing pecuniary losses, regardless of their size.86

81. See, e.g., Sangero, supra note 51, at 150–65 (“In every one of the legal systems that were examined . . . there is a requirement for some kind of immediacy”); Fletcher, supra note 58, at 18–28. 82. See infra pp. 153–55. 83. See infra pp. 151–60. 84. For a discussion of these considerations in the context of curtailing speech, see infra pp. 196–201. 85. See supra pp. 86–87. 86. For a discussion of reasons for the impermissibility of killing to protect property see, e.g., Leverick, supra note 49, at 131–42. See also David Luban, Preventive War, 32 Phil. & Pub.


If one excludes all but very few types of harm, the threshold function may be considerably simplified. Suppose that the only benefit that may ever justify killing a person is the saving of human lives. In that case, x would not represent the terror attack’s entire possible harm but only the number of persons that would be killed. If y, the harm that the preemptive measure is likely to inflict, also represents a number of lives, then one may construct a threshold function without confronting difficulties of incommensurability. (b) Marginal Net Benefit and Alternative Courses of Action In deciding whether to take a certain anti-terrorist measure, one should compare the measure not only to inaction but also to other courses of action that either do not infringe any deontological constraint or whose net benefit is large enough to render the infringement permissible. In particular, the decision-maker should consider those alternatives that mitigate the risk of a terror attack through less harmful acts.87 The comparison to alternative, deontologically permissible courses of action requires redefining the relevant variables of the threshold function according to their marginal values. Specifically, the action’s relevant benefit is only the difference between the risk of a terror attack if the alternative course of action is pursued and the magnitude of this risk under the current one. Similarly, the relevant cost is only the added harm that the infringement inflicts in comparison to the alternative. Thus, the inquiry is whether the more harmful infringement is justified given its marginal net benefit. As discussed below, this qualification may well result in a change in the values of K and K ′.88 This analysis affects the content of the “minimal impairment” requirement, also known as the requirement that the infringement will be narrowly

Aff. 207, 233 (2004) (“the need to formulate the doctrine of preventive war narrowly suggests that the only threat justifying a preventive war is that of an armed attack against the basic rights of a state’s people, not its economic interest in maintaining a level substantially beyond the fulfillment of basic rights.”). 87. Cf. McCann v. UK, (1995) 21 EHRR 97 (holding that killing terrorists was justified as a means to thwart an immanent risk of a deadly terrorist attack, yet that the “anti-terrorist operation as a whole” was unjustified, since the government could have used alternative measures to thwart the risk, such as arresting the suspects at an appropriate stage). 88. See infra pp. 155–56, 169–70.

149

150


tailored to achieve its aim. According to some suggestions in the legal literature, an infringement meets this requirement if it is the “least-harmful” among all possible infringements that are equally effective in thwarting the risk.89 In contrast, we suggest that the evaluation of an infringement must also include a comparison to less harmful alternatives that are possibly less effective in tackling the risk, in order to determine whether the action’s marginal net benefit is large enough to justify the harsher measure, given the value of the parameters K and K ′.

5. The Threshold (a) General The distinctive mark of deontologically constrained CBA is that whenever an act or a rule infringes a deontological constraint, such infringement is only permissible if its (relevant) net benefit surpasses a certain threshold. This threshold can be interpreted as reflecting the proportionality requirement. According to this requirement, common to the criminal law doctrine of self-defense and to human rights law, the scope of the harm that the infringement inflicts should be proportional to the act’s benefit (that is, to the terrorist attack’s expected harm). The size of the threshold is determined by the moral severity of the infringement. As mentioned, Posner and Vermeule argue that a central deficiency of moderate deontology is that it “builds in an arbitrary threshold,”90 and thus “[it] is quite mysterious . . . why the sheer catastrophic size of the threatened harm should matter.”91 The following discussion demonstrates that even though there is room for discretion in setting the threshold level, this process is far from arbitrary. As already discussed, a central factor determining the size of the threshold is whether the infringement involves an intended harm or merely a foreseen one. An infringement of the former type, which is the focus of the present section, entails a substantially higher threshold than the latter. We also

89. See, e.g., the decision of the Canadian Supreme Court in R. v. Oakes, [1986] 1 S.C.R. 103. See also Guy Davidov, Separating Minimal Impairment from Balancing: A Comment on R. v. Sharpe (B.C.C.A.), 5 Rev. Const. Stud. 195 (1999). 90. Posner & Vermeule, supra note 10, at 188. 91. Id. at 188.


submit that the threshold should be substantially higher when the infliction of harm by the state aims at deterrence rather than preemption.92 The type of harm is of course also relevant in setting the threshold.93 We shall now discuss two additional factors that may affect the size and shape of the threshold—the level of certainty that the targeted person will attack if not thwarted and his culpability. (b) Probability of the Terrorist Attack A central element in determining the permissibility of the infringing act is the expected result absent such an act. These results are a function of p, the probability that the targeted person(s) will succeed in executing the terror attack unless thwarted, and x, the expected harm of the terrorist attack. As discussed above, one may exclude low-probability risks from the calculation of the action’s net benefit.94 In addition—and this is the focus of the current discussion—the probability of the risk may affect the size of the threshold. Arguably, instances of imminent threat give rise to a unique type of justifiable infringement upon the aggressor’s right to life. While we do not share the view that when the harm is imminent, the preemption does not infringe the constraint against active/intended infliction of harm due to the aggressor’s culpability,95 we believe that in these cases the required threshold may be relatively low. In contrast, where the likelihood that the aggressor will attack is low, the threshold should arguably be very high, such that an infringement can be justified only in catastrophic or near-catastrophic circumstances.96 This proposition rests on two grounds. First, it serves important secondorder, rule-consequentialist considerations. When the risk is remote, it is

92. See supra pp. 140–43. 93. However, the ranking of different types of harms is not as simple as it may seem, as illustrated by the debate whether committing torture is worse than killing. See infra p. 162. 94. See supra pp. 89–91. 95. See supra pp. 136–39. 96. This distinction may also apply to the assessment of the efficacy of the infringement. Arguably, the threshold should be set at a low level only if the harm that the state inflicts is (reasonably) effective in preventing (or mitigating) the danger. See Daniel Statman, On the Success Condition for Legitimate Self-Defense, 118 Ethics 659 (2008).

151

152


normally less likely that the infringement is necessary to prevent the attack.97 The concern is that if the size of the threshold is independent of the probability of the attack, preemptive measures such as targeted killings, torture, or detention will become standard policy measures, implemented on a regular basis.98 In addition, when the risk is remote, there is a high danger of erroneous, and sometimes biased, evaluation of the legitimacy of the infringement. The more remote the expected harm, both in time and in likelihood, the more difficult it is to assess its probability and magnitude. One may expect systematic overestimation of threats, driven by cognitive bias99 and by the government’s possible tendency to give a higher weight to one type of error (inaction that results in a terror attack) over the other (inflicting harm on innocent persons).100 Setting probability thresholds may serve as a necessary corrective mechanism. Setting the size of the threshold according to the imminence of the risk is also grounded in deontological considerations that would hold true even absent any of the above considerations. Arguably, taking one’s life, dignity, or liberty in response to an imminent risk is qualitatively different from inflicting these harms in other circumstances. The deontological constraint against actively/intentionally killing a person reflects not only the moral value of life but also a normative judgment about the meaning of errors. Faced with uncertainty regarding the threat imposed by a suspect, a decision-maker must weigh the risk of a “false negative,” that is, inaction that would result in a terror attack, against that of a “false positive,” namely, an unnecessary infringement when the suspect would not have attacked anyhow. When punitive measures are concerned, providing a probability threshold (the requirement that the guilt of the accused be proven “beyond a reasonable doubt”) reflects a normative judgment regarding the appropriate weight of each of these errors. It embodies the deontological notion that the harms generated by false convictions of

97. Leverick, supra note 49, at 87–89. 98. For the use of this terminology in the broader context of “preventive war,” i.e., the use of force prior to the existence of an imminent attack see Michael Walzer, Just and Unjust Wars 77-79 (3d ed. 2000); Luban, supra note 86, at 225. 99. See, e.g., W. Kip Viscusi, Valuing Risks of Death from Terrorism and Natural Disasters, 38 J. Risk & Uncertainty 191 (2009) (reporting that nationally representative sample values preventing terrorism deaths at about the same level as preventing deaths from traffic accidents, although the latter poses a much greater personal risk). 100. Jonathan S. Masur, Probability Thresholds, 92 Iowa L. Rev. 1293 (2006–2007).


innocent people greatly exceed the harms generated by acquitting guilty criminals.101 Applying a probability threshold in the case of preemptive measures reflects a related rationale. The moral weight of the harm that results from a false positive—an active and intended infliction of harm—exceeds that of harms generated by a false negative—a passive, unintended failure to prevent harm. According to the doing/allowing and intending/foreseeing distinctions, as well as the distinction between deontological options and deontological constraints, a state’s failure to eliminate threats of terrorism is morally less significant than an erroneous active/intended harming of people.102 The view that one type of errors (false positive) is morally worse than the other (false negative) entails that the government can justifiably act only when the probability of a terror attack is sufficiently high, such that the likelihood of a false positive is lower than that of a false negative. In the context of the criminal law justification of self-defense, the “imminence” requirement typically contains a temporal, in addition to a probabilistic element. A threat is considered imminent only if it is expected to materialize within a very short period of time.103 However, this more stringent approach to the requirement is questionable. An often-cited example in the criminal law literature is that of a battered woman who kills her abusive partner when the threat to her life is not immediate but still inevitable, as she lacks any meaningful alternatives to the use of deadly force.104 In the international relations context, a typical example is that of launching an attack against a hostile state or organization that does not yet possess weapons of mass destruction but is highly likely to gain such weapons and to wield them against the state in the near future.105 In the fight against terrorism, it is often the case that a terrorist can be preempted only before the harm becomes immediate (or that the inevitable harm to innocents at that point in time is significantly lower). While some scholars insist that in

101. See, e.g., Robert Nozick, Anarchy, State and Utopia 96–108 (1974); Alex Stein, Foundations of Evidence Law 141–53, 172–83 (2005). 102. See supra pp. 41–46, 60–70. 103. See, e.g., Leverick, supra note 49, at 87–89. 104. Leverick, supra note 49, at 89–93; Richard A. Rosen, On Self-Defense, Imminence, and Women Who Kill Their Batterers, 71 N.C. L. Rev 371 (1993); Kimberly Kessler Ferzan, Defending Imminence: From Battered Women to Iraq, 46 Arizona L. Rev. 213, 231–37 (2004). 105. Ferzan, id. at 218–31; Alan M. Dershowitz, Preemption—A knife that Cuts Both Ways 76–104 (2006).

153

154


such cases the self-defense justification does not apply, as the risk is not imminent,106 others argue that in some or all of these examples, the infringement of the aggressors’ right to life may well be justified.107 According to this view, the imminence requirement should be subsumed within necessity (the so-called “immediately necessary” standard),108 so that the morally relevant issue is not the immediacy of the harm but the immediacy of the response necessary to thwart it.109 In addition, one may argue that the requirement of imminence in the criminal law self-defense justification primarily aims at restricting the use of “self help.” It reflects the presumption that when the risk is not imminent, one may call the police for help. This rationale does not apply to the case of state action.110 Accordingly, the crucial question is not how close to completion the threat is, but rather when the preemptive measure should be taken. Sometimes, to be effective, one must take the preventive measures before the threat is imminent in the temporal sense. Otherwise, the would-be victim would bear the risk that when the risk is imminent, it will be too late to take an effective measure to preempt the harm.111 The decisive factor is thus the probability of the attack in the absence of preemption rather than its timing.112 This is not to say that the temporal element of the risk is completely irrelevant. Ignoring the temporal element may raise the problem of cumulative likelihood over time. Occasionally, even relatively improbable attacks can attain a high cumulative likelihood within a few years.113 It may also raise

106. Walzer, supra note 98, at 79; Uniacke, supra note 49, at 159; Ferzan, supra note 104, at 255–62. 107. See, e.g., Rosen, supra note 104; Paul H. Robinson, Criminal Law Defenses § 131(c) (1984); Fletcher, supra note 81, at 21; Anthony Clark Arend, International Law and the Preemptive Use of Force, 26 Wash. Q. 89, 98 (2003). 108. Robinson, supra note 107, at 77. 109. Id. at 77; Abraham D. Sofaer, On the Necessity of Pre-emption, 14 Eur. J. Int’l L. 209 (2003). A middle ground, that is available only in the context of criminal liability, is that in the absence of imminence, killing an attacker may be subject to a (partial) excuse defense. Leverick, supra note 49, at 108; Larry Alexander, A Unified Excuse of Preemptive SelfProtection, 74 Notre Dame L. Rev. 1475 (1999). 110. Assuming, that is, that an action by the United Nations Security Council under Chapter VII of the Charter of the United Nations is not a viable option. 111. Jeremy Horder, Killing the Passive Abuser: A Theoretical Defece, in Criminal Law Theory: Doctrines of the General Part 283, 292 (S. Shute & A.P. Simester eds., 2002); Dershowitz, supra note 105, at 105–52. 112. Luban, supra note 86, at 233–34. 113. Id. at 234.


difficult evidentiary problems.114 If so, the assessment of imminence should take into account both the likelihood of harm and the period for which this likelihood is calculated. On these grounds, a moderate deontologist may wish to divide the spectrum of risks into two categories and set different thresholds for each. One category consists of cases in which there is absolute certainty that the suspect will launch a terrorist attack if not preempted, as well as cases in which the probability is sufficiently high to be considered a “moral certainty.”115 The other category consists of cases in which the risk is lower than the probability threshold of “moral certainty.” K should be considerably greater in the latter category. To incorporate this consideration, one may introduce another parameter, a probability threshold denoted as p*. Using a multiplier form of T, two distinct levels of K are defined, denoted as Kim (where the risk is imminent) and Kr (where the risk is remote), such that Kim is much lower than Kr . ⎧ px − yK im if (4) T = ⎨ ⎩ px − yK r if

p ≥ p* p < p*

In evaluating the permissibility of an infringement according to such a threshold function, it is essential to measure the marginal value of p, that is, the likelihood of an attack if the state takes an alternative, less harmful measure. To illustrate, assume that the targeted person poses an imminent risk (e.g., p = 95%) of killing others. Consequently, the threshold that should be met to justify targeting the aggressor will be set according to Kim, and may thus be rather low. Assume, however, that the government can also try to capture the terrorist and that the probability of success of this measure is 80% (in comparison to the assumed certainty that the targeted killing will thwart the attack). Thus, there is a 19% chance that the terrorist will escape

114. W. Michael Reisman & Andrea Armstrong, The Past and Future of the Claim of Preemptive Self-Defense, 100 Am. J. Int’l L. 525, 526 (2006) (arguing that in anticipatory self-defense the “interpretive latitude of the unilateralist becomes wider, yet the nature and quantum of evidence that can satisfy the burden of proof resting on the unilateralist becomes less and less defined and is often . . . extrapolative and speculative”); Michael Bothe, Terrorism and the Legality of Pre-emptive Force, 14 Eur. J. Int’l L. 227 (2003). 115. Frank Jackson & Michael Smith, Absolutist Moral Theories and Uncertainty, 103 J. Phil. 267, 275 (2006). See also Ron Aboodi, Adi Borer & David Enoch, Deontology, Individualism, and Uncertainty: A Reply to Jackson and Smith, 105 J. Phil. 259 (2008).

155

156


and carry out her terrorist plans.116 Trying to capture the terrorist would not be justified if it involves too high a risk to bystanders and to the soldiers. If so, the targeted killing should be evaluated in comparison to the alternative of inaction, as discussed above. If, however, the attempt to capture the terrorist is permissible given the values of the relevant variables and parameters, then one should only consider the marginal benefit of the targeted killing. In this case, the variable p of the threshold function, according to which one should evaluate the targeted killing, is the marginal one, namely 19%. Assuming that this marginal probability is lower than the probability threshold p*, the result is not only a possibly lower net benefit resulting from this act in comparison to the alternative of inaction (since the expected risk of a terror attack is now lower, as only its marginal value is relevant), but also a higher threshold Kr.117 Setting different thresholds for justifying anti-terror measures depending on the probability of the terrorist attack raises another thorny question, namely what is the relevant probability when an anti-terror measure indiscriminately affects both suspected terrorists and innocent people. Consider, for example, a case in which it is known that only one person out of a group of one hundred intends (in “moral certainty”) to commit a terrorist attack. Suppose further, that the only way to preempt the would-be perpetrator is by curtailing the freedom of movement of the entire group. Hence, limiting the freedom of movement of the whole group will, in moral certainty, save lives. However, the probability that limiting the freedom of movement of each member of the group will yield this benefit is 1%, way below the probability threshold. Which threshold function is appropriate in such a case? Defining the action under consideration as inflicting harm on the group would imply that this is a case of imminent danger, or at least a case that meets the probability threshold, such that a relatively low K is appropriate. In contrast, assessing the infringement on an individualized basis would mean that the risk is very low (only 1%), hence calling for a very large K. While some scholars seem to endorse the former approach,118 we find it

116. The chances that the would-be terrorist will escape (20%) and launch a terror attack (95%), assuming that the two events are independently distributed, are 0.2x0.95=0.19. 117. Therefore, as discussed earlier, we suggest that the evaluation of an infringement must include a comparison to less harmful alternatives that are possibly less effective in tackling the risk, in order to determine what the appropriate value of K is. 118. See, e.g., Jackson & Smith, supra note 115.


rather problematic.119 The application of the former approach results in circumventing the deontological constraint against preempting a person when the risk posed by that person is lower than the relevant probability threshold.120 The question of indiscriminate measures is closely connected to the issue of profiling. Without delving into this complex issue, it should be noted that in addition to the collective/individual threat issue, when profiling is based on religious or ethnic classification, it raises the additional concern of discrimination and the indirect harms associated with it.121 In this respect, the fight against terrorism raises a dilemma that has long been considered moot in “classic” warfare. In the latter context, the laws of armed conflict proscribe only targeting civilians,122 and thus do not subject the legitimacy of killing combatants to an inquiry of the level of risk posed by each enemy combatant. While the justification of this approach is not self-evident,123 it is an established convention. The problem is how to classify suspected terrorists. On the one hand, terror attacks, which violate the laws of war, are not privileged as acts of war and should thus be assessed in the usual way as culpable homicides. Accordingly, to justify preemptive measures

119. For a related view see Aboodi et al., supra note 115. 120. See e.g., Edmond v. Goldsmith, 183 F.3d 659 (7th Cir. 1999), aff ’d, City of Indianapolis v. Edmond, 531 U.S. 32 (2000). In this case, the court enjoined the Indianapolis police department from setting up roadblocks to catch drug offenders. Judge Richard Posner stated that if the roadblocks were assessed “at the level of the entire program . . . these roadblocks probably are legal, given the high ‘hit’ rate and the only modestly intrusive character of the stops.” However, he held that the reasonableness of a search must be determined by the existence or nonexistence of “individualized suspicion of wrongdoing,” since if the court were to adopt a program-level analysis, then the court would perform a cost-benefit analysis. Id. at 661. See also Bernard E. Harcourt, Judge Richard Posner on Civil Liberties: Pragmatic Authoritarian Libertarian, 74 U. Chi. L. Rev. 1723 (2007). 121. See, e.g., Murad Hussain, Defending the Faithful: Speaking the Language of Group Harm in Free Exercise Challenges to Counterterrorism Profiling, 117 Yale L.J. 920 (2008) (Counterterrorism measures that scrutinize conduct and behavior that are presumed to be probative of terrorist activity, may inflict pervasive dignitary and stigmatic harms when such conduct-based profiling specifically targets activity that is also expressive of Muslim identity); Daniel Moeckli, Human Rights and Non-Discrimination in the ‘War on Terror’ (2008). 122. Article 48 to the First Protocol to the Geneva Conventions, supra note 71: “Basic rule: In order to ensure respect for and protection of the civilian population and civilian objects, the Parties to the conflict shall at all times distinguish between the civilian population and combatants and between civilian objects and military objectives and accordingly shall direct their operations only against military objectives.” See also id. Article 51. 123. See Jeff McMahan, The Ethics of Killing in War, 114 Ethics 693, 722–73 (2004).

157

158


against a suspected terrorist, one must establish the target’s “individual dangerousness,” that is, to show that the risk that this person poses, given his specific activities, capabilities, and intentions, is sufficiently high. On the other hand, one may wonder whether the fact that terrorists violate the laws of war should provide them a wider protection from preemptive measures than the one accorded to “legal” combatants. The rule set forth in this respect in the First Additional Protocol to the Geneva Conventions is ambiguous. It provides that “civilians shall enjoy the protection afforded by this section, unless and for such time as they take a direct part in hostilities.”124 According to one view, a civilian might be a legitimate target for attack only if the attack against him is carried out while he is directly participating in the hostile activities, such as when he shoots or plants a bomb.125 According to a competing approach, all persons “performing the function[s] of combatants” are legitimate targets, including those who plan an attack.126 Moreover, this approach comes close to classifying terrorists as (illegal) combatants, by referring to a person’s active “membership” in a terror organization as a sufficient basis for denying him the protection accorded civilians.127

124. Article 51(3) of the First Protocol to the Geneva Conventions, supra note 71. 125. See, e.g., Kristen E. Eichensehr, On Target? The Israeli Supreme Court and the Expansion of Targeted Killings, 116 Yale L.J. 1873, 1877 (2007) (“This narrow reading assures that the target poses an immediate threat (the timing of the attack signals the proximity of the threat)”); Georg Nolte, Preventive Use of Force and Preventive Killings: Moves into a Different Legal Order, 5 Theoretical Inq. L. 111 (2004). 126. The Israeli Supreme court decision in Public Comm. against Torture in Israel v. Government of Israel, supra note 43, at § 35. See also Michael N. Schmitt, Direct Participation in Hostilities and 21st Century Armed Conflict, in Krisensicherung und Humanitärer Schutz–Crisis Management and Humanitarian Protection: Festschrift für Dieter Fleck 505 (Horst Fischer et al eds., 2004); Orna Ben-Naftali & Keren R. Michaeli, ‘We Must Not Make a Scarecrow of the Law’: A Legal Analysis of the Israeli Policy of Targeted Killings, 36 Cornell Int’l L.J. 233 (2003); Daniel Statman, Targeted Killing, 5 Theoretical Inq. L. 179 (2004); Tamar Meisels, Combatants—Lawful and Unlawful, 26 Law & Phil. 31, 51–63 (2007). 127. The Israeli Supreme court decision in Public Comm. Against Torture in Israel v. Government of Israel, id. at § 39: “A civilian who has joined a terrorist organization which has become his ‘home,’ and in the framework of his role in that organization . . . commits a chain of hostilities, with short periods of rest between them, loses his immunity from attack ‘for such time’ as he is committing the chain of acts.” In the context of detentions, the United States Supreme Court came close to this approach, by finding legal authorization for the detention of at least those persons who were “part of or supporting forces hostile to the United States or coalition partners” in Afghanistan and who “engaged in armed conflict against the United States there.” Hamdi v. Rumsfeld, 542 U.S. 507, 516 (2004). For a critique see Matthew C. Waxman, Detention as Targeting: Standards of Certainty and Detention of Suspected Terrorists, 108 Colum. L. Rev. 1365 (2008).


In our opinion, the appropriate approach is an intermediate one. One should not rule out the legitimacy of targeting persons who violate the laws of war, even when the action is not taken while they actually commit a terror attack. The relevant inquiry should not focus on the temporal element but rather on the magnitude of risk that such persons pose. At the same time, one should not regard a person’s mere “membership” in a terror organization, or his past involvement in hostilities, as sufficient and conclusive evidence that this person poses a sufficiently high risk justifying targeting him. Unlike “legal” combatants who are characterized as such only if, inter alia, they wear a fixed distinctive emblem recognizable at a distance and carry arms openly,128 “illegal” combatants cannot be easily identified. The risk of targeting innocent civilians thus requires determining what level of risk each person, who is not a “legal” combatant, poses. Such determination should take into account all available information and not rest on the mere classification of persons as (illegal) combatants.129 To sum up, a deontologically constrained CBA of anti-terror measures should plausibly set different thresholds for the net benefit that may justify the pertinent infringements, depending on the expected probability of the terrorist attack absent those measures. When the probability reaches “moral certainty,” the threshold need not be extremely high. When the probability is lower, anti-terror measures may be justified only if a very high threshold is met.130 When an anti-terror measure entails indiscriminately harming both suspected terrorists and innocent people, the pertinent probability is the one referring to the risk posed by each person individually and not by the group as a whole (which leaves the door open to indiscriminate measures only if their expected

128. Article 1 of the 1907 Hague Regulations concerning the Laws and Customs of War on Land, annex to the Hague Convention (IV) respecting the Laws and Customs of War on Land, The Hague, 18 October 1907. See George Fletcher, Romantics at War— Glory and Guilt in the Age of Terrorism 108 (2003). 129. David Kretzmer, Targeted Killing of Suspected Terrorists: Extra-Judicial Executions or Legitimate Means of Defence?, 16 Eur. J. Int’l L. 171, 211 (2005) (“given the inherent uncertainty in identification, in order to justify targeting a suspected terrorist, not only must the authorities have clear and convincing evidence of that person’s active involvement in terrorist activities. They must also have credible evidence that the person poses a real and concrete danger to the lives of others that cannot be thwarted without attacking him at the time he is attacked”). See also McMahan, supra note 123, at 722–23. 130. For a related argument see Luban, supra note 86, at 230 (“preventive war can be justified, if at all, only against . . . states that exhibit clear evidence of a military build-up with aggressive intentions. [In addition] If a state seems likely to develop WMD and give them to terrorists, the case readily assimilates to the restricted doctrine of preventive war.”).

159

160


net benefit is extremely high). A person’s “membership” in a terrorist organization should not be sufficient to make him a legitimate target. (c) The Aggressor’s Culpability Another factor affecting the size of the threshold K is the preempted person’s moral responsibility for creating the danger. A very high threshold applies when the state actively/intentionally harms an innocent person who unintentionally or negligently poses a threat to others. In contrast, a considerably lower threshold should apply when a preemptive measure is taken against a morally blameworthy person, at least in a context closely related to his blame. If a terrorist has made it inevitable that either he or the potential victim(s) will be harmed, fairness requires that it be the terrorist.131 One may also argue that the aggressor’s responsibility for the situation justifies discounting the value of his interests. Such discounting may legitimize selfdefense even if the number of targeted aggressors exceeds the number of their intended victims.132 A threshold function can reflect this consideration by setting two or more different levels of the multiplier K, depending on the preempted person’s moral responsibility. A more nuanced inquiry would also take into account the degree of certainty of the preempted person’s guilt. (d) Summary To recapitulate, K plausibly depends on the type of anti-terrorist measure, the likelihood that the terrorist act will be carried out absent the preemption, and the targeted person’s degree of culpability. The parameter K expresses the notion that the expected harm if the government does not apply the

131. See, e.g., Michael S. Moore, Torture and the Balance of Evils, 23 Isr. L. Rev. 280, 321–22 (1989); David Wasserman, Justifying Self-Defense, 16 Phil. & Pub. Aff. 356 (1987); Shlomit Wallerstein, Justifying the Right to Self-Defense: A Theory of Forced Consequences, 91 Va. L. Rev. 999, 1027–32 (2005); Jeff McMahan, Torture, Morality, and Law, 37 Case West. Res. J. Int’l L. 241 (2006). 132. See, e.g., Jeff McMahan, Self-Defense and the Problem of the Innocent Attacker, 104 Ethics 252, 261 (1994) (“[A]ccording to commonsense morality, an Innocent Victim is permitted to kill a [culpable aggressor] irrespective of differences in age, quality of life, or usefulness to society. . . . She may kill any number of [culpable aggressors] if this is necessary for self-defense.”); George P. Fletcher, Rethinking Criminal Law 857–58 (1978); Paul H. Robinson, A Theory of Justification: Societal Harm as a Prerequisite for Criminal Liability, 23 UCLA L. Rev. 266, 272–73 (1975).


preemptive measure, px (inactive/unintentional harm) and the harm that the measure inflicts, y (active/intentional harm) are not symmetric. While the multiplier K, reflecting the magnitude of the deontological constraint against actively/intentionally harming the suspected terrorists, will often be quite large, there may be cases in which it would be equal to or even less than 1. This may be the case when the probability p exceeds the threshold level of moral certainty, p*, and the active/intentional harm is inflicted only on people who are morally responsible for executing the terror attack and causing its harm (x) absent the preemption. In such cases, the difference between a standard CBA and a deontologically constrained one may be rather small (though the two may still diverge in terms of the scope of costs and benefits each one takes into account—the former being much more inclusive). In fact, a deontologically constrained CBA may even be more permissive than a standard CBA, since the former but not the latter may justify killing even a large number of culpable aggressors to save a fewer number of intended victims. Under most circumstances, however, K would substantially exceed 1, and thus a deontologically constrained CBA will be more restrictive than a standard one.

6. Torture Torture poses one of the most troubling dilemmas in the fight against terrorism. Is it ever legitimate to use coercive interrogation techniques in order to force a person to reveal information that can save lives? By “torture,” we denote the intentional infliction of extreme physical suffering on a nonconsenting, defenseless person, to obtain a confession or information, or to coerce the sufferer or others to act in certain ways, by breaking the victim’s will.133 The Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment requires signatory states to criminalize all acts of torture and take effective measures to prevent

133. Cf. the definition of torture in Article 1 of the Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, adopted Dec. 10, 1984, S. Treaty Doc. No. 100-20 (1988), 1465 U.N.T.S. 113. For classification of different interrogation techniques that violate one’s right to dignity, see the decision of the Israeli Supreme Court HCJ 5100/94 Public Comm. against Torture in Israel v. State of Israel, 53(4) PD 817. English trans. available at 38 I.L.M. 1471 (1999) (Isr.). See also Jeremy Waldron, Cruel, Inhuman, and Degrading Treatment: The Words Themselves (working paper, 2008), available at http://ssrn.com/abstract=1278604.

161

162


acts of torture in all territories under their jurisdiction.134 Numerous national legal systems lay down comparable prohibitions.135 Torture clearly infringes a deontological constraint against actively inflicting intentional physical and psychological harm. It reduces a person to his corporality, as the tortured person is reduced to a reflex produced within his body. Assuming, however, that under certain circumstances, it is permissible to kill a person who poses an imminent risk to the lives of others, should not the same analysis apply to the torture of at least culpable persons in similar circumstances? Legal economists argue that killing is clearly worse than torture, as the victim of torture can at least survive.136 They thus claim that if it is permissible to kill a person in certain circumstances, coercive interrogation should be permissible a fortiori.137 In contrast, advocates of an absolute prohibition of torture argue that torturing a person may be, at least in some circumstances, worse than killing him. Torture violates the prohibition against assault upon the defenseless.138 Moreover, inducing a person to act by harming him is worse than harming one to prevent him from acting because it uses him as a means.139

134. Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, supra note 133. 146 states have ratified the Convention. See generally Jeremy Waldron, Torture and Positive Law: Jurisprudence for the White House, 105 Colum. L. Rev. 1681 (2005). 135. For a general survey, see, e.g., Waldron, supra note 133. In the United States, see 18 U.S.C. §§ 2340–2340A (2000) (outlawing torture); Detainee Treatment Act of 2005, Pub. L. No. 109-148, § 1003, 119 Stat. 2739, 2739. However, following the enactment of the latter law, President George W. Bush declared that he considers it an unconstitutional encroachment on his authority, so that he is not bound by it. The Department of Justice has contrived an interpretation of the law according to which a conduct does not count as cruel, inhuman, or degrading if it is undertaken on behalf of a legitimate governmental interest such as intelligence gathering. See a letter from William E. Moschella to Senator Patrick Leahy, available at http://www.scotusblog.com/movabletype/archives/CAT%20 Article%2016.Leahy-Feinstein-Feingold%20Letters.pdf. See also David Luban, Torture and the Professions, 2 Crim. Justice Ethics 58 (2007). 136. Posner & Vermeule, supra note 10, at 189; Posner, supra note 10, at 82. 137. Posner & Vermeule, supra note 10, at 185–87; Posner, supra note 10, at 81. See also Sanford Levinson, “Precommitment” and “Postcommitment”: The Ban on Torture in the Wake of September 11, 81 Tex. L. Rev. 2013, 2032 (2003). 138. Henry Shue, Torture, 7 Phil. & Pub. Aff. 124, 125 (1978); David Sussman, What’s Wrong with Torture?, 33 Phil. & Pub. Aff. 1, 3, 15 (2005) (arguing that a unique moral aspect of torture separates torture from other types of violence); Michael Davis, The Moral Justification of Torture and other Cruel, Inhuman, or Degrading Treatment, 19 Int’l J. App. Phil. 161, 165 (2005). 139. Sherry F. Colb, Why is Torture “Different” and How “Different” is it, 30 Cardozo L. Rev. 1411 (2009); Christopher Kutz, Torture, Necessity, and Existential Politics, 95 Cal. L. Rev. 235,


Another line of argument, central to the academic debate on torture, focuses on second-order considerations. Several scholars have claimed that the very legitimization of the use of torture, even if limited to extreme circumstances of the so-called “ticking bomb scenario,” would induce interrogators to use this measure in other circumstances as well.140 Some go one step further, arguing that the very discussion of the possible legitimacy of the use of torture is dangerous in this respect and should be avoided.141 Here, too, the question is whether justifying torture is different from legitimizing killing. We shall not discuss here these second-order considerations in any detail. The slippery slope argument, as well as the concern for an abuse of power, rest on questionable empirical assumptions. Even if this argument holds, it does not necessarily exceed a purely consequentialist framework. The argument that the mere discussion of torture, let alone the formulation of administrative guidelines for its execution or the enactment of laws regulating it, is incompatible with due respect for human dignity, was discussed in chapters 2 and 5.142 In what follows, we concentrate on the first-order arguments. It stands to reason that, while the end result of killing a person (for example, in an armed conflict) is worse than torturing him, the act of torturing a defenseless person may be considered more disrespectful to his dignity as an autonomous human being. The assessment that at least under some

275 (2007); Seth F. Kreimer, Too Close to the Rack and the Screw: Constitutional Constraints on Torture in the War on Terror, 6 U. Pa. J. Const. L. 278, 294–95 (2003); Waldron, supra note 134; David Luban, Liberalism, Torture, and the Ticking Bomb, 91 Va. L. Rev. 1425, 1429–32 (2005). 140. Henry Shue was influential in arguing that while torture may be justified when it is the least harmful means available to secure a supremely important aim, it should nevertheless be strictly prohibited, since “[a]ny practice of torture one set in motion would gain enough momentum to burst any bonds and become a standard operating procedure.” Shue, supra note 138, at 141. See also Hugo M. Mialon, Sue H. Mialon & Maxwell B. Stinchcombe, Torture in Counterterrorism: Agency Incentives and Slippery Slopes (working paper, 2009), available at http://userwww.service.emory.edu/~hmialon/ TortureInCounterterrorism.pdf (arguing that legalizing torture can significantly increase torture of innocent individuals through a slippery slope mechanism); Yuval Shany, The Prohibition against Torture and Cruel, Inhuman and Degrading Treatment and Punishment: Can the Absolute be Relativized under Existing International Law?, 56 Cath. U. L. Rev. 837 (2007). 141. Slavoj Žižek, Welcome to the Desert of fhe Real!: Five Essays on September 11 and Related Dates 103–04 (2002); Crocker, supra note 25, at 606–10; Luban, supra note 139, at 1440; Henry Shue, Torture in Dreamland: Disposing of the Ticking Bomb, 37 Case W. Res. J. Int’l L. 231 (2006). 142. See supra pp. 52–53, 117–22. See also supra pp. 70–75.

163

164


circumstances, torture is worse than killing does not warrant an absolute deontological constraint against torturing a terrorist (as even the active/ intended killing of an innocent person may be morally permissible under extreme circumstances). It seems that much of the difference between preemption and inflicting intentional harm against persons to force them to act does not lie in the conditions in which the infringement is justified but in the likelihood that these conditions are satisfied in reality. As we saw, a crucial factor in setting K, the threshold that has to be met to justify the infringement, is the interrogated person’s blame. Setting aside for a moment the problem of incomplete information, one should distinguish between two types of cases. One type of blame may be the interrogated person’s refusal to reveal information that can save lives, e.g., by enabling the authorities to preempt terror attacks.143 Another type of blame refers to the interrogated person’s activities outside the interrogation, such as taking part in terror attacks. The latter context raises the difficult question of how to classify the case of such a blameworthy person, who is tortured to force his collaborators to surrender. In what follows, we put this question aside as well, and focus on the more typical, and potentially justified, case of torture, where the state does attribute blame to the interrogated suspected terrorist at least for his refusal to reveal life-saving information.144 Employing threshold function (4) in the present context, a central element in determining the magnitude of the threshold is whether the likelihood p that the infringement is indeed necessary to generate the desired benefit meets the requirement of moral certainty, that is, whether the value of p exceeds p*. In the present context, the probability p represents the likelihood that all of the following three conditions are met: (1) the tortured person possesses information that can save lives and refuses to reveal it; (2) torture will be effective in obtaining reliable information and preventing the loss of lives; and (3) there is no other, less harmful way to prevent the loss of lives. Arguably, if p < p*, the threshold K should be set at a very high level, and torture can be justified only if its expected net benefit is saving the lives of

143. Sussman, supra note 138, at 18. This case is different from the one in which the state actively inflicts intentional harm on a person to force some other person to avoid the attack or to otherwise surrender or provide information. In the latter case, the state does not attribute blame to the tortured person for refusal to act in a way that can save lives, and the threshold K is extremely high. 144. Another difficult question is how to classify the case in which the interrogated person is not a suspect, who obtained the information as a collaborator in the terror attack, but an innocent person, who gathered it incidentally.


hundreds or even thousands. Only when p ≥ p* would the threshold K be set at a lower level. However, it is highly unlikely that the condition p ≥ p* is ever met in real-life scenarios. It is in this respect that torture may be different from preemptive targeted killing. To justify targeted killing the relevant likelihood is that the target will be involved in the terror attack if not preempted, and that there is no other effective way to prevent the terrorist attack. However, the efficacy of the preemption is typically self-evident. In contrast, the efficacy of torture is typically much more dubious. To justify torture, one must establish that executing the terror attack and killing people does not require the detainee’s involvement (thus preemption through detention is ineffective), that the likelihood of attack is very high and that no alternative preventive means are available, that getting the information from the detainee will enable the authorities to prevent the attack, and that torture will induce the detainee to reveal the information.145 It is notoriously difficult to verify all of these conditions. Only rarely would the probability that all these conditions are met be sufficiently high to justify setting the threshold K at a less than extremely high level.146 This is not to say, however, that circumstances in which torture could be morally permissible, or even required, are unimaginable.147

145. For a discussion on the efficacy of torture in obtaining reliable information see, e.g., Commission of Inquiry into the Methods of Interrogation of the General Security Services regarding Hostile Terrorist Activity, 23 Isr. L. Rev. 146 (1989); Jeannine Bell, Behind this Mortal Bone: The (In)Effectiveness of Torture, 83 Indiana L.J. 339 (2008). 146. See, e.g., Davis, supra note 138, at 170; Luban, supra note 139, at 1440–45; Colb, supra note 139 (the uncertainty that inevitably clouds real-world interrogation warrants the complete ban of torture); James B. White, Law, Economics, and Torture, in Law and Democracy in the Empire of Force 265 (H.J. Powell, & J.B. White eds., 2009); Yuval Ginbar, Why Not Torture Terrorists? Moral, Practical and Legal Aspects of the “Ticking Bomb” Justification for Torture (2008). 147. A possible example is the German kidnapping case, in which a person accused of kidnapping a child was arrested while picking up the ransom, refused to reveal information concerning the victim’s location, and there was a real risk to the child’s life. The European Court of Human Rights decided that threatening to use force against the kidnapper if he would not reveal the relevant information subjected him to “inhuman treatment,” prohibited by Article 3 of the European Convention on Human Rights. 22978/05 Gäfgen v. Germany (June 30, 2008), §§ 60–70. See also Kai Ambos, May a State Torture Suspects to Save the Life of Innocents?, 6 J. Int’l Crim. Just. 261, 262-3 (2008). For other examples see Seumas Miller, Torture, Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/entries/torture/ (last revised Jan. 3, 2009). See also Moore, supra note 131; Alan Dershowitz, Tortured Reasoning, in Torture: A Collection 257 (Sanford Levinson ed., 2004).

165

166


Finally, to justify an act of torture, it is not enough that the practice in general is justifiable. Suppose that a general policy of torturing suspected terrorists under certain circumstances, which results in torturing 10 people a year, reduces the scope of terrorist activity and saves 1,000 potential victims from death or severe injuries annually. Assume further that there is no other, less harmful, way to save these potential victims, and that the ratio of 1:100 is large enough to justify the policy. This information, in and of itself, is insufficient to justify a specific instance of torture that complies with the criteria set by the general policy. Presumably, the aim of the infringement is to achieve some concrete information, which differs from one instance to another. Even if the practice of torture can be justified, as it provides, on average, information that is valuable enough, one must also establish that in each and every case the expected outcome of the infringement meets the threshold.148

E. Constrained Economic Analysis of Unintended Harm •

1. General Anti-terrorist activities often involve risking the life, body, and property of innocent people. As discussed above, even if actively/intentionally harming innocent people is not absolutely prohibited, its permissibility is subject to an extremely high threshold. However, the use of anti-terrorist measures may also harm innocent persons unintentionally, that is, as a mere side effect. In these cases, the harmful effect does not provide the state with a reason for the action or an explanation thereto.149 Under what circumstances is the government allowed to endanger the lives of, or even knowingly kill, innocent civilians as a side effect of an act aimed at protecting the lives of other civilians or soldiers? Answering this question through standard CBA requires one to compare the expected direct and indirect consequences of an action against terrorists with the consequences of inaction, including their expected effects on the

148. See supra p. 143. 149. See supra pp. 41–46, 63–70.


motivation of future terrorists and on demoralization.150 Standard economic analysis does not differentiate between active infliction of harm by the government on a particular group of innocent people and the harm suffered by an as of yet unspecified group of innocent people due to the terrorist activity and the government’s inaction. This view conflicts with commonsense morality, deontological ethics, and prevailing legal norms. Once again, we will use the 2006 judgment of the German Constitutional Court, in which it struck down a law authorizing the air force to shoot down aircrafts that are wielded as deadly weapons, as our point of departure. The court reasoned that killing innocent people for the sake of saving others violates basic human rights to life and human dignity.151 The court noted that an assessment that the passengers are doomed anyway (because the terrorists plan to crash the aircraft into their target) would not change the outcome because human life and dignity enjoy the same constitutional protection regardless of the duration of an individual human being’s physical existence.152 We submit that both the absolutist’s unqualified opposition to authorizing state officials to shoot down the airplane (endorsed by the German court) and a pure consequentialist CBA of such action are unsatisfactory. On the one hand, state officials should be permitted—and even required—to shoot down an airplane carrying a few innocent passengers to prevent it from crashing into a dam or a nuclear plant and killing tens of thousands of people (and causing secondary environmental damage and other losses).153 Contrary to the court’s depiction, such action does not treat the passengers as a means to save the lives of others. Killing the innocent passengers under these circumstances is merely a side effect of shooting down the airplane to save the lives of the intended victims. While such killing infringes a constraint, this infringement is sometimes justified. On the other hand, it seems immoral to shoot down an airplane carrying 400 passengers to save the lives of 401 other innocent people, at least if it is

150. See, e.g., Sandler & Enders, supra note 8, at 309–11. 151. Supra note 42. The court added that the law is invalid because in the case of an aircraft hijacking, the factual situation and the extent of the threat are typically not sufficiently clear to justify the taking of innocent lives. Id. at § 125–29. 152. Id. at § 130. 153. See, e.g., David Ormerod, Smith & Hogan Criminal Law 322 (11th ed. 2005); Michael Bohlander, Of Shipwrecked Sailors, Unborn Children, Conjoined Twins and Hijacked Airplanes—Taking Human Life and the Defence of Necessity, 70 J. Crim. L. 147, 158 (2006).

167

168


expected that otherwise the passengers have a reasonable chance of survival. As indicated, an act that unintentionally inflicts harm infringes a deontological constraint, but this constraint is different from the one against inflicting intended harm, and consequently the applicable threshold is less stringent, both quantitatively and qualitatively. Qualitatively, in order for us to justify an act, it must serve some legitimate aim. In addition, there must be a reasonable proportion between the act’s expected benefit and its cost.154 One should ask whether the side effect is a proportional or reasonable price to pay for achieving the aim. In other words, an active infliction of foreseen (though unwanted) harm on some people is permissible only if this harm is inflicted to avoid sufficiently greater harm to others. For instance, killing noncombatants as a mere side effect is justifiable only if the noncombatant deaths are proportional to the intended desirable consequences of bombing a terrorists’ camp, unrelated to those deaths. In the words of the First Additional Protocol of the Geneva Convention, an attack is prohibited if it is “expected to cause incidental loss of civilian life . . . which would be excessive in relation to the concrete and direct military advantage anticipated.”155 As in the case of intentionally inflicting harm, here too one may require that the type of harm that the activity aims to preempt is not lexically inferior to the harm that it inflicts, even as a mere side effect. In addition, to justify the action, one must show that its adverse side effect is unavoidable. Finally, one may also take into account the probability that the harmful side effect will materialize. When the probability that the action will inflict harm is sufficiently low, it seems that the action would infringe a constraint only if the harm occurs; whereas when this probability is higher, the mere exposure to risk plausibly infringes a constraint.156 Related issues pertain to uncertainties over the identity of the persons who might be harmed as a result of the action. Whereas from a consequentialist viewpoint, an act that would harm one unknown person out of a group is identical to an act that would inflict the same harm on a specific person (since the result in both

154. See, e.g., Alison McIntyre, Doing Away with Double Effect, 111 Ethics 219, 221–23 (2001); Cryer & Simester, supra note 40, at 34. 155. First Protocol to the Geneva Conventions, supra note 71, Article 51(5)(b). See also, e.g., Judith Gail Gardam, Proportionality and Force in International Law, 87 Am. J. Int’l. L. 391, 406–10 (1993); Jean-Marie Henckaerts & Louise Doswald-Beck, Customary International Law, Vol. 1: Rules 49 (2005). 156. Cf. David McCarthy, Rights, Explanation, and Risks, 107 Ethics 205, 208–12 (1997). See also supra p. 95.


cases is harm to one person),157 deontology may well distinguish between the two acts and find the latter more objectionable.158

2. Constructing the Threshold Function The following analysis takes the hijacked plane scenario as paradigmatic. Assume, first, that absent the preventive measure, the passengers are expected to survive and that the government action will certainly kill the passengers and prevent the harm to others.159 Under these assumptions, one may use the following threshold function: (5) T = ( px – qy) – (Ku + qyKu′) where x is the expected number of people who will be killed or severely wounded as a result of the terrorist attack; y is the expected number of innocent people killed or severely wounded as a result of the preemptive action; p is the probability that the terrorists will successfully carry out their plan if no preemptive action is taken; q is the probability that the preemptive action will kill the passengers; and Ku and Ku′ are an additive and a multiplier threshold, respectively, reflecting the constraint against unintended harm. This function assumes that in determining the permissibility of the preemptive act, only lost lives and severe physical harms to innocent people count, thus excluding the well-being of the terrorists. This assumption is relaxed in the following section. The last factor in this function, qyKu′, represents the view that the threshold should be positively correlated with the magnitude of the risk q that the activity will result in killing persons. The incorporation of the additive threshold Ku, in addition to the multiplier one, may reflect the view that the action infringes a constraint even if no harm materializes. For instance, merely exposing persons to the risk of unintended harm may infringe a constraint. As indicated, one may wish to divide the

157. See, e.g., T.C. Schelling, The Life You Save May Be Your Own, in Problems in Public Expenditure Analysis 127 (Samuel B. Chase, Jr. ed., 1968). 158. See, e.g., Charles Fried, The Value of Life, 82 Harv. L. Rev. 1415 (1969); Sanford H. Kadish, Respect for Life and Regard for Rights in the Criminal Law, 64 Cal. L. Rev. 871, 893–94 (1976); Frank Ackerman & Lisa Heinzerling, Priceless: On Knowing the Price of Everything and the Value of Nothing (2003) (arguing that the public is more likely to object to bear a risk which is faced by an identifiable community). 159. See Bohlander, supra note 153, at 158 (describing such possible scenarios).

169

170


spectrum of risks into several categories and set different thresholds for each, such that when the probability q that the action will indeed inflict harm, is sufficiently low, Ku = 0. Function (5) disregards the nationality and other characteristics of the civilians involved. It excludes lesser bodily harms to innocent people, damage to property, and monetary losses. It further excludes long-term effects of the preemptive action, such as its deterrent effect on future terrorist attacks. All, or most, of these considerations may, however, be considered when choosing among the deontologically permissible alternatives (including the alternative of doing nothing). The threshold function need not exclude all of the above factors. A more complex function may differentiate between different innocent people according to their affinity to the terrorists. It can also consider any type of physical harm, where each harm is weighted according to its severity. Finally, instead of making the choice between various permissible actions using standard CBA or lexical priorities between competing values, one may use a more sophisticated function to assess the marginal net benefit of taking one infringing act rather than another, along the lines discussed in the previous section.

3. Killing Persons Who are Doomed It may be that the passengers and crew are likely to die irrespective of the preemptive action because the terrorists plan to crash the airplane into their target. Contrary to the German court’s ruling, the prevailing view among English-speaking philosophers and jurists is that killing a person who is fated to die anyway, as a side effect of saving others, is permissible or at least less objectionable than killing a person who is not similarly doomed.160

160. See, e.g., Baruch A. Brody, Abortion and the Sanctity of Human Life 12–25 (1975); Frances M. Kamm, Morality, Mortality, Vol. II: Rights, Duties, and Status 54 (1996) (“killing someone who is under a threat of death may be better than killing someone who is not, when the person is certainly doomed anyway (no one can help him)”); id. at 248–49 (discussing “the doomed victim”); John P. Reeder, Jr., Killing and Saving 58-63, 164–68 (1996); Moore, supra note 131, at 302–04; Michael Bohlander, In Extremis—Hijacked Airplanes, “Collateral Damage” and the Limits of Criminal Law, Crim. L. Rev. 579, 580 (2006); Bohlander, supra note 153, at 158. Cf. James A. Montmarquet, On Doing Good: The Right and the Wrong Way, 79 J. Phil. 439, 445–54 (1982) (discussing the permissibility of killing people who are already threatened, to save other threatened people).


In this case, the threshold function may be modified to capture the judgment that the victims are doomed by adding variable r, the probability that the passengers will survive absent the preventive act, as follows: (6) T = ( px – rqy) – (Ku + rqyKu ′) In the extreme case in which it is certain that all passengers will die anyway (r = 0), a consequentialist would hold that (all else being equal) shooting down the airplane is justified even for the sake of saving one person on the ground. In contrast, according to the threshold function (6), the preemptive action will be permissible only if px – Ku > 0. This function requires that the net benefit of shooting down the airplane will exceed a certain threshold Ku, to overcome the deontological constraint against actively/ intentionally killing innocent people, as opposed to letting them die. In addition to the probability that the harmed people will die anyway, one may wish to consider such factors as qualitative differences between the inflicted death and the otherwise expected one (such as violent vs. peaceful death) and the time period by which death is accelerated.161 While these factors do not loom large in the context of a terrorist attack, they may be significant in other contexts.162

4. Victims’ Moral Responsibility and Nationality In setting the pertinent thresholds, one may take into account the moral responsibility of the civilians who are harmed as a side effect of the antiterrorist measure. In one type of cases, while the persons who might be harmed are themselves, in a way, victims of the terror attack, they acted negligently by ignoring warnings to avoid certain areas. It seems that ordinarily such faults should not substantially affect the size of the threshold. A more

161. See, e.g., Freid, at 1433–37 (discussing the relevance of “how one dies”). 162. For example, in the famous case of the conjoined twins, the British Court of Appeal authorized the separation of twins—a procedure which was known to cause the death of one of them—when otherwise both were expected to die within three to six months. In Re A (Children) (Conjoined Twins: Surgical Separation), [2001] Fam. 147 (C.A. Civ.) (U.K.). Despite the fact that the inflicted death shortened the lifespan of one of the twins by several months, a central ground of the court’s decision was that she was doomed to die anyway. Id. at 196–97. See also Sabine Michalowski, Sanctity of Life—Are Some Lives More Sacred than Others?, 22 Legal Stud. 377 (2002); Bohlander, supra note 153, at 155–57; Ormerod, supra note 153, at 321–22.

171

172


difficult issue is whether the government should distinguish between unintentionally harming enemy nationals and harming its own nationals. A typical scenario is the one in which terrorists hide among a civil population or launch their attacks from within a populated area. A preliminary issue in the context of the fight against terrorism is that it is not always clear whether and to what extent a certain population is affiliated with the terrorists. The fact that innocent civilians belong to the terrorists’ ethnic group does not, in and of itself, justify treating them as the “enemy.” Even when there is a close association between the terrorists and the civil population, it is often unclear whether the civilians cooperate with the terrorists willfully or are coerced to cooperate with them.163 Setting these difficulties aside, consider the case in which the persons who might be unintentionally harmed by the anti-terrorist measure cannot be blamed for willfully collaborating with the terrorists. They are, however, members of the terrorists’ national or ethnic group—the group whose interests (as perceived by the terrorists) the terrorists strive to further. There may be significant second-order considerations for imposing particular restrictions on harming enemy civilians, even unintentionally. Armies, in particular ones that have airborne and long-range artillery capabilities, often regard their own safety and the success of their military mission as paramount, while underestimating the risk to enemy civilians.164 Only when the expected victims of the military action are state nationals can one expect that their interests will be appropriately taken into account by decision-makers.165 As far as deontological considerations go, some argue that the threshold that has to be met to justify unintended harm to state nationals (as in the hijacked plane scenario) should be substantially higher than the one applying to comparable harm of enemy civilians (as in many instances of targeted killings). According to this view, while the state must subject enemy civilians to as little harm as possible, it is not required to take such precautions

163. A related question, which we addressed above, is what types of willful activities should be considered as sufficient to deny a person from the protection afforded to “innocent persons” from inflicting intended harm. See supra pp. 157–60. 164. Benvenisti, supra note 72, at 94–95. 165. In addition, state nationals are typically entitled to monetary compensation for their losses that result from a state military action, whereas aliens are often denied such compensations. See Yaël Ronen, Avoid or Compensate? Liability for Incidental Injury to Civilians Inflicted During Armed Conflict, 42 Vand. J. Trans. L. (forthcoming).


that would risk the lives of its own nationals, including its soldiers.166 This position does not attribute guilt to the enemy civilians, in violation of the principle of individuality, which rejects collective responsibility and states that individuals are responsible only for their own actions.167 Neither does this position assign different values to people’s lives according to their nationality. Rather, it rests on a distinction between the state’s duty to ensure and protect the rights of its own nationals and its lesser duty to respect enemy civilians. Such respect mandates that the state not target the latter and should strive to reduce the harms it inflicts on them. According to this stance, however, a state’s duty toward its own people takes precedence over its duty toward enemy civilians.168 Consequently, it is argued that imposing risks on combatants is justified only to the extent that the risk is necessary to secure the interests of the state or the combatants but not to protect the lives of enemy civilians.169 This conclusion goes too far. The sound premise that a state is under a duty to further the interests of its people and that it bears no such duty (or bears drastically lesser duties) toward aliens, let alone enemy civilians, does not yield the conclusion that a state is not subject to constraints when dealing with enemy civilians. While the premise accurately depicts the scope of the state’s deontological options or lack thereof, the alleged conclusion refers to deontological constraints. Assuming—as we do at this stage—that there is a constraint against unintentionally killing persons, this constraint holds regardless of the lack of special duties of the sort characterizing a state’s relationships toward its people.170 There is no reason to assume that the duty to protect the lives of state nationals, including its soldiers, is lexically superior to the constraint against actively inflicting harm on aliens. To justify unintended harming of enemy civilians, the expected net benefit of the act (primarily in terms of lives saved) must surpass a certain threshold.

166. Benvenisti, supra note 72, at 87–90. 167. For a view that the ideas of collective action, collective intention, and collective guilt all have a sound grounding in Western culture see George P. Fletcher, The Storrs Lectures: Liberals and Romantics at War: The Problem of Collective Guilt, 111 Yale L.J. 1499 (2002). 168. Benvenisti, supra note 72, at 89. 169. Id. at 89. According to Benvenisti, “this does not preclude the moral duty of combatants to consider taking some risks to reduce the harm to enemy civilians, but not a duty to actually risk themselves,” since “imposing risks on [combatants] to protect enemy civilians means using them as mere tools for the benefit of others.” Id. at 90. 170. See supra pp. 60–63.

173

174


The state may thus be required to risk the lives of its combatants, whenever doing so is necessary to protect its own interests without killing, even unintentionally, enemy civilians.171 At the same time, unlike the consequentialist who assesses costs and benefits impartially, a deontologist decision-maker can straightforwardly set a lower threshold when the foreseen, unintended victims of an anti-terrorist action are enemy civilians, rather than its own people, based on agent-relative considerations.172

F. Measures Involving Both Intended and Unintended Harm •

Anti-terror measures often inflict both intended and unintended harm. Paradigmatically, targeting a terrorist may also have unavoidable, harmful side effects. In the case of targeted killings, these side effects may include deaths, injuries, and psychological traumas suffered by innocent people. In such cases, an act infringes both the constraint against actively/intentionally harming people and the constraint against actively, though unintentionally, harming (other) people. It should thus be subject to a two-tier assessment. First, one should evaluate whether the infringement of the constraint against intentional harm is permissible given the relevant threshold. At this stage, the adverse side effect is taken into account in calculating the activity’s net benefit as one of its costs but not as an infringement of a constraint. The threshold function employed at this first stage should embody the distinction between intended harm and additional adverse side effects. For instance, we may add variables to denote the distinction between the intended harm yi (assuming it is certain), and the prospect of unintended harms yu, that will occur in probability qu: (7) T = px – (quyu + yi) – (Ki + yiKi ′)

171. Cf. the Israeli Supreme Court decision in Public Comm. against Torture in Israel v. Government of Israel, supra note 43, at § 46 (“the state’s duty to protect the lives of its soldiers and civilians must be balanced against its duty to protect the lives of innocent civilians harmed during attacks on terrorists”). See also Walzer, supra note 98, at 151 (“soldiers are supposed to accept (some) risks in order to save [enemy] civilian lives”). 172. See supra pp. 41–48. This may not be true when enemy civilians become subject to the state’s effective control, as prisoners of war, or “protected persons” in occupied territories. See, e.g., Benvenisti, supra note 72, at 88.


where Ki, and Ki′ are an additive and a multiplier threshold, respectively, representing the judgment that the action infringes the constraint against inflicting intended harm. If the infliction of the intended harm is permissible, the action should then be subject to an additional review, which addresses the infringement of the constraint against inflicting unintended harm: (8) T = px – (quyu + yi) – (Ku + quyuKu′) The parameters Ku, and Ku′ reflect the threshold for the constraint against exposing persons to the risk of unintended harm. Note that this second stage does not necessarily collapse into the first. The action may pass the first scrutiny but not the second when the threshold for infringing the constraint against inflicting intended harm is lower than that of infringing the other constraint (Ki > Ku). This possibility is a plausible one since typically, those who suffer the unintended harm are innocent persons, while the target of the intended harm may be blameworthy, and the action is required to preempt an imminent terror attack. In addition, even if Ki > Ku, if the expected unintended harm is substantially greater that the expected intended harm (qu yu > yi), the action may again be permissible as far as its intended consequences are concerned, but impermissible due to the infringement of the constraint against inflicting unintended harm.

•

G. Conclusion

Standard CBA of the fight against terrorism rests on the notion that the end of maximizing aggregate human welfare justifies all means. Two central characteristics of this fight lend credence to this notion: the potentially catastrophic consequences of failing to win the struggle and the widespread violations of moral constraints and legal norms by terrorists. Second-order considerations that often support respecting moral constraints and obeying the restraints imposed by the laws of war, such as the expectation (and fear) of reciprocity, seem irrelevant in the context of the fight against terrorism. Standard CBA views inflicting harm on innocent persons as a “cost” that the state should minimize. However, it does not differentiate between harms that the state inflicts by implementing various anti-terrorist measures and those that (other) innocent persons will suffer as a result of unthwarted terrorist attacks. Actions and inactions, intended harms and unintended ones, are all accorded similar weight in the moral and legal calculus.

175

176


This chapter sought to demonstrate that an economic approach, which aims at maximizing welfare, can and should incorporate deontological constraints. The incorporation of constraints makes the consideration of all relevant factors much more nuanced, as it requires one to take into account various pertinent distinctions, such as between action and inaction, intended harm and unintended harm, the purpose of inflicting harm, and so forth. At the same time, the integration of constraints does not entail ignoring social welfare, including the state’s duty to protect the lives of its citizens and promote their welfare. Constrained CBA operationalizes the judgment that when enough good or bad consequences are at stake, a democracy is justified in infringing the relevant constraints. It is in this sense that a constrained economic analysis of the fight against terrorism reflects the deepest commitment to the basic values of a democratic society, which include both protecting the lives of its citizens and respecting deontological constraints.

•

seven

Freedom of Speech

•

A. General

freedom of speech is considered a basic human right, often enjoying a

special status. Philosophers, jurists, and economists nevertheless constantly debate its justifications and scope. The literature on freedom of speech is abundant,1 and the legal doctrine, particularly in the United States, is rich and complex.2 This chapter neither purports to resolve the numerous disputes revolving around the regulation of speech nor to fully reflect their complexity. In particular, we shall mostly sidestep the important institutional aspects of speech regulation, including the appropriate scope of judicial review of legislative and administrative discretion. We shall similarly avoid discussing the sometimes crucial choice between various sanctions and remedies (criminal, administrative, or civil) and between prior restraints of speech and after-the-fact sanctions or remedies. A relatively abstract discussion, coupled with specific illustrations, should suffice to demonstrate how a deontologically constrained CBA of freedom of speech may contribute to the normative and policy analysis of this important issue. As do other chapters of this book, this one strives to establish that moderate deontology best accounts for the current doctrine (with American law as our central

1. For useful, general monographs and collections of essays, see, e.g. Eric Barendt, Freedom of Speech (2d ed. 2005); Thomas I. Emerson, The System of Freedom of Expression (1971); Frederick Schauer, Free Speech: A Philosophical Enquiry (1982); Freedom of Speech, Vol. 1: Foundations; Vol. 2: Doctrine (Larry Alexander ed., 2000). 2. For a detailed analysis of American law, see, e.g., Geoffrey R. Stone et al., The First Amendment (2d ed. 2003); Laurence H. Tribe, American Constitutional Law 785– 1061 (2d ed. 1988). For comparative accounts, see, e.g., Barendt, supra note 1; Fredrick Schauer, Freedom of Expression Adjudication in Europe and the United States: A Case Study in Comparative Constitutional Architecture, in European and US Constitutionalism 49 (Georg Nolte ed., 2005); Michel Rosenfeld, Hate Speech in Constitutional Jurisprudence: A Comparative Analysis, 24 Cardozo L. Rev. 1523 (2003).

177

178


example) and that a constrained CBA fruitfully highlights the pertinent normative factors and their interrelations. Section B of this chapter provides a brief, comparative overview of the legal doctrine, focusing on American law. Section C critically discusses the most sophisticated economic analysis of free speech to date, offered by Richard Posner. Section D argues that while freedom of expression is expected to promote good outcomes, a satisfactory justification for this freedom must include a deontological element. Section E then demonstrates how the integration of a deontological constraint against suppression of free speech with CBA improves the analysis of this subject both descriptively and normatively.

•

B. Doctrinal Background

Freedom of speech is recognized as a fundamental human right in both domestic bills of rights and international human rights conventions. Yet, even where the prohibition on curtailing free speech is expressed in absolute terms (as in the First Amendment of the United States Constitution), speech is not accorded unqualified protection. Legal systems use two central techniques to delimit the constitutional protection of free speech: exclusion of some types of speech from the ambit of protection, and incorporation of a limitation clause, authorizing decision-makers to curtail free speech when necessary to protect compelling interests. The two techniques are often employed in tandem. Accordingly, there is broad consensus among legal systems that such verbal communications as involved in criminal conspiracy, blackmail, the warning of a criminal to escape from the police, and initiation of a restraint of trade are not constitutionally protected.3 Obscenity, sexual harassment, and the provocation of racial hatred are similarly unprotected in some systems. The lines drawn between protected and unprotected speech vary from one legal system to another and are often contested.

3. See Frederick Schauer, Categories and the First Amendment: A Play in Three Acts, 34 Vand. L. Rev. 265, 267–82 (1981); Kent Greenawalt, Speech, Crime, and the Uses of Language 57–71 (1989) [hereinafter Greenawalt, Speech] (distinguishing between expressions deserving of free speech protection and “situation-altering utterances”); Kent Greenawalt, “Clear and Present Danger” and Criminal Speech, in Eternally Vigilant: Free Speech in the Modern Era 96 (Lee C. Bollinger & Geoffrey R. Stone eds., 2002) [hereinafter Greenawalt, Criminal Speech].

freedom of speech

A complementary question of delineation pertains to expressive, or allegedly expressive, activities that do not involve verbal or linguistic utterances. Lawmakers have to decide what human activities are to be considered “speech” or “expression.” While it is quite obvious that painting, photography, and ballet are all forms of expressions despite the absence of verbal communication, it is more difficult to categorize such actions as burning an object (a flag, a draft card) or appearing naked in public, which may or may not have some expressive purpose. Potential criteria for classifying such actions as protected speech are the actor’s intent, the way the action is perceived by its audience, and the government’s reasons for regulating it.4 At least under some circumstances, such symbolic acts do merit protection.5 The other, often complementary technique of a limitation clause is employed, for example, by the Canadian Charter of Rights and Freedoms. Section 2 of the charter provides that everyone has the freedom of “opinion and expression.” Yet, under Section 1, this and other fundamental freedoms are subject “to such reasonable limits prescribed by law as can be demonstrably justified in a free and democratic society.”6 Competing values that may justify setting limits to free speech include national security, privacy, and prevention of imminent violence. Different balancing formulae are often set for different categories of speech, such as political, commercial, and artistic.7 In the same vein, sexually explicit material and commercial advertising enjoy lesser protection than other categories of speech under American law.8

4. Joseph Raz, Free Expression and Personal Identification, 11 Oxford J. Legal. Stud. 303, 303 (1991) (the actor’s intent); Cass R. Sunstein, Words, Conduct, Caste, 60 U. Chi. L. Rev. 795 (1993) (the actor’s intent and the audience’s understanding); Larry Alexander, Free Speech and Speaker’s Intent, 12 Const. Comm. 21 (1995) (government’s reasons for regulation). See also the reference to “Track Two” analysis below, and infra pp. 213–18. 5. See, e.g., R.A.V. v. City of St. Paul, 505 U.S. 377 (1992) (ruling that a statute proscribing the placing of a burning cross which is known to arouse anger, alarm, or resentment in others on the basis of race or similar bases is content-based and thus unconstitutional under the First Amendment); Dean Alfange, Jr., Free Speech and Symbolic Conduct: The Draft-Card Burning Case, 1968 Sup. Ct. Rev. 1 (criticizing United States v. O’Brien, 391 U.S. 367 (1968)). 6. 2 Peter Hogg, Constitutional Law of Canada 36–10, 38-2–38-7 (5th ed. 2006). See also Art. 5 of the Basic Law for the Federal Republic of Germany, which provides that every person shall have the right freely to express herself in speech, writing and pictures, and to inform herself from generally accessible sources; yet that these rights may be limited to the extent necessary to protect the young and the right to personal honor. 7. On categorization of speech types, see generally Schauer, supra note 3, at 282–96; infra pp. 218–21. 8. See, e.g., Jeffrey M. Shaman, The Theory of Low-Value Speech, 48 S.M.U. L. Rev. 297 (1995). See also infra pp. 218–21.

179

180


The very distinction between excluding some types of expressions and activities from the ambit of free speech protection and using limitation clauses or some balancing techniques is not always clear. This vagueness is reflected in the way the phrases “protected speech” and “unprotected speech” are used in the United States. The latter phrase encompasses both “expressions” that lie outside of free speech doctrine (such as criminal conspiracy or a political murder) and conduct that lies at the heart of First Amendment law, when its regulation is considered constitutional (such as accounts of troop movements).9 Put differently, the term “unprotected speech” is used to denote both cases in which a regulation does not infringe on a person’s freedom of speech and cases in which it infringes on this freedom justifiably. To get a better view of the vast landscape of free speech doctrine, it is useful to further subdivide it. The following taxonomy, though constantly contested, is nevertheless methodologically valuable. One distinction is between private and public suppression of speech. Examples of private suppressions include an employer who forbids the expression of certain political views in the workplace or a shopping center that forbids picketing on its premises. Public suppression includes prohibitions imposed by the government, including local government. Protecting free speech against private suppression often requires the state to take positive measures. A related distinction is thus between the government refraining from curtailing speech and taking positive steps to protect this freedom. Since our discussion will focus on public suppression of speech, we shall neither discuss private curtailment nor inquire whether and to what extent the government should ensure equal access to expressive media. Neither will we discuss the extent to which the public should have access to information held by the government that may be conducive for an open public debate. Another basic distinction is between content-based and content-neutral restrictions of speech. Content-based restrictions are directed at the communicative impact of speech. They restrict certain expressions because their content is thought to be harmful or because it is believed that the very discussion of certain issues in public is undesirable. Note that although restrictions of the latter type are apparently viewpoint-neutral, by foreclosing open debate on certain issues, they tend to shelter existing policies and practices.10

9. See, e.g., Near v. Minnesota, 283 U.S. 697, 716 (1931). Cf. New York Times v. U.S., 403 U.S. 713, 726–27 (1971). 10. Schauer, supra note 3, at 284–85.

freedom of speech

Such restrictions are nevertheless deemed acceptable in some contexts, such as in public schools and in the army. Subject to these special contexts, content-based restrictions of speech are strictly scrutinized. Under American law, they are deemed constitutional only if they serve an exceptionally important purpose and are the minimal means necessary to attain this purpose.11 Content-neutral restrictions typically regulate the place, timing, or manner of expression due to its noncommunicative impacts. Such impacts may include excessive noise in residential neighborhoods, litter, and traffic congestion. These restrictions are subject to judicial review but are likely to be upheld if they are at all reasonable.12 Following Laurence Tribe, content-based restrictions and their judicial review are commonly labeled “Track One,” while content-neutral regulations of the time, place, or manner of expression are known as “Track Two.”13 Another useful distinction applies to content-based restrictions and focuses on how the speech is expected to cause harm. Roughly, harm may be caused in one step or two. Revelation of secrets or intimate information, defamation, and verbal sexual harassment are all instances of one-step harm. In such cases, the very expression causes the harm, even if it does not affect anybody’s behavior. Often, however, the ultimate harm is brought about in two steps. The speaker incites, persuades, or provides useful information to people who might then do harmful things. Calling for the overthrow of the government, instigating violence, or otherwise advocating unlawful actions are paradigmatic.14 Admittedly, the borderline between one- and two-steps harms is sometimes blurred, and even when it is clear, the same speech may cause both types of harms. Focusing on content-based regulation, which lies at the core of freedom of speech doctrine, different legal systems take very different positions. These differences may be illuminated by briefly describing two central types of such regulations: prohibitions against incitement to violence or other unlawful conduct and the suppression of hate speech.

11. See, e.g., Perry Education Ass’n v. Perry Local Educators’ Ass’n, 460 U.S. 37, 45 (1983) (“For the State to enforce a content-based exclusion it must show that its regulation is necessary to serve a compelling state interest and that it is narrowly drawn to achieve that end”); United States v. Eichman, 496 U.S. 310, 318 (1990); Tribe, supra note 2, at 798–99, 832–41. 12. Tribe, supra note 2, at 791, 977–1010. 13. Id. at 789–804. 14. Larry Alexander, Freedom of Speech, in 2 Encyclopedia of Applied Ethics 299, 303 (1998).

181

182


The 1966 United Nations Covenant on Civil and Political Rights not only allows states to proscribe incitement to violence but actually requires them to do so. Article 19(2) of the Convention provides that “everyone shall have the right to freedom of expression,” and Article 19(3) sets forth a general limitation clause (“the exercise of the right . . . [may] be subject to certain restrictions, but these shall only be such as are provided by law and are necessary”). However, Article 20(1) provides that “[a]ny propaganda for war shall be prohibited by law,” and Article 20(2) adds that “[a]ny advocacy of national, racial or religious hatred that constitutes incitement to discrimination, hostility or violence shall be prohibited by law.” Other international conventions carry the same message.15 According to this approach, one need not assess the specific consequences of suppressing such expressions in any particular circumstances. Many national legal systems follow in the footsteps of the international conventions. The Canadian Supreme Court held constitutional a statute criminalizing the willful promotion of hatred against racial, religious, or ethnic groups.16 Similarly, the German courts have validated the repression of hate speech against ethnic and other groups.17 In contrast, under American constitutional law, determining the constitutionality of silencing incitements to violence or hatred requires a case-bycase assessment of the expected consequences of the government’s action or inaction. In the typical category of cases where the harm is expected in two steps—incitement to violence—the central element that determines the legitimacy of the regulation is the risk of harm. According to the old doctrine, as expressed by Justice Holmes in Schenck v. United States (1919), “[t]he question in every case is whether the words are used in such circumstances and are of such a nature as to create a clear and present danger that they will bring about the substantive evils that Congress has a right to prevent.”18 While the test of “clear and present danger” sounds extremely protective of free speech, it was criticized for being employed to legitimize

15. See, e.g., Art. 4 of the International Convention on the Elimination of All forms of Racial Discrimination (1965). On this Article, its legislative history, and its incompatibility with the constitutional protection afforded to freedom of speech in some legal systems, see Natan Lerner, The U.N. Convention on the Elimination of all Forms of Racial Discrimination 43–53 (2d ed. 1980). 16. Regina v. Keegstra, [1990] 3 S.C.R. 697. See generally Hogg, supra note 6, at 43-1–43-12. 17. Rosenfeld, supra note 2, at 1548–54. 18. Schenck v. United States, 249 U.S. 47, 52 (1919).

freedom of speech

far-reaching restrictions of free speech.19 Eventually, in Brandenburg v. Ohio (1969), the Supreme Court formulated a new test that is even more stringently protective of free speech.20 It held that proscribing “advocacy of the use of force or of law violation” is unconstitutional “except where such advocacy is directed to inciting or producing imminent lawless action and is likely to incite or produce such action.”21 This test, which is still in force, results in an almost absolute protection of the right to publicly advocate violence or other violation of the law, even when doing so entails considerable risks to the targets of the speech and to the peaceful coexistence of different parts of society.22 The scope of free speech protection is hence considerably broader in the United States than in other liberal democracies.23 This difference is also reflected in the related issue of offensiveness. Under American law, offensiveness is not a sufficient basis for curtailing speech. As the Supreme Court made clear in Texas v. Johnson (1989), “[i]f there is a bedrock principle underlying the First Amendment, it is that the government may not prohibit the expression of an idea simply because society finds the idea itself offensive or disagreeable.”24 Accordingly, in the famous case of the march of neo-Nazis in Skokie, Illinois, the court held that the high intensity of the offense to the feelings of Holocaust survivors was not sufficient to prohibit the march.25 Narrow exceptions to this rule are recognized in such cases as individual defamation, situations of “captive audience,” and obscenity.26

19. See, e.g., Martin H. Redish, Advocacy of Unlawful Conduct and the First Amendment: In Defense of Clear and Present Danger, 70 Cal. L. Rev. 1159, 1166 (1982). 20. 395 U.S. 444 (1969). 21. Id. at 447. Brandenburg, a Ku Klux Klan leader, made a speech in a KKK gathering, where a large wooden cross was burned. He spoke in a derogatory manner about “niggers” and Jews, and said that if the authorities will continue “to suppress the white, Caucasian race, it’s possible that there might have to be some revengeance taken.” The speech was filmed by a television reporter invited to the meeting and was later broadcasted. Brandenburg was convicted under a statute that prohibited “advocat[ing] . . . the duty, necessity, or propriety of crime, sabotage, violence, or unlawful methods of terrorism as a means of accomplishing industrial or political reform.” The Supreme Court reversed the conviction, holding that the statute was unconstitutional. See also R.A.V., 505 U.S. 377 (described supra note 5). 22. See, e.g., Greenawalt, Crime Speech, supra note 3. 23. See Rosenfeld, supra note 2; Schauer, supra note 2. 24. Texas v. Johnson, 491 U.S. 397, 414 (1989) (holding that flag burning cannot be criminalized). 25. See Collin v. Smith, 578 F.2d 1197 (7th Cir. 1978), cert. denied, 439 U.S. 916 (1978). See also R.A.V., 505 U.S. 377 (described in supra note 5). 26. Tribe, supra note 2, at 849–56, 904–28.

183

184


In contrast, other western democracies consider offensiveness a sufficient basis for curtailing speech, mainly when the expression is directed at minorities. In Canada, this approach is founded on the values of multiculturalism and group-regarding equality, whereas in Germany the limitation on hate speech is grounded in respect for human dignity.27 It should be noted that the actual protection of free speech in any country depends not only on the legal doctrine but also on how the courts interpret and implement it, which in turn reflects the particular social, cultural, and political environment. At times, the actual practice and court rulings are quite at odds with the legislative phrasing and judicial rhetoric.28

C. Cost-Benefit Analysis of Free Speech and Its Critique •

A considerable body of literature analyzes freedom of expression, or the “speech market,” from an economic point of view.29 Some economists, pointing to similarities between the market for goods and the market for ideas, have called for the deregulation of both.30 Others point to the special characteristics of ideas and symbols as public goods, which absent regulation might be underproduced due to their positive externalities.31 At the same time, it is argued that some speech should be limited due to its negative externalities.32 In the same vein, since listeners often find it difficult to evaluate the quality of ideas, regulation may be necessary to avoid adverse selection of ideas.33 Most notably, Richard Posner has proposed a formula for an economic

27. Rosenfeld, supra note 2. 28. See, e.g., supra notes 18, 19 and accompanying text. 29. For an overview of the literature, see Richard A. Posner, Economic Analysis of Law 727–43 (7th ed. 2007); Hugo M. Mialon & Paul H. Rubin, The Economics of the Bill of Rights, 10 Am. L. & Econ. Rev. 1, 6–15 (2008). 30. See, e.g., R.H. Coase, The Market for Goods and the Market for Ideas, 64 Am. Econ. Rev. Papers & Proceedings 384 (1974). 31. See, e.g., Daniel A. Farber, Free Speech without Romance: Public Choice and the First Amendment, 105 Harv. L. Rev. 554 (1991); Eric Rasmusen, The Economics of Desecration: Flag Burning and Related Activities, 27 J. Legal Stud. 245 (1998). 32. Mialon & Rubin, supra note 29, at 6. 33. Albert Breton & Ronald Wintrobe, Freedom of Speech vs. Efficient Regulation in Markets for Ideas, 17 J. Econ. Behav. & Org. 217 (1992).

freedom of speech

evaluation of any regulation banning free speech.34 According to Posner’s formula, a speech should be allowed if and only if: (1) B ≥

pH +O − A (1 + d )n

where B is the total (social, scientific, aesthetic, or other) benefit of the challenged speech, H – the speech’s expected harm; p – the probability that the harm will actually be materialized if the speech is allowed; d – some perperiod (e.g., one year) discount factor for future harms; n – the number of periods between the time that the speech takes place and the harm from it materializes; O (for offensiveness) – the disutility experienced by disinterested people from merely being cognizant of the speech (e.g., from knowing that pornography is being sold or atheism is being propagated); and A – the administrative costs of the regulation. In this formula—which peculiarly sets the necessary conditions for allowing speech, rather than for justifying its repression—B is presumably measured according to people’s preferences and not by any objective assessment of the intrinsic value of speech.35 This benefit is compared to the speech’s harm, weighted by its probability and discounted for its futurity. Just as the formula takes into account any benefit, so it takes into account any harm, whatever its kind, probability, or chronological remoteness. The formula intentionally disregards the government’s motivation in suppressing any particular speech.36 It also deliberately rejects any categorization of types of speech (such as political, scientific, or artistic). Posner separately discusses cases in which B = 0, but applies the same formula to these cases as well.37 B may even be negative, as in the case where fringe presidential candidates

34. Richard A. Posner, Free Speech in an Economic Perspective, 20 Suffolk U. L. Rev. 1 (1986) [hereinafter Posner, Free Speech]; Richard A. Posner, Frontiers of Legal Theory 62 (2001) [hereinafter Posner, Frontiers]. The following analysis focuses on Posner’s 2001 formula. 35. Cf. Rasmusen, supra note 29. 36. Posner, Frontiers, supra note 34, at 70. Posner subsequently conceded that the regulation’s purposes and motivation may be instrumentally important for assessing its consequences. Richard A. Posner, Pragmatism Versus Purposivism in First Amendment Analysis, 54 Stan. L. Rev. 737, 745 (2002). 37. Posner, Frontiers, supra note 34, at 74–75; Posner, Free Speech, supra note 34, at 9–12. Posner rationalizes the dissimilar judicial attitude to different categories of speech on the basis of their different characteristics in terms of supply, demand, and externalities. See Posner, Frontiers, supra note 34, at 85; Posner, Free Speech, supra note 34, at 19–24; infra pp. 218–21.

185

186


participate in a televised debate, thus reducing the time available for frontrunners to present their plans, which are more valuable to the audience.38 Like other economists, Posner does not distinguish between the costs inflicted on people who are directly affected by a speech and the “costs” to disinterested people who merely read or hear about its existence (“a cost is a cost”).39 The offensiveness of any speech is assumed to be immediate and certain.40 The economic analysis of free speech has been criticized for not being truly “economic,” as it cannot use market prices to guide the analysis.41 Critics further point to the inherent limitations of abstract modeling;42 argue that while such values as free speech and public safety are comparable, they are incommensurable;43 and claim that the discontinuity of outcomes of protecting/curtailing free speech undermines the possibility of trade-off between the formula’s variables.44 Posner concedes that employing his formula faces formidable difficulties due to the indeterminacies characterizing the field, yet he believes—and we share his belief—that the formula can serve as a fruitful way of framing and thinking about the issue.45 We also share Posner’s implied assumption that the problems of incommensurability and discontinuity do not render CBA of constitutional issues impossible or fruitless.46 At the same time, we join those who reject Posner’s assertion that—unlike “the moral approach” to freedom of speech, which he portrays as “spongy and arbitrary”—his analysis “skirts

38. Posner, Frontiers, supra note 34, at 68–69. 39. Id. at 69. 40. Id. at 70. 41. Michael Rushton, Economic Analysis of Freedom of Expression, 21 Ga. St. U. L. Rev. 693, 715–19 (2005). 42. Peter J. Hammer, Note, Free Speech and the “Acid Bath”: An Evaluation and Critique of Judge Richard Posner’s Economic Interpretation of the First Amendment, 87 Mich. L. Rev. 499, 508–11, 514–15 (1988). 43. Id. at 512–14. On the distinction between incomparability and incommensurability, see generally supra pp. 108–16. 44. Hammer, supra note 42, at 511–12, 514. 45. Posner, Frontiers, supra note 34, at 68. In an earlier contribution, Posner proposed to address (some of) the difficulties of implementing a CBA of freedom of speech by explicitly including in the formula a variable E (for error), denoting the legal-error costs incurred in trying to distinguish between valuable and invaluable information. See Posner, Free Speech, supra note 34, at 8, 24–29. 46. See supra pp. 108–16.

freedom of speech

contentious moral and ideological issues.”47 Economic analysis is as valueladen and potentially as manipulable as other normative analyses.48 A conspicuous difficulty of Posner’s formula, at least insofar as it strives to interpret and rationalize extant legal norms, is that it requires policy-makers (as well as courts reviewing legislation and administrative policies and actions) to assess the expected benefit of expressions B.49 Such assessment is incompatible with the basic notion of evaluative neutrality underlying the protection of free speech in liberal democracies. According to this notion, regulators must not silence any speech based on their assessment of its value, merit, truth, or desirability.50 From a deontological point of view, two additional inadequacies of Posner’s formula are the absence of any constraint against intentionally/ actively suppressing free speech and its nondifferential consideration of any cost inflicted by any speech. Posner’s analysis does not attribute any priority to people’s freedom of expression and freedom of access to information, and thus sets no constraint on the state’s power to limit them. This disregard for the intrinsic value of dignity, autonomy, and liberty is incompatible with the common perception of freedom of speech as a universal, basic human right (rather than merely a derivative status).51 Furthermore, this position is likely to result in underprotection of free speech.52 For example, whenever a

47. Posner, Frontiers, supra note 34, at 62–63. 48. See Hammer, supra note 42, at 531–32; supra pp. 105–08. Additional critiques leveled against Posner’s analysis are that courts are not institutionally competent to conduct CBA, and that CBA tarnishes First Amendment’s unique symbolic status. See Hammer, supra note 42, at 528–30 and 533–34, respectively. The critique also targeted specific conclusions Posner derived from his general analysis. See id. at 517–28 (challenging the proposition that localities should be afforded greater authority to suppress speech than the federal government and the claim that externality analysis can guide government policies towards the regulation of speech). 49. Cf. Richard Posner, Not a Suicide Pact: The Constitution in a Time of National Emergency 113 (2006). 50. Larry Alexander, Is There a Right of Freedom of Expression? 11–12 (2005) (describing evaluative neutrality as “the core of any conception of freedom of expression”); David Kretzmer, Freedom of Speech and Racism, 8 Cardozo L. Rev. 445, 458 (1987); Miller v. Civil City of South Bend, 904 F.2d 1081, 1098 (7th Cir. 1990) (Posner J., concurring) (cautioning against the dangers of “letting judges play art critic”). See also infra pp. 218–21. 51. Cf. Alexander, supra note 50, at 133–34, 180. 52. Jed Rubenfeld, The First Amendment’s Purpose, 53 Stan. L. Rev. 767, 780–82, 832 (2001) (sharply criticizing the suggestion that speech may be curtailed whenever “its harmful consequences are thought to outweigh its expressive value”). Cf. Posner, supra note 49,

187

188


certain speech has no benefit or a negative benefit (B ≤ 0), then even a small, improbable, and chronologically remote harm would suffice to warrant its suppression, subject only to the administrative costs of such suppression. This is a far cry from currently prevailing moral and legal norms, especially in the United States. According to these norms, advocacy of unlawful conduct may be silenced only if it is likely to produce “imminent lawless action,”53 or only to prevent “clear and present danger” of “substantive evils.”54 The concern that Posner’s formula would result in underprotection of free speech is exacerbated by its unqualified consideration of any cost of any speech, regardless of its type, magnitude, probability, chronological remoteness, and the way in which it is brought about. Inter alia, it gives full weight to the disinterested feelings/preferences of people annoyed by the mere knowledge of the speech.55 At the same time, the formula inconsistently disregards the disinterested feelings/preferences of people annoyed by the mere knowledge of limitations being imposed on free speech. The formula also ignores the very frustration experienced by people who are silenced, beyond the loss of benefit they (and society at large) expect to derive from the speech. Finally, focusing on the interests of the audience, Posner justifies the silencing of some speech (e.g., messages delivered by fringe presidential candidates) to facilitate the delivery of more valuable information (the messages of front-runners).56 To avoid the troubling outcome of underprotection of free speech, Posner resorts to second-order and institutional considerations, particularly the difficulties of estimating the costs and benefits of free speech.57 In line with

at 113 (suggesting that “advocacy of a holy war against the United States, the West generally . . . Western values, and modernity in general has no redeeming social value.” Such viewpoints may thus be considered as “off the agenda” of fruitful debate, and the modest limitation of their expression due to induced self censorship resulting from FBI surveillance is therefore legitimate). 53. Brandenburg, 395 U.S. at 447. 54. Schenck, 249 U.S. at 52. 55. See infra pp. 206–10. 56. Posner, Frontiers, supra note 34, at 68–69. Admittedly, while some of these weaknesses characterize Posner’s formula, they are not inevitable features of CBA of freedom of speech. 57. Posner, Frontiers, supra note 34, at 70. This is even more conspicuous in Posner’s earlier analysis: Posner, Free Speech, supra note 34, at 24–29 (discussing the various sources of errors in estimating the costs and benefits of speech). Cf. supra pp. 24–27.

freedom of speech

his general distrust of the government,58 he approves of restrictions of freedom of speech only if it may be shown “with some degree of confidence” that the benefits of the restrictions exceed their costs.59 Indeed, taking these epistemological and institutional arguments to the extreme, one may conclude that the state should rarely, if ever, regulate speech. However, such an extreme position rests on very questionable assumptions, is likely to lead to underprotection of free speech when government intervention is necessary to actively protect this freedom and to underregulation in other instances, and hardly contributes to clarifying the relevant normative dilemmas. Posner’s analysis is also incompatible with both deontology and commonsense morality in its disregard of the motivation underlying the regulation. Contrary to Posner’s analysis, there seems to be a widespread intuition that there is an important difference between curtailment of free speech as a regrettable side effect of attaining a desirable outcome (such as preventing traffic congestion) and measures aiming at silencing specific views.60 Attaining a better grasp of the pertinent normative factors and legal issues, as well as reaching intuitively more acceptable conclusions, thus require one to modify the standard CBA of free speech, primarily by incorporating a deontological constraint against suppressing free speech.61 The next section establishes that such a constraint is indeed warranted.

•

D. The Constraint Against Suppressing Free Speech

Standard CBA of free speech presupposes that limitations on the power of government to regulate free speech rest on consequentialist justifications. Indeed, many of the theories of free speech focus on its desirable outcomes.

58. See, e.g., Posner, Frontiers, supra note 34, at 75 (rejecting the notion that the government may be trusted in some spheres more than in others, he declares: “The government cannot be trusted, period”). 59. Id. at 71. See also Ronald H. Coase, Advertising and Free Speech, 6 J. Legal Stud. 1 (1977); Richard A. Epstein, Property, Speech, and the Politics of Distrust, 59 U. Chi. L. Rev. 41 (1992). 60. See supra pp. 180–81; infra pp. 213–18. 61. The incorporation of such a constraint would not overcome other difficulties inherent in standard economic analysis, such as the determination of the goodness of outcomes according to people’s preferences. In the present context, determining the political, scientific or aesthetic value of an expression (B) by aggregating people’s preferences seems inappropriate even from a consequentialist point of view. See supra pp. 11–18, 30–32; infra pp. 218–21.

189

190


Kent Greenawalt has identified six types of goods promoted by free speech:62 the discovery of truth,63 interest accommodation and social stability, exposure and deterrence of abuses of authority,64 development of independent judgment and personal autonomy,65 facilitation of liberal democracy,66 and the promotion of tolerance.67 These goods vary in many respects. Some of them are relevant to public life whereas others pertain to both public and private spheres. Some focus on the interests of speakers, others on the interests of the audience, and yet others on the interests of people who are neither speakers nor listeners.68 At least one of these goods, truth discovery, may be considered important regardless of people’s preferences or its contribution to their happiness, and perhaps even regardless of its effect on human well-being. Some theories of freedom of speech are pluralistic in the sense that they view this freedom as promoting some or all of these goods. Others are monistic: they justify the special status of freedom of speech as aiming at promoting one goal only, such as liberal democracy or individual autonomy.69 The monistic theories either ignore other goods or view their promotion as instrumental in maximizing the one good that they ultimately value. As different categories of speech vary in their contribution to the promotion of

62. Kent Greenawalt, Free Speech Justifications, 89 Colum. L. Rev. 119, 130–47 (1989) (critically discussing these rationales). 63. This rationale for freedom of speech has been powerfully advocated by Mill. See John Stuart Mill, On Liberty (1859), reprinted in On Liberty and Other Essays 5 (Oxford Univ. Press, 1991). For critical discussions of this rationale, see also Lawrence Byard Solum, Freedom of Communicative Action: A Theory of the First Amendment Freedom of Speech, 83 Nw. U. L. Rev. 54, 68–72 (1989); Alexander, supra note 50, at 128–30. 64. See also Vincent Blasi, The Checking Value in First Amendment Theory, 1977 A.B. Found. Res. J. 521. 65. See also Martin H. Redish, The Value of Free Speech, 130 U. Pa. L. Rev. 591 (1982); Alexander, supra note 50, at 130–32. 66. See also Alexander Meiklejohn, Political Freedom: The Constitutional Powers of the People (1960); Solum, supra note 63, at 72–77; Alexander, supra note 50, at 136–46. 67. See also Lee C. Bollinger, The Tolerant Society: Freedom of Speech and Extremist Speech in America (1986). For a critique of this view, see, e.g., Alexander, supra note 50, at 132–33; David A. Strauss, Why Be Tolerant?, 53 U. Chi. L. Rev. 1485 (1987) (reviewing Bollinger, id.). 68. On the different interests involved in freedom of speech, see, e.g., T.M. Scanlon, Freedom of Expression and Categories of Expression, 40 U. Pitt. L. Rev. 519, 520–28 (1979). 69. See, e.g., Meiklejohn, supra note 66 (liberal democracy); Redish, supra note 65 (individual self-realization).

freedom of speech

different goods, the various theories differ as to the range of expressive activities they deem worthy of constitutional protection.70 These theories propose justifications for constitutional protection of free speech. In that sense, they may be described as providing consequentialist justifications for deontological constraints on the factoral level. Standard economic analysis of free speech, however, seems to be consequentialist on the factoral level as well. Thus, Posner proposes that speech can legitimately be suppressed for the sake of promoting more speech.71 He subsequently pointed to the advantages of setting general legal rules;72 yet his endorsement of rules rests on purely pragmatic grounds and is therefore unlikely to yield a truly deontological moral theory of free speech on the factoral level. Deontologists do not endorse free speech merely because it is expected to maximize such values as truth, tolerance, or government accountability. Rather, they recognize a constraint against preventing people from expressing themselves or listening to others. This constraint may be seen as a particular manifestation of the more general constraint against actively/ intentionally harming people’s autonomy and dignity. Respect for people’s autonomy and dignity requires everybody, including the state, to refrain from silencing other people or preventing them from listening to others. Such respect underlies the liberal theory of the state. Citizens may be legitimately expected to respect the community’s collective decisions—that is, to obey the law—only if the community respects them as autonomous and rational people of equal worth.73 While respect for human dignity requires one to refrain from inflicting other types of harm as well, there is an especially close

70. For example, Martin Redish has advocated an inclusive monistic theory, asserting that the only value underlying free speech is “individual self-realization,” yet that other values such as facilitation of liberal democracy and truth discovery are important because they contribute to individual self-realization. Redish, supra note 65. Cass Sunstein has argued that only political speech deserves stringent protection, but his notion of political speech encompasses any speech which is “both intended and received as a contribution to public deliberation about some issue.” Cass R. Sunstein, Democracy and the Problem of Free Speech (1993). 71. Posner, Frontiers, supra note 34, at 68–69. Interestingly, feminists have made the comparable argument that since pornography silences women, it should be curtailed to promote overall free speech. See, e.g., Catherine A. MacKinnon, Pornography, Civil Rights, and Speech, 20 Harv. C.R.-C.L. L. Rev. 1 (1985). 72. Posner, supra note 36, at 740–41. 73. C. Edwin Baker, Human Liberty and Freedom of Speech 47–69 (1989) [hereinafter Baker, Human Liberty]; C. Edwin Baker, Scope of the First Amendment Freedom of Speech, 25 UCLA L. Rev. 964, 990–92 (1978) [hereinafter Baker, First Amendment].

191

192


connection between expressions of ideas and feelings and one’s inner thoughts and emotions. Interpersonal communication is essential for the development and fulfillment of one’s rational faculties, emotions, and artistic creativity. It is essential to self-realization.74 Hence, suppression of speech is particularly harmful to one’s dignity. In the same way, treating people as autonomous human beings necessitates letting them shape their own lives, and such shaping entails free access to any information, view, or argument. Suppressing speech that aims to convince people of the truthfulness of certain beliefs or persuade them to do certain things via rational arguments adversely affects people’s autonomy because it interferes with their control over their reasoning process.75 Just as inducing a person to act by lying to her is disrespectful of her autonomy—using her as a means to realizing the liar’s ends—so too does governmental prevention of access to information and arguments, in order to affect people’s beliefs and conduct, violates their autonomy.76 From this perspective, inhibiting a certain behavior by lying to people or by suppressing speech is more harmful to their autonomy than an outright prohibition of the same behavior.77 Silencing some views rather than others also violates the requirement to treat people with equal respect.78 The deontological justification for free speech often rests on contractarian theories of the state. People entrust the government with the protection of interests they cannot safeguard by themselves, such as protecting their lives, liberty, and property; but they do not authorize the state to interfere in other spheres.79 To regard themselves as autonomous, people must retain

74. Solum, supra note 63, at 79–81. 75. David A. Strauss, Persuasion, Autonomy, and Freedom of Expression, 91 Colum. L. Rev. 334, 354 (1991). 76. Id. at 353–59 (advocating this proposition and discussing counter-arguments). Note that this deontological rationale of free speech may justify restrictions on the power of the state to prevent people’s access to available information without necessarily requiring the government or anybody else to actively provide people with information. As indicated before, we do not discuss here the latter issue. 77. Id. at 359–60. 78. David A.J. Richards, Toleration and the Constitution 166–74 (1986); D.F.B. Tucker, Law, Liberalism, and Free Speech 34–56, 63 (1985). See also infra pp. 218–21. 79. Thomas Scanlon, A Theory of Freedom of Expression, 1 Phil. & Pub. Aff. 204, 215 (1972) (arguing that a defensible principle of freedom of speech should rest on the view that “the powers of the state are limited to those that citizens could recognize while still regarding themselves as equal, autonomous, rational agents”). For a subsequent revision of his theory, see Scanlon, supra note 68. See also David A. J. Richards, Free Speech and Obscenity

freedom of speech

sovereignty over their beliefs and ability to rationally weigh competing reasons for action. At the same time, even autonomous and rational people may legitimize certain limitations on freedom of speech as a means of protecting themselves from harm inflicted by others or even as a self-paternalistic measure. Thus, freedom of speech need not be absolute.80 A comparable argument can be made regarding contractarianism as a theory of private morality. In and of itself, a contractarian ethical or political theory is neither a sufficient nor a necessary condition for the recognition of a constraint against curtailing free speech. Contractarianism is a foundational normative theory, while our focus is on the factoral level. A deontological factoral theory may possibly be justified by a consequentialist foundational theory (such as rule-consequentialism), and a contractarian foundational theory may conceivably yield a consequentialist factoral theory.81 As the above analysis implies, a plausible version of the deontological constraint against suppressing free speech should not rest solely on respect for the autonomy and dignity of the listeners nor focus exclusively on the autonomy and dignity of the speakers. On the one hand, a speaker-centered theory faces difficulty explaining, for example, a constraint against preventing the reading of works of a dead author. It may also fail to justify a constraint against denying access to ideas produced by entities that may not be considered autonomous human beings, such as young children or corporations, or access to naturally existing information.82 An audience-centered theory, on the other hand, might legitimize silencing a person whose views nobody wishes to hear, as potential listeners can plausibly waive their right to receive certain information. Similarly, if a person desires to express her support for a well-known opinion, in circumstances where this support adds nothing to the audience’s deliberation, silencing her would not, presumably,

Law: Toward A Moral Theory of the First Amendment, 123 U. Pa. L. Rev. 45, 59–70 (1975); Richards, supra note 78, at 165–87, Tucker, supra note 78, at 34–56. 80. Scanlon, supra note 68. See also Greenawalt, supra note 3, at 32–33. 81. See generally Shelly Kagan, Normative Ethics 17–22, 189–94, 223–39, 294–303 (1998); supra pp. 26–27. See also Greenawalt, supra note 62, at 128 (arguing that, even if nonconsequentialist reasons for protecting free speech may ultimately be justified as producing the best effects overall, for the purpose of practical thought such reasons should nevertheless count as nonconsequentialist). 82. Alexander, supra note 50, at 8–9; Roger A. Shiner, Freedom of Commercial Expression 163–91 (2003).

193

194


infringe the audience’s rights in any way.83 Proponents of each of these competing theories may perhaps reply to these critiques. We do not believe, however, that one must choose between the two. Rather, the two theories may be combined. Accordingly, suppression of free speech sometimes infringes the rights of speakers, sometimes the rights of the audience, and typically the rights of both.84 The deontological conception of free speech carries various implications, many of which will be discussed in greater detail in the following section. Inter alia, from a moderate deontology perspective, there is a substantial difference between silencing speech because the government disapproves of the speaker’s views or because it prefers that certain issues will not be discussed in public, and silencing speech as a mere side effect of pursuing other goals, such as preventing litter and traffic congestion. The very existence of a threshold constraint implies that even worthless speech deserves protection as long as the costs it inflicts do not surpass the threshold. Unlike consequentialism, moderate deontology is likely to differentiate between different harms which speech may cause. It may discount, or exclude altogether, harms caused by people being rationally persuaded to do certain things. It may similarly ignore the displeasure experienced by disinterested people. Deontology may also exclude chronologically remote, low-probability, and very small harms caused by any speech. Some clarifications are in order before proceeding to discuss how these implications may affect a constrained CBA of freedom of speech. First, as alluded to already, consequentialism and promoting personal autonomy are not incompatible. A consequentialist moral theory may incorporate autonomy into its theory of the good.85 Yet, while a consequentialist theory would aim at maximizing overall autonomy, only a deontological factoral theory treats respect for autonomy as a constraint on promoting the good. At any rate, since welfare economics rests on a preference-satisfaction theory of human well-being, it attributes no intrinsic value to autonomy, and thus this path is not open to it. Second, the usefulness of a constrained CBA of free speech does not assume that respect for dignity and autonomy is the sole rationale for this principle.

83. Solum, supra note 63, at 79. 84. See, e.g., Redish, supra note 65. Cf. Scanlon, supra note 68, at 520–28 (analyzing the interests of participants, audience, and bystanders). 85. See supra pp. 30–32.

freedom of speech

The ensuing analysis is perfectly compatible with a pluralistic view that justifies freedom of speech in large part by its desirable consequences. One may be a pluralist and still recognize the necessity of setting constraints on limiting freedom of speech.86 Finally, as explained above, while a constrained CBA assumes that moderate deontology is the most appropriate moral theory on the factoral level, this factoral theory may rest on consequentialism at the foundational level (such as, possibly, rule-consequentialism).

•

E. Constrained Cost-Benefit Analysis 1. The Regulation’s Net Benefit (a) General

Deontologically constrained CBA deviates from standard CBA in two central respects: it may exclude or discount certain types of costs and benefits, and it prohibits any act or rule infringing a deontological constraint unless its net benefit meets a certain threshold.87 Both aspects of threshold deontology are manifest in the legal thinking of freedom of speech and may thus be fruitfully illuminated through a constrained CBA. This subsection discusses the first aspect, and the following one will discuss the second, paying heed to the interrelations between them. In considering the benefit which may justify infringement of the constraint against suppressing free speech, moderate deontology is unlikely to take into account just any cost or benefit. A moderate deontologist may hold that some types of harm can never justify the curtailment of free speech, and thus the benefit from their elimination should not be part of a constrained CBA of speech regulation. Alternatively, she may maintain that avoiding such harms can only justify speech regulation in extreme cases, thus either substantially discounting their weight, or setting a high threshold for the justification of regulation aimed at preventing such harms. In what follows, we discuss both exclusion and discounting of harms.

86. On the strengths and weaknesses of pluralistic theories of freedom of speech, see Solum, supra note 63, at 82–85. 87. See generally supra pp. 84–96.

195

196


A deontologically constrained CBA of speech regulation may exclude or discount chronologically remote, low-probability, or insubstantial harms of the relevant speech, harms resulting from the behavior of legally responsible listeners who were rationally persuaded to act in a certain way, and the offensiveness of some expressions. A moderate deontologist may also wish to distinguish between infringing a constraint in order to prevent the decrease in human welfare expected from a certain speech and infringing a constraint to increase the current level of welfare. A deontologist may discount additional harms, or only some (or none) of the above harms. None of these excluders is self-justificatory. They all require careful normative examination. Note, also, that excluding or discounting these harms is tantamount to excluding or discounting the benefits from suppressing the harmful speech. Thus, a view that one must not infringe freedom of speech in order to prevent a certain type of harms is translated to the exclusion of these harms in calculating the benefit from suppressing the harmful speech. (b) Chronologically Remote Harms Standard CBA employs a discount rate to determine the present value of future benefits and costs. This feature is evident in Posner’s formula discussed above (though, somewhat surprisingly, while the speech’s harm is discounted for its futurity, no discounting expressly applies to the speech’s benefit). Deontologically constrained CBA may treat future outcomes in the same way, but it may also completely exclude or radically discount them. Various rationales have been offered for the exclusion of chronologically remote harms. One rationale rests on second-order considerations: the further the expected harm, the more difficult it is to assess its probability and magnitude, and thus the greater the danger of erroneous, counterproductive regulation. This argument is instrumental and contingent. As such, it may comfortably fit into a standard CBA.88 Another rationale is that chronological remoteness may serve as a proxy for low probability. For instance, in the context of advocacy of lawless action, ordinarily the longer the period between the advocacy and the action, the less likely it is that the action

88. Cf. Posner, Free Speech, supra note 34, at 8–9, 24–29 (proposing an economic model for CBA of speech regulation which includes a variable E for the costs of regulators’ errors).

freedom of speech

will occur.89 Various changes in circumstances and intervening factors may render such advocacy ineffective, thus reducing its probability. This argument is, however, contingent as well. It does not hold whenever some harm is both chronologically remote and highly probable.90 At any rate, it is parasitic on the minimal probability requirement that will be discussed below, and thus arguably redundant. A third argument rests on a strong belief in the functioning of the “marketplace of ideas.” In such a market, false and harmful ideas, such as those of hate speech, are doomed to failure, and truthful and beneficial ideas will eventually prevail. Whenever there is enough time for the free market to restrain harmful ideas, state regulation is unnecessary.91 This argument serves not only to explain the exclusion or discount of chronologically remote harms, but also—and primarily—to substantiate an opposition to any regulation of speech, save for expressions that are expected to cause an immediate harm. It is, however, susceptible to powerful critiques. The “marketplace of ideas” is saturated with market failures.92 It is not clear that absent regulation, truth will necessarily prevail over falsehood. It may well be that demagogy, shortsightedness, and irrational sentiments will have the upper hand.93 Even within communities committed to the rational pursuit of truth, such as those of scientists and academics, the making of arguments is subject to considerable restrictions, thus reflecting skepticism regarding the efficacy of the “market” to eliminate false claims. Similarly, in judicial proceedings, the law imposes significant limitations on the form and substance of admissible evidence.94 Finally, if the crux of the marketplace of ideas metaphor is that free deliberation is the best way to weed out

89. See Posner, Frontiers, supra note 34, at 65; Larry Alexander, Incitement and Freedom of Speech, in Freedom of Speech and Incitement Against Democracy 101, 109 (David Kretzmer & Francine Kershman Hazan eds., 2000). 90. Posner, supra note 49, at 121–22. 91. Posner, Frontiers, supra note 34, at 65. 92. See supra notes 31–33 and accompanying text. Posner concedes that the market in ideas cannot be expected to function perfectly, yet he believes that the “very thing that makes this market inefficient—the extremely high costs of information [about the quality of ideas]—makes the regulation of them inefficient.” Posner, Frontiers, supra note 34, at 84–85. Strauss argues that the very notions of production and consumption, selling and buying, and supply and demand, do not meaningfully capture the processes of forming and disseminating political, moral, or aesthetic ideas. Strauss, supra note 75, at 348–49. 93. Baker, First Amendment, supra note 73, at 976–78; Harry H. Wellington, On Freedom of Expression, 88 Yale L.J. 1105, 1130 (1979). 94. Alexander, supra note 50, at 128.

197

198


false and harmful ideas, then this process seems unnecessary whenever the falsehood and harmfulness of a speech is a given, as in the case of advocacy of violence against innocents. Contrary to the marketplace of ideas rationale, under prevailing moral and legal norms, speech is sometimes protected although it is known to be worthless and even damaging.95 We then turn to a fourth basis for excluding/discounting chronologically remote harms. According to this distinctively deontological argument, even if the net benefit of suppressing dangerous expressions is great enough to warrant curtailing free speech, the constraint against silencing people may only be infringed if there are no alternative means to bring about similar good outcomes without such an infringement.96 The common feature of this rationale and the marketplace of ideas argument is that both maintain that, whenever it is possible to prevent certain harms without curtailing speech, it is impermissible to curtail speech. The difference between them lies in that the present rationale does not rest on a belief in the effectiveness of the unregulated market of ideas. The fact that a harm is expected to materialize only after some time plausibly implies that the state will have the opportunity to prevent it without curtailing free speech. For instance, the police may take measures to prevent the unlawful conduct without silencing those who advocate it. At other times, harmful advocacy may be counteracted by education.97 This argument echoes the deontological sentiment that only in emergency situations may a constraint be justifiably infringed.98 Moderate deontology need not, however, set so high a threshold; it is compatible with a more flexible chronological threshold. We shall focus on the deontological basis for excluding chronologically remote harms, but it should be stressed that a rule-consequentialist may endorse comparable exclusions for the instrumental reasons discussed above, thus reaching similar conclusions on the factoral level. Plausibly, the exact shape of any excluder depends on its rationale.

95. Redish, supra note 19, at 1162; Kertzmer, supra note 50, at 468–76. See also infra pp. 218–21. 96. See supra pp. 81–83; infra pp. 222–23. 97. Cf. Whitney v. California, 274 U.S. 357, 377 (1927) (Brandeis, J., concurring) (“If there be time to expose through discussion the falsehood and fallacies, to avert the evil by the process of education, the remedy to be applied is more speech, not enforced silence”). For a critique of the more speech argument, see Strauss, supra note 75, at 347–48. 98. See supra pp. 116–17; Whitney, 274 U.S. at 377 (holding that “only an emergency can justify repression” of free speech).

freedom of speech

The availability of effective alternative, noninfringing preventive measures against chronologically remote dangers vary from one setting to another. Such measures are more likely to succeed when the speech is directed at a relatively small group of identified people and advocates a specific unlawful behavior. In contrast, when a speech advocating lawless action is directed at the general public or at an unidentified group of people; when it produces its harmful effect gradually, over a long period of time; and when its adverse effect is the product of the cumulative effect of numerous similar expressions—then speech regulation may be the only effective way to thwart the danger.99 Under such circumstances, even a deontologist might not exclude chronologically remote harms from a constrained CBA. Within a deontologically constrained CBA of speech regulation, the fourth argument may take one of two forms. It may justify the exclusion of chronologically remote harms when determining the permissibility of the regulation. Alternatively, it may enter the analysis when choosing among different permissible courses of action. Here we examined the first possibility, and the second will be discussed below.100 (c) Low-Probability Harms Alongside the tendency to exclude or drastically discount chronologically remote harms, a deontologist may argue that it is never permissible to curtail freedom of speech if the probability that the speech will actually bring about the harm is miniscule.101 To illustrate, under many circumstances, the likelihood that a call to undemocratically overthrow the government would result in anybody taking actual steps in this direction is very small, thus rendering the silencing of the speaker unjustifiable. As a matter of fact, the law often permits the curtailment of free speech only if its expected harm is certain or “clear,” thus excluding low-probability harms.102

99. See Clay Calvert, Hate Speech and Its Harms: A Communication Theory Perspective, 47 J. Communication 4 (1997); Kretzmer, supra note 50, at 462–67. Cf. Alexander, supra note 89, at 109. 100. See infra pp. 222–23. 101. See supra pp. 89–91. See also supra pp. 147–48; infra pp. 344–45. 102. See supra pp. 182–83; Redish, supra note 19; David G. Barnum, The Clear and Present Danger Test in Anglo-American and European Law, 7 San Diego Int’l L.J. 263 (2006).

199

200


Like the exclusion of chronologically remote harms, the present excluder may rest on purely consequentialist or instrumental grounds, such as the risk of miscalculation by a regulator or by a court reviewing governmental decisions.103 Absent such bases, a deontologist may still insist that respect for people’s autonomy and dignity mandates that no amount of improbable harms ever justifies the silencing of people.104 While the chronological remoteness and probability of harms are analytically separate, they are interrelated. In assessing the probability of a certain harm, one must pay heed to the temporal dimension. The probability of a certain harm materializing within a week may be very low. At the same time, the probability of the same harm materializing within a year or a decade may be much higher.105 The assessment of imminence should therefore take into account both the likelihood of harm and the period for which this likelihood is calculated. In setting the minimal probability threshold, one should also take into account the type and magnitude of the speech’s expected harm. Thus, while it may be justified to exclude the danger of political unrest if the probability that a certain speech will actually bring it about is, say, less than 5 percent, one might not wish to exclude the danger of a mass killing of innocent people even if the probability that some speech will bring it about is considerably smaller. A deontologist may hold that some types of expression, such as the advocacy of genocide, should be prohibited even if the likelihood that they will ever bring about the advocated behavior is quite small. In fact, the United Nations Genocide Convention criminalizes incitement to genocide regardless of the probability that it would actually bring it about.106 A different way to justify this prohibition, when the likelihood of ensuing genocide is very small, does not rest on the fear of genocide. According to the alternative justification, advocacy of genocide should be suppressed due

103. See Jonathan S. Masur, Probability Thresholds, 92 Iowa L. Rev. 1293 (2007) (arguing that the exclusion of low-probability risks is a corrective to a systematic overestimation of speech-based threats due to people’s tendency to overstate low-probability, emotionally salient dangers, as well as due to informational asymmetries favoring the government). 104. For a related argument in the context of the fight against terrorism, see supra pp. 147–48. 105. Cf. David Luban, Preventive War, 32 Phil. & Pub. Aff. 207, 234 (2004). 106. Convention on the Prevention and Punishment of the Crime of Genocide, Dec. 9, 1948, 78 U.N.T.S. 277, Art. 3 (“The following acts shall be punishable: . . . (c) Direct and public incitement to commit genocide”). On this provision and on the United States’ opposition to it due to the lack of requirement regarding the probability that the incitement would actually result in genocide, see William A. Schabas, Genocide in International Law: The Crime of Crimes 307, 319–34 (2009).

freedom of speech

to its additional harmful effects, whose probability may be very high. Such likely effects include ethnic hatred, interracial hostility, minor breaches of the peace, and widespread emotional distress among the targeted group. Assuming that some or all of these harms should be taken into account in calculating the benefit of speech regulation (an assumption examined in the next subsection), they may justify infringing the constraint against suppressing free speech even if the danger of genocide is excluded due to its low probability. Moreover, a moderate deontologist may possibly justify the suppression of advocacy of genocide even if one disregards both the prevention of genocide (due to its low probability) and the elimination of the other adverse effects, such as emotional distress (due to their type or magnitude). Such suppression may be justified on the ground that the speech itself, regardless of its expected effect, has an objectively-determined, intrinsic negative value large enough to meet the pertinent threshold. In this respect, advocacy of genocide or similar horrendous crimes is different from the advocacy of other lawless actions. The latter either have some positive value of their own, or their disvalue is not large enough to justify the infringement of the deontological constraint against suppressing speech. This possibility exceeds the boundaries of constrained CBA as analyzed in this book because it rests on a theory of the good that takes into account nonwelfarist components.107 (d) Small Harms Both standard CBA and a deontologically constrained CBA would object to regulating a speech that is expected to cause certain and immediate harm, if the magnitude of the total expected harm is too small.108 A constrained

107. For the argument that some expressions lie outside the scope of speech protection altogether, see, e.g., Alon Harel, Bigotry, Pornography, and the First Amendment: A Theory of Unprotected Speech, 65 S. Cal. L. Rev. 1887 (1992). This position rejects contentneutrality, at least to some extent. 108. Standard CBA may reach this conclusion if the net benefit of the regulation is smaller than its administrative costs. A deontologist may reach the same conclusion even absent any administrative costs, as long as the benefit of the regulation is lower than the threshold of the pertinent constraint. The requirement of minimal total harm is reflected in United States law. See, e.g., Bridges v. California, 314 U.S. 252, 263 (1941) (interpreting the Schenck’s formula, quoted in p. 182 above, the court ruled that “the substantive evil must be extremely serious”); Greenawalt, Speech, supra note 3, at 191; Redish, supra note 19, at 1179–80.

201

202


CBA may, however, add another condition, namely that only harms whose magnitude surpasses a certain threshold would count in calculating the total expected harm, thus excluding small harms.109 For example, even if certain expressions may be curtailed due to their offensive nature—an issue we will soon address—a deontologist may hold that no amount of minor offenses or moderate feelings of resentment ever suffices to justify the suppression of speech. Whether such adverse feelings are experienced by few or by millions, they are an unavoidable cost of living in a tolerant, free society, and should therefore be ignored. Similarly, a political protest calling on people not to file their tax returns on time may be considered protected speech whatever the number of taxpayers expected to follow this call.110 Unlike the exclusion of chronologically remote harms, which is primarily relevant to harms caused in two steps, the exclusion of small harms is relevant to harms caused in one step as well (such as offensiveness). The minimal magnitude of the harm necessary to render it morally relevant may depend on the type of harm involved and the way it is brought about.111 (e) Harms Brought About Through Rational Persuasion Inasmuch as the deontological constraint against curtailing free speech is grounded in respect for the autonomy of potential listeners, it may be argued that harms resulting from rational persuasion of the listeners to act in a certain way must not be taken into account in calculating the net benefit of silencing the speech. Even if the government may legitimately prohibit certain harmful activities, it should not proscribe the advocacy of such activities. Respect for autonomy requires letting people hold a belief or desire, and certainly consider its merits, even if they cannot act on it.112 Such a

109. See supra pp. 79–80. 110. Cf. Whitney, 274 U.S. at 377–78 (“[I]t is hardly conceivable that this Court would hold constitutional a statute which punished as a felony the mere voluntary assembly with a society formed to teach that pedestrians had the moral right to cross unenclosed, unposted, waste lands and to advocate their doing so, even if there was imminent danger that advocacy would lead to a trespass”). 111. See infra pp. 206–10 and 202–06, respectively. A more difficult question is whether the minimal magnitude of the prevented harm should vary from one category of speech to another (cf. infra pp. 218–21). 112. Strauss, supra note 75, at 359–60. A related argument is that, since the ultimate harm is caused by the actions of the listeners who rationally considered the pertinent factors, the speaker is not morally responsible for the harm.

freedom of speech

position does not necessarily rely on the “more speech” argument, namely that whenever there is enough time for the government to counteract the effects of the harmful speech by exposing its fallacy, it must not be silenced.113 Neither does it rest on the optimistic belief in the unregulated “marketplace of ideas.”114 Even if one cannot expect more speech or the market of ideas to counteract the harmful speech, if its harmful effects are brought about through rational persuasion, those effects should be disregarded or discounted when considering the benefit of suppressing the speech. This is, roughly, what David Strauss dubs “the persuasion principle,”115 and Scanlon earlier termed “the Millian Principle.”116 Several issues need to be clarified before proceeding to examine how this principle may be incorporated into a threshold function. The first has to do with the relationship between rational persuasion and the exclusion or discounting of chronologically remote harms. While there is considerable overlap between these two possible excluders, the overlap is not complete. In two types of cases, the two excluders potentially diverge, and thus it makes a difference whether a threshold function adopts one or the other (or both). The first category consists of expressions that rationally persuade people to immediately or imminently engage in harmful activities. The second category consists of expressions that induce people to engage in chronologically remote, harmful activities through false factual statements, threats, or other manipulations, rather than through rational persuasion. The elimination or discounting of chronologically remote harms would not exclude the benefits of silencing the first type of expressions. The elimination of harms brought about through rational persuasion would not exclude the benefits of suppressing the latter. Another clarification pertains to the relationship between the rational persuasion principle and the incidence of the constraint against curtailing free speech. Arguably, to the extent that the constraint rests on respect for people as autonomous and rational human beings, it should not protect outright factual lies, brainwashing, threats, and other manipulations that do

113. See supra note 97 and accompanying text. 114. Strauss, supra note 75, at 348–49. See also supra pp. 197–98. 115. Strauss, supra note 75. 116. Scanlon, supra note 79. Scanlon and Strauss’s principles resonate with Solum’s theory, resting on Habermas’s distinction between communicative action and strategic action. See Solum, supra note 63.

203

204


not affect people’s behavior through reasoning and reflection.117 According to this argument, suppressing modes of expression that aim at bypassing the process of rational deliberation does not infringe any constraint. This restrictive conception of free speech is problematic, however. Even expressions that strive to affect people’s behavior in a manipulative way or that directly harm people without trying to affect anyone’s behavior must not be restricted unless the benefit of the restriction surpasses a certain threshold. This position is supported by pragmatic considerations, such as the difficulty in distinguishing between rational persuasion and manipulation118 and the potential chilling effect of imposing legal liability for false expressions. But its primary justification is principled. Unless enough bad outcomes are at stake, respect for people’s autonomy requires letting them express themselves and hear others express themselves, even if the pertinent expressions do not consist of rational arguments or otherwise valuable information or ideas. At the very least, respect for the speaker’s dignity and autonomy requires refraining from silencing her so long as she sincerely believes in the truthfulness and soundness of her claims. Similarly, under at least some circumstances, respect for the autonomy of listeners may require letting them judge for themselves the accuracy and cogency of arguments. Accordingly, the legal suppression of racial insults and bigotry, for instance, may well be justified; but its justification lies not in denying that it infringes freedom of expression but rather in determining that enough bad is avoided by such measures to justify the infringement.119 Finally, one may possibly endorse the rational persuasion principle without denying the unique status of situation-altering communications, such as those involved in criminal conspiracy or the initiation of an unlawful restraint of trade.120 The fact that in hiring a professional killer, one “rationally persuades” another person to commit murder does not require disregarding

117. Cf. Solum, supra note 63, at 91–93, 107–09, 114–15, 119–26 (discussing strategic, as opposed to communicative, action). 118. For instance, differentiating between making false factual assertions (manipulation) and expressing “false” opinions (persuasion) presupposes a distinction between facts and opinions. This distinction is, however, very problematic. See Fredrick F. Schauer, Language, Truth, and the First Amendment: An Essay in Memory of Harry Canter, 64 Va. L. Rev. 263, 276–81 (1978); Alexander, supra note 50, at 70–74. 119. Cf. Alexander, supra note 50, at 135 (criticizing a narrow conception of the deontological constraint against suppressing speech encompassing only rational persuasion). 120. See supra note 3 and accompanying text.

freedom of speech

or discounting the expected harm of such communication. It should be conceded, however, that the borderline between situation-altering and other communications is often blurred, and that this blurriness may have to do with the weakness of the very notion that harms brought about through rational persuasion should not be considered (or should be discounted) in judging the permissibility of curtailing speech. Arguably, the fact that a rational, autonomous agent is morally and legally responsible for directly harming other people should not exculpate the person who has rationally persuaded her to inflict the harm (whether the harm is a particular criminal act or a general practice of racial discrimination).121 Having clarified these points, and assuming that despite the critique, one holds that the present excluder is warranted, we can examine how threshold functions may incorporate the persuasion principle. The first choice to be made is between complete disregard for harms brought about through rational persuasion and mere discounting of their weight. Since we believe that even when the injurious effects of an expression are the result of rational persuasion, there may be cases in which the expression’s expected harm is great enough to warrant an infringement of the constraint against actively/intentionally suppressing speech, we do not endorse total exclusion of such harm.122 Apparently, reducing the effect of harms caused through rational persuasion may be done in two ways. One possibility is to discount harms resulting from rational persuasion (that is, discount the corresponding benefits of curtailing the harmful speech). Alternatively, whenever the negative effects of a certain speech are brought about through rational persuasion, one may set a particularly high threshold for the justification of suppressing speech. This choice is not merely formal. Often, a single expression brings about different harms in different ways. The speaker may rationally persuade some people to act in a harmful way and at the same time instigate violence by people who will act without any rational deliberation. The expression may also involve direct, one-step harms, such as those resulting from the disclosure of confidential information or defamation. The relative weight of the different harmful outcomes—those that are brought about through

121. See, e.g., Robert Amdur, Scanlon on Freedom of Expression, 9 Phil. & Pub. Aff. 287 (1980). Cf. supra p. 73. 122. Cf. Strauss, supra note 75, at 360–61 (conceding that “the persuasion principle can be overridden if the consequences of permitting the speech are sufficiently harmful”); Solum, supra note 63, at 119–23.

205

206


rational persuasion and those that are brought about otherwise—may vary. The latter formulation requires setting different formulations to each of these cases, whereas the former one enables to treat them through a unified formula. Compare, for example, three cases in which the regulation’s total expected net benefit is similar: In case A, all of the avoided harms are the product of rational persuasion; in case B, half of the avoided harms are the product of such persuasion; and in case C—none. If the same threshold function applies to all three cases (which may be warranted for methodological or pragmatic reasons), then a function that discounts rational persuasion would differentiate between those cases, while setting a very high threshold without such discount would treat those cases similarly. Since the persuasion principle calls for such differentiation, one should either use the former formulation or construct different functions for different scenarios, based on the expected relative weight of the harms brought about through rational persuasion.123 Later on, we shall examine and support the use of different threshold functions for dissimilar bases of regulation and dissimilar categories of speech. A moment’s reflection reveals that none of the conventional distinctions— such as between content-based and content-neutral regulation or between political and commercial speech—necessarily coincides with the persuasion principle. For this reason, such categorization does not substitute the discounting of harms resulting from rational persuasion. (f) Offensiveness The term offensiveness has numerous meanings. In this subsection, we use it to denote an adverse psychological effect or state of mind, such as disgust, revulsion, anger, or shock, experienced by people whose deeply held religious beliefs, moral convictions, or social values are offended by being exposed to, or merely knowing about, such expressions as antireligious epithets and blasphemy, hard core pornography and obscenity, flag burning, racial and ethnic hate speech, and chauvinist insults.124 Drawing the line between these

123. Similar considerations pertain to the exclusion/discounting of other harms discussed above and below, whenever the same speech is expected to cause different types of harm. 124. On additional meanings of offensiveness and on the considerable difficulty of drawing the line between offensiveness and harm, see generally Joel Feinberg, The Moral Limits of the Criminal Law, Vol. 2: Offense to Others (1985); Harlon L. Dalton, “Disgust” and Punishment, 96 Yale L.J. 881, 886–89 (1987) (reviewing Feinberg).

freedom of speech

types of offensive expression, on the one hand, and abusive language aimed predominantly at inflicting severe psychological wounds on the listener, fighting words intended to provoke immediate violence by the addressees, and other situation-altering utterances, on the other hand, may be extremely complicated.125 Another complication stems from the fact that, in addition to preventing adverse mental states, suppressing offensive speech sometimes yields additional benefits. This is the case when the silenced speech is likely to result in outbreaks of racial or religious riots in which people may be injured or killed. The question of whether such benefits justify speech regulation raises concerns of its own.126 This subsection puts aside these complications. It focuses instead on the core of offensive expressions and asks whether their prevention may be considered part of the net benefit justifying the curtailment of free speech. Standard CBA of speech regulation takes into account all costs and benefits, including the offense taken by people who are not directly affected by the speech. In fact, to the extent that utilities and disutilities are measured according to people’s preferences, even the preferences of people who do not experience any adverse feeling, but merely prefer that some expressions be suppressed, should presumably be given full weight. There are, though, second-order consequentialist reasons to disregard the offensiveness of speech, let alone the mere preferences of disinterested people. For one thing, the great difficulty of determining the existence and quantifying the intensity of such feelings and preferences renders the analysis rather manipulative.127 Unqualified consideration of such feelings also raises the concern of confusing preferences with normative judgments.128 Legitimizing speech regulation on grounds of offensiveness may thus be too risky, as it would enable the government to use offensiveness as a pretext to suppressing speech where

125. See Chaplinsky v. State of New Hampshire, 315 U.S. 568, 572 (1942) (holding that constitutional protection does not apply to “insulting or ‘fighting’ words—those which by their very utterance inflict injury or tend to incite an immediate breach of the peace”). While the court described such fighting words as a “well-defined” class of speech, demarcating this class is rather difficult. See, e.g., Kent Greenawalt, Fighting Words: Individuals, Communities, and Liberties of Speech 47–58 (1995). 126. See, e.g., Tribe, supra note 2, at 849–56; Robert Post, Religion and Freedom of Speech: Portraits of Muhammad, 14 Constellations 72 (2007). 127. Cf. supra p. 28. Posner, Frontiers, supra note 34, at 78. Moreover, while certain expressions offend some people, it is at least possible that some people are deeply upset by the suppression of those expressions. 128. See supra p. 28.

207

208


the real grounds are illegitimate.129 Furthermore, history reveals that ideas, artistic works, and certain behaviors that were deemed deeply offensive in the past (and are still so regarded in some societies) have become perfectly acceptable and even conventional. The very exposure to challenging expressions may alter people’s sensibilities. The short-term offense experienced by some (or many), may in the long run be outweighed by the greater good produced by the challenging and stimulating effect of the offensive speech.130 These consequentialist rationales for excluding or discounting offensiveness are contingent and somewhat speculative. The weight of the concern that offensiveness be used as a pretext for ill-motivated regulation varies considerably from one context to another. Similarly, there is no particular reason to believe that people will cease to be offended by some types of speech in the long run. Moreover, the long-term effects argument may be turned on its head, as it is not clear that losing certain sensibilities is necessarily a good thing (and, in any case, it raises questions regarding the intergenerational allocation of welfare). Be that as it may, a deontologist may hold that offensiveness should be excluded or radically discounted in calculating the net benefit of suppressing speech on the basis of the notion of tolerance. Respect for people’s dignity requires people to withstand the adverse emotional impact caused by offensive speech. The very fact that a person is annoyed by a radical political caricature, by watching or hearing about flag burning, by pornography, or by blasphemous expressions, is not a sufficient reason to curtail people’s freedom to engage in such expressive activities.131 Rather than excluding offensiveness altogether or conversely taking any offense into account, a moderate deontologist may adopt a more nuanced position. In deciding whether an offense should fully or partially count in calculating the net benefit of the regulation, various characteristics of the offense should be considered. Offensiveness may be excluded altogether unless either or both its intensity and duration surpass certain (high) thresholds.132

129. Strauss, supra note 75, at 342. 130. Posner, Frontiers, supra note 34, at 78–79. 131. Bollinger, supra note 67; Tucker, supra note 78, at 128–36; Cf. Fienberg, supra note 86, at 38–39 (excluding offensiveness as a ground for limiting the expression of opinions). 132. This is an application of the small harms excluder discussed supra pp. 201–02. Cf. Feinberg, supra note 124, at 27–30, 34–35 (discussing the intensity, duration, and extent of offense as determining its magnitude, which in turn is one factor determining its seriousness).

freedom of speech

Offense taken by overly sensitive people may be similarly disregarded.133 Another pertinent factor is whether the offended person could have reasonably avoided the offensive speech.134 For instance, it is very easy to avoid watching offensive movies or reading offensive books but more difficult to avoid pornographic posters in one’s workplace. Whenever avoidance is feasible, it may be seen as a preferable alternative to silencing. This distinction rests on the (contestable) assumption that offense resulting from bare knowledge of the existence of some speech should not be taken into account. Alternatively, if offense resulting from such bare knowledge is deemed relevant, one may still exclude such offense whenever the offended people can easily avoid not only the speech itself but also the knowledge of its existence.135 Yet another factor is the manner in which an opinion or claim is made. It is one thing to severely criticize a certain religion in an academic article and another to publicly ridicule its sacred symbols. Not only is the latter typically more offensive, suppressing the former is potentially more harmful to freedom of thought and of rational deliberation, and thus its offensiveness may be disregarded.136 One may or may not accept the normative judgments underlying the full or partial exclusion or discounting of offensiveness in assessing the net benefit of restricting speech according to these criteria. Our main point is that insofar as one accepts these judgments, they can and should inform a deontologically constrained CBA of speech regulation. As a matter of fact, liberal democracies do tend to disregard mere offensiveness as a ground for

133. Cf. Post, supra note 126, at 81–82; Feinberg, supra note 124, at 33–35 (discussing the “discounting of abnormal susceptibilities” factor in determining the seriousness of offense). Another possibility would be to exclude offense to people whose sensitivity is considered immoral, such as the offense taken by homophobic or racist people due to their homophobia or racism. See infra p. 254. 134. See, e.g., Tucker, supra note 78, at 135; Scanlon, supra note 68, at 542–50 (analyzing the problem of unwilling audience in the context of pornography). See also Feinberg, supra note 124, at 32–33, 34–35 (discussing the “standard of reasonable avoidability” and the “volenti maxim” in determining the seriousness of an offense). 135. As alluded to earlier, the conclusion may change whenever the bare knowledge of the offensive speech is likely to result in violence by overly sensitive, intolerant people, in which innocent people may be injured or killed. 136. It may, however, be argued that while the public ridicule is more offensive, it is actually more important for freedom of thought, because it calls the attention of more people to the issue and does so in a more powerful and vivid way. For a critical discussion of the distinction between different manners of making a claim, see, e.g., Post, supra note 126, at 80–82.

209

210


speech regulation, and they often resort to some or all of the distinctions considered.137 (g) Combining Excluders Once the normative judgments regarding the inclusion, exclusion, or discounting of various effects of speech regulation are made, one has to define their interrelations and integrate them into a constrained CBA. A rather extreme position may be that only absolutely certain, immediate, and very serious harms (excluding, for instance, mere offensiveness), may justify the infringement of the constraint against suppressing free speech.138 A less extreme possibility is to set independent thresholds for the probability, futurity, and magnitude of the speech’s harms. For instance, it may be held that only harms which are more likely to occur than not to occur ( p > 0.5), expected within a period of one month, and whose magnitude is greater than some value are deemed relevant in assessing the permissibility of suppressing free speech. According to this formulation, if the harm is not expected within the designated period, it will not be taken into account whatever its expected magnitude and probability. Similarly, if the probability that the harm will materialize is lower than the threshold, it will be disregarded whatever the timing of its expected occurrence and its scope; and the same applies to small harms. Alternatively, a threshold function may take into consideration only very serious harms (whatever their probability or expected timing) and harms which are immediate and certain (whatever their type or magnitude). Under either of these two alternatives, the various thresholds may be strictly set, as in the above example, but they may also be less rigid, leaving some discretion to the decision-maker. There is a vast number of possible combinations of different excluders and discounts, with varying degrees of discretion. Once one excludes (or drastically discounts) those harms that do not meet the minimal probability, immediacy, magnitude, or other thresholds, another issue is whether the remaining harms should be weighted according to their probability and chronological remoteness. One may avoid such weighting and hold that at this stage only the aggregate harm counts, regardless of

137. As described in supra p. 183, American law takes an extreme position in this regard. See also Greenawalt, supra note 125. 138. Justice Holmes’s famous passage in Schenck (quoted in p. 182 above) at least literally represents such a position.

freedom of speech

its exact probability and timing. If, for instance, a certain expression is expected to bring about serious violence within a relatively short period and with a sufficiently high probability, it would make no difference whether its probability is, say, 65 or 95 percent or whether it will materialize in two days or two weeks. According to another possibility, the harms that meet the thresholds should be multiplied by their probability, and their present value should be determined using some discount rate. While courts and legal scholars unsympathetic to formal analysis rarely formulate their criteria in such a way, it seems that such a formulation captures commonsense morality and the legal doctrine in many jurisdictions.139

2. The Threshold (a) General The distinctive mark of deontologically constrained CBA is that whenever an act or a rule infringes a deontological constraint, such infringement is permissible if its (relevant) net benefit surpasses a certain threshold. Silencing a speaker is preferable to a blood bath,140 and oftentimes considerably less good (or bad) is sufficient to justify the curtailment of speech. Much of free speech doctrine and theory revolves around setting the appropriate thresholds. Without getting into intricate doctrinal and theoretical issues, this section discusses three general questions: should the threshold depend on the severity of the harm inflicted by the regulation, should different thresholds be set for different bases of regulation, and should different thresholds be set for different categories of speech. (b) The Threshold’s Shape In chapter 4, we discussed three possible shapes of threshold functions: additive, multiplier, and combined.141 In accord with our general tendency

139. Some courts do in fact employ comparable formulations. See, e.g., Dennis v. United States, 341 U.S. 494, 510 (1951) (“In each case (courts) must ask whether the gravity of the ‘evil,’ discounted by its improbability, justifies such invasion of free speech as is necessary to avoid the danger”). See also Greenawalt, Crime Speech, supra note 3, at 116. 140. Tribe, supra note 2, at 853. 141. See supra pp. 93–96.

211

212


to favor combined threshold functions, we believe that a combined function is most appropriate in the present context as well. Thus, the amount of (relevant) net benefit necessary to render suppression of speech permissible should be a function of the actual or probable harm inflicted by the infringing act; yet the threshold would be positive even if infringing the constraint causes no harm, such that: (2) T = (B – H) – (K + K ′H) where B is the total (social, scientific, aesthetic, or other) benefit of the challenged speech: H is the speech’s relevant harm, discounted according to the applicable excluders discussed above; and K and K ′ are the additive and multiplier thresholds, respectively. The harm to people’s autonomy and dignity caused by speech regulation varies from case to case. Moderate limitations on election campaign finance inflict smaller harm on people’s autonomy and dignity than a total banning of election campaign or harsher restrictions on its financing. The multiplier component of a combined function, K ′, formalizes the notion that the stricter the limitations on campaign finance, the larger the net benefit necessary to render them permissible. Likewise, zoning ordinances restricting the location of establishments hosting nude dancing and other adult entertainment adversely affect freedom of expression to a lesser extent than a complete ban on such establishments. The less stringent the zoning restrictions, the smaller the net benefit necessary to justify them.142 In the same vein, limiting the time or place of a political demonstration, due to its noncommunicative impact, may leave the protestors with more or less reasonable alternatives. If, for instance, the protesters strive to influence a decision that is expected within forty-eight hours, postponing the demonstration for a longer period would make it pointless, thus seriously harming freedom of expression. Under different circumstances, the same postponement may only entail minor inconvenience, thus causing a much smaller harm. A multiplier or a combined threshold would reflect the judgment that the net benefit of forcing the postponement of the demonstration, necessary to justify it, should be much higher in the first scenario.

142. Despite the fact that these restrictions pertain to the place of expression, and notwithstanding judicial reasoning focusing on their “secondary effects,” rather than their communicative impact, these restrictions are content-based. Tribe, supra note 2, at 952.

freedom of speech

At least theoretically, speech regulation may cause very little or no harm at all. Prohibiting the dissemination of information whose possessors do not wish to disseminate anyway, or whose potential audience already knows, may be deemed unharmful. Plausibly, even such regulation should only be permitted if its net benefit surpasses a certain threshold. This notion may be formalized using the additive component K, of the combined threshold function. (c) Different Thresholds for Different Bases of Regulation According to Posner’s formula, in line with standard CBA, the purpose or motivation underlying a legislative or administrative curtailment of free speech is basically inconsequential. It does not affect the legitimacy or constitutionality of speech regulation.143 When a law or an action brings about undesirable results, such as silencing criticism of the government, it adds nothing to note that its motivation was probably inappropriate (selfishness in this example). This position skirts the debate concerning the observability, verifiability, and intrinsic significance of the mental states of governmental bodies,144 and avoids the difficulty of handling multipurpose actions and legislation.145 This feature of standard economic analysis of speech regulation squarely contradicts the legal doctrine in the United States and elsewhere. According to prevailing legal norms, reflecting commonsense morality, there is a fundamental difference between Track One, content-based limitations on speech which aims to suppress certain points of view or an open debate of certain issues, and Track Two, content-neutral limitations, aiming at reducing traffic congestion, excessive noise in residential neighborhoods, and so forth.146 Content-based regulation (including an apparently neutral but ill-motivated regulation) is subject to a stringent standard of judicial review known as “strict scrutiny.” Such regulation would only be deemed constitutional if it is

143. Posner, Frontiers, supra note 34, at 70. See also Posner, supra note 36. 144. See supra pp. 63–70. 145. Posner, supra note 36, at 751–52. Cf. supra p. 145. 146. Elena Kagan, Private Speech, Public Purpose: The Role of Governmental Motive in First Amendment Doctrine, 63 U. Chi. L. Rev. 413 (1996).

213

214


“narrowly tailored” to further a “compelling state interest.”147 In contrast, regulation of the noncommunicative impact of expressions—its time, place, and manner—which affects freedom of speech only incidentally, is subject to a much less exacting review, employing a balance-of-interests test.148 This remarkable gap between economic analysis and prevailing norms may be narrowed to some extent, as there are instrumental reasons to consider the regulator’s motives. Ordinarily, there is a positive correlation between the purpose of actions and their effects. Hence, a law aiming to silence speech is much more likely to have this effect than a law aiming to regulate the safety of transportation, which as a side effect restricts the dissemination of pamphlets in certain places, for example.149 In addition, along with its direct effect on the ability of people to express their views, a content-based regulation sends a message regarding the legitimacy and worth of those views. No such message is sent when expressions are silenced as a by-product of regulation which truly aims at attaining other goals.150 Moreover, consequentialist justifications of free speech emphasize the importance of diversity of expressions for the advancement of truth discovery, deterrence of abuses of authority, facilitation of liberal democracy, and so forth. In that respect, the costs of content-based regulation are much greater than those of a content-neutral one. Content-neutral regulation is more likely to have an indiscriminate effect on the expression of different views.151 A law forbidding the burning of the American flag to protest governmental policies and one forbidding the burning of any object on city streets for safety reasons, similarly proscribe a politically motivated flag burning. However, allowing the former is much more harmful to speech diversity.152 Another argument favoring the more lenient judicial approach to content-neutral restrictions, compatible with CBA, is that a truly undiscriminatory restriction saves the court the task of assessing B, each expression’s benefit, as it

147. See supra note 11 and accompanying text. 148. See supra p. 181. 149. Posner, supra note 36, at 745. 150. See Raz, supra note 4, at 313–16. On the expressive effect of legislation, see generally supra pp. 117–22. 151. Posner, Free Speech, supra note 34, at 17. Sometimes, though, content-neutral regulation does have a disparate effect on different people. For instance, banning the distribution of handbills has a much greater effect on the poor than on the affluent, who can afford access to the electronic media. Tribe, supra note 2, at 979–81. 152. Alexander, supra note 50, at 38–51.

freedom of speech

equally applies to all expressions. In contrast, a content-based regulation necessitates such assessment, which in turn increases the risk of judicial error (and, one should add, the risk of error by the regulator).153 The last argument raises the general concern (mentioned earlier and discussed in some detail below) that CBA requires one to assess the benefits of different expressions. Putting this concern aside for a moment, it is doubtful that any of the above instrumental, second-order, or institutional arguments successfully accounts for the centrality and significance of the distinction between Track One and Track Two restrictions of free speech under current legal norms. Thus, for example, suppression of all political speech does not discriminate between different views and thus does not require weighing their specific value. More important, to the extent that the above arguments provide consequentialist justifications for treating intended and unintended restrictions of speech differently, they do not deny its importance. On the contrary, they provide foundational justifications for a deontological factoral normative theory of free speech embracing the Track One/Track Two distinction, which basically corresponds to the well-known intending/foreseeing distinction. To be sure, a consequentialist can take the intended/foreseen harm distinction into account without transcending act-consequentialism. In addition to the use of purposes as proxies for expected outcomes, a consequentialist may plausibly maintain that people’s well-being is decreased to a greater extent when they are silenced intentionally, due to their views, than when they cannot express themselves due to Track Two restrictions.154 An improved theory of human well-being (and even more so, a sophisticated theory of the good that takes actions and motivations into account) may thus accommodate this distinction. It is doubtful whether standard economic analysis, resting on a preferencesatisfaction theory of human welfare, can adopt such a theory of the good.155 But even if it could, this is not the way the legal doctrine—viewing freedom of speech as a basic human right—is structured. We concede that Track Two jurisprudence may be understood as either reflecting a consequentialist, balance-of-interests approach (where “speech interests” are accorded greater

153. Posner, Free Speech, supra note 34, at 17. It does not, however, save the court the need to assess the overall value of the expressions silenced by the regulation. 154. Cf. Daphna Lewinsohn-Zamir, Beyond the Bottom Line: The Complexity of Outcome Assessment (working paper, available at http://papers.ssrn.com/sol3/papers. cfm?abstract_id=1479051). 155. See supra pp. 30–32.

215

216


weight) or as embodying a moderate deontological approach setting a relatively low threshold that has to be met to justify the restriction. As further explained below, we favor the latter understanding but do not deny that the former is conceivable as well. However, when it comes to the core, Track One analysis of freedom of expression, at least in the United States, the jurisprudence takes a distinctively deontological shape. It does not primarily aim at maximizing total speech but rather sets a constraint against suppressing speech.156 From a deontological perspective, inasmuch as the constraint against suppressing free speech is a particular manifestation of the more general constraint against actively/intentionally harming people—in the present context, harming their autonomy and dignity—then the distinction between Track One, intended silencing of speech, and Track Two, merely foreseeable silencing, is clearly justified. In fact, the central role this distinction plays in the legal analysis of free speech lends support to the proposition that freedom of speech is grounded, to a considerable extent, in deontological morality. We thus maintain that a fruitful analysis of speech regulation should establish different thresholds for different bases of regulation. The threshold for Track One, content-based, intended restrictions of speech based on its communicative impact should be high. The threshold for Track Two, content-neutral, unintended restrictions, based on the noncommunicative impact of speech, should be much lower. From a moderate deontology perspective, the fact that some harm is actively inflicted as a mere side effect of attaining other goals does not necessarily render it permissible. Even in such a case, there is a requirement of proportionality.157 A sincere motivation to keep the streets clean or avoid traffic congestion does not justify each and every curtailment of free speech. The difference between Track Two and Track One regulation is not that the former is not subject to a deontological constraint against actively suppressing speech but rather that the size of the threshold is very different. This conception is reflected in the legal doctrine. As aptly described by Harry Kalven, in weighing the conflicting considerations bearing on the permissibility or constitutionality of Track

156. See, e.g., Tribe, supra note 2, at 835–36 (“While the government may foster the values of free expression found in the first amendment, it is precluded by the amendment from compelling expression or suppressing expression, even where government would justify such intrusion on personal liberty as a pursuit of first amendment values”). 157. See supra p. 160; infra p. 282.

freedom of speech

Two regulation, one should put a thumb on the speech side of the scale.158 Putting a thumb on the scale is tantamount to setting a modest threshold, such that even if the regulation’s total benefit exceeds its total cost, it would still be impermissible unless its net benefit surpasses some positive, though relatively low, threshold.159 This analysis implies that the same restriction of speech may be justified if brought about as a side effect of pursuing a legitimate goal, yet unjustified if intended. Following the discussion in chapter 3, this framework assumes that governmental motivations and purposes are in principle observable, verifiable, and morally significant. Accordingly, an apparently content-neutral regulation, whose true purpose is to repress certain expressions due to their content, is subject to Track One analysis.160 What legislative and administrative acts should be subject to a Track Two examination? As forcefully demonstrated by Larry Alexander, practically any legal norm or governmental action has some effect on the expression or reception of some information by some people.161 Track Two thus arguably includes not only restrictions on the use of amplifying devices in residential neighborhoods or the burning of draft cards. It includes income tax on composers and broadcasting stations, which reduces the incentives to produce and broadcast music; speed limitations denying people the experience of driving faster; and the entire body of property, contract, and tort law, as any law may affect who says what to whom, and with what effect. Presumably, the permissibility of each and every legislation and administrative action should therefore be subject to a moderate deontological threshold.162 This conclusion is however unacceptable from both a moral and a legal perspective. From a moral point of view, it would imply an implausible expansion of the deontological constraint against curtailing free speech. Such a constraint does not apply when those remote, indirect effects have not occurred to the governmental decision-maker, and plausibly not even

158. Harry Kalven, Jr., The Concept of the Public Forum: Cox v. Louisiana, 1965 Sup. Ct. Rev. 1, 28. See also Tribe, supra note 2, at 791. 159. An alternative, instrumental explanation for putting a thumb on the scale (in both Track One and Track Two cases) is that the vague nature of the benefits of speech, as compared to its direct costs, justifies a “strategy of forward defense.” Posner, Frontiers, supra note 34, at 81–82, 84. 160. Tribe, supra note 2, at 794–804. 161. Alexander, supra note 50, at 13–19. See also Posner, supra note 36, at 743–44. 162. Cf. Alexander, supra note 50, at 20–37.

217

218


when they have occurred to her. From a legal policy perspective, examining the constitutionality of every legal norm or administrative action based on its effect on free speech would be unreasonable. Both morally and legally, the only cases to which Track Two analysis should apply are those that have actually been subject to such analysis: truly content-neutral restrictions on the time, place, or manner of expressive activities. It may be added that, although it is sometimes difficult to draw the exact line of demarcation between those cases to which a Track Two threshold function should apply and those to which it should not, this difficulty is not significant. As explained above, the smaller the adverse effect of regulation on free speech, the lower the pertinent threshold. When the adverse effect of a Track Two regulation on free speech is indirect, remote, and hardly noticeable, the appropriate threshold would thus be correspondingly trivial. (d) Different Thresholds for Different Categories of Speech As already described, standard CBA of speech regulation does not distinguish between different categories of speech: political, commercial, artistic, scientific, and so forth.163 Rather than assuming that certain types of speech— such as political or academic—are necessarily more socially valuable than others—such as commercial or pornographic—the costs and benefits of each speech should be specifically assessed. To the extent that the law forbids regulators to engage in a case-by-case evaluation of the expected benefit of each expression, this is, according to standard CBA, only because regulators cannot be trusted to accurately and impartially make such evaluations and because such assessments are costly. Inasmuch as the law affords greater protection to certain types of speech than to others, this differentiation arguably rests on economic grounds. Commercial speech deserves less protection because unlike other speakers, the commercial speaker “is expected to recoup the full economic value of his speech.” In contrast, other types, such as political speech, may have considerable positive externalities, and thus in the absence of legal protection and encouragement, are likely to be underproduced.164

163. See supra p. 185. 164. Posner, Frontiers, supra note 34, at 73–77, 85–86; Posner, Free Speech, supra note 34, at 19–24.

freedom of speech

This position is hardly compatible with prevailing moral intuitions and extant legal doctrine.165 Two distinctive features of freedom of speech law in the United States and elsewhere are a commitment to evaluative neutrality and the application of different standards of judicial review to different categories of speech. Political speech pertaining to governmental issues, as well as ordinary literature and art (even containing sexually explicit material), enjoy a high degree of protection. Commercial speech, in contrast, enjoys lesser protection, and obscenity is hardly protected at all.166 Grounding evaluative neutrality solely in regulators’ incompetence and partiality is hardly convincing. Even if one does not trust the judgment of regulators, there are normative judgments—such as aversion to racism and to illegal violence—to which legal systems are committed regardless of the judgment of any particular regulator. If speech advocating racial hatred or lawless action deserves any protection at all, it cannot plausibly be due to its expected positive externalities. A competing rationale for both evaluative neutrality and treating different categories of speech differently is based on deontological morality. Deontologically motivated protection of free speech does not rest exclusively on the positive outcomes of free speech. At least to some extent, it rests on equal respect for people’s autonomy and dignity as speakers and listeners. Thus, the decisive criterion need not be the benefit expected from any speech but rather the harm to people’s autonomy and dignity inflicted by silencing speakers or denying listeners access to information. This conception is compatible with both evaluative neutrality and with categorization of speech. According varying degrees of protection to different political views or to different works of arts is incompatible with the notion that people deserve equal respect.167 Hence, the regulator need not—and must not—assess the value of each political speech or work of art. But for cases in which the expected costs from any speech surpass a certain threshold, speech should not be suppressed even if it has no benefit or has negative benefit. At the same time, it stands to reason that the harm to people’s autonomy and dignity from being silenced or denied access to information varies from one category of speech to another. A deontologist may reasonably hold, for

165. Schauer, supra note 3, at 282–96. 166. Tribe, supra note 2, at 890–928; Shaman, supra note 8; Sunstein, supra note 70, at 8–11, 121–65. 167. See supra p. 192.

219

220


example, that expressing oneself through art, literature, and politics is more important to one’s autonomy than expressing oneself through commercial speech. Similarly, one’s autonomy is harmed to a greater extent if one is denied access to political arguments or works of art than if one is denied access to commercial speech or obscenity.168 These judgments are not uncontestable. One may argue that commercial speech and pornography are not less important to the speaker’s autonomy, dignity, and self-realization than artistic or political expressions.169 We nevertheless find this argument quite compelling, compared to the alternatives. Inasmuch as one focuses on the speaker’s dignity (rather than on the listener’s autonomy), this analysis implies that the threshold may be lower whenever the speaker is not a human being but an organization or a corporation (or, even that in this case, there is no deontological constraint against speech regulation).170 A deontologist need not argue that these differences constitute the only rationale for the dissimilar treatment of different types of speech.171 There may be additional reasons for such differentiation, related to the various goods promoted by free speech. Since, however, the stringency of the constraint against suppressing speech (that is, the size of its threshold) should be a function of the harm inflicted by such suppression,172 it makes sense to provide greater protection to expressions and information in those spheres of life where suppression is expected to be more harmful to people’s autonomy and dignity. Importantly, this approach does not require differentiating between different expressions or different categories of speech according

168. Baker, Human Liberty, supra note 73, at 194–224 (“[Commercial speech] lacks the crucial connections with individual liberty and self-realization that are central to justifications for the constitutional protection of speech.” Id. at 196). 169. See, e.g., Redish, supra note 65, at 60–68 (arguing that commercial speech significantly contributes to listeners’ autonomy and thus deserves constitutional protection), 68–76 (criticizing the exclusion of obscenity from the ambit of constitutional protection); Leslie Green, Pornographies, 8 J. Pol. Phil. 27 (2000) (arguing that homosexual pornography significantly contributes to homosexuals’ self-realization and autonomy). 170. Shiner, supra note 82, at 163–91. Shiner also argues (id. at 192–38) that constitutional protection of corporations’ commercial speech can neither be derived from hearers’ rights. For different positions on whether corporations should have a constitutionally protected right of free speech in political matters and on the scope of this right, see First National Bank of Boston v. Bellotti, 435 U.S. 765 (1978). 171. On the difficulties facing a justification of the categorization of types of speech on the sole ground of their different contribution to speakers’ or listeners’ autonomy, see Sunstein, supra note 70, at 137–44. 172. Supra pp. 211–13.

freedom of speech

to their social value—a differentiation that is arguably at odds with the basic premises of freedom of speech.173 This line of reasoning implies that in deciding whether to curtail an expression, rather than assessing the social benefit of each expression, or the benefit from categories of expression, a policy-maker should focus on the speech’s expected harm. Only if the expected harm is large enough to meet the threshold (H > K)—a threshold whose size varies from one category of speech to another based primarily on the contribution of each category to people’s autonomy—may a speech be justifiably suppressed. Within the context of deontologically constrained CBA, this means that the policy-maker may focus on the speech’s costs while disregarding its expected benefits (thus, so to speak, replacing CBA with CA). Instead of examining whether the net benefit of suppressing a certain speech (that is, the benefit of suppressing the speech minus the benefit of the speech itself) meets the threshold, one should examine whether its benefit (that is, the benefit of suppressing the speech) meets the threshold. Such modification also has an important methodological advantage. Standard CBA, as well as deontologically constrained CBA, ordinarily assesses costs and benefits by aggregating people’s preferences. This method is particularly problematic in the present context, because it implies that unpopular views would get lesser protection than popular ones, since less people value them, and some actually detest them. This implication contradicts the very notion of freedom of speech. Most probably, given the great demand for pornography, using aggregation of preferences to assess the value of speech would also result in granting pornography particularly strong protection, compared to poetry, for example. (e) Summary Following the discussion of the relevant benefits that should affect the permissibility of suppressing speech (in the previous subsection), this subsection examined the factors that should determine the size and shape of the threshold. We conclude that different thresholds should apply to different speech regulations according to their basis (intended or unintended) and

173. See, e.g., Steven H. Shiffrin, The First Amendment, Democracy, and Romance 39–44 (1990); Kenneth L. Karst, Equality as a Central Principle in the First Amendment, 43 U. Chi. L. Rev. 20, 30–35 (1975).

221

222


according to the extent to which they adversely affect the free expression and reception of ideas and information. Different thresholds should apply to different categories of speech, primarily on the basis of each category’s contribution to people’s autonomy and dignity. Within each category, evaluative neutrality is mandated by (among other things) the duty to treat people fairly and with equal respect. These are but preliminary, tentative observations on how threshold functions could and should be constructed. A detailed normative analysis, based on extensive inquiry into the exact circumstances of different areas of expressive activities in different societies, is required to determine the incidence and exact shape and stringency of the appropriate threshold functions.

3. Choosing Among Permissible Courses of Action In our discussion of chronologically remote harms, we mentioned that a common justification for the imminent or present danger requirement is that future harms may be avoided by persuasion, education, prevention, and other means that do not entail the curtailment of free speech. 174 Whether or not this argument justifies the imminence requirement (and whether one endorses this requirement at all), it is important to position this argument within the broader framework of the consequentialism/deontology debate. Within standard CBA, all costs are taken into account, and the choice between silencing dangerous speech and combating it by more speech or by other means depends on the cost effectiveness of each means. If the expected harms of a certain speech are greater than its expected benefits, and if—all things considered—it is cheaper to prevent these harms by silencing the speech than by providing counterarguments or taking preventive measures, then suppression is warranted. Deontology, in contrast, may require that (1) some of the costs and benefits of the speech will be ignored in judging the permissibility of its suppression, (2) the (net) benefit of suppressing the speech will meet a certain threshold, and (3) the choice between an infringing act that meets the threshold and a noninfringing act (or an act infringing another constraint) will reflect the lexical priority of some values over others. In chapter 4, we illustrated this

174. Supra p. 198.

freedom of speech

point using a hypothetical choice between saving a sufficiently large number of lives by deliberately killing one person and saving the same number of people by spending $10,000,000.175 While moderate deontology need not hold that killing the one person is impermissible whatever the alternative monetary cost, it may reasonably hold that the existence of the alternative of attaining the same benefit without infringing any constraint renders an otherwise permissible infringement impermissible. Similarly, whenever “more speech” is expected to effectively counteract a harmful speech, its availability may render an otherwise justifiable suppression of speech impermissible because “more speech” infringes no constraint. Another example is speech that is expected to bring about imminent spectator violence. The prevention of impending disorder and violence by a hostile audience angered by a speech may sometimes justify the silencing of otherwise protected speech. However, such silencing would be impermissible if violence and disorder may be prevented by other means, such as reasonably employing police powers to control the audience and maintain order.176 Finally, assuming—as we do—that content-neutral restrictions on expression infringe a deontological constraint (with a relatively low threshold), even restrictions whose net benefit meets the threshold may be deemed impermissible if the good produced by the restrictions may be obtained by other means. Thus, while the costs of certain expressive activities in terms of litter, noise, and traffic congestion may warrant limitations of free speech in the absence of alternative ways to cope with these adverse effects, such limitations are unacceptable if the costs of controlling traffic or cleaning up handbills are not unreasonable. As demonstrated elsewhere in the book, this notion of lexical priority between infringing and noninfringing acts and between acts infringing different constraints, may either be considered separately (after determining the permissibility of infringing acts using a threshold function that disregards alternative measures) or using a threshold function that examines the marginal net benefit of the infringing act, given alternative courses of action.177

175. See supra pp. 81–83. 176. Note, Hostile-Audience Confrontations: Police Conduct and First Amendment Rights, 75 Mich. L. Rev. 180 (1976). See also David G. Barnum, Freedom of Assembly and the Hostile Audience in Anglo-American Law, 29 Am. J. Comp. L. 59 (1981). 177. See supra pp. 149–50, 155–56, and 169–70; infra pp. 346–47.

223

224


F. Conclusion

Constitutional protection of freedom of speech enhances social welfare in various ways. Neither ordinary morality nor prevailing legal norms, however, permits curtailing free speech merely because such curtailment would enhance social welfare even further. There is a moral and legal constraint on suppressing speech. A consequentialist analysis of free speech thus misses an important aspect of the issue. The analytical framework offered in this chapter highlighted this aspect. Due to the vast complexity of free speech doctrine, this chapter did not provide a comprehensive analysis of the topic. The discussion did not even exhaust the pertinent substantive dilemmas and methodological choices involved in constructing threshold functions to determine the permissibility of speech regulation. Hopefully, the discussion of some of these dilemmas and choices demonstrated the fruitfulness of a deontologically constrained economic analysis of the issue. This discussion revealed that while a purely consequentialist approach is likely to yield underprotection of freedom of speech, our framework need not result in more stringent protection of free speech than is currently provided in any liberal democracy. Deontologically constrained CBA of free speech leaves much room for normative deliberation, which may lead to different conclusions, taking into account the truly important factors and their interrelations.

•

eight

Antidiscrimination Laws

•

A. Introduction

in recent decades, liberal democracies have adopted a growing number

of prohibitions—enforced by criminal sanctions, civil remedies, or both—on various discriminatory practices of private actors. An important landmark in this respect was the entrance into force in 1969 of the International Convention on the Elimination of All Forms of Racial Discrimination, which requires member states to take measures to prevent discrimination by private actors.1 In the United States, federal laws prohibiting discrimination by private actors were enacted prior to the Convention. For example, Title VII of the Civil Rights Act of 1964 proscribes discrimination in the labor market on the basis of race, color, religion, sex, or national origin. The Equal Pay Act of 1963 prohibits discrimination between employees on the basis of sex. Age discrimination is prohibited by the Employment Act of 1967, and the Americans with Disabilities Act of 1990 (ADA) prohibits discrimination on grounds of physical or mental limitations. Similarly, in 2000, the European Community enacted two Directives (pursuant to Article 13 of the Treaty of Amsterdam of 1997) aimed at protecting people from discrimination on grounds of race, ethnic origin, religion or belief, disability, age, or sexual orientation.2 Discrimination is prohibited in other spheres as well, including, in the United States, publicly funded education programs (Title VI of the

1. Article 3(1) of the International Convention on the Elimination of All Forms of Racial Discrimination, available at http://www.unhchr.ch/html/menu3/b/d_icerd.htm. 2. Council Directive 2000/43/EC of 29 June 2000, Implementing the Principle of Equal Treatment Between Persons Irrespective of Racial or Ethnic Origin; Council Directive 2000/78/EC of 27 November 2000, Establishing a General Framework for Equal Treatment in Employment and Occupation. These provisions add to the prohibition on sex discrimination that was set forth in the Treaty of Rome in 1957, Article 119. All EU Member States were due to have transposed these Directives into national laws by 2003. This process has not been uniformly applied, and in 2007, the European Community has initiated infringement

225

226


Civil Rights Act), housing (the Fair Housing Act of 1968), and lending (the Equal Credit Opportunity Act of 1974).3 While the arguments in this chapter do apply to the latter areas, we will demonstrate our thesis mainly with regard to employment antidiscrimination laws. Employment antidiscrimination laws often prevent employers from expressing their preferences toward members of certain groups, limiting their freedom to run their business as they see fit. These norms also proscribe employers from applying certain profit-maximizing business practices and even require them to take active, often costly, measures to “accommodate” the needs of certain (actual and potential) employees. Some argue that antidiscrimination laws constitute an invasion of what Richard Epstein labels as “forbidden grounds,”4 violating the private-public distinction which excludes the private sphere from the domain of legitimate state intervention. Milton Friedman goes so far as to argue that antidiscrimination laws are “similar in principle [to] Hitler Nuremberg Laws,”5 sacrificing liberty in the name of what he sees as the majority’s preference for equality. The enactment of antidiscrimination laws clearly rejected this position, but it did not render the debate moot. The ongoing interpretation and application of these laws continue to reflect the debate concerning their legitimacy. The extensive academic discussion on the legitimate scope of antidiscrimination laws includes two central schools of thought: economic analysis and deontological inquiry. As in other areas discussed in this book, we explore here the feasibility and desirability of integrating these two modes of inquiry. Standard economic analysis is based upon the implicit normative baseline that prohibiting a discriminatory practice may be justified only if the relevant practice is inefficient. As such, it does not assign intrinsic value to individual freedom, autonomy, human dignity, or equality. Attempts to consider these values in roundabout ways result in confusion and indeterminacy.

procedures against the noncomplying states. See http://ec.europa.eu/employment_social/ fundamental_rights/legis/lginfringe_en.htm. 3. In the European Union, while the Framework Directive that prohibits discrimination on wide range of grounds applies only to employment, the Race Directive applies to various spheres of activity, including social security, health care, education, housing, and the supply of goods and services which are available to the public. 4. Richard A. Epstein, Forbidden Grounds: The Case Against Employment Discrimination Laws (1992). 5. Milton Friedman, Capitalism and Freedom 113 (1962).

antidiscrimination laws

The indeterminacy is partly due to disagreements regarding what constitutes social welfare: should the satisfaction of racist or otherwise objectionable preferences be taken into account in calculating social welfare (so-called “laundering preferences”), and should the preferences of people who neither suffer directly from discrimination nor engage in discriminatory practices (“disinterested preferences”) be taken into account. In fact, one may get the impression that the methodological choices regarding the measure of social welfare occasionally reflect a “pick and choose” strategy, as if the normative economic analysis is used to substantiate predefined, nonconsequentialist convictions about liberty and equality. Deontological analysis of antidiscrimination laws maintains that discrimination—treating people less favorably on the basis of such characteristics as race or gender—is inherently wrong. Accordingly, treating people equally is not merely “a majority’s preference,” as suggested by Friedman, but a moral duty. Assessment of any antidiscrimination legislation requires, first, to prioritize between two conflicting constraints. On the one hand, there is the constraint against harming people by discriminating against them, thereby diminishing their autonomy, offending their dignity, and depriving them of various material and nonmaterial goods. On the other hand, there is the constraint against limiting people’s freedom, including the freedom to choose with whom to interact and contract. When the latter constraint trumps the former—such as in the sphere of intimate relationships— proscribing discrimination is unjustified. Where the constraint against discrimination takes priority—such as, plausibly, in the labor market—an additional question arises, namely whether the costs involved in securing equal treatment are high enough to override the constraint. This chapter proposes an analytical framework for addressing this last question. Clearly, employment discrimination is an extremely complex issue. We do not purport to resolve all the intricate normative questions it raises. In particular, while we shall discuss the relationships between the deontological constraint against discrimination and attributing intrinsic value to equality as an element of a theory of the good, we shall not analyze the latter in detail. Our modest aim is to point out the inherent difficulties of applying unconstrained CBA to this issue and to demonstrate how deontologically constrained CBA may contribute to understanding the pertinent dilemmas. We suggest that such analysis not only better captures the relevant normative questions but also coheres with current legal doctrines. The reminder of this chapter is comprised of four parts. Section B sets the stage for the theoretical discussion by providing a brief, general description

227

228


and typology of antidiscrimination laws. Section C then utilizes positive economic analysis to identify possible reasons for the continued existence of discriminatory practices in competitive markets. Section D critically discusses the standard, unconstrained economic analysis of antidiscrimination laws. Section E then incorporates deontological constraints into the economic analysis. This section delineates the constraint against discrimination and its proper scope given conflicting constraints. It then examines the relationships between this constraint and the distributive goals of antidiscrimination norms. Finally, it demonstrates how current legal norms actually incorporate threshold constraints. Following much of the economic analysis of antidiscrimination laws, this chapter does not formalize the various arguments mathematically.

•

B. Current Legal Norms

The basic antidiscrimination requirement is that businesses and employers refrain from acting on the basis of racist, sexist, and other irrational sentiments based on stereotypes of certain groups. Often, antidiscrimination laws go further and proscribe also “rational” discriminatory practices, that is, practices that would increase the business profits. Sometimes, antidiscrimination laws not only ban intentional discrimination but go one step further and require employers to avoid a seemingly neutral policy that results in an unequal outcome. The law imposes requirements akin to affirmative action, requiring employers to bear the costs of integrating members of particular groups. These requirements include avoiding business practices that cause a “disparate impact” on certain groups,6 making workplace accommodations to enable individuals to qualify for the job,7 and ensuring that members of

6. See, e.g., 42 U.S.C. § 2000e-2(k); Griggs v. Duke Power Co., 401 U.S. 424, 432 (1971) (finding that “good intent or absence of discriminatory intent does not redeem employment procedures or testing mechanisms that operate as ‘built-in headwinds’ for minority groups and are unrelated to measuring job capability”); Elizabeth Bartholet, Application of Title VII to Jobs in High Places, 95 Harv. L. Rev. 945, 958–59 (1982); Reva B. Siegel, Discrimination in the Eyes of the Law: How “Color Blindness” Discourse Disrupts and Rationalizes Social Stratification, 88 Cal. L. Rev. 77, 95–96, 101–02 (2000). 7. In the United States, these include the provision set forth in Title I of the ADA, requiring employers to make “reasonable accommodations to the known physical or mental limitations of an otherwise qualified individual with a disability,” and Title VII of the Civil Rights Act’s protection of religion, including an implicit duty to accommodate an employee’s religious observance or practice. See, e.g., Stewart J. Schwab & Steven L. Willborn, Reasonable


certain groups are “fairly represented” in the relevant workforce.8 These requirements may be seen as a cure to (past or present) hidden or even unconscious forms of discrimination against some groups, or as a response to illegitimate favoritism of other groups. However, they apply irrespective of actual proof of such hidden or unconscious discrimination. Some judges and legal scholars have therefore argued that the accommodation requirement extends beyond the antidiscrimination principle.9 Others see it as fundamentally similar to the classic discrimination models.10 Antidiscrimination laws limit and qualify the prohibitions on discrimination in various ways. In the United States, the federal employment antidiscrimination laws apply only to firms employing at least fifteen workers.11

Accommodation of Workplace Disabilities, 44 Wm. & Mary L. Rev. 1197 (2003); Michael A. Stein, The Law and Economics of Disability Accommodations, 53 Duke L.J. 79 (2003). 8. Title VII of the Civil Rights Act explicitly excludes such a requirement by stating that “[n]othing contained in this subchapter shall be interpreted to require any employer . . . to grant preferential treatment to any individual or to any group because of the race, color, religion, sex, or national origin of such individual or group on account of an imbalance which may exist with respect to the total number or percentage of persons of [such group] employed by any employer . . .” (42 U.S.C. § 2000e-2(j)). However, such a requirement is applied in other countries, such as Israel (Equality to Persons with Disabilities Act, 1998) and Northern Ireland (under the Fair Employment Act, 1989, employers are required, in certain conditions, to secure fair participation in employment by members of the Protestant or the Roman Catholic communities. See Catherine Barnard, EC Employment Law 336 (3d ed. 2006)). 9. See, e.g., Erickson v. Board of Governors, 207 F.3d 945 (7th Cir. 2000), cert. denied, 531 U.S. 1190 (2001) (Judge Easterbrook holding that Title I of the ADA is not within Congress’s power under Section 5 of the Fourteenth Amendment, since it “requires employers to consider and to accommodate disabilities, and in the process extends beyond the antidiscrimination principle”). See also Stewart & Willborn, supra note 7; Mark Kelman, Market Discrimination and Groups, 53 Stan. L. Rev. 833 (2001); Pamela S. Karlan & George Rutherglen, Disabilities, Discrimination, and Reasonable Accommodation, 46 Duke L.J. 1 (1996). 10. Michael Ashley Stein, Same Struggle, Different Difference: ADA Accommodations as Antidiscrimination, 153 U. Pa. L. Rev. 579 (2004); Samuel R. Bagenstos, “Rational Discrimination,” Accommodation, and the Politics of (Disability) Civil Rights, 89 Va. L. Rev. 825 (2003); Christine Jolls, Antidiscrimination and Accommodation, 115 Harv. L. Rev. 642 (2001). See also infra pp. 246–51. 11. 42 U.S.C. § 2000e(b); 42 U.S.C. § 12111. On these provisions see generally Jeffrey A. Mandell, The Procedural Posture of Minimum Employee Thresholds in Federal Antidiscrimination Statutes, 72 U. Chi. L. Rev. 1047, 1060–67 (2005). The U.S. Supreme Court decided, however, that Title VII also applies to partnerships and governs their decisions of whom to include as partners. Hishon v. King & Spalding, 467 U.S. 69 (1984). The EC Directives do not include exemption based on a minimum number of employees. Moreover, the European Court of Justice held that even an exemption for employers who do not employ more than five workers is inconsistent with the European Directive on sex discrimination.

229

230


Some exemptions are granted to certain types of organizations, most notably religious institutions12 and “bona fide” private membership clubs.13 Then there are limits to the burden a business is expected to bear in meeting the antidiscrimination requirements. For instance, in the United States, Title VII of the Civil Rights Act sets three types of such limits. First, direct discrimination on the basis of religion, sex, or national origin (but not race) is permitted when such characteristic “is a bona fide occupational qualification reasonably necessary to the normal operation of that particular business or enterprise;”14 Second, an employment practice that causes a disparate impact on the basis of race, color, religion, sex, or national origin is legitimate if the practice is “job related for the position in question and consistent with business necessity.”15 Third, the employer is required to “reasonably accommodate” an employee’s religious observance and practice, as far as the required accommodation does not impose “undue hardship on the conduct of the employer’s business.”16 The last exemption is also applied under the ADA—an employer is exempt from the duty to make “reasonable accommodations to the known physical or mental limitations of an otherwise qualified individual with a disability” if making it “impose[s] an undue hardship on the operation

See Case 165/82 Commission of the European Communities v. United Kingdom of Great Britain and Northern Ireland, [1983] ECR 3431. 12. See 42 U.S.C. § 2000e-1(a) (“This subchapter shall not apply to . . . a religious corporation, association, educational institution, or society with respect to the employment of individuals of a particular religion to perform work connected with the carrying on by such corporation, association, educational institution, or society of its activities”); 42 U.S.C. § 2000e-2(e); Corp. of the Presiding Bishop of the Church of Jesus Christ v. Amos, 483 U.S. 327 (1987). 13. 42 U.S.C. § 2000e(b). See also EEOC v. Chicago Club, 86 F.3d 1423 (7th Cir. 1996). 14. 42 U.S.C. § 2000e-2(e)(1) (emphasis added). See also 1 Charles A. Sullivan Et Al., Employment Discrimination: Law And Practice § 3.05[A] (3d ed. 2002); Michael J. Frank, Justifiable Discrimination in the News and Entertainment Industries: Does Title VII Need a Race or Color BFOQ?, 35 U.S.F. L. Rev. 473 (2001); William R. Bryant, Justifiable Discrimination: The Need for a Statutory Bona Fide Occupational Qualification Defense for Race Discrimination, 33 Ga. L. Rev. 211 (1998). 15. 42 U.S.C. § 2000e-2(k)(1)(A)(i) (emphasis added). The Equal Pay Act permits employers to pay a person wages that are different than those that are paid to employees of the opposite sex for equal work, “where such payment is made pursuant to (i) a seniority system; (ii) a merit system; (iii) a system which measures earnings by quantity or quality of production; or (iv) a differential based on any other factor other than sex.” 29 U.S.C. § 206(d). 16. 42 U.S.C. § 2000e-(j) (emphasis added). See also Karen Engle, The Persistence of Neutrality: The Failure of the Religious Accommodation Provision to Redeem Title VII, 76 Tex. L. Rev. 317 (1997).


of the business.”17 Other legal systems include similar provisions in their antidiscrimination laws.18

•

C. Motivations for Discrimination

To assess the desirability of antidiscrimination legislation, one should first understand the underlying motivations for discriminatory practices. At first glance, discrimination in the marketplace conflicts with the goal of profit maximization and is therefore irrational. However, direct discrimination can also be considered a rational response to employees’ or customers’ preferences, or alternatively, as a rational reliance on statistical correlation between membership in certain groups and relevant economic factors. In the following lines, we briefly present these three possible rationales for discrimination.19 Discriminatory Preferences of the Business Owner In his influential 1957 book, The Economics of Discrimination, Gary Becker presented a positive economic analysis of workplace discrimination based

17. 42 U.S.C. § 12112(b)(5)(A) (emphasis added). The ADA defines “undue hardship” as “an action requiring significant difficulty or expense, when considered in light of the factors set forth in subparagraph (B)” (42 U.S.C. § 12111(10)(A)). These factors include “the effect on expenses and resources, or the impact otherwise of such accommodation upon the operation of the facility; . . . the type of operation or operations of the covered entity, including the composition, structure, and functions of the workforce of such entity; the geographic separateness, administrative, or fiscal relationship of the facility or facilities in question to the covered entity.” 18. For instance, under the EC Directive Establishing a General Framework for Equal Treatment in Employment and Occupation (supra note 2), “an apparently neutral provision, criterion or practice” which have a disparate impact on members of a protected group is nevertheless permitted if it is “objectively justified by a legitimate aim and the means of achieving that aim are appropriate and necessary” (Article 1(2)(b)). See also Evelyn Ellis, EU Anti-discrimination Law (2005); Mark Bell, Anti-discrimination Law and the European Union (2002). The duty of reasonable accommodation to people with disabilities requires employers to take “appropriate measures, where needed . . . to enable a person with a disability to have access to, participate in, or advance in employment . . . unless such measures would impose a disproportionate burden on the employer” (id. Article 5). The question of whether direct, intentional discrimination can be similarly justified is debated. See, e.g., Barnard, supra note 8, at 322–23. 19. For an overview of economic theories of discrimination, see John J. Donohue III, The Law and Economics of Antidiscrimination Law, in 2 Handbook of Law and Economics 1387 (A.M. Polinsky & Steven Shavell eds., 2007); Jacob E. Gersen, Markets and Discrimination, 82 N.Y.U. L. Rev. 689, 696–713 (2007).

231

232


on the presumption that employers who discriminate are motivated by their animus to those who are discriminated against, as a manifestation of prejudice, racism, or chauvinism.20 Becker pointed out that an employer who hires a disfavored worker incurs not only monetary costs (the worker’s wage) but also a psychic penalty. As a result, such an employer is only willing to pay a disfavored worker a wage which is lower than the worker’s marginal productivity. This psychic “tax” lowers both the demand for disfavored workers and their earnings. This explanation implicitly assumes that the relevant market is not “contestable,”21 that is, it rules out the possibility of entry of new employers, who do not act on the basis of discriminatory preferences against the disfavored group of workers.22 Subsequent scholars pointed out that if this assumption is relaxed, and one instead assumes that the market is contestable, animus-based discrimination must disappear, at least in the long run.23 Some economists (most notably, Milton Friedman) have argued that since the market should be expected to solve this problem, legal prohibition on discrimination is unnecessary.24 However, the empirical evidence of persistent discrimination in competitive markets (such as the enduring exclusion of blacks from entire industries in southern states prior to the legislation of the Civil Rights Act of 1964,25 and the wage gaps between men

20. Gary S. Becker, The Economics of Discrimination 14–17 (1957, 2d ed. 1971). 21. A “contestable market” is a market in which entry and exit are free. Thus, “even in the absence of price-taking behavior in markets with relatively few firms, perfect contestability provides an ‘invisible hand’ that guides market equilibrium to a competitive-type result [in which wage equals marginal productivity]”: Walter Nicholson, Microeconomic Theory 661 (6th ed. 1995). See also William J. Baumol et al., Contestable Markets and the Theory of Industry Structure (1982); William. J. Baumol, Contestable Markets: An Uprising in the Theory of Industry Structure, 72 Am. Econ. Rev. 1, 5 (1982); Chris Stefanadis, Sunk Costs, Contestability, and the Latent Contract Market, 12 J. Econ. & Mgmt. Strategy 119 (2003). 22. Becker made this assumption explicit in a later work: Gary Becker, Discrimination, economic, in 4 International Encyclopedia of the Social Sciences 208–10 (D.L. Sills ed., 1968) (arguing that the shortage of entrepreneurial skill prevents the elimination of the discriminating employers). An alternative reason for the survival of animus-based discrimination even in a competitive market is search costs. See Dan A. Black, Discrimination in an Equilibrium Search Model, 13 J. Lab. Econ. 309 (1995). 23. Ronald G. Ehrenberg & Robert S. Smith, Modern Labor Economics: Theory and Public Policy 407–32 (9th ed. 2006). 24. Friedman, supra note 5, at 108–15; Epstein, supra note 4, at 480–94. 25. James J. Heckman & Brook S. Payner, Determining the Impact of Federal Antidiscrimination Policy on the Economic Status of Blacks, 79 Am. Econ. Rev. 138 (1989); John J. Donohue III,


and women)26 refute this prediction. As pointed out by Kenneth Arrow, in contestable markets, Becker’s theory “predicts the absence of the phenomenon it was designed to explain.”27 This led to alternative explanations for discriminatory practices in the marketplace. Catering to Prejudiced Employees and Customers Discriminatory practices may respond to the racial or sexist preferences of the firms’ current workforce or customers.28 Prejudiced employees or customers who are required to interact with a disfavored worker or salesperson face a psychic “tax” that affects their decisions. A prejudiced employee would require a higher wage to agree to work with a disfavored coworker,29 and a buyer would only be willing to pay a lower price when purchasing the good from a disfavored salesperson.30 A related motivation may have to do with the enhancement of cooperation and harmony within the firm. Arguably, employers benefit from establishing a homogeneous work force, thus facilitating the establishment of informal norms, improving communication within the organization, and mitigating possible misunderstandings.31 Profit maximizing firms can thus be expected to either segregate workers by group characteristics or avoid hiring people based on their group membership. Unlike the employer’s animosity, this set of motivations can explain an enduring pattern of discrimination.32

The Search for Truth: In Appreciation of James J. Heckman, 27 Law & Soc. Inquiry 23 (2002). 26. For a report of a comprehensive empirical study and a discussion of others, see Gersen, supra note 19. 27. Kenneth Arrow, The Theory of Discrimination, in Discrimination in Labor Markets 3, 10 (Orley Ashenfelter & Albert Rees eds., 1973). 28. Cf. Richard H. McAdams, Cooperation and Conflict: The Economics of Group Status Production and Race Discrimination, 108 Harv. L. Rev. 1005 (1995) (suggesting that discriminatory practices may serve the interest of workers or customers in increasing the prestige and status of their group by subordinating other groups). 29. Barry R. Chiswick, Racial Discrimination in the Labor Market: A Test of Alternative Hypotheses, 81 J. Pol. Econ. 1330 (1973). 30. See, e.g., George J. Borjas & Stephen G. Bronars, Consumer Discrimination and SelfEmployment, 97 J. Pol. Econ. 581 (1989). 31. Epstein, supra note 4, at 60–69. 32. For a discussion of the effect of market structure on the prevalence of discrimination see, e.g., Major G. Coleman, Racial Discrimination in the Workplace: Does Market Structure Make a Difference?, 43 Inds. Rel. 660, 676–86 (2004); Gersen, supra note 19; Product

233

234


“Statistical Discrimination” Discriminatory business practice may reflect an actual or perceived statistical correlation between the applied classification (such as race or gender) and relevant characteristics of (potential) employees or customers. The saving in information costs may render the classification profitable, even when the correlation between the classification and the relevant characteristics is weak or questionable.33 Similarly, a firm’s choice to employ a selection procedure that has a disparate impact on members of a certain group may not aim to discriminate but rather to reduce the selection costs.34 In economic terms, dissimilar treatment that is based on the fact that on average members of one group differ from members of another group in relevant elements (such as productivity) is justified by the higher costs of a more personalized evaluation method. Note that while competition may reduce discrimination based on employer animus, it may actually enhance discrimination based on a statistical correlation between the applied classification and relevant characteristics. This positive economic analysis of discrimination in the marketplace will serve as a basis for the normative analysis in the following section.

•

D. Standard Normative Economic Analysis

Given the possible reasons just described for engaging in discriminatory practices, should these practices be proscribed? Economists committed to efficiency and to the free market tend to answer this question negatively. They assume that the only justification for state intervention is evident market failures, believe that such intervention should primarily aim at

Market Structure and Labor Market Discrimination (John S. Heywood & James H. Peoples eds., 2006). 33. See, e.g., Katherine V.W. Stone, The New Psychological Contract: Implications of the Changing Workplace for Labor and Employment Law, 48 UCLA L. Rev. 519, 599 (2001); Richard A. Posner, The Economics of Justice 362–63 (1981); Epstein, supra note 4, at 32–41, 52–53; Edmund S. Phelps, The Statistical Theory of Racism and Sexism, 62 Am. Econ. Rev. 659 (1972); Arrow, supra note 27, at 24 (noting that “[s]kin color and sex are cheap sources of information” for distinguishing between different groups of workers); Andrea Moro & Peter Norman, A General Equilibrium Model of Statistical Discrimination, 114 J. Econ. Theory 1 (2004). 34. Thomas A. Lambert, The Case Against Private Disparate Impact Suits (Environmental Racism), 34 Ga. L. Rev. 1155, 1161 (2000).


correcting those failures, and are generally skeptical of the ability of governmental intervention to improve matters. Arguably, in the absence of barriers to bargaining, the parties’ freedom of contract seems to be the best way to maximize their preference satisfaction. For example, an employer and a disabled person can be expected to contract for efficient accommodations by agreeing on a lower wage that would compensate the employer for her accommodation costs. Compelling the employer to make accommodations exceeding those she would have made otherwise results in an inefficient and hence unjustified outcome.35 In the same vein, Richard Posner argues that, were discriminatory practices inefficient, those who suffer from them would have “purchased” from the bigots their freedom to discriminate. Bigots are willing to make less profits (or earn lower wages or pay higher prices, in the case of employees and customers) in exchange for the freedom to discriminate. The endurance of discrimination hence indicates—so the argument goes—that the discriminators value their freedom to discriminate more than the members of protected groups value the entitlement not to be discriminated against. Therefore, antidiscrimination laws are inefficient.36 Scholars committed to economic efficiency, but who nonetheless find this conclusion morally unacceptable, try to demonstrate that antidiscrimination norms are, in fact, efficient. Their arguments generally follow the strategies for defending consequentialism against the deontological critique, described in chapter 1.37 In what follows, we briefly review these arguments. Following the long-term and indirect effects strategy discussed in subsection 1.D.(1), one line of argument points to the incompleteness of the analysis purporting to establish the efficiency of discriminatory practices and the inefficiency of prohibiting them. The claim that the parties can bargain to avoid inefficient discrimination ignores a crucial problem of externalities,

35. See, e.g., Ron A. Vassel, The Americans with Disabilities Act: The Cost, Uncertainty and Inefficiency, 13 J.L. & Com. 397, 406–10 (1994); Mark A. Schuman, The Wheelchair Ramp to Serfdom: The Americans with Disabilities Act, Liberty, and Markets, 10 St. John’s J. Legal Comment. 495, 508–09 (1995); Epstein, supra note 4, at 59–78. 36. Posner, supra note 33, at 351–63; Richard A. Posner, The Efficiency and the Efficacy of Title VII, 136 U. Pa. L. Rev. 513 (1987). 37. See supra pp. 21–33. These arguments do not establish that discrimination practices are necessarily inefficient. They only aim to refute the contention that such practices that survive in competitive markets are inevitably efficient, and thus call for a detailed CBA in each context.

235

236


i.e., costs that are not internalized through bargaining due to transaction costs or other market failures. These costs may well exceed the practice’s benefits.38 Those who suffer from discrimination cannot be expected to purchase the bigots’ freedom to discriminate because the benefits of such a bargain have the characteristics of a public good: transaction costs in coordinating employees to make such a bargain might prevent the attainment of the efficient outcome.39 In addition, the very act of bribing sexist or racist firms might “undermine the self-esteem that is necessary to make the move . . . welfare-enhancing.”40 Moreover, employment discrimination undermines the ex ante incentives for investment in human capital by would-be workers who perceive that they would be treated as average members of their group rather than according to their specific qualities. Consequently, the stereotypes become a self-fulfilling prophecy.41 In the same vein, commentators have indicated various positive externalities of employing persons with disabilities.42 At the same time, antidiscrimination laws also have various adverse long-term and indirect effects that should be taken into account. Some scholars argue that the laws actually decrease the incentive to employ disabled people and members of other protected groups,43 and consequently

38. See, e.g., John J. Donohue III, Advocacy Versus Analysis in Assessing Employment Discrimination, 44 Stan. L. Rev. 1583 (1992); J. Hoult Verkerke, Free to Search, 105 Harv. L. Rev. 2080, 2086 (1992); Edward M. Iacobucci, Antidiscrimination and Affirmative Action Policies: Economic Efficiency and the Constitution, 36 Osgoode Hall L.J. 293, 300–01 (1998). 39. John J. Donohue III, Prohibiting Sex Discrimination in the Workplace: An Economic Perspective, 56 U. Chi. L. Rev. 1337, 1352 (1989). 40. Id. at 1352. Indeed, the ADA and Title VII, as well as other rigidities in the labor market, make it illegal or impractical for employees to bargain by accepting lower wages. 41. David A. Strauss, The Law and Economics of Racial Discrimination in Employment, 79 Geo. L.J. 1619, 1640 (1991); Stein, supra note 7, at 155–61; Stewart J. Schwab, Is Statistical Discrimination Efficient?, 76 Am. Econ. Rev. 228 (1986); Jeffery G. MacIntosh, Employment Discrimination: An Economic Perspective, 19 Ottawa L. Rev. 275 (1987). Stereotypes may also adversely affect the willingness of members of minority group to invest resources in negotiations. See Ian Ayres, Fair Driving: Gender and Race Discrimination in Retail Car Negotiations, 104 Harv. L. Rev. 817 (1991). 42. See, e.g., Stein, supra note 7, at 104–08 (outlining several positive externalities such as public cost savings, including the reduction of disability-related public assistance obligations); Samuel R. Bagenstos, The Americans with Disabilities Act as Welfare Reform, 44 Wm. & Mary L. Rev. 921 (2003) (suggesting that the ADA may save welfare payments); J. Hoult Verkerke, Is the ADA Efficient?, 50 UCLA L. Rev. 903 (2003). 43. See, e.g., Vande Zande v. Wis. Dep’t of Admin., 44 F.3d 538, 545 (7th Cir. 1995) (Posner, J.) (noting that placing the employer that already invested in accommodating a disabled worker under any further obligation would ultimately “hurt rather than help disabled workers”).


reduce the group’s employment level or wages.44 In particular, legislation that makes it illegal for employees with disabilities to bargain for lower wages or other employment differentials may impede the provision of workplace accommodations.45 Another line of argument—somewhat resembling the move to ruleconsequentialism discussed in subsection 1.D.(2)—has been to interpret Becker’s analysis as justifying the prohibition of employment discrimination, since its persistence is a result of obstructions to the free market. According to this argument, at least in the short-run, markets are only rarely fully “contestable,” and antidiscrimination legislation serves as a tool to mimic the outcome that would have resulted in such markets.46 A third argument—following the “Preferences for Constraints” argument discussed in subsection 1.D.(3)—is that social welfare analysis should consider individuals’ preferences for equal treatment. According to this claim, third parties suffer psychic harm merely from living in a society that tolerates discriminatory practices. For instance, John Donohue argues that for Title VII to be efficient, it is sufficient that every American is willing to pay as little as five dollars annually to live in a society that limits racial discrimination.47 Such preferences cannot be satisfied through voluntary bargaining since the benefits of such a bargain have characteristics of a public good.48 The fourth strategy for bringing the consequentialist analysis closer to commonsense morality—people’s feelings of virtue when they “do the right thing” and remorse when they do not (subsection 1.D.(4))—is particularly unhelpful in the present context: at least some instances of market discrimination stem from prevalent prejudiced sentiments of business’s owners,

44. Christine Jolls, Accommodation Mandates, 53 Stan. L. Rev. 223 (2000); Daron Acemoglu & Joshua D. Angrist, Consequences of Employment Protection? The Case of the Americans with Disabilities Act, 109 J. Pol. Econ. 915, 931 (2001); Stein, supra note 7, at 112–13; Schuman, supra note 35, at 506. But see John J. Donohue III, Understanding the Reasons for and Impact of Legislatively Mandated Benefits for Selected Workers, 53 Stan. L. Rev. 897, 909–12 (2001) (questioning parts of Jolls’s work). 45. Schwab & Willborn, supra note 7, at 1271; Amy L. Wax, Disability, Reciprocity, and “Real Efficiency”: A Unified Approach, 44 Wm. & Mary L. Rev. 1421, 1423–24 (2003) (arguing that minimum wage and equal pay legislation prevent employers from hiring and retaining workers with disabilities). 46. John J. Donohue III, Is Title VII Efficient?, 134 U. Pa. L. Rev. 1411, 1431 (1986). 47. John J. Donohue III, Further Thoughts on Employment Discrimination Legislation: A Reply to Judge Posner, 136 U. Pa. L. Rev. 523, 531 (1987). See also Donohue, supra note 39; Schwab & Willborn, supra note 7, at 1218. 48. Donohue, supra note 39, at 1352.

237

238


managers, employees, or customers. To address this difficulty, some commentators have suggested that antisocial preferences be excluded, thus resorting to the fifth strategy of replacing actual preferences with ideal preferences (subsection 1.D.(5)). Opponents of antidiscrimination norms argue that the discriminatory preferences of employers, coworkers, and customers should be included in the social welfare calculation since they are typically reflected in a significant Willingness To Pay (WTP).49 In contrast, supporters of antidiscrimination norms argue that such preferences should be ignored for several reasons. First, racist, sexist, and similar preferences are morally and socially unacceptable.50 Second, such preferences are cognitively irrational. Thus, statistical discrimination often reflects irrational stereotypes and cognitive biases. Discriminatory practices perpetuate prevailing beliefs about the attributes of subordinated groups that are hard to change. Eliminating such beliefs is especially difficult due to the cognitive bias to confirm a priori stereotypes and to insufficient interaction with members of the subordinated groups.51 It was accordingly suggested that “if we were to force some members of the currently excluded group into the market, employers would eventually learn that a given level of education implies more talent for that group.”52 It has been further argued that these preferences will be eliminated once an antidiscrimination legislation is implemented, thanks to the law’s

49. Epstein, supra note 4, at 486–88; Jason Zarin, Beyond the Bright Line: Consideration of Externalities, the Meaning of Undue Hardship, and the Allocation of the Burden of Proof under Title I of the Americans with Disabilities Act, 7 S. Cal. Interdisc. L.J. 511, 521 (1998). 50. See, e.g., William Landes, The Economics of Fair Employment Laws, 76 J. Pol. Econ. 507, 548 (1968) (arguing that “psyche losses to whites should not be deducted from benefits [generated by proscribing discrimination against blacks], because by passing a fair employment law society is saying, in effect, that the psyche income from discrimination that accrues to whites should not enter society’s social welfare function”); Stein, supra note 7, at 122; Schwab & Willborn, supra note 7, at 1216. 51. Several studies accentuated the last factor by showing that differences in the reliability of generalized information about members of different groups may result in unequal treatment to groups with equal productivity. See Dennis J. Aigner & Glen G. Cain, Statistical Theories of Discrimination in Labor Markets, 30 Indus. & Lab. Rel. Rev. 175 (1977); Gerald S. Oettinger, Statistical Discrimination and the Early Career Evolution of the Black-White Wage Gap, 14 J. Lab. Econ. 52 (1996). 52. Michael Spence, Market Signaling: Informational Transfer in Hiring and Related Screening Processes 99 (1974). See also Thomas F. Pettigrew & Linda R. Tropp, A Meta-Analytic Test of Intergroup Contact Theory, 90 J. Person. & Soc. Psych. 751 (2006); Samuel R. Bagenstos, The Structural Turn and the Limits of Antidiscrimination Law, 94 Cal. L. Rev. 1, 5–7 (2006); Anthony G. Greenwald & Linda Hamilton Krieger, Implicit Bias: Scientific Foundations, 94 Cal. L. Rev. 945 (2006); Jerry Kang, Trojan Horses of Race, 118 Harv. L. Rev.


“expressive” effect.53 If satisfaction of irrational discriminatory preferences does not enhance—but rather decreases—the discriminators’ well-being (or the well-being of their prejudiced workers or customers) then antidiscrimination laws may actually increase their well-being. At the end of the day, none of these responses are satisfactory. The market failure and externalities arguments are contingent and context dependent. If one takes into account the positive external effects of antidiscrimination measures, the negative effects should be considered as well (for instance, workplace accommodations for one worker may make coworkers worse off, objectively or at least subjectively). Extending the scope of analysis by considering long-term and indirect effects does not, therefore, necessarily alter the conclusion of the economic analysis. More fundamentally, the primary problem with the claim that the parties can bargain to avoid discrimination is not transaction costs or externalities (the fact that purchasing the right to discriminate from the bigots has the characteristics of a public good and thus faces great coordination difficulties). The primary problem is that, just as we do not expect potential victims of rape or deception to purchase from rapists and deceivers the latter’s “freedom” to engage in these activities, so one should not expect members of the discriminated-against groups to purchase bigots’ freedom to discriminate. Relying on disinterested preferences for equal treatment is equally unsatisfactory. Not only is such reliance highly manipulable, it is quite clear that in at least some societies, disinterested racist or sexist preferences are more prevalent and powerful than preferences for equal treatment. Finally, the move to ideal preferences and the argument that discriminatory preferences and prejudices should be ignored, as well as the suggestion that people’s adaptive preferences, rather than their ex ante ones, should be

1489, 1497–1535 (2005); Nilanjana Dasgupta, Implicit Ingroup Favoritism, Outgroup Favoritism, and Their Behavioral Manifestations, 17 Soc. Just. Res. 143, 147–48 (2004). 53. Amitai Etzioni, The Moral Dimension: Towards a New Economics 31–32 (1988) (arguing that preferences change as the constraints change); Schwab & Willborn, supra note 7, at 1217–18 (“there is considerable evidence that Title VII has changed existing preferences about the proper role of women and African Americans in the workplace”); Donohue, supra note 39, at 1348–55; John J. Donohue III, Discrimination in Employment, in 3 The New Palgrave Dictionary of Economics and the Law 615, 619–20 (Peter Newman ed., 1998); Aaron S. Edlin, The New Palgrave: Surveying Two Waves of Economic Analysis of Law, 2 Am. L. & Econ. Rev. 407, 419 (2000). Cf. Epstein, supra note 4, at 306 (admitting that preferences are malleable but objecting to their molding by the government).

239

240


considered—all arguably contain “a fatal circularity.”54 They rely on the effects of antidiscrimination laws and their underlying moral justification. On the whole, it seems that the various arguments discussed in this section do not bridge the gap between standard economic analysis and commonsense morality. Rather, they contribute to the indeterminacy of standard economic analysis of this issue and demonstrate the difficulties it faces due to its denial of moral constraints.55

E. Integrating Deontological Constraints with Economic Analysis •

Having exposed the limitations of standard economic analysis of employment discrimination, this section proposes to improve the analysis by incorporating a deontological constraint against discrimination into the economic analysis. Subsection 1 delineates the constraint and its appropriate incidence given the conflicting constraint against curtailing the autonomy of discriminators. Subsection 2 explores the relationships between the moral constraint against discrimination and redistributive goals, and particularly the extent to which one or both of these normative factors underlie the antidiscrimination laws described in section B. Focusing on the moral constraint against discrimination, subsection 3 then examines the integration of this constraint into economic analysis of antidiscrimination laws.

1. The Constraint Against Discrimination and Its Incidence The Constraint against Discrimination One could imagine a general moral duty to treat other people fairly or rationally, i.e., to make any decision affecting other people’s well-being on the basis of

54. Donohue, supra note 39, at 1343–44. 55. On the indeterminacy of the normative economic analysis of antidiscrimination laws see, e.g., Edward J. McCaffery, Slouching Towards Equality: Gender Discrimination, Market Efficiency, and Social Change, 103 Yale L.J. 595, 648–51 (1993) (arguing that efficiency considerations alone do not justify or oppose Title VII); Thomas A. Cunniff, Note, The Price of Equal Opportunity: The Efficiency of Title VII after Hicks, 45 Case W. Res. L. Rev. 507, 523–41, 546–48 (1995) (arguing that the efficiency of Title VII depends on the allocation of burden of proof of the empirical evidence).


all relevant factors and those factors only. Such a general duty, however, seems too demanding. It would excessively curtail people’s freedom and raise difficult questions regarding the relevance of various factors.56 For our purposes, it suffices to acknowledge a more modest duty, limited in terms of the people to which it applies, the people affected by the decisions, and the grounds on which the decision is made. We focus on possible grounds for employers’ decisions that adversely affect employees or would-be employees. A decision negatively affecting an employee or a job seeker, such as not hiring, not promoting, or firing her, infringes a constraint if it is based on a trait of that person which constitutes a prohibited ground for distinction. These traits vary. Some, like race, age, and sex, are not chosen and cannot be altered (at all, or at a reasonable cost). However, the fact that a certain trait is unchosen or unalterable does not render it forbidden ground for employment decisions. The fact that a genetically unintelligent person cannot do much about it does not make intelligence an illegitimate ground for selection.57 At the same time, the possibility of converting to a different religion does not make discrimination based on religion legitimate. Some traits (such as sexual orientation) constitute an important element of one’s identity,58 and some (such as religion) involve important commitments. Once again, however, the importance of the trait does not characterize or explain all of the forbidden grounds for discrimination. The fact that religion does not constitute an important element of some people’s identity or does not involve any commitment does not legitimize a decision not to hire these people because of their religion.59

56. See Richard J. Arneson, What is Wrongful Discrimination?, 43 San Diego L. Rev. 775, 783–85 (2006); John Gardner, On the Ground of Her Sex(uality), 18 Oxford J. L. Stud. 167, 168–69 (1998) (analyzing the problems of resting antidiscrimination norms on a general duty to act rationally, including the fact that some forms of discrimination are rational). 57. See, e.g., Bernard Boxill, Blacks and Social Justice 12–18 (1992); Mark Kelman & Gillian Lester, Jumping the Queue: An Inquiry into the Legal Treatment of Students with Learning Disabilities 185–87 (1997). 58. See Robert Wintemute, Sexual Orientation and Human Rights: The United States Constitution, The European Convention, and the Canadian Charter (1995) (discussing, among other things, the use of “fundamental choice arguments” to justify and advance sexual orientation antidiscrimination norms); Gardner, supra note 56, at 169–74. 59. See Sophia Moreau, What is Discrimination? 17–18 (March 2009) (unpublished manuscript, available at http://law.usc.edu/faculty/documents/WhatIsDiscriminationMoreau2009.pdf).

241

242


Sometimes a decision based on a prohibited ground not only deprives a person of a certain benefit, such as promotion, but also causes a psychological injury. It may injure one’s dignity and involve humiliation and degradation. But the moral constraint against employment (and marketplace) discrimination seems to be considerably broader as it is not even conditioned upon the person discriminated against being aware of the discrimination, let alone suffering a psychological injury.60 Characterization of the forbidden grounds is therefore more difficult than meets the eye.61 The notion of personal autonomy can be helpful in this regard.62 Following Joseph Raz, we assume that leading an autonomous life requires an environment in which individuals have an adequate range of valuable options and the opportunities to choose among them.63 Depriving people of such opportunities on the basis of such traits as sex, religion, and race, either by the state or by entities controlling these opportunities (including employers, landlords, and shopkeepers) constitutes a forbidden harm.64 If every individual has “a right to a certain set of important freedoms of contract,” including the right to have access to jobs, goods, and services, then employers and providers of such goods and services have a duty to reasonably accommodate these rights.65 Traits such as race and sex must not affect the freedom of contract of their bearers.66 Identifying the forbidden grounds is not done in a vacuum. It depends on social, historical, and cultural factors. The fact that for many generations, certain groups have been consistently discriminated against in a vast array

60. Cf. John Gardner, Liberals and Unlawful Discrimination, 9 Oxford J. L. Stud. 1, 6 (1989). 61. See also Arneson, supra note 56, at 796 (pointing out that the “various suspect classifications pose radically separate and distinct questions of justice that require remedies specifically attuned to each type”). 62. Autonomy is a notoriously complex concept. For a critical survey of its various meanings, see generally John Christman, Autonomy in Moral and Political Philosophy, in Stanford Encyclopedia of Philosophy (2003), http://plato.stanford.edu/entries/autonomy-moral. 63. Joseph Raz, The Morality of Freedom 369–95 (1986). 64. Gardner, supra note 60, at 17–21. 65. Moreau, supra note 59, at 14–25. 66. In addition, depriving people of employment opportunities on the basis of sex, color, or sexual orientation often threatens their pride in their identity, which is also essential for leading an autonomous life. See Raz, supra note 63, at 254; John Gardner, Private Activities and Personal Autonomy: At the Margins of Anti-Discrimination Law, in Discrimination and the Limits of Law, 148, 155 (Bob Hepple & Erika M. Szyszczak eds., 1992).


of contexts makes the continued discrimination against members of these groups particularly harmful to their autonomy. In that sense, an irrational decision not to hire a woman or a black person is very different from a capricious decision not to hire someone because his name begins with the letter E. The latter decision is indeed irrational, but its effect on that person’s autonomy is typically negligible.67 According to this understanding of the constraint against discrimination, an infringement of the constraint does not require a specific mental state or motivation on the part of the discriminator. Since the focus is on the effect of the decision on the autonomy of those discriminated against, the constraint is infringed even if the employer does not aim to harm the people she discriminates against and does not think of them as morally inferior (for example, in instances of “disparate impact”).68 This conclusion may rest on the argument that some actions are morally impermissible regardless of the actor’s intentions.69 However, even if one embraces the intending/foreseeing distinction,70 the fact that the harm inflicted by discriminatory practices is often merely foreseen, rather than intended (e.g., in the cases of statistical discrimination and catering to prejudiced customers), does not remove its immorality. The intending/foreseeing distinction is crucial (according to its advocates) in justifying an active infliction of foreseen (though unwanted) harm on some people only if this harm is inflicted to avoid countenancing greater harm to others. This is not ordinarily the case with discrimination. The discriminator may believe, rightly or wrongly, that the harm caused by the discrimination is smaller than the costs involved in not discriminating (to herself or to her prejudiced customers, for example). However, neither the monetary costs involved in using nondiscriminatory screening methods nor the psychic costs borne by racist customers, are “harms” in the

67. See also Andrew Koppelman, Justice for Large Earlobes! A Comment on Richard Arneson’s “What is Wrongful Discrimination?,” 43 San Diego L. Rev. 809, 814 (2006) (“There is nothing in the structure of deontology that demands that it be indifferent to the issues of social context”); Andrew Altman, Civil Rights, in Stanford Encyclopedia of Philosophy (2003), http://plato.stanford.edu/entries/civil-rights (rejecting the alleged dichotomy between individualistic and systemic accounts of discrimination). 68. Cf. Moreau, supra note 59, at 22–24. 69. See, e.g., Judith Jarvis Thomson, Self-Defense, 20 Phil. & Pub. Aff. 283, 292–96 (1991) (advocating the “irrelevance-of-intention-to-permissibility thesis”). 70. On this distinction, on the related distinction between harming a person as a side effect of aiding other people and using a person as a means to aiding others, and on the trolley problem, see generally supra pp. 41–46.

243

244


relevant sense.71 Avoiding these costs cannot, therefore, justify infringing the constraint against harming people by discriminating against them, even if the harm is merely foreseen. Unlike intended harm, foreseen harm is not necessarily immoral, but it may well be. We next turn to the doing/allowing distinction.72 Arguably, an employer who does not hire someone on forbidden grounds merely refrains from assisting her, and thus, according to the doing/allowing distinction, does not infringe the constraint against doing harm. This argument is likewise unpersuasive. Helping each and every person would indeed be impossible or overly demanding, but the constraint against discrimination only extends to one’s employees and job applicants. If, for example, there are several candidates for one opening, choosing one of them does not infringe the constraint against harming the others as long as the choice does not rest on forbidden grounds (and certainly does not infringe any constraint against harming anybody else, at least under normal circumstances). Conflicting Constraints Thus far, we have established a plausible deontological constraint against discrimination. Before incorporating this constraint into CBA, one must delineate its incidence in light of conflicting constraints. The central conflicting constraint is against interfering with the autonomy and freedom of the discriminator. A moral theory committed to the intrinsic value of autonomy should recognize not only the autonomy of the people discriminated against but also that of the discriminators.73 In resolving this conflict, a distinction is commonly drawn between different spheres of relationships. It is generally accepted that one’s freedom to choose his or her spouse and intimate friends, even on the basis of race, color, gender, and religion, should not be curtailed. In this sphere, the deontological constraint against

71. The fact that these costs do not account as harm does not necessarily imply that they should not be taken into consideration within a deontologically constrained CBA. While the psychic costs borne by a racist may be excluded altogether (see infra p. 254), the monetary costs involved in using more expensive screening methods may be relevant in deciding whether there is enough good (or bad) outcomes to justify an infringement of the threshold constraint, and they may certainly be relevant in choosing among different options that do not infringe the constraint or that meet its threshold. 72. On this distinction and its role within deontology, see generally supra pp. 41–46. 73. Gardner, supra note 66, at 155; Moreau, supra note 59, at 21–22.


limiting people’s freedom (a constraint that applies to the state)—including the freedom to choose with whom to interact and associate—outweighs the constraint against discrimination (applying to individuals).74 Personal autonomy requires a space for “truly spontaneous and self-expressive activities and relationships.”75 This is not to say that racial or sexist decisions in the intimate sphere are moral,76 or that such decisions necessarily involve smaller harm to the excluded people.77 The freedom to make such decisions does not even rest on the assumption that the psychic costs borne by people who are forced to have intimate relationships with people they irrationally hate or feel superior to is greater than the psychic costs borne by racist employers (although this may well be the case).78 The point is that in the private, intimate sphere, the moral duty not to discriminate on such grounds is outweighed by the moral duty not to interfere in people’s autonomy.79 When it comes to commercial and market settings, and in particular to firms employing dozens of workers (or retailers serving many customers), the constraint against discrimination takes precedence—by most accounts— over the constraint against limiting the freedom of the discriminators. Indeed, some scholars object to employment antidiscrimination laws on libertarian grounds,80 but this extreme view is unpersuasive,81 and at any rate can hardly contribute to the analysis of legal norms it squarely opposes.

74. See Kelman, supra note 9, at 867–71; Larry Alexander, What Makes Wrongful Discrimination Wrong? Biases, Preferences, Stereotypes, and Proxies, 141 U. Pa. L. Rev. 149, 163 (1992). See also Donohue, supra note 39, at 1342 (“the courts have . . . grant[ed] exceptions to the antidiscrimination laws when significant privacy interests are at stake. . . . Furthermore, Title VII’s inapplicability to firms with fewer than 15 employees also represents an attempt . . . to accommodate privacy and associational concerns.”). 75. Gardner, supra note 66, at 154–55. 76. Arneson, supra note 56, at 778. 77. Kelman, supra note 9, at 868. 78. Cf. Kelman, supra note 9, at 870 (pointing out that the cost to the discriminator from interacting with people she would not like to interact with, is larger in the personal sphere than in the impersonal one). 79. Alexander, supra note 74, at 156 (explaining that under most moral theories, people “have a moral right to do what is morally wrong”). 80. Epstein, supra note 4. 81. See, e.g., Kelman, supra note 9, at 867–71; Verkerke, supra note 38 (reviewing Epstein, supra note 4); Peter Siegelman, Shaky Grounds: The Case against the Case against Antidiscrimination Laws, 19 Law & Soc. Inquiry 725 (1994); Donohue, supra note 38. For further critical discussions of Epstein’s position see the symposium published in 31 San Diego L. Rev. 1–262 (1994).

245

246


Other scholars interpret the limitations on the duty of accommodation (e.g., the “undue hardship” exemption in the ADA) as embodying a constraint against interfering with the discriminators’ freedom of contract.82 As further elaborated in subsection 3 below, however, we maintain that these limitations are better understood as setting a threshold above which the amount of good involved is high enough to overcome the constraint (rather than as a balance between two moral constraints). Drawing the line between intimate relationships and the market sphere is sometimes difficult. Some legal systems do not forbid discrimination in hiring household workers, while others view such exemption as too broad.83 Legal systems also differ regarding the minimal number of employees above which the prohibition on discrimination applies.84 Notwithstanding such inevitable delimitation issues, however, there is a broad consensus regarding the validity of this distinction in both morality and law.

2. Moral Constraints and Redistributive Goals Outside standard economic analysis, it is widely acknowledged that other factors besides efficiency bear on the morality of marketplace discrimination. So far we have focused on one such factor: the moral constraint against discrimination. Another major factor of intrinsic value is the (re)distribution of well-being, power, or wealth. According to a prevailing view, at least some antidiscrimination norms are result-oriented. They aim at enhancing equality, narrowing socioeconomic gaps between different communities, and fostering social integration (hereinafter redistributive goals). Some scholars argue that all antidiscrimination laws are redistributive. Both theoretical analysis and empirical studies indicate that all antidiscrimination

82. Moreau, supra note 59, at 23–24. 83. Compare Title VII of the Civil Rights Act, 42 U.S.C. § 2000e(b) (applying the antidiscrimination norms only to “person[s] engaged in an industry affecting commerce.”), with Case 165/82, supra note 11 (holding that a British statute providing that the prohibition of sex discrimination does not apply to employment in a private household is inconsistent with the European Council Directive concerning employment sex discrimination). 84. While in the United States federal antidiscrimination laws apply only to employers who employ at least 15 workers (Title VII, 42 U.S.C. § 2000e(b); ADA, 42 U.S.C. § 12111), the European Court of Justice held that even an exemption for employers who do not employ more than five workers is inconsistent with the European directive on sex discrimination (Case 165/82, supra note 11).


laws—and in fact, even universal norms applying to all employees, such as the requirement of overtime pay—have significant distributive effects.85 These effects are not incidental. They are, so the argument goes, the true basis of these laws. “Antidiscrimination law aims at a wholesale, not a retail, injustice.”86 The law prohibits discrimination “not because it seeks to enforce a uniform norm of ethical conduct on individual employers, but because . . . discrimination contributes to a pattern of social and economic subordination that has intolerable effects on our society.”87 A more common position draws a distinction between different antidiscrimination norms. Accordingly, any antidiscrimination legal norm belongs to one of two distinct categories: those resting on the moral constraint against prejudice-based discrimination (what Mark Kelman labels “simple discrimination”) and those aiming at redistribution (“accommodation” in Kelman’s terminology).88 The first category prohibits discrimination on irrational grounds, such as skin color or sex. The discriminator typically intends to harm members of a certain group or believes in their inherent inferiority.89 The second category characteristically proscribes discrimination based on the employer’s rational and legitimate desire to maximize profits. Accommodation duties and the disparate impact doctrine are paradigmatic examples. Scholars endorsing this classification tend to view antidiscrimination norms resting on the moral constraint against discrimination as more justifiable than those aiming at redistribution, for two reasons. First, redistributive norms impose significant monetary costs on blameless employers acting on perfectly legitimate and rational grounds, rather than on society at large.

85. Sharon Rabin-Margalioth, Anti-Discrimination, Accommodation, and Universal Mandates—Aren’t They All the Same?, 24 Berkeley J. Empl. & Lab. L. 111 (2003); Jolls, supra note 10, at 688–97. 86. Bagenstos, supra note 10, at 837. 87. Id. 88. Kelman, supra note 9; Jolls, supra note 10, at 643 (observing that many commentators view accommodation requirements as “fundamentally distinct from, broader than, and often less legitimate than” antidiscrimination); Bagenstos, supra note 10, at 828 (describing “a near-consensus in the literature” that there is “a fundamental normative difference between antidiscrimination requirements and accommodation mandates”). 89. See, e.g., Arneson, supra note 56, at 779; Kelman, supra note 9, at 841–42. The requirement of intention to discriminate on forbidden grounds does not entail a certain motive. Thus, simple (or direct) sex discrimination may be motivated by “chivalry” or “courtesy.” (Gardner, supra note 60, at 4).

247

248


Second, such norms assist some groups rather than others. It is not selfevident—so the argument goes—that mentally or physically disabled people, for example, are more deserving than, say, people of low intelligence, who also fare badly in the employment market.90 A major problem facing the distinction between the two types of antidiscrimination norms is that it is often difficult to draw the line between them. For instance, Kelman argues that prohibition of discrimination based on customers’ or coworkers’ prejudice rests on moral constraints, since in this case the employer merely serves as the customers’ and coworkers’ agent.91 This argument is problematic, for the employer acts on purely rational grounds, namely maximizing her profits.92 At the same time, Kelman classifies prohibitions of statistical discrimination as distributive because foregoing such discrimination entails real monetary costs due to the use of more expensive screening methods.93 It is hard to see why statistical discrimination—which often rests on a weak correlation between a candidate’s color or sex and her qualifications, and which may be only marginally cheaper than other screening methods—violates no moral constraint, whereas reluctantly catering to prejudiced customers does. Note that according to this distinction, even a prohibition of discrimination on grounds of pregnancy and childbearing is redistributive, since employing women of childbearing age is more costly.94

90. See generally Kelman, supra note 9 (arguing that, while simple discrimination should be regarded as a violation of a side constraint, accommodation is merely a policy argument competing with other social resource claims); Gardner, supra note 60, at 1–17 (highlighting the theoretical difference between direct discrimination, resting on individual liability for harming other people and indirect discrimination (also known as disparate impact), grounded in forward-looking, redistributive considerations). See also Mark Kelman, Strategy or Principle? 92–93 (1999) (analyzing the accommodation requirements of the ADA as a tax which creates inappropriate incentives and deadweight loss). 91. Kelman, supra note 9, at 847–49. Alternatively, one could argue that catering to prejudiced customers violates a moral constraint because the employer fails to act against the prevailing wrongful norm, and thus contributes to its maintenance. Arneson, supra note 56, at 789–90. 92. For this and other critiques of the view that catering for prejudiced customers and coworkers is inherently immoral, see Alexander, supra note 74, at 173–76, 193 (concluding that, “[a]lthough the [customers’] reactions may often express intrinsically immoral preferences, taking them into account [by the discriminator] does not”). See also Jolls, supra note 10, at 686 (arguing that discrimination based on customer or coworker attitudes is similar to requirements of accommodation). 93. Kelman, supra note 9, at 850–52. 94. See, e.g., Reva B. Siegel, Employment Equality Under the Pregnancy Act of 1978, 94 Yale L.J. 929 (1985); Jolls, supra note 10, at 651, 660–65.


Moreover, the classification problem does not pertain to borderline cases only. First, most discriminatory practices arguably rest on mixed motives, and correspondingly all antidiscrimination laws have mixed goals.95 Second, seemingly rational decisions often conceal irrational ones, and therefore insisting on proof of the latter would vitiate the efforts to extinguish them.96 Rather than a clear dichotomy between different antidiscrimination norms, there seems to be a continuum between prohibitions that are primarily based on the discriminator’s blameworthiness and those that principally aim at redistribution.97 A more radical critique of the common distinction contends that all antidiscrimination laws are grounded in moral constraints. If one understands the constraint against discrimination as a constraint against harming people’s autonomy and their protected set of freedoms of contract; and if (as argued in the preceding subsection), this constraint is infringed regardless of whether the discriminator intends to discriminate members of certain groups and whether he views them as morally inferior—then this constraint can underlie all antidiscrimination laws.98 This understanding of the moral constraint is broader than the one advocated by those who believe that only

95. See Paul Brest, In Defense of the Antidiscrimination Principle, 90 Harv. L. Rev. 1, 7 (1976) (“[R]ace-dependent decisions that are rational and purport to be based solely on legitimate considerations are likely in fact to rest on assumptions of differential worth of racial groups or on the related phenomenon of racially selective sympathy and indifference”); Alexander, supra note 74, at 167–73, 176–83 (discussing seemingly rational discrimination based on irrational proxies, rational discrimination tainted by associations with biases, stereotypes, and ideologies, and discrimination stemming from unconscious bias); Linda Hamilton Krieger, The Content of Our Categories: A Cognitive Bias Approach to Discrimination and Equal Employment Opportunity, 47 Stan. L. Rev. 1161 (1995) (arguing that many prejudiced, employment-based decisions result from categorization-related decision errors). 96. See George Rutherglen, Disparate Impact under Title VII: An Objective Theory of Discrimination, 73 Va. L. Rev. 1297, 1309–11 (1987) (“Only rarely will plaintiffs find direct evidence of discriminatory intent. . . . A direct inquiry into the elusive intent of the defendant is usually fruitless. . . . The theory of disparate impact . . . addresses the difficulty of proving pretextual discrimination. . .”); David A. Strauss, Discriminatory Intent and the Taming of Brown, 56 U. Chi. L. Rev. 935, 1010–14 (1989) (“The failures of the discriminatory intent standard in the area of employment discrimination . . . strongly support the use of the disparate impact standard . . .”). 97. Interestingly, while Gardner initially insisted that the harm principle (corrective justice) and distributive justice are “radically discontinuous,” and thus a single legal norm cannot coherently rest on both of them (Gardner, supra note 60, at 12–17), he later retracted this claim and conceded that the same practice may be wrong for two different reasons, and that legally proscribing it may thus rest on both of them (John Gardner, Discrimination as Injustice, 16 Oxford J. L. Stud. 353, 366–67 (1996)). 98. Moreau, supra note 59.

249

250


“simple discrimination” rests on such moral grounds, namely that only illwilled discrimination infringes a moral constraint. Also, as indicated in the preceding subsection, the present position does not ignore the cultural background of discriminatory practices, as this background has a direct effect on the way discrimination reduces autonomy. All of the positions described so far share the assumption that distributive goals and moral constraints are two fundamentally different grounds for prohibiting discrimination. There is, however, a way of looking at distributive goals and moral constraints as two sides of the same coin. This position follows Raz’s unitary ideal of autonomy and his theory of perfectionist liberalism.99 According to Raz, the state is duty-bound to promote people’s autonomy. Contrary to the traditional “harm principle,” Raz’s theory allows the state “to use coercion both in order to stop people from actions which would diminish people’s autonomy and in order to force them to take actions which are required to improve peoples’ options and opportunities.”100 This morality of autonomy denies the distinction between culpability-based and end-state justifications for antidiscrimination norms. “The ‘harm principle’ and ‘distributive justice’ are not discontinuous principles but are subsumed within a single, continuous . . . principle.”101 A thorough examination of the various positions described above exceeds the scope of our discussion. For our purpose, the central questions are, first, how do these positions affect the cost-benefit analysis of antidiscrimination laws, and second, how are they related to the consequentialism/ deontology distinction. Clearly, all positions imply that standard CBA, ignoring both distributive concerns and moral constraints, is off the mark. The positions focusing on distributive justice entail that the social welfare function of standard CBA should be altered. Rather than simply aiming at maximizing aggregate utility, its distribution should count as well. Positions focusing on moral constraints would require incorporating constraints into the analysis. If distribution and moral constraints are both normatively important, then a complex formula, taking both of them into account, may be required.

99. Raz, supra note 63. 100. Id. at 416. Cf. Canadian Human Rights Act (R.S. 1985, c. H-6), section 2 (“The purpose of this Act is to . . . give effect . . . to the principle that all individuals should have an opportunity equal with other individuals to make for themselves the lives that they are able and wish to have and to have their needs accommodated . . .”). 101. Gardner, supra note 60, at 18–21.


The distributive concerns discussed above pertain to the goodness of outcomes brought about by antidiscrimination laws. As such, they may be part of either a consequentialist or a deontological factoral theory, depending on whether the goodness of outcomes is the only factor taken into account. Improving standard CBA by incorporating distributive concerns is both justifiable and feasible. Since it exceeds the scope of this study, however, we will not pursue this possibility further but rather focus on moral constraints. Our discussion proceeds on the assumption that moral constraints are pertinent to all antidiscrimination norms. Accordingly, all such norms may fruitfully be analyzed using a deontologically constrained CBA. If, contrary to this assumption, such constraints underlie only some of these norms, then the proposed analysis is useful for those norms only.

3. Integrating Threshold Constraints with Economic Analysis The moral constraint against discrimination entails that prohibiting discrimination on such forbidden grounds as race and sex can be justified even if the prohibition imposes a social cost that exceeds its benefit. This is the case not only in the context of simple discrimination (that is, when acting on the basis of a forbidden ground) but also with regard to accommodation requirements. As clearly stated by Judge Richard Posner, “it would not follow . . . that an accommodation would have to be deemed unreasonable if the cost exceeded the benefit however slightly.”102 The constraint against discrimination mandates that an accommodation may well be required even if its net social benefit is negative. This is not to say that welfare analysis is irrelevant in evaluating the desirability of antidiscrimination legislation and in implementing concrete

102. Vande Zande, 44 F.3d 538, 542 (7th Cir. 1995). Judge Calabresi adopted a similar approach, noting that an accommodation is reasonable “if its costs are not clearly disproportionate to the benefits that it will produce”: Borkowski v. Valley Central School District, 63 F.3d 131, 138 (2d Cir. 1995). See also Stein, supra note 7, at 178 (willing to endorse the view that “society ought to look beyond economics and instead be motivated by concerns for human dignity and well-being,” thus providing even “wholly inefficient accommodations”).

251

252


legal provisions.103 A sufficiently high net cost may well justify an infringement of the moral constraint against discrimination and correspondingly a limitation of the incidence of the antidiscrimination legal norm (or an exception to it). As indicated in section B, under current legislation, employers are sometimes permitted to employ an otherwise discriminatory practice if the practice is “reasonably necessary” or required as a “business necessity.”104 Similarly, employers are exempted from the duty of accommodation if it would be “unreasonable” or would impose on them “undue hardship.”105 Interpretation and implementation of these provisions may greatly benefit from a careful cost-benefit analysis subject to a deontological constraint. To assess the permissibility of a given type of discrimination, one needs a threshold function that determines the size of the threshold and the pertinent types of costs and benefits. The size of the threshold may depend on the form of discrimination, which serves as a proxy for the size of the harm. Plausibly, the extent of nonpecuniary harm inflicted when an employer uses an employment practice that causes a “disparate impact” on the basis of

103. For such an extreme position see, e.g., Gregory S. Crespi, Efficiency Rejected: Evaluating “Undue Hardship” Claims Under the Americans With Disabilities Act, 26 TULSA L.J. 1 (1990). See also Cass R. Sunstein, Cost-Benefit Analysis without Analyzing Costs or Benefits: Reasonable Accommodation, Balancing, and Stigmatic Harms, 74 U. Chi. L. Rev. 1895, 1898–99 (2007) (raising the possibility that unlike the “undue hardship” exception, the “reasonable accommodation” requirement be read as adopting a cost-blind approach, or as allowing the employer to opt for the most cost-effective measure from among the reasonable ones). 104. According to Title VII of the Civil Rights Act, acting on the basis of religion, sex or national origin is permitted when such characteristic “is a bona fide occupational qualification reasonably necessary to the normal operation of that particular business or enterprise.” 42 U.S.C. § 2000e-2(e)(1). See also Frank, supra note 14; Bryant, supra note 14. In addition, an employment practice that causes a disparate impact on the basis of race, color, religion, sex, or national origin is legitimate if the practice is “job related for the position in question and consistent with business necessity.” 42 U.S.C. § 2000e-2(k)(1)(A)(i). 105. Under the ADA, an employer is exempt from the duty to make “reasonable accommodations” to the known physical or mental limitations of an otherwise qualified individual with a disability if making it “impose[s] an undue hardship on the operation of the business.” 42 U.S.C. § 12112(b)(5)(A). The ADA defines “undue hardship” as “an action requiring significant difficulty or expense, when considered in light of the factors set forth in subparagraph (B).” 42 U.S.C. § 12111(10)(A). Similarly, under Title VII, an employer is required to “reasonably accommodate” an employee’s religious observance and practice only as far as the required accommodation does not impose “undue hardship on the conduct of the employer’s business.” 42 U.S.C. § 2000e-(j).


race, color, religion, sex, or national origin is significantly lower than when she refuses to hire or otherwise discriminates against an individual directly on one of these grounds. This variation can justify a difference in the threshold that has to be met in each of these cases. Accordingly, the threshold is highest in cases of simple discrimination. As discussed in section B, under Title VII, acting on the basis of religion, sex, or national origin (but not race) is permitted when such characteristic “is a bona fide occupational qualification reasonably necessary to the normal operation of that particular business or enterprise.” Even in cases of racial discrimination, a threshold may be appropriate in exceptional cases. For example, as Kelman rightly points out, “an employer might temporarily segregate workers in an otherwise impermissible fashion to avert severe racial violence, if no less rights-violative alternative were available.”106 In such a case, the employer would have to prove that the costs of observing the antidiscrimination norm were sufficiently great to meet a rather high threshold. A deontologically constrained CBA also requires setting the structure of the threshold function. As discussed in chapter 4, the constraint may be expressed by adding to the act’s harms, C, some factor K (where K > 0), such that an infringing act is permissible only if its total benefit, B, exceeds the sum of C and K; alternatively, the infringement can be expressed by multiplying the act’s harms by some factor K ′, such that an infringing act is permitted only if its total benefit B, exceeds its weighted harms, CK ′ (where K ′ > 1).107 In the present context, the language of thresholds set forth by Title VII and the ADA seems to reflect the former approach, since it determines the threshold irrespective of the actual harm that the discriminatory decision or practice inflicts on the person who is discriminated against. Yet, the “necessity” and “undue hardship” standards may also be interpreted as laying down a multiplier threshold whose size is determined by the extent of the harm inflicted by the infringing act. This approach finds support in Judge Posner’s statement that the assessment of whether an accommodation

106. Kelman, supra note 9, at 836, n.7. Cf. Washington v. Lee, 263 F. Supp. 327 (M.D. Ala. 1966), aff ’d per curiam, 390 U.S. 333 (1968)) (striking down an Alabama statute requiring permanent and complete racial segregation of prisoners, yet recognizing that “there is merit in the contention that in some isolated instances prison security and discipline necessitates segregation of the races for a limited period); Wittmer v. Peters, 87 F.3d 916, 919 (7th Cir. 1996) (suggesting that “separation of the races in a prison that was undergoing a race riot would not violate the Constitution”). 107. A third possibility is a combined, additive and multiplier, threshold, See supra pp. 84–86, 93–96.

253

254


imposes an “undue burden” under the ADA should be made “in relation to the benefits of the accommodation to the disabled worker as well as to the employer’s resources.”108 As regards the types of costs and benefits relevant to measuring whether the threshold is met, a central question is whether discriminatory preferences and prejudices of either the employer or the firm’s customers or other employees should be taken into account. As we have seen, disregard for antisocial preferences is possible even within a purely consequentialist theory, and has in fact been advocated by some economists.109 However, this view is more likely to be part of a deontologically constrained CBA. The constraint against discrimination implies that satisfying racial or sexist preferences should not constitute part of the social good. In a sense, this view is reflected in current antidiscrimination laws. The express limitations on the duties of accommodation discussed earlier do not apply to antidiscrimination norms that do not entail out-of-pocket expenses. By limiting the application of these limitations to accommodation duties, the law implicitly delegitimizes other costs.110 Finally, within the deontological threshold function, chronologically remote and uncertain costs and benefits—such as the concern that some antidiscrimination measures might in the long-run create disincentives to employing disabled people111—may be discounted.112

108. Vande Zande, 44 F.3d at 543. See also Borkowski, 63 F.3d at 139 (Calabresi, J.) (“[The concept of undue hardship] looks not merely to the costs that the employer is asked to assume, but also to the benefits to others that will result”). 109. See supra pp. 238–39. 110. See, e.g., Gerdom v. Continental Airlines, Inc., 692 F.2d 602 (9th Cir. 1982) (holding that customer preference for slim female flight attendants did not justify a discriminatory policy where weight was unrelated to job performance), cert. denied, 460 U.S. 1074 (1983); Bradley v. Pizzaco of Nebraska, Inc., 7 F.3d 795, 799 (1993) (holding that customer preference for no-beard on the face of a delivery is not a colorable business justification of the employer’s no-beard policy, which had a disparate impact on African American men). As mentioned in supra note 106 and the accompanying text, however, even in cases of simple discrimination, one could imagine scenarios in which the net cost of complying with the prohibition (excluding the “psychic costs” experienced by bigots) would be so high as to justify noncompliance with, or exception to, the prohibition. 111. See, e.g., Vande Zande, 44 F.3d at 545 (Posner, J. noting that placing the employer that already invested in accommodating a disabled worker under any further obligation would ultimately “hurt rather than help disabled workers”); Christine Jolls, supra note 44 (accommodation mandate may reduce a given group’s employment level or wages); Acemoglu & Angrist, supra note 44, at 931. 112. On chronologically remote and uncertain costs and benefits, see also supra pp. 89–91, 147–49, and 196–201; infra pp. 344–45.


Given a certain threshold for accommodation duties, the efficacy of antidiscrimination norms may be considerably enhanced by spreading the costs involved in such accommodations amongst all employers or even amongst all members of society.113 Indeed, some countries do employ mechanisms for such cost spreading, thereby considerably lowering the actual costs borne by any individual employer. These include direct subsidies and indirect tax credits and tax deductions.114 Such mechanisms effectively reduce discrimination in circumstances where imposing the costs of accommodation on the single employer would be unreasonable and involve undue hardship. Deontological morality supports such mechanisms not only because they enhance the efficacy of antidiscrimination norms but also because they enhance their legitimacy. Deontology is less demanding than consequentialism. Rather than mandating everyone to always do what would maximize the overall good, it allows for options.115

•

F. Conclusion

This chapter analyzed employment discrimination through the prism of deontologically constrained CBA. It sought to expose the limits of standard

113. Some scholars suggest that the state should fund all necessary accommodations, while others argue that it should fund only those reasonable accommodations that “would impose an undue hardship” on the employer. For the first view, see, e.g., Scott A. Moss & Daniel A. Malin, Public Funding for Disability Accommodations: A Rational Solution to Rational Discrimination and the Disabilities of the ADA, 33 Harv. C.R.-C.L. L. Rev. 197 (1998); Sue A. Krenek, Beyond Reasonable Accommodations, 72 Tex. L. Rev. 1969, 2009–13 (1994). For the more modest suggestion, see, e.g., Stein, supra note 7, at 174–77 (suggesting that when both the individual worker with a disability and society in general benefit, but the employers lose, the state should compensate losing employers for making the accommodations). See also Michael A. Stein, Empirical Implications of Title I, 85 Iowa L. Rev. 1671, 1684 (2000) (“Providing for extra-reasonable accommodations could overcome existing market inequities borne by the most stigmatized among the disabled”); Samuel R. Bagenstos, The Future of Disability Law, 114 Yale L.J. 1 (2004). 114. Direct subsidies are granted, for example, in Israel. See section 49 of the National Insurance Law [Consolidated Version], 1995, S.H. 1522; Equal Rights of Persons with Disabilities (State Participation in Financing Accommodations) Regulations, 2006, K.T. 6480, 754. Tax credits and deductions are used, for instance, in the United States. See sections 44 and 190 of the Internal Revenue Code, 26 U.S.C §§ 44, 190. 115. See supra pp. 19–20, 33–40, 41, 47, and 98–103; Alexander, supra note 74, at 172 (criticizing “moral propositions that demand considerable individual sacrifice for the social good”). Clearly, state funding of accommodations may rest also on distributive concerns—this time the distribution of the burden among different employers.

255

256


economic analysis of this issue and to demonstrate the fruitfulness of incorporating a deontological constraint into the economic analysis. We demonstrated two of the strengths of our proposal and one of its limitations. First, constrained CBA better captures the pertinent normative issues, including those not adequately addressable within standard CBA. Second (and relatedly), constrained CBA is much more compatible with actual legal norms, which often embody threshold constraints. At the same time, constrained CBA is particularly useful in analyzing issues where the two primary normative factors are the maximization of human well-being and deontological constraints. Employment discrimination is problematic in that sense, due to the pertinence of a third factor: equality. We conceded that our proposal is useful to the extent that antidiscrimination norms actually rest on the constraint against discrimination, rather than on redistribution of power, wealth, or well-being. While deontologically constrained CBA does not necessarily exhaust all pertinent moral factors, it does facilitate a more comprehensive normative analysis than standard CBA, and it structures the analysis in a way that better explains and corresponds with current legal norms.

•

nine

Contract Law

•

A. Introduction

there are very few legal fields in which economic analysis has gained

such prominence as in contract law. Ideologically and methodologically, economic analysis is especially apt for analyzing market interactions, which is hardly surprising given that this is the domain in which economic theory originated. Voluntary transactions are the building blocks of the market; and the free, competitive market is the ideal prototype of an economically efficient mechanism. In a perfectly competitive market, aggregate utility is maximized when every person rationally pursues her own ends. Market transactions are not only Kaldor-Hicks efficient but also Pareto efficient, since presumably no one would make a contract unless she expects it to improve her lot. Market transactions do not ordinarily involve infringements of deontological constraints; hence, the tension between standard economic analysis and prevailing moral convictions is less conspicuous in this area than in others. Finally, since market transactions usually involve money or money equivalents, monetization poses no serious difficulty for economic analysis of contract law, compared to such spheres as family or criminal law. We nevertheless maintain that certain deontological constraints apply to contracting behavior and that incorporating deontological constraints with economic analysis of contract law may be fruitful. Following a brief outline of the economic analysis of contract law in section B, section C will briefly survey the deontological constraints pertinent to contract law and critically examine the standard economic response to them. Section D will then demonstrate how deontological constraints may be integrated with economic analysis of the contracting stage, focusing on the doctrines of mistake and misrepresentation. Section E will highlight the differences between economic and deontological analyses of contract performance and breach, and discuss the difficulties facing integration of deontological

257

258


constraints with the economic analysis of contract remedies, given the current state of the pertinent theories.

•

B. Economic Analysis of Contract Law

Contracts are the legal means to effectuate voluntary transactions, which normative economics favors enthusiastically. Such transactions facilitate the transfer of resources (goods, services, money) from people who value them less to people who value them more, thereby increasing the overall welfare derived from these resources. Contracts also allocate risks and prospects involved in any transaction, including market price fluctuations and increases or decreases in the promisor’s costs of performance or the promisee’s benefit from performance. Risk allocation is particularly important in forward-looking transactions, but some risks (such as a hidden defect in the sales object) also characterize instantaneous transfers of resources. Voluntary transactions facilitate division of labor in society, which is the key to higher productivity at lower costs. Each production unit (from a power plant to a law professor) specializes in producing certain products or services, and the outcomes are then traded in complex, mutually beneficial exchanges. From an economic point of view, the role of contract law is to facilitate and encourage these voluntary exchanges of resources. The very availability of legal enforceability prevents a prisoner’s dilemma situation, in which each party strives to receive the counter-performance without doing her share, and fears that the other party will do the same—thus resulting in mutual nonperformance. Legal enforceability facilitates reliance on the expected performance by the other party, which is particularly essential for the execution of long-term and complex projects. Other ways in which contract law facilitates voluntary transactions include, for example, the imposition of precontractual disclosure duties (preventing inefficient transactions due to information problems) and the reduction of transaction costs through the provision of default rules (thus facilitating transactions that would otherwise be prohibitively costly and increasing the joint surplus of other transactions). Standard economic analysis of contracts and contract law assumes that people are cognitively and motivationally rational.1 Each party aims to

1. On the two components of economic rationality, see supra pp. 11–16; infra pp. 323–27.

contract law

maximize her own utility and is able to optimally calculate the expected outcomes of her decisions, paying heed to the similarly rational decisions of the other party. From an economic perspective, contract law should provide the parties with optimal incentives to act efficiently, that is, in a way that would maximize the joint surplus from the transaction, throughout the contractual process. Chronologically, this process begins by acquiring relevant information and searching for an appropriate partner, continues in negotiating and formulating the contract, moves on to investing in performance efforts by the promisor and in reliance by the promisee, deciding whether to perform or breach, and finally, taking measures in response to a possible breach. The complexity of economic analysis of contract law stems, among other things, from the realization that any rule may have different effects on the parties’ behavior in various stages of the transactions. Thus, for example, the measure of damages for breach of contract affects not only the promisor’s decision whether to perform or breach the contract, but also the extent of the promisee’s reliance on the expected performance and the scope of information the parties share prior to contracting. Very often, some of the incentive effects of any rule are desirable while others are undesirable, making the formulation of an efficient law of contract almost impossible given the absence of reliable data on the magnitude of each effect. This complexity is reflected in the fact that legal economists disagree on practically every aspect of contract law. Most legal economists do, however, share the consequentialist denial of any constraint on promoting social utility. Thus, economic analysis does not recognize a constraint against promise breaking but rather celebrates contract breaching whenever the breach produces slightly more net benefit than performance. Likewise, it judges truth telling and information disclosure only according to their effect on the transaction’s efficiency and the incentives that disclosure duties create for obtaining the information in the first place. In the same vein, compensation for harm is only due if, and to the extent that, such compensation would bring about an efficient outcome.2 It is this common feature of economic analyses of contract law that we challenge. In line with our general thesis, we aim to demonstrate that the denial of constraints weakens economic efficiency both as an interpretative and as a normative theory of contract law.

2. See, e.g., Robert E. Scott & George G. Triantis, Embedded Options and the Case against Compensation in Contract Law, 104 Colum. L. Rev. 1428 (2004); Robert Cooter & Ariel Porat, Anti-Insurance, 21 J. Legal Stud. 203 (2002).

259

260


C. Deontological Constraints and Contract Law 1. The Pertinent Constraints: An Overview

Two major categories of deontological constraints are relevant to contracts: those pertaining to the contracting stage and those applying to the performance stage. At the formation stage, the overarching principle is that of free will. A contracting party must not lie, threaten, or use force to induce contracting. Lying and deceiving adversely affect the interests of the other party and fail to respect her capacity for reasoned decision-making. Some deontologists maintain that lying is merely a special case of the general constraint against actively/intentionally harming other people. Others believe that there is an independent constraint against lying, which is violated even if the lie does not harm the deceived and even if it benefits her (or, alternatively, believe that every lie ipso facto harms the deceived).3 A related question is whether the relationship between people negotiating a contract gives rise to a duty to actively disclose information to one another. Inducing contracting through force or threats likewise offends the liberty of the threatened party. Drawing the line between a legitimate use of one’s bargaining advantages and an illegitimate exploitation of the other’s weaknesses is, however, a difficult task.4 At the performance stage, just as promisors must keep their promises, a contracting party must not breach her contractual obligations. The questions revolving around the existence and exact scope of this constraint are a primary source of controversy among consequentialists and deontologists, as well as between different brands of deontological theories. It is unclear whether there is an independent moral duty to keep one’s promises; and if no such independent duty exists, whether the very breach of a promise necessarily harms the promisee.5 If the very act (including omission) of breach constitutes harm, then the controversy is less important. Theoreticians who do not recognize an independent constraint against breaking promises and who do not accept that every promise breaking ipso facto involves harm, would not regard just any contract breach as infringing a constraint.

3. Shelly Kagan, Normative Ethics 106–13 (1998). See also infra pp. 274–77. 4. See, e.g., Charles Fried, Contract as Promise: A Theory of Contractual Obligation 92–103 (1981); Michael J. Trebilcock, The Limits of Contract 78–101 (1993). 5. Kagan, supra note 3, at 118–20.

contract law

In fact, these theoreticians are likely to deny that a mere promise or agreement, absent any reliance by the promisee or enrichment by the promisor, gives rise to contractual liability in the first place.6 Another crucial difference between these two understandings of the constraint against breaking promises refers to the significance of paying damages. Assuming that monetary damages can fully compensate the promisee for her losses, is there an infringement of a constraint if the breaching promisor fully compensates the promisee? Those who do not admit of an independent constraint against promise breaking may conclude that in such a case no constraint is infringed because there is no harm or because the harm is repaired. Conversely, those who argue for an independent constraint against promise breaking may insist that while paying damages mitigates the immorality of the breach, it does not eliminate it altogether.7 In addition to the two main categories of constraints, other deontological constraints may pertain to contractual relations. In particular, the relationships created between contractual parties (or even between people negotiating a contract) may give rise to special moral obligations of consideration, cooperation, and mutual assistance.8 In the present context, it is immaterial whether these obligations are considered independent constraints, derivatives of the abovementioned constraints, or mere manifestations of the general constraint against harming other people. The constraint against harming other people may also be relevant when contracts adversely affect third persons, as in many instances of illegal contracts.9

6. See, e.g., Patrick S. Atiyah, Promises, Morals, and Law 177–215 (1981). 7. The latter view seems to better reflect prevailing moral intuitions. See David Baumer & Patricia Marschall, Willful Breach of Contract for the Sale of Goods: Can the Bane of Business Be an Economic Bonanza, 65 Temp. L. Rev. 159, 164–67 (1992); Tess Wilkinson-Ryan, Jonathan Baron, Moral Judgment and Moral Heuristics in Breach of Contract, 6 J. Empirical L. Stud. 405 (2009). But see Daniel Markovits, Contract and Collaboration, 113 Yale L.J. 1417, 1504 (2004) (“A promisor who pays expectation damages continues to collaborate with her promisee, in spite of her breach”). 8. On moral constraints rooted in special relationships, see generally Kagan, supra note 3, at 125–37. 9. As elaborated in chapter 8, a person may infringe the constraint against harming other people not only through her behavior in the contracting and performance stages but also by refusing to contract with some people on racial, chauvinist, or similar grounds. Prohibiting such discrimination may, however, raise a concern about infringing on people’s autonomy by the state. We shall not discuss such constraints here.

261

262


2. The Economic Response and Its Critique In previous chapters, when discussing other issues involving deontological constraints, such as the killing of innocent people to save the lives of others or discrimination in the marketplace, we noticed that consequentialists, including legal economists, expend great effort to demonstrate that, contrary to appearances, consequentialism does not entail the violation of those constraints.10 Such efforts are less noticeable when it comes to contract law. For example, legal economists straightforwardly advocate the breach of contracts when the costs, including opportunity costs, of performance to the promisor are greater than the benefits to the promisee.11 There may be two explanations for this unapologetic stance of economic analysis of contract law. First, the deontological constraints against promise breaking, lying, and withholding information are usually perceived as much weaker than the constraints against infringing upon people’s basic liberties, such as the right to life or to human dignity. Breaking a promise or lying is very different from actively/intentionally killing a person or discriminating against her. Hence, the conclusions of economic analysis of contract law are less morally repugnant.12 Second, despite the fundamental differences between deontology and consequentialism, the practical implications of deontology and of consequentialism for contract law are often similar. At the contracting stage, both deontology and consequentialism ordinarily insist that contracts should be voluntary, that is, inter alia, free of deceit and duress. Deontologists tend to insist on the voluntariness of contracts because of the intrinsic value they attribute to people’s freedom and autonomy. Consequentialists insist on voluntariness because it is the best guarantee for the bargain’s efficiency: only if both parties voluntarily make the contract can one be assured that the contract is Pareto efficient. At the performance stage, most deontologists and consequentialists agree that, at least as a point of departure, contract remedies should protect the promisee’s expectation interest, rather than merely her restitution or reliance interests. Deontologists may come to this

10. See generally supra pp. 21–33. See also supra pp. 234–40. 11. See, e.g., Robert Cooter & Thomas Ulen, Law and Economics 254–61 (4th ed. 2004); Steven Shavell, Damage Measures for Breach of Contract, 11 Bell J. Econ. 466 (1980). 12. Though even this generalization is subject to exceptions. See, e.g., Oren Bar-Gill & Omri Ben-Shahar, Credible Coercion, 83 Tex. L. Rev. 717 (2005).

contract law

conclusion because they view expectation damages as equivalent to the promised performance,13 or because only expectation remedies reflect the parties’ forward-looking commitment to affirmatively treat each other as ends in themselves—a commitment that underlies their collaborative relationships.14 Consequentialists may reach the same conclusion because remedies protecting the promisee’s expectation interest ensure that the promisor will fully internalize the costs of her breach, thus inducing her to perform the contract if and only if performance is efficient.15 While it is true that the deontological constraints ordinarily applicable to contractual issues are relatively weak, the convergence of consequentialism and deontology in this sphere must not be overstated. Deontological and economic theories of contract law do sometimes diverge not only with regard to the normative underpinnings of contract law but also in their policy recommendations. This divergence stems from the fact that economic analysts view private, free will as very important, but only for instrumental reasons. Private will is an object of people’s preferences, and so having more freedom enhances one’s well-being according to a preference theory of the good. A minimal degree of liberty is also a means to attaining and safeguarding other components of people’s welfare. Finally, the will also provides economic analysis with a criterion for measuring well-being. The value of entitlements is usually measured according to people’s willingness to pay for them, and willingness to pay is (partially) determined by people’s will.16 Thus, in situations of market failure, for example, economic analysis may endorse involuntary transfers of assets whenever the benefits of such transfers, all things considered, are greater than their costs. In the same vein, economic analysis would approve of a breach of contract whenever the net benefit of the breach, all things considered, exceeds the net benefit of performance. The central role of the will in contract law enables economists to defend the morality of economic analysis of contract law even when it conflicts with commonsense morality. For instance, Steven Shavell has argued that contracts are usually incomplete: even when they lay down an obligation to

13. Fried, supra note 4, at 17–21; Ernest J. Weinrib, Punishment and Disgorgement as Contract Remedies, 78 Chi.-Kent L. Rev. 55, 65–70 (2003) (deriving expectation damages from Kant’s notions of contract and promise). 14. Markovits, supra note 7, at 1497–1515. 15. A. Mitchell Polinsky, An introduction to Law and Economics 33–36 (3d ed. 2003). For a more nuanced analysis, see infra pp. 294–97. 16. See also infra p. 318.

263

264


do something, they do not explicitly specify whether this obligation would apply under every possible contingency. Since it stands to reason that the parties would have exempted the promisor from her duty to perform whenever the costs of performance exceed the value of performance to the promisee, and since the promisor ordinarily knows that this is the case, then there is nothing immoral in breaching a contract so long as the breacher compensates the injured party according to the latter’s expectation interest. A breach is therefore immoral only if damages are undercompensatory (which, admittedly, they often are).17 Interestingly, Shavell’s attempt to reconcile economic analysis of contract breach with prevailing moral intuitions (that is, roughly, with moderate deontology) resembles the attempt of absolutist deontologists to square the duty to keep promises with these intuitions. Suppose that keeping a mundane promise would entail unexpectedly huge monetary costs for the promisor or not saving the life of another person. A moderate deontologist would readily admit that the constraint against promise breaking has thresholds, that in these cases the threshold is met, and that the infringement is therefore permissible. An absolutist cannot endorse this answer, yet she may argue that any promise is tacitly subject to various unexpressed background conditions. Both parties understand that a routine promise is subject to the condition that keeping it would not entail enormous monetary costs to the promisor or the abandoning of a person whose life the promisor can save. Hence, nonperformance under these circumstances infringes no constraint. The parties’ presumed will underlies both the absolutist’s justification of nonperformance on the grounds that the promise was never intended to be kept under extreme circumstances and the economic argument that contractual obligations are not ordinarily meant to apply when nonperformance produces more good than does performance. The similarity between the two arguments is further highlighted by Shavell’s choice of examples. Practically all the illustrations he brings (and those used in his survey) involve nonperformance due to unexpected hardship for the promisor or unexpected lack of interest in the performance by the promisee. These cases verge on impossibility, impracticability, or frustration of purpose. Shavell makes almost no reference to nonperformance due to the promisor’s desire to strike an alternative, more lucrative bargain—the paradigm of efficient breach in the economic literature. This choice of examples camouflages the first

17. Steven Shavell, Is Breach of Contract Immoral?, 56 Emory L.J. 439 (2006).

contract law

difficulty in Shavell’s argument: While many people might justify nonperformance in impossibility-like situations, it is doubtful that they would justify it when nonperformance is not meant to avoid losses but rather to facilitate larger profits. Committed consequentialists would justify nonperformance in both cases, and, in fact, deny that the two are different. But this is beside the point, as Shavell aims to demonstrate that the efficient breach doctrine is not immoral under commonsense morality. Shavell employs what he describes as a “simple and natural” definition of morality: “performance is morally required in a contingency if and only if the parties did specify, or would have specified, performance in that particular contingency.”18 While Shavell—like the absolutist deontologist—may be right that the parties would have specified that performance is not required in impossibility-like circumstances, we doubt that they would feel the same way in cases of nonperformance due to the promisor’s desire to accrue greater profits and keep them for herself.19 The comparison to the absolutist’s argument helps to reveal another difficulty with the efficient breach theory. In the unexpectedly high monetary costs and life-saving examples, absolutist and moderate deontologists may differ regarding the question of compensation. Compensating the promisee for her losses seems appropriate in at least some instances of permissible infringements of constraints.20 If, in contrast, the promisor has not infringed any constraint because she did not undertake to perform under certain circumstances in the first place (the absolutist’s argument), then presumably no promise was broken, and no compensation is owed. Returning to the efficient breach argument, assuming we face a contingency in which the parties would not have specified performance, why should the promisor pay damages when she was not expected to perform anyway?21

18. Id. at 440. 19. See Baumer & Marschall, supra note 7, at 164–67 (providing empirical evidence that businesspeople view a deliberate breach as unethical); Wilkinson-Ryan & Baron, supra note 7; Daphna Lewinsohn-Zamir, Beyond the Bottom Line: The Complexity of Outcome Assessment (working paper, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_ id=1479051). We doubt that the promisee would have authorized a unilateral breach even subject to receiving some share of the extra profits. Most probably, the parties would have agreed that if such an opportunity comes up, they would renegotiate nonperformance and sharing of the extra gains. 20. Kagan, supra note 3, at 128–30. 21. Seana Shiffrin, Could Breach of Contract Be Immoral?, 107 Mich. L. Rev. 1551, 1556–59 (2009). There may be cases (primarily when the promisee is risk-averse and the promisor

265

266


Given the hypothetical nature of the parties’ presumed intentions, economists may reply that in cases where the parties would have exempted the promisor from performance because it would be inefficient, they would nevertheless require the promisor to pay a sum of money equal to the promisee’s expectation interest, thus inducing only efficient breaches. Put differently, the promisor’s obligation should be interpreted as an option to perform or pay damages.22 Even this reply is, however, unpersuasive in most cases. In this hypothetical agreement, the promisor’s nonperformance could only be considered moral if she exercised the option, that is, if she concomitantly offered to pay damages (unless there is a bona fide disagreement regarding the existence of breach or the loss to the promisee). This is seldom the case, especially in adjudicated disputes. Therefore, even if court-awarded damages put the promisee in the same monetary position she would have occupied absent the breach (which is not ordinarily the case), nonperformance would still be immoral.23 In fact, the immorality of the breach would not disappear even if the court would compel the promisor to perform her obligations through the award of specific performance. The final—and most powerful—critique of Shavell’s argument also echoes the moderate/absolutist debate. Nobody denies that if a contract explicitly specifies that performance is not required under certain circumstances, or that a substitute payment is optional, nonperformance in those circumstances does not constitute promise breaking. Even absent such explicit specification, if that is a reasonable interpretation or supplementation of the contract, no promise breaking is involved. In such cases, not only is

risk-neutral) where the optimal arrangement would be to allow for nonperformance subject to the payment of a sum of money equivalent to expectation damages, as a sort of insurance (see infra p. 295). But this is not the case where the promisee is risk-neutral or both parties are risk-averse. 22. See, e.g., Shavell, supra note 17; Jody S. Kraus, The Correspondence of Contract and Promise, 109 Colum. L. Rev. (forthcoming). This understanding of contract has a respectable pedigree. See Oliver Wendell Holmes, The Path of the Law, 10 Harv. L. Rev. 457, 462 (1897). But see Joseph M. Perillo, Misreading Oliver Wendell Holmes on Efficient Breach and Tortious Interference, 68 Fordham L. Rev. 1085 (2000). 23. Cf. Seana Valentine Shiffrin, The Divergence of Contract and Promise, 120 Harv. L. Rev. 708, 728 (2007); Kraus, supra note 22. Kraus emphasizes the difference—overlooked by economic analysis—between promising to do X or pay damages for breaching the promise to do X, and promising to do X or pay a similar sum of money, where payment of money constitutes performance. Most promisors would charge more to make the first promise, because in addition to the costs of doing X or paying money, it also imposes the reputational and personal costs of breaching a moral obligation if the promisor does not do X.

contract law

nonperformance not immoral, it is not even a breach of promise. Thus, while both absolutist deontologists and consequentialists can justify nonperformance in these cases, they actually have nothing to justify. The economic approach advocates efficient breach—a breach whose benefits exceed its costs—even when the contract clearly and explicitly does call for performance and excludes substitutive money payment. In such cases, the efficient breach doctrine favors breach, notwithstanding that under Shavell’s own definition, “performance is morally required.” Here lies the crucial difference between deontological theories attributing inherent value to private will and the economic theory which values free will only instrumentally.24 We conclude, that to make economic analysis of contract law morally acceptable without relinquishing the insights gained from economic methodology, it must be integrated with deontological constraints. In the next section, we demonstrate such integration in analyzing the doctrines of mistake and misrepresentation, and in the following section in the context of performance and breach.

D. Constrained Economic Analysis of Mistake and Misrepresentation •

1. A Brief Doctrinal Background The rules pertaining to mistake and misrepresentation in the contracting process are rather complex.25 All modern legal systems let the misinformed party avoid the contract under some, but not all, circumstances, and sometimes grant her monetary relief (even in circumstances that do not render

24. For further critique of Shavell’s argument, see Shiffrin, supra note 21. See also Steven Shavell, Why Breach of Contract May Not Be Immoral Given the Incompleteness of Contracts, 107 Mich. L. Rev. 1569 (2009). While Shavell argues that standard economic analysis is compatible with commonsense morality, others try to provide an economic rational for prevailing moral and legal intuitions concerning breaches and remedies. See, e.g., Oren Bar-Gill & Omri Ben-Shahar, An Information Theory of Willful Breach, 107 Mich. L. Rev. 1479 (2009). 25. See generally Restatement (Second) of Contracts §§ 151–73 (1981); 1 E. Allan Farnsworth, Farnsworth on Contracts §§ 4.9–4.15, at 465–501 (3d ed. 2004); 2 Farnsworth, id. §§ 9.2–9.4, at 589–623; Hein Kötz & Axel Flessner, European Contract Law, Vol. 1: Formation, Validity, and Content of Contracts; Contract and Third Parties 171–208 (trans. by Tony Weir, 1997) (providing a broad comparative analysis of the subject); David K. Allen, Misrepresentation (1988).

267

268


the contract voidable). Distinctions are commonly drawn between a common mistake and a unilateral one; and within unilateral mistakes, distinctions are made between mistakes actively induced by the other party, mistakes the other party knowingly or recklessly failed to correct, and mistakes the other party could not have reasonably prevented. False assertions are usually treated more severely than nondisclosure, which in turn requires classifications of nonverbal gestures and distinctions between half-truths (inducing false inferences) and mere silence.26 All legal systems also take into account the seriousness of the mistake. Unless the other party fraudulently deceived the misinformed party, misinformation renders a contract voidable only if the mistake refers to a fundamental or material fact affecting the decision to conclude the contract or to contract under the agreed terms.27 Further distinctions are drawn between different types of transactions, from speculative, commercial transactions— where disclosure duties are minimal—to consumer transactions—where the supplier’s duties are maximal.28 Fiduciary relationships give rise to stricter precontractual disclosure duties. Numerous additional factors, such as whether the uninformed party had equal access to the information, whether the informed party took active measures to conceal information, and the mental state of the party who provided the incorrect information or who failed to provide the correct one, play a significant role as well.29

26. See, e.g., Edwin Peel, Treitel on the Law of Contract §§ 9-123–9-154, at 424–40 (12th ed. 2007); Allen, supra note 25, at 21–23; Donald C. Langevoort, Half-Truths: Protecting Mistaken Inferences by Investors and Others, 52 Stan. L. Rev. 87 (1999). 27. Kötz & Flessner, supra note 25, at 178–83; Peel, supra note 25, §§ 9-013–9-016, at 367–69. 28. See, e.g., Sheldon Gardner & Robert Kuehl, Acquiring an Historical Understanding of Duties to Disclose, Fraud, and Warranties, 104 Comm. L.J. 168, 170–88 (1999); Geraint Howells & Stephen Weatherill, Consumer Protection Law 63–72, 395–435 (2d ed. 2005). Cf. Jules L. Coleman, Douglas D. Heckathorn & Steven M. Maser, A Bargaining Theory Approach to Default Provisions and Disclosure Rules in Contract Law, 12 Harv. J. L. Pub. Pol’y 639, 689–707 (1989) (arguing that where merchants “have ample endogenous resources for private contracting,” court intervention in the form of imposing legal disclosure duties is unnecessary). In developed economies, special rules apply to disclosure and deception in the stock market, where transactions are usually done anonymously through institutional intermediaries. We shall mostly leave this body of law outside of our discussion. 29. See, e.g., Restatement (Second) of Contracts §§ 160–162; Peel, supra note 26, §§ 9-026– 9-041, 9-129–9-130, 9-133–9-145, at 374–83, 427–29, 430–434; Jacques Ghestin, The Precontractual Obligation to Disclose Information: French Report, in Contract Law Today: Anglo-French Comparisons 151 (Donald Harris & Denis Tallon eds., 1989); Barry Nicholas, The Pre-contractual Obligation to Disclose Information: English Report, in Contract Law Today, id. at 166.

contract law

While there are considerable variations among legal systems, the general trend in the past century has been to move away from notions of caveat emptor toward greater solidarity in negotiation, meaning stricter prohibitions against deceit and increased disclosure requirements.30 The ambiguities resulting from this trend, the vast complexity of the rules, and the major practical importance of errors and misrepresentations in the contracting stage, have sparked lively theoretical discussions. We shall first describe the main contributions of standard economic analysis in this area, and then discuss the deontological perspective.

2. Standard Economic Analysis The economic efficiency of contracts presupposes that the parties are optimally informed about the costs, benefits, risks, and prospects involved in the contract. Only under this presumption may the contract be expected to benefit both parties (and society at large).31 It follows that a misrepresentation by one party is likely to detract from the contract’s efficiency, and that whenever one party possesses information that may affect the other party’s decision whether to contract and under what terms, a disclosure duty is warranted. Deliberate misrepresentation is inefficient because it disseminates misinformation. Were it allowed, contracting parties would never be able to rely on the information provided by the other party and would have to waste considerable resources in discovering the truth for themselves. As for disclosure duties, even if initially the mistaken party is better able to prevent the mistake, once the other party becomes aware of the mistake or has reason to know about it, she is often able to rectify it more cheaply. Disclosure duties prevent duplicate searches when one party already possesses the relevant information. Ordinarily, the cost of transferring the correct information to the uninformed party is small and its benefit clear. Accordingly, prohibitions against deceit and precontractual disclosure duties are prima facie efficient.32

30. Patrick S. Atiyah, The Rise and Fall of Freedom of Contract 771–78 (1979); 1 Farnsworth, supra note 25, § 4.11, at 470–78; Gardner & Kuehl, supra note 28; Ghestin, supra note 28. 31. Trebilcock, supra note 4, at 102. 32. Anthony T. Kronman, Mistakes, Disclosure, Information, and the Law of Contracts, 7 J. Legal Stud. 1, 2–8 (1978); Trebilcock, supra note 4, at 112. It should be noted, however,

269

270


This initial conclusion must, however, be qualified. At least four types of considerations may militate against antideception rules and disclosure duties. First, such prohibitions and duties may be superfluous if people can infer from the behavior of the other party what information the latter holds. Second, such rules may adversely affect the incentive to acquire information in the first place. Third, disclosure duties and antideception rules may hinder, rather than encourage, the dissemination of information in the market. Finally, dishonesty may facilitate efficient transactions when full candor would thwart contracting. Disclosure duties and prohibitions on deception may be unnecessary where concealment of information is not feasible or where sharing of accurate information is expected without legal rules. Sometimes the uninformed party can infer from the behavior of the other party what information the latter holds. For example, if a buyer has made inquiries about the sale object, and the seller knows about those inquiries because they required her cooperation, the seller may learn about the buyer’s information from the content of the buyer’s offer or from her mere willingness to negotiate.33 Furthermore, when information is verifiable and fraud is forbidden, economic incentives may suffice to induce disclosure.34 Disclosure duties may thus be unnecessary.

that communication costs are not always trivial. Choice of communication’s form and substance is significant especially when the relevant information is complex, and the attention span of the uninformed party is limited. See Richard Craswell, Taking Information Seriously: Misrepresentation and Nondisclosure in Contract Law and Elsewhere, 92 Va. L. Rev. 565 (2006). 33. See, e.g., Asher Wolinsky, Prices as Signals of Product Quality, 50 Rev. Econ. Stud. 647 (1983). Similarly, when consumers know that a manufacturer acquired information about the risk of a product that it manufactures, or even that such acquisition of information is feasible, they can infer from the manufacturer’s omission to disclose the results of the research that the product is of high risk. See, e.g., Steven A. Matthews & Andrew Postlewaite, Quality Testing and Disclosure, 16 Rand J. Econ. 328 (1985). 34. Suppose that prior to contracting only the seller knows the quality of her product and that the buyer will observe the quality thereafter. Now, if the product is of the highest quality, the seller will certainly tell the buyer, in order to charge a higher price. Otherwise, the buyer would infer from the seller’s silence that the object is not of the highest quality and will be willing to pay less for it. If the object is of second-best quality, the seller knows that the buyer will infer from her silence not only that the object is not of the highest quality (in which case the seller would not have kept silent), but also that it is not of the second-best quality (for a similar reason). A seller of a second-best product would therefore also reveal this information, and so forth. See Sanford J. Grossman, The Informational Role of Warranties and Private Disclosure of Product Quality, 24 J. L. & Econ. 461 (1981); Paul R. Milgrom, Good News and Bad News: Representation Theorems and Applications, 12 Bell J. Econ. 380 (1981).

contract law

Social norms and moral convictions can reinforce honesty and sharing of information as well. Disclosure duties and prohibition on lies may not only be unnecessary, but actually undesirable, having an adverse effect on the incentive to acquire information in the first place. Anthony Kronman has famously argued that while sharing information is likely to promote contracts’ efficiency, a disclosure duty weakens the bargaining power of the information holder and reduces her share of the contractual pie. It therefore dilutes her incentive to invest in acquiring the information ex ante. This argument distinguishes between deliberately and casually acquired information, and entails that disclosure duties should apply to the latter only.35 Arguably, to eliminate the disincentive for information gathering, sellers should not be allowed to ask prospective buyers whether they possess any information that might increase the value of the sales object. As keeping silent in response to such a question inevitably signals that such information does exist, efficiency may require that buyers be allowed to lie in such circumstances.36 Kronman’s argument disregards the important distinction between productive, or socially beneficial information, which enhances aggregate utility (such as technological innovations or information about the location of useful minerals), and purely distributive information, which does not contribute to the total welfare, but only enables its holder to get a greater share of the contractual surplus (such as mere foreknowledge).37 From a social point of view, cost-effective investments in acquiring productive information are desirable, and thus no disclosure duty should apply to such information (or, at the very least, such duties should be imposed with great caution). Conversely, while investment in acquiring distributive information is perfectly

35. Kronman, supra note 32, at 9–18. See also Matthews & Postlewaite, supra note 33 (arguing that imposing disclosure duties on manufacturers could discourage product testing). 36. Saul Levmore, Securities and Secrets: Insider Trading and the Law of Contracts, 68 Va. L. Rev. 117, 137–42 (1982); Randy E. Barnett, Rational Bargaining Theory and Contract: Default Rules, Hypothetical Consent, the Duty to Disclose, and Fraud, 15 Harv. J.L. & Pub. Pol’y 783, 794–801 (1992); infra p. 286. But see Michael J. Borden, Mistake and Disclosure in a Model of Two-Sided Informational Inputs, 73 Mo. L. Rev. 667, 688–96 (2008) (pointing out that sometimes, disclosing the fact that a party possesses deliberately-acquired, secret information does not necessarily deprive her of most of her bargaining power, as the information—such as a scientific discovery—is inaccessible to the uninformed party). 37. Jack Hirshleifer, The Private and Social Value of Information and the Reward to Inventive Activity, 61 Am. Econ. Rev. 561 (1971); Cooter & Ulen, supra note 11, at 279–86; Steven Shavell, Acquisition and Disclosure of Information Prior to Sale, 25 Rand J. Econ. 20 (1994).

271

272


rational from the point of view of the prospective bargainer, it is socially wasteful. An effective way to eliminate the incentive to invest in acquiring such information is to impose a disclosure duty. Further complications stem from the fact that often the same piece of information has both socially beneficial and distributive effects. The proposition that no disclosure duty should apply to deliberately acquired, productive information requires further qualification. Depending on the structure of the market, a strong incentive to acquire productive information may stem not only from the prospect of gaining informational advantage over one’s counterpart but also from a desire to gain such advantage over one’s competitors. For example, a bidder may acquire information about a prospective project to outbid her competitors. In such cases, the concern that disclosure duties might excessively reduce the incentive to acquire beneficial information may be overstated.38 More obvious examples would be information a landowner deliberately obtains about her property and medical information a person gets from undergoing a health check. The incentive to effectively use one’s property and to look after one’s health may be strong enough to secure the obtaining of the relevant information, even if a precontractual disclosure duty is imposed on land vendors and on people purchasing life insurance.39 Having discussed the arguments that disclosure duties and antideception rules may be unnecessary and may adversely affect incentives to gather information in the first place, we now turn to the claim that disclosure duties may actually reduce, rather than enhance, the dissemination of information in the market. In a well-functioning market, prices reflect the supply and demand for any asset, thus creating accurate incentives for their production and consumption. When a participant in the market possesses favorable or unfavorable information regarding a certain asset or expected market fluctuations, the mere buying or selling of considerable amounts of that asset, based on privately held information, increases the demand or the supply of that asset. This, in turn, brings the asset’s price closer to its true value given the new information. Arguably, disclosure duties are likely to obstruct this process. Facing a choice between disclosing information and refraining from making the contract, information holders may well opt for the latter, thus

38. Ofer Grosskopf & Barak Medina, A Revised Economic Theory of Disclosure Duties and Break-up Fees in Contract Law, 13 Stan. J.L., Bus. & Finance 148 (2007). 39. Kronman, supra note 32, at 25.

contract law

depriving the market of valuable information. This consideration applies to both deliberately and casually acquired information, and it legitimizes the nondisclosure of mere foreknowledge.40 In response to this argument, it has been pointed out that when information pertains to a specific asset in a market of heterogeneous assets, nondisclosure does not send the right signal to the market. Suppose that a buyer, who has a reason to believe that a certain parcel contains oil, does not disclose this information to the seller and buys the parcel at its regular market price. If anything, this transaction signals that there is no oil in that parcel. Moreover, even if the price paid is somewhat higher than the regular market price, in the absence of information about the oil, landowners of comparable parcels (but with no prospects for oil) may falsely infer that their parcels are more valuable than they are.41 Even when commodities are homogenous and the privately held information pertains to the entire market, the social benefit from the transaction may be smaller than the costs of obtaining the foreknowledge ex ante. This would be the case if the information is expected to become public shortly (as in the famous Laidlaw case),42 or if the information holder is only capable of buying or selling a small quantity which is unlikely to affect market prices (which might have well been the case in Laidlaw).43 The appropriate scope of the argument that nondisclosure facilitates efficient dissemination of information thus merits careful delineation. The fourth and final type of qualification to the prima facie efficiency of antideception rules and disclosure duties refers to situations in which nondisclosure and even lies may facilitate efficient transaction that would otherwise fail. Assume, for example, that a certain project requires assembling a considerable number of parcels. Revealing the entrepreneur’s plan to prospective sellers would result in a market failure known as a “holdout” or an “assembly problem.”44 Exploiting her monopoly power (without the consent of all landowners, the entire project will not come to fruition), each landowner might demand an excessive price, which could frustrate the

40. Barnett, supra note 36, at 794–801; Christopher T. Wonnell, The Structure of a General Theory of Nondisclosure, 41 Case W. Res. L. Rev. 329, 340–46, 351–60 (1992). 41. Melvin A. Eisenberg, Disclosure in Contract Law, 91 Cal. L. Rev. 1645, 1672 (2003). 42. Laidlaw v. Organ, 15 U.S. (2 Wheat.) 178 (1817). 43. Eisenberg, supra note 41, at 1669–73. 44. Lloyed Cohen, Holdouts and Free Riders, 20 J. Legal Stud. 351 (1991).

273

274


entire project. Concealing, or even lying about, the purpose of the purchase may thus facilitate efficient transactions.45 In the same vein, lying about one’s reservation price may arguably facilitate successful bargaining in bilateral monopoly scenarios.46 These analyses have appreciably contributed to our understanding of the potential effects of disclosure duties and antideception rules on the contracting parties and on society at large. Some of the economic insights have become accepted wisdom even outside of the law and economics literature.47 At the same time, some normative implications of these analyses—including the unqualified endorsement of lies in certain circumstances—are problematic. These analyses also largely fail to account for extant legal doctrines.48 It is therefore worth examining the deontological perspective and its possible contribution to economic analysis of these issues.

3. Deontology and Deception A vast philosophical literature debates the definition and the moral wrongness of lies and deceptions. Generally speaking, deceiving is defined as intentionally causing another person to have a false belief. Lying is a special case of deceiving: making an assertion the speaker believes to be false with the intention to deceive another person about the content of that assertion.49 Thus, while lying requires an active assertion, deception requires no assertion

45. Levmore, supra note 36, at 141–44. 46. Larry Alexander & Emily Sherwin, Deception in Morality and Law, 22 Law & Phil. 393, 441–45 (2003). See also infra pp. 280–81. 47. See, e.g., James Gordley, Mistake in Contract Formation, 52 Am. J. Comp. L. 433, 451 (2004) (advocating a disclosure duty but agreeing that “there should be an exception if one of the parties has expended money or effort to acquire the information”); Principles of European Contract Law, Parts I and II, prepared by the Commission on European Contract Law 232 (Ole Lando & Hugh Beale eds., 2000). 48. Kimberly D. Krawiec & Kathryn Zeiler, Common-Law Disclosure Duties and the Sin of Omission: Testing the Meta-Theories, 91 Va. L. Rev. 1795, 1818–21, 1856–73 (2005) (arguing, on the basis of a large-scale statistical survey of American cases, that courts are not more likely to require disclosure when the information is casually acquired as opposed to deliberately acquired). See also Eric A. Posner, Economic Analysis of Contract Law After Three Decades: Success of Failure?, 112 Yale L.J. 829 (2003) (arguing that economic analysis of contract law in general has largely failed to explain, or to justify reform in, contract law). 49. See generally James E. Mahon, The Definition of Lying and Deception, in Stanford Encyclopedia of Philosophy (2008), http://plato.stanford.edu/entries/lying-definition.

contract law

(one may say things or behave in a way she knows will cause others to draw false inferences, without asserting anything) and may involve mere silence. Lying and other forms of deception that actively/intentionally inflict harm on the deceived infringe the deontological constraint against harming other people. It may be argued that the very act of lying, or of actively/ intentionally deceiving, inevitably infringes the constraint against harming others since it adversely affects the autonomy of the victim (according to an objective theory of well-being), frustrates her typical preference not to be deceived (according to a preference theory), and is likely to make her unhappy if and when truth is revealed (according to a mental state theory).50 In response, one may contend that at least some of these harms are either uncertain or insignificant, and that some lies are actually benevolent. According to this view, if lying/deceiving is wrong only because it infringes the constraint against harming other people, then not every lie—and certainly not every deception—infringes a constraint.51 Whether or not active/intentional deception necessarily harms the deceived, many ethicists believe that lying is inherently wrong even if it causes no harm. Thus, Kant, who advocated an absolute prohibition on false assertions, held that lying destroys the human dignity of the liar, does not respect the rationality of the deceived, and fails to treat her as an end.52 While physical coercion uses someone’s person as a means to the coercer’s ends, lying manipulates one’s reason. It thus directly violates her autonomy, her capacity to decide for herself on the basis of true facts what ends she would like to pursue.53 Whatever their effects in any particular instance, lies are said to violate a constitutive rule of language use. According to this universal rule, an assertion implies truth telling. Lying is therefore always

50. Kagan, supra note 3, at 108–13. 51. For the view that deception is often beneficial, and in fact essential for healthy human relationships (and hence truth telling is morally overrated), see David Nyberg, The Varnished Truth: Truth Telling and Deceiving in Ordinary Life (1993). 52. Immanuel Kant, The Metaphysics of Morals 182 (Roger Sullivan ed., Mary Gregor trans., 1996); Immanuel Kant, On a Supposed Right to Tell Lies from Benevolent Motives, in Kant’s Critique of Practical Reason and Other Works on the Theory of Ethics 361 (Thomas K. Abbott ed. & trans., 6th ed. 1923). On the negative effects of lying on the liar and her self respect, see Sissela Bok, Lying: Moral Choice in Public and Private Life 23–28 (1978). 53. Christine M. Korsgaard, The Right to Lie: Kant on Dealing with Evil, 15 Phil. & Pub. Aff. 325, 330–37 (1986); Charles Fried, Right and Wrong 62–69 (1978). As mentioned above, those who deny that there is an independent constraint against lying may insist that this disrespect for autonomy is harm in itself.

275

276


morally wrong.54 Relatedly, every lie, regardless of its concrete outcomes, may be seen as promise breaking (the implicit promise to tell the truth).55 Every lie undermines the mutual faith and trust underlying social life.56 Since absolute deontology holds that one must not lie even for the sake of avoiding terrible consequences (such as the murder of an innocent person), absolutists have made considerable efforts to differentiate lying from other forms of deception, that are not considered equally immoral. For instance, lying may be distinguished from saying things that are technically true but are known or even meant to induce false inferences (such as half-truths). Lying may similarly be distinguished from acting in a manner that is known or meant to create a false impression but that involves no assertion, and from keeping silent where others are likely to draw incorrect conclusions from one’s silence.57 To eschew the counterintuitive implications of an absolute constraint against lying, philosophers have further narrowed the definition of lying by excluding false assertions made to people who arguably have no right to receive true information, such as aggressors and liars.58 Consequentialists, in contrast, need not struggle with such delicate distinctions and delimitations. To be sure, many consequentialists grant that

54. On this view, see Alasdair MacIntyre, Truthfulness, Lies, and Moral Philosophers: What Can We Learn from Mill and Kant?, 16 Tanner Lectures on Human Values 307, 311–12 (Grete B. Peterson ed., 1995); Fried, supra note 53, at 68. 55. Fried, supra note 53, at 67. For a critique of this view, see Thomas E. Hill, Jr., Autonomy and Self-Respect 27–28 (1991). 56. See, e.g., Bok, supra note 52, at 18–20; Robert C. Solomon, What a Tangled Web: Deception and Self-Deception in Philosophy, in Lying and Deception in Everyday Life 30, 36–41 (Michael Lewis & Carolyn Saarni eds., 1993); MacIntyre, supra note 54 (discussing offense against trust as one of the two central traditional bases for the immorality of lying, the other one being offense against truth); Geoffrey J. Warnock, The Object of Morality 83–85 (1971). 57. Alexander & Sherwin, supra note 46, at 400–04. See also Roderick M. Chisholm & Thomas D. Feehan, The Intent to Deceive, 74 J. Phil. 143 (1977) (identifying eight different ways of deception, based on the distinctions between commission and omission, between causing a person to have a false belief and not having a true belief, and between causing a person to continue having a false belief (or continue not having a true one) and not giving up a false belief (or continue not having a true belief). The distinction between active deception and nondisclosure follows the basic deontological distinction between commission and omission and reflects widely held moral intuitions. See, e.g., Jonathan Haidt & Jonathan Baron, Social Roles and the Moral Judgments of Acts and Omissions, 26 Eur. J. Soc. Psychol. 201 (1996) (finding that subjects judge active lying as morally worse than intentional withholding the truth). 58. Bok, supra note 52, at 14–15; MacIntyre, supra note 54, at 341–42; Fried, supra note 53, at 69–75.

contract law

lying is usually undesirable and therefore endorse a strong presumption against lying.59 Since, however, the morality of any presentation, misrepresentation, or anything else is determined by its overall outcomes, exact classifications are immaterial.60 Moderate deontology can avoid the extreme implications of absolutism without overly narrowing the definition of lying. When the outcomes of compliance with the imperative not to lie are bad enough, the moderate may openly admit that the constraint against lying is infringed and yet hold that the infringement is justified. Still, to give meaning to a threshold constraint against lying/deceiving, one must draw the contours of the constraint. As we are not interested in normative ethics for its own sake, we will not try to delineate the scope of this constraint in the abstract. Rather, we shall focus on the specific context of precontractual interactions, and will further restrict ourselves to that part of morality that contract law may reasonably take into account, given the institutional and other limitations of the law. The next section outlines the pertinent constraint and integrates it with the economic analysis of precontractual mistake and misrepresentation.

4. Constrained Economic Analysis (a) Integrating Constraints A fully developed framework for analyzing precontractual mistakes and misrepresentations would require a detailed analysis of the pertinent moral factors, characterization of different contractual environments, and a careful study of institutional factors and of the interplay between the law, nonlegal norms, and people’s behavior. Deriving legal rules from such analysis would also require consideration of the relationships between the contract law doctrines of mistake and misrepresentation and other legal doctrines, such as the torts of fraudulent and negligent misrepresentation

59. See, e.g., John Stuart Mill, Utilitarianism, reprinted in On Liberty and Other Essays 129, 154–55 (John Gray ed., Oxford Univ. Press, 1991) (1859). 60. Jeremy Bentham, An Introduction to the Principles of Morals and Legislation 205, Ch. XVI, sec. 1, para. XVI.31) ( J.H Burns & H.L.A Hart eds., 1982) (“Falsehood, take it by itself, consider it as not being accompanied by any other material circumstances, nor therefore productive of any material effects, can never, upon the principle of utility, constitute any offence at all”); Bok, supra note 52, at 49.

277

278


and the rules of contract breach and remedies.61 Formulating such fullfledged theory of precontractual deception and disclosure duties, including the available sanctions, exceeds the scope of the present discussion. Based on the above analysis of the economics and morality of precontractual deception and on our general framework for deontologically constrained CBA, we shall only describe how the permissibility of precontractual deception can be analyzed using threshold functions, and what choices should be made in this regard. These functions could then be integrated into an otherwise standard economic analysis of precontractual deception and disclosure duties. We propose that, to better understand and assess the legal doctrines of mistake and misrepresentation, the constraint against precontractual deception should have the following basic characteristics: (1) it should encompass only deception adversely affecting the interests of the deceived (excluding “harmless deception”); (2) it should not cover deception pertaining to one’s reservation value; (3) it should only cover intentional deception, thus excluding inadvertent and negligent misrepresentation or nondisclosure; (4) it should prohibit all forms of deception, including false assertions, half-truths, and silence; yet (5) different thresholds should apply to different forms of deception; (6) the size of the threshold should be a function of the harm inflicted by the deception; and (7) all types of costs and benefits should in principle be taken into account within the threshold functions (though some costs and benefits may possibly be discounted when some forms of deception are involved). In the next subsection, we shall also propose that— contrary to standard CBA—the doctrines of mistake and misrepresentation should reflect options not to promote social utility. Excluding Harmless Deception To normatively analyze precontractual deception, one need not decide whether lying (or any other form of deception) is immoral per se or only when it adversely affects the interests of the other party. For the law, unlike morality, rarely, if ever, deals with “harmless lies,” that is, lies which only offend

61. Willful and negligent nondisclosure gives rise to tort liability in many legal systems. When a seller conceals a hidden defect in the sale object, the buyer may either seek to avoid the contract for misrepresentation or sue for the removal of the defect or for damages for a breach of a warranty.

contract law

the dignity of the liar and the autonomy of the deceived.62 The centralist, coercive nature of legal enforcement, its various institutional limitations, and its high operational costs mandate this restraint. Thus, even if there is a deontological constraint against lying that is independent of the general constraint against harming other people, this constraint should not be incorporated into the analysis of precontractual deception. The general constraint against harming other people does not apply to any interest a person may have, but usually only to violations of fundamental rights.63 In contrast, the constraint against harmful precontractual deception may in principle refer to any harm. Deception causing even a moderate monetary loss infringes the constraint. This broad incidence is justified by the immorality of the manner in which the harm is inflicted, namely the deception. To infringe the constraint, the magnitude of the harm to the deceived must not, however, be too minor. If knowing the true facts would not have appreciably affected the decisions of the deceived, the deception may be deemed too insignificant to infringe any constraint. Plausibly, the minimally required harm should vary according to the forms of deception, as further discussed below. Arguably, an exception to the exclusion of harmless deception should be recognized when deceiving does not affect the interests of the other party (for example, because she knows the truth), yet it reveals the dishonesty and unreliability of the deceiver. In fact, no exception is needed. When deceiving does not affect the interests of the other party, the constraint against precontractual deception would not apply. However, save for very short and trivial transactions, contracts entail reliance on the honesty and trustworthiness of the other party (inter alia, because the legal enforcement mechanism is rather weak). Blatant dishonesty between contracting parties, as in some cases of insincere promises, may thus palpably harm the interests of the deceived as it reduces the probability that the promisor will perform her obligations.64 Deception indicating that a negotiating party is dishonest and unreliable is therefore not necessarily “harmless.”

62. Alexander & Sherwin, supra note 46, at 432 (“[L]egal rules do not track, even loosely, arguments based on the liar’s violation of natural law, his own humanity, or the rules of language. . . . Nor does law correspond to moral theories that emphasize the effect of deception on the victim’s autonomy.”). 63. See supra p. 42. 64. Ian Ayres & Gregory M. Klass, Insincere Promises: The Law of Misrepresented Intent (2005).

279

280


Excluding Deception as to One’s Reservation Value Conventions regarding legitimate bargaining techniques vary from one market to another, yet it is generally accepted that people need not disclose— and may even lie about—their reservation value. A seller is not required to disclose the minimal price at which she will be willing to sell and in some markets may even mislead the buyer to believe that selling at a lower price would drive her out of the market. A buyer may create the false impression that she has lost interest in the contract because she can get a better bargain elsewhere.65 One may conjecture that in some contexts such deceptions are efficient. A buyer who believes that the seller’s reservation price is higher than it actually is and a seller who believes that the buyer’s reservation value is lower than it really is, may be more likely to reach an agreement than parties who know for certain that there is a huge gap between their reservation values.66 At the same time, lack of information, or misinformation, about the other party’s reservation price, and thus about the size of the surplus, may obstruct contracting due to wrong expectations and misunderstandings. Assuming that such practices infringe the constraint against precontractual deception, it is difficult to determine whether enough good outcomes are produced by these practices to render them permissible. There is, however, a prevailing notion that at least some of these deceptions (as well as mere puffs) are permissible regardless of the good they produce (if at all). This notion may rest on grounds of consent and fairness. Often, an otherwise impermissible active/intentional harming of another person is deemed permissible if the latter consents to the harm. Such consent is plausibly implicit when people knowingly and voluntarily participate in bargaining where these forms of deception are deemed legitimate and not incompatible with standards of fair dealing. Just as a boxer cannot legitimately complain when her opponent punches her, bargainers cannot complain about the rules of the game in which they have freely chosen to participate.67 Since neither of the parties ordinarily wishes to expose her reservation valuation, she cannot legitimately expect the other party to expose

65. See, e.g., Marcel Kahan, Games, Lies, and Securities Fraud, 67 N.Y.U. L. Rev. 750, 753–60 (1992). 66. Alexander & Sherwin, supra note 46, at 444–47. 67. Bok, supra note 52, at 129–33. As Bok points out, this argument is weaker the less one’s choice to bargain under these rules is voluntary. Cf. Fried, supra note 4, at 80–85.

contract law

hers, and should therefore agree to mutual deception in this regard.68 Note that none of these grounds commits one to the dubious claim that the very fact that the other party is a liar eliminates the prohibition on deceiving. It is also worth noting that our analysis of reservation value cannot be generalized to all precontractual misrepresentations. There is no reason to believe that negotiating parties expect their partners to misrepresent material facts pertinent to contracting decisions. Empirical studies do not point to a dichotomy between people’s expectations regarding behavior in the marketplace and behavior in other contexts.69 Intentionality The deontological constraint against deception in precontractual relations does not cover inadvertent or negligent misleading. There may be good reasons for the law to render contracts made on the basis of mistaken assumptions voidable even if the mistake was not induced by one of them or was induced inadvertently or negligently. There may similarly be sound reasons to impose tort liability for negligent misrepresentations. These reasons, however, have nothing to do with deontological constraints as ordinarily conceived.70 To infringe the constraint against deception, a negotiating party should thus intentionally, that is, at least knowingly, deceive the other party. It follows that to analyze the desirability of rules governing inadvertent or negligent misrepresentations and nondisclosure one may use standard CBA, free of deontological constraints. Intentionality certainly covers both rare cases in which the deceiver wishes to harm the other party per se, and the more common cases in which harming the other party serves as a means to attaining a legitimate goal,

68. One may also argue that deceiving about one’s reservation price is harmless (and thus, as explained above, not covered by the constraint against precontractual deception). This is because people do not, and should not, believe such representations, or because no one has a legitimate expectation to any particular division of the transaction’s surplus. This argument is, however, problematic. People are sometimes misled by representations about the other party’s reservation price (otherwise, nobody would have bothered making them), and agreeing to a worse bargain than one would have agreed to absent the deception is a harm under our definition. 69. See, e.g., Stewart Macaulay, Non-contractual Relations in Business: A Preliminary Study, 28 Am. Soc. Rev. 55, 58 (1963) (describing the inclination of businesspeople to rely on “common honesty and decency”). 70. Heidi M. Hurd, The Deontology of Negligence, 76 B.U. L. Rev. 249 (1996); Larry Alexander, Negligence, Crime, and Tort: Comments on Hurde and Simons, 76 B.U. L. Rev. 301, 301–02 (1996).

281

282


such as increasing one’s profits. We tend to think that it also covers cases in which the harm is a mere side effect.71 For instance, a person may refuse to reveal, or even lie about, her sexual orientation for reasons of privacy, even in circumstances where this information is not deemed legally irrelevant. The fact that disadvantaging the other party (who made the contract on the basis of the false information) is an expected side effect of attaining another goal does not, in and of itself, entail that no constraint is infringed. Only if attaining the other goal produces enough good outcomes or avoids enough bad ones, is the threshold met.72 At the same time, the fact that the foreseen harm to the other party is a mere side effect—rather than a goal or a means to attaining some legitimate goal—does considerably lower the threshold. Encompassing All Forms of Deception Deontological theories, particularly absolutist theories that prohibit even “harmless lies,” often draw delicate distinctions between lies and other forms of deception. For instance, they may refrain from condemning telling half-truths, behaving in a way intended to make others draw false inferences, and misleadingly staying silent. Since the constraint against precontractual deception that we propose applies only to harmful and intentional deception, there seems to be no compelling reason to restrict its application to lies (that is, intentionally misleading assertions), or to active deception (including half-truths and nonassertive misleading behavior). Both in terms of the blameworthiness of the deceiver and harm to the deceived, deception by omission, by nonassertive utterances or behavior, or by half-truths, may be as severe as a lie. Arguably, a constraint against intentionally deceiving by omission raises concerns similar to those raised by a general constraint against not aiding other people, which underlies the basic doing/allowing distinction.73

71. On the distinction between harming a person as a side effect of aiding or saving other people, and using a person as a means to aiding or saving others, see supra pp. 44–46 and references therein. 72. To use a more familiar example, diverting a trolley to a track on which one person will be killed as a mere side effect of saving the life of an animal standing on the other track violates a constraint. On the requirement of proportionality between the good produced by the intended goal and the harm inflicted as a side effect, see Shelly Kagan, The Limits of Morality 151–65 (1989); Alison McIntyre, Doing Away with Double Effect, 111 Ethics 119, 221–23 (2001). 73. See supra pp. 43, 45–57. See also supra pp. 60–63.

contract law

However, several limitations on the scope of the present constraint assuage this fear of overdemandingness. First, the constraint covers only intentional, harmful omissions and not inadvertent or negligent ones. Second, the constraint does not cover any intentional, harmful omission but only intentional, harmful deception. Third, the constraint applies only between negotiating parties. Once people are negotiating a contract, they are no longer strangers. Some degree of decency and good faith toward each other may be expected of them. Arguably, by voluntarily entering into negotiations, the parties implicitly agree to abide by the rules governing these relations. Fourth, as noted above, the constraint against deception applies only if the harm inflicted by the deception is large enough. The minimal harm varies from one form of deception to another. Thus, in the case of deception by silence, it would not suffice that the information holder knows that the other party lacks some piece of information; it must be a significant piece of information that could reasonably influence her decision to contract under certain terms.74 Finally, within the framework of moderate deontology, even an infringement of this relatively narrow constraint may be permissible if the good produced by such an infringement exceeds a certain threshold. Different Thresholds for Different Forms of Deception While the restriction of the proposed constraint to harmful deception and the move from absolutist to moderate deontology considerably reduce the importance of the distinctions between different forms of deception, these distinctions are nevertheless important because different forms of deception entail different threshold functions.75 Threshold functions applying to different forms of deception may vary in one or more of the following aspects: (1) the minimum harm necessary for the constraint to be infringed; (2) the threshold that has to be met to render the infringement permissible; and (3) the type of costs and benefits that are deemed relevant in determining whether the net benefit of the deception meets the threshold. From a deontological point of view, precontractual deception may be classified into the following categories: (1) spontaneous lies, (2) lies in response to a question, (3) half-truths, (4) nonassertive misleading utterances and behavior, (5) evading questions, and (6) mere silence. In addition,

74. An additional requirement might be that the information holder not only knows of the other person’s mistake but also that the mistake is significant in the described sense. 75. On setting the size of the threshold, see generally supra pp. 93–96.

283

284


within each category, the morality of the deception depends on whether deceiving the other party is the actor’s intended goal, a means to attain her goal, or merely a foreseen side effect. Spontaneous lies seem more reproachful than lies in response to a direct question because in the latter case it may well be that absent the question, the deceiver would have remained silent. There have been suggestions in the philosophical and legal literature that some questions are illegitimate and therefore a lie in response to such a question does not infringe any constraint.76 We doubt this view. The two oft-quoted examples are those of the murderer asking about the whereabouts of his intended victim and the landowner asking a purchaser, who invested a lot of money in finding subsurface minerals in the seller’s land, whether she got any information that might affect the value of the land. In both cases, so it seems to us, lying would infringe the constraint against deception. Lying is nevertheless clearly permissible in the first case because it prevents horrific outcomes, thus overcoming the constraint. It is possibly permissible in the second case as well, because it enables the investor to reap the fruits of her investment, thereby maintaining the incentive to obtain socially beneficial information.77 Many ethicists consider lies, whether spontaneous or in response to a question, to be morally worse than half-truths and nonassertive misleading utterances and behavior. A common justification offered for this distinction is that the responsibility for the false inference rests, at least partially, upon the deceived who has drawn the false inference.78 This rationalization is not wholly persuasive. If we take the speakers’ meaning to be what the speaker invites hearers to understand from what she says, then half-truths and other knowingly misleading utterances are not necessarily different from lies, even if the wrongness of the latter lies in the violation of the basic

76. See, e.g., Fried, supra 53, at 73–78; Barnett, supra note 36, at 799–801. Consequentialists may resort to a similar argument. See, e.g., Henry Sidgwick, The Methods of Ethics 318 (7th ed. 1907). 77. To defend the claim that no constraint is infringed in lying to the murderer, one may argue that when a constraint is justifiably infringed to advance good outcomes, the person who was harmed usually deserves compensation, an apology, or at least an explanation—none of which seems necessary in the present case. However, while compensation or an apology is indeed appropriate when an innocent person is harmed to promote good outcomes, this is not necessarily the case when the infringement is induced by the culpable behavior of the harmed person. Hence, the fact that a person deserves no compensation, apology, or explanation does not necessarily imply that no constraint has been infringed. 78. See, e.g., MacIntyre, supra note 54, at 336–37. See also Langevoort, supra note 26, at 89.

contract law

rule of language use.79 Even noncommunicative deception (such as putting on a swimsuit in order to mislead others into believing that one is about to go swimming), may be seen as an active breach of trust, though it does not necessarily invite a false inference and does not violate a rule of language use. Jonathan Adler has argued that by choosing other forms of deception rather than simply lying, the deceiver implicitly undertakes lesser responsibility for the truthfulness of the hearer’s implicature, and conveys a message of moral effort to avoid outright lies. The availability of this choice is desirable as a means to enhance “social harmony, lessening the strains of commitment, and facilitating the exchange of information,” even if the deceiver’s choice is made for other reasons.80 Evading a question seems morally worse than mere silence for at least two reasons. First, in the case of silence, the deceiver may repress the need to consciously decide whether to reveal privately held information, whereas a question posed by the other party forces one to make this decision (and evasion means a decision to deceive). Second, while in the case of mere silence a person may not know for sure whether the other party possesses the relevant information or whether the information is important for her, once the question is posed, it is quite clear that the other party lacks the relevant information and that she thinks it is material. At the same time, evading a question seems less morally wrong than lying or telling a half-truth. It is passive and it does not similarly invite reliance on one’s assertions. It may even signal to the other party that she should undertake an independent inquiry into the relevant issue. While we insist that half-truths and other forms of deception, including silence, infringe the constraint against precontractual deception, we believe that there are morally significant differences between different forms of deception. These differences entail different thresholds. As one gradually moves from a spontaneous lie to a mere silence, the amount of positive outcomes necessary to overcome the constraint gradually decreases.81 In addition, within each category, the threshold should be lower for deception which is merely an expected side effect of attaining another goal.

79. Jonathan E. Adler, Lying, Deceiving, or Falsely Implicating, 94 J. Phil. 435, 444 (1997). 80. Id. at 448–52. As Adler concedes, however, some people may abuse the distinction. 81. Cf. Langevoort, supra note 26, at 91–92 (suggesting to situate half truths “on a continuum roughly half way between the duty to avoid affirmative misrepresentation and the more controversial and contingent duty to reveal hidden private information”).

285

286


To be sure, some legal economists argue that while the imposition of disclosure duties should be subject to rigorous CBA, lies are simply inefficient and hence forbidden. For instance, Richard Posner has stated that since the liar “makes a positive investment in manufacturing and disseminating misinformation,” and since this investment is socially wasteful, lies are “naturally” prohibited.82 This statement is problematic. Whereas discovering or producing accurate information often involves considerable investments, lying is usually very cheap. For example, discovering subsurface minerals in the vendor’s land may necessitate considerable efforts by the prospective purchaser, but lying about this discovery only requires one to say “no” in response to the landowner’s inquiry whether the purchaser has any special information about the land, or to spontaneously say: “I need your land for pasture.” As Saul Levmore has demonstrated, from an efficiency perspective, in the absence of other adequate incentives, permitting nondisclosure might not be enough to encourage the obtaining of socially beneficial information; permission to lie is necessary as well.83 To the extent that one accepts that there is a normative difference between various forms of deception—as the law seems to acknowledge—this difference cannot persuasively be sustained on efficiency grounds. The only way to bridge the gap between standard CBA on the one hand, and commonsense morality and extant legal doctrine on the other hand, is by integrating deontological constraints with CBA of precontractual deception. Threshold’s Dependence on the Magnitude of the Harm The amount of harm caused by precontractual deception varies from case to case. It differs both in its absolute magnitude and in its magnitude relative to the scope of the contract. Similar harms in similar contracts may have very different adverse effects on the well-being of the deceived, depending on her overall situation. As in other contexts, we believe that the size of

82. Richard A. Posner, Economic Analysis of Law 111 (7th ed. 2007). See also Kronman, supra note 32, at 19, n.49; Borden, supra note 36, at 692 (pointing out the long-term, adverse effects of permitting lying). 83. Levmore, supra note 36, at 137–42. See also Kim Lane Scheppele, Legal Secrets: Equality and Efficiency in the Common Law 164 (1988) (pointing to inconsistency in Kronman’s approval of nondisclosure of deliberately acquired information and his objection to all fraud).

contract law

the threshold should be a function of the size of the expected harm. The greater the expected harm, the higher the threshold should be.84 Types of Costs and Benefits In other contexts we have suggested that from a deontological perspective, some types of costs and benefits—such as very small benefits, benefits and costs that are lexically inferior to the harm prohibited by the constraint, chronologically remote costs and benefits, and probabilistic ones—should not bear on the permissibility of infringing certain constraints, or should be subject to a discount rate.85 Since the harms intentionally inflicted by precontractual deception are often solely or primarily pecuniary, most of the conceivable excluders of types of costs and benefits seem inappropriate. Intentionally lying or withholding information is very different from killing, torturing, or even discriminating against people on the basis of their skin color. In particular, we maintain that the threshold function should take into account the effects of allowing deception on the incentives to acquire socially beneficial information ex ante. Although such long-term effects are not certain and may be chronologically remote, this is not a sufficient reason to disregard them. We are less sure whether such effects should be excluded, or at least further discounted, when their probability is very low. Possibly, low-probability benefits should be excluded from the threshold function for lies but not from the functions for deceptively evading questions and keeping mum. In some contexts, there is a strong, common intuition that an infringement of a constraint may be permissible to avoid sufficiently large bad consequences, such as the premature death of innocent people, but not to promote any amount of good consequences, such as prolonging people’s ordinary life expectancy.86 This distinction is also relevant to deception.87 Typically, in contractual settings, if deception produces any positive outcomes at all, these outcomes are of the latter category: incentivizing people to ex ante obtain socially beneficial information, disseminating information in the market, or facilitating efficient contracting. We do

84. See supra pp. 93–96. See also supra pp. 150, 155, 169–71, 174–75, 211–13, and 252–54; infra p. 342. 85. See supra pp. 86–93, 147–49, 195–211, and 254; infra pp. 342–46. 86. See supra pp. 91–93. 87. Bok, supra note 52, at 78–81.

287

288


not suggest excluding such outcomes from the threshold functions. Still, the fact that the constraint against precontractual deception is ordinarily infringed to promote good outcomes rather than to eliminate bad ones, weighs in favor of subjecting these benefits to a discount rate (or, alternatively, in favor of a higher threshold). This means that, to render deception permissible, its net benefit should be quite sizable (and larger net benefit would be required as one moves from mere silence to half-truths and to false assertions). There are also atypical scenarios in which precontractual or preagreement deception seeks to avoid bad outcomes. To mention but two examples, a person who desperately needs a loan to finance a life-saving medical treatment may conceal facts relevant to her (in)ability to repay the loan; and interrogators of a suspected terrorist may falsely promise to let her free if she would provide information about the location of a ticking time bomb.88 Clearly, precontractual deception is more easily justifiable in such cases than in the typical ones. Hence, the good produced by eliminating bad outcomes should not be subject to any discount rate. If, indeed, deception is permissible in these cases, the resulting contracts/agreements are not voidable for mistake or misrepresentation. In the loan example, the conclusion that the lender cannot avoid the contract does not mean that the borrower need not repay the loan. On the contrary, the validity of the contract implies that she should repay it and will be liable for breach of contract if she does not. Similarly, in the interrogation scenario, the agreement is not voidable for misrepresentation. After revealing the information, the suspected terrorist is actually the one who insists that the agreement should be kept. In this case, the state may be relieved of its obligation on the grounds that its promise was given under an illegal threat or according to a doctrine authorizing the termination of governmental agreements on grounds of public considerations (surely the state cannot avoid the agreement by arguing that it never intended to keep it).89

88. Another example is provided by a judgment of the Israeli Supreme Court, dealing with the validity of agreement to end a mutiny in a military prison, made between the army authorities and the rebellious inmates, which included an undertaking by the authorities not to pursue charges against the inmates. Kogan v. Military Attorney General, 51(5) P.D. 67 (1997). 89. Both arguments were made by the army authorities in Kogan v. Military Attorney General, and the latter was adopted by the court. Id. at 77–96.

contract law

(b) Integrating Options This book primarily discusses deontological constraints. However, moderate deontology also diverges from consequentialism in allowing people to (sometimes) further their own interests or the interests of their loved ones or their community, rather than promote the overall good. Deontology allows for agent-relative options.90 Plausibly, options (like constraints) have thresholds: when a sufficiently large good (or bad) is at stake, the option is overridden by the requirement to maximize the overall good. The size of the threshold may (or may not) depend on the size of the sacrifice needed on the part of the agent.91 The notion of options can help in understanding the law of precontractual disclosure duties, including cases that do not involve deception, such as negligent nondisclosure. Since economic analysis centers on incentives for future behavior, economic analyses of precontractual nondisclosure and deception have naturally paid a great deal of attention to cases where disclosure duties and prohibitions on deception may affect people’s ex ante behavior. These are primarily the cases in which information is acquired through deliberate and costly efforts. These cases, however, are not typical. Estimating the exact percentage of these cases would require an extensive empirical study of reported and unreported cases. Yet, extensive reading of Israeli case law and unsystematic reading of case law from around the world reveals that adjudicated disputes revolving around deliberately acquired information are uncommon. Most precontractual mistake and misrepresentation cases refer to casually acquired information about the particular transaction or the transaction’s object.92

90. See supra pp. 41, 46–47, 60–63, and 98–103. See also supra pp. 33–40. 91. Kagan, supra note 3 at 161–70; supra pp. 98–103. 92. See also Craswell, supra note 32, at 568 (stating that cases “where one side makes a costly investment to acquire information . . . are relatively rare”). Our (and Craswell’s) assessment is different from the findings of Krawiec and Zeiler’s study, according to which only 80% of the cases in their dataset (371 out of 466) involved casually acquired information (Krawiec & Zeiler, supra note 48, at 1838). The remaining 20% include both cases in which the deliberately acquired information was intrinsic to the specific transaction (such as latent defects or hidden potential of the property), and extrinsic information (particularly about market conditions) (id. at 1800–02). Only 14 cases in the entire dataset involved extrinsic information and out of these 14, only 6 pertained to deliberately acquired information—the fact pattern of the famous Laidlaw case and much of the ensuing economic literature (id. at 1860). Contrary to the authors’ characterization, we found that all of the 6 cases (which the authors have kindly directed us to), actually involved intrinsic

289

290


With regard to casually acquired information—which constitutes the lion’s share of the cases—economic analysis tends to enjoin broader disclosure duties than most legal systems actually impose. Take, for example, material information that is equally or almost equally accessible to both parties but that only one of them actually possesses. Even if the informed party is unsure whether the other party possesses the pertinent information, since the costs of disclosure are usually trivial and their contribution to assuring the contract’s efficiency is significant, a disclosure duty would be efficient. Such a duty, it is true, would reduce the incentives of the other party to attain the same information by herself, but this is not a genuine cost, as the search for information that one party already possesses is wasteful. The concern that a general disclosure duty would inhibit the incentive of both parties to attain the pertinent information (as each party would expect the other party to get the information and share it) is not particularly worrisome either. Each party still has a substantial incentive to obtain information that could strengthen her bargaining position. Since in the cases under discussion (which constitute the great majority of cases) at least one party has casually obtained the information, no incentive for deliberate inquiry is necessary. Economists who resent market regulation may find efficiency grounds for restraining disclosure duties even with regard to casually acquired information. For instance, Richard Posner opposes a general disclosure duty regarding the characteristics of consumer goods, pointing out that some information is readily available by casual inspection at the time of purchase. Even if such inspection is impracticable, if the product is inexpensive and is purchased repeatedly, the cost to the consumer of ascertaining the product’s characteristics by using it is still very low—merely the cost of the first purchase. Sometimes the conformity of the product to the consumer’s needs and taste can be determined by her alone, and hence no disclosure would do. Even if none of these arguments holds, “governmental intervention” in the form of imposing disclosure duties may be unnecessary because competitive pressures would make sellers offer warranties, which are even better than disclosure duties.93 To these arguments, one may add that providing information may be costly, especially if the information is complex.

information, and in at least 5 of them the information was casually acquired (one case was unclear on this point). We have not examined other cases in the dataset. 93. Posner, supra note 82, at 113.

contract law

These arguments are more persuasive in some contexts than in others. At any rate, it seems that the law approves of much more nondisclosure than standard economic analysis could accept. It seems that the primary ground for this approval lies in the law’s recognition that it should not (and cannot) require everybody to always do what would maximize total utility. Just as the recognition of constraints results in prohibiting efficient deception when the amount of good outcomes produced by the deception does not meet the threshold, the recognition of options results in allowing inefficient nondisclosure when the loss of welfare due to such nondisclosure does not meet the option’s threshold. This conception can explain the significance of the accessibility of the information to the other party. When information is equally, or almost equally, accessible to both parties, it is easier to legitimize the informed party’s choice to look after her own interests rather than maximize the good, because the other party is in some way responsible for not getting the information.94 To conclude: while economic analysis of precontractual deception and disclosure duties has greatly contributed to our understanding of the incentive and allocative effects of conceivable legal rules, it neglected some of the crucial normative factors. At the same time, much of the rich philosophical literature on the morality of lying is not directly useful for legal policy-making. We submit that incorporating deontological constraints against precontractual deception (and deontological options not to share information) with economic analysis would improve the normative and interpretative analysis of this complex issue.95

94. The relative accessibility of the information features prominently in Scheppele’s theory of precontractual disclosure. Scheppele, supra note 83, at 111–78. Krawiec & Zeiler (supra note 48, at 1813–15, 1852–55) indeed found out that “courts are significantly more likely to require disclosure when the transaction was one in which the parties had unequal access to information,” and that what drives case outcomes is the combination of the casual acquisition of information and the unequal access (but see supra note 92). 95. We have not tried to formulate a complete theory of precontractual deception and disclosure. It is worth noting, however, that some such theories fall in line with our general framework. Thus, in his comprehensive analysis of disclosure in contract law, Melvin Eisenberg has proposed a detailed set of rules to govern precontractual disclosure duties (assuming lies are forbidden). Eisenberg’s set of rules is guided by what he calls the Disclosure Principle. According to this principle, “the law should require disclosure of material facts except in those classes of cases in which a requirement of disclosure would entail significant efficiency costs.” Eisenberg, supra note 41, at 1648. This principle “puts a thumb on the scale . . . in favor of disclosure, because of the efficiency and moral reasons that support disclosure. To overcome this presumption, it is not enough that in a given class of cases a requirement of disclosure would entail some relatively slight efficiency costs. Instead, the

291

292


E. Remedies for Breach of Contract 1. A Brief Doctrinal Background

Remedies for breach of contract may schematically be divided into three types: enforced performance, substitutionary monetary relief, and termination of the contract (ordinarily coupled with recovery of property transferred or money paid).96 Putting termination aside, there are roughly three models regarding the relationships between enforced performance and monetary damages: in some legal systems, such as in Germany, enforced performance is the rule while monetary substitute is the exception; in others, primarily Common Law systems, damages are the standard remedy and specific performance the exception; and in yet other systems, such as the French, enforced performance is the standard remedy for breach of some types of obligations and monetary damages the remedy for other types of obligations. Notwithstanding these dissimilar points of departure, legal systems largely converge in terms of the remedies actually sought by aggrieved parties and awarded by courts. In practice, specific performance mostly refers to the transfer of unique objects (primarily real estate), and much less so to fungible goods, personal services, or to complex projects where compliance is hardly observable or verifiable.97 The declared goal of damages in practically all legal systems is to bring the aggrieved party as close as possible to the position she would have been in had the contract been duly performed (thus protecting her expectation interest). Ordinarily, damages for breach of contract do not aspire to put the aggrieved party in the position she would have been in had she not made the contract at all (that is, protecting her reliance interest). Yet, when it is difficult to establish the aggrieved party’s expectation interest (often

presumption is overcome only if disclosure would entail significant efficiency costs.” Id. at 1655–56. 96. For general surveys of the law of contract remedies, see G.H. Treitel, Remedies for Breach of Contract: A Comparative Account (1988); Principles of European Contract Law, supra note 47, at 359–459. On American law, see 3 Farnsworth, supra note 25, §§ 12.1–12.20c, at 148–383. 97. See generally Treitel, supra note 96, at 43–74; Louis J. Romero, Specific Performance of Contracts in Comparative Law: Some Preliminary Observations, 27 Les Cahiers de Droit 785 (1986); Henrik Lando & Caspar Rose, On the Enforcement of Specific Performance in Civil Law Countries, 24 Int’l Rev. L. & Econ. 473 (2004).

contract law

due to the speculative nature of the question how much she would have profited from the contract had it been performed), the reliance measure may serve as a minimal approximation of her expectation interest. This approximation assumes that people make profitable contracts; hence, the contract benefits are at least as large as the costs. At the same time, practically all legal systems deny the right to reliance damages when the breacher proves that reliance surpasses expectation.98 Rarely does a legal system entitle the aggrieved party to the extra profits made, or losses avoided, by the breacher as a result of the breach, when these gains exceed the aggrieved party’s expectation interest (thus denying protection of the disgorgement interest).99 Sometimes, especially when the contract has been terminated for the breach, the aggrieved party is entitled to protection of her restitution interest, that is, to have the benefits she has conferred on the breaching party restored to her. In most legal systems, under most circumstances, the aggrieved party can get restitution in excess of expectation.100 Legal systems differ regarding the availability of damages for nonmonetary, emotional harms; yet, in general, such damages are less obtainable than damages for monetary losses.101 Punitive damages are not ordinarily awarded for breach of contract.102 To some extent, the parties may contract around ordinary remedy rules. The parties can, for example, stipulate the damages for breach. Such stipulations, however, as well as other agreed deviations from ordinary remedy rules, are subject to judicial scrutiny, and some of them are simply ineffective.

98. See Restatement (Second) of Contracts § 349; 3 Farnsworth, supra note 25, § 12.16, at 284–88; Torsten Schackel, Der Anspruch auf Ersatz des negativen Interesses bei Nichterfüllung von Verträgen, 2001 Zeitschrift Europ. Privatrecht 248, 249–50 (German and Austrian law). 99. See generally James Edelman, Gain-Based Damages: Contract, Tort, Equity and Intellectual Property 149–89 (2002). 100. See, e.g., 1 George E. Palmer, The Law of Restitution §§ 4.1-4.6, at 363–427 (1978); Hanoch Dagan, The Law and Ethics of Restitution 282–89 (2004); Andrew Skelton, Restitution and Contract (1998); Dan B. Dobbs, Dobbs Law of Remedies §§ 12.7(1), 12.7(5), at 162, 178 (1993); 3 Farnsworth, supra note 25, § 12.20, at 334; Principles of European Contract Law, supra note 47, at 422–26; infra pp. 303–05. A fifth goal of contract remedies is to restore the contractual equivalence by adapting the aggrieved party’s obligations to the actual performance by the breaching party. See Eyal Zamir, The Missing Interest: Restoration of the Contractual Equivalence, 93 Va. L. Rev. 59 (2007). 101. Treitel, supra note 96, at 194–201. 102. Id. at 78–79.

293

294


Freedom of contract is thus more restricted in the domain of remedies than in other aspects of the contract’s content and performance.103

2. Standard Economic Analysis From an economic point of view, remedies for breach of contract are primarily evaluated according to the incentives they create for the parties at different stages of the contractual process: prior to and at the time of contracting (for example, the decision whether to enter into a contract, with whom, and under what conditions; how much information to gather before contracting and what information to share with the other party); after contracting (for the promisor, how much effort and what precautions to take to ensure performance and how much to look for alternative opportunities; for the promisee, to what extent to rely on the contract and whether to prepare for its potential breach); at the performance stage (whether to perform or to breach), and even later (what measures to take to mitigate the loss in case of breach, whether to sue for the breach, etc.). In any of these stages and with regard to both parties, economic analysis endorses rules creating incentives for behavior that would maximize aggregate social utility, which, in the absence of externalities, means maximization of the joint contractual surplus. Focusing on the performance of the contract and the promisor’s precautions, economic analysis begins with the notion that remedy rules should induce the promisor to perform and take precautions to avoid breach as long as performance and such precautions are efficient, and to breach and avoid such precautions if the breach is efficient. This is the well-known efficient breach theory.104 A perfect protection of the promisee’s expectation interest through damages supposedly creates an optimal incentive in that sense. Full expectation damages make the promisee indifferent between performance and breach, while at the same time make the promisor (and society at large) better off. Expectation damages are necessary because they force

103. Dobbs, supra note 100, § 12.9, at 245–73; 3 Farnsworth, supra note 25, § 12.18, at 300 (maintaining that the parties’ “power to bargain over their remedial rights is surprisingly limited”). 104. For early versions of the theory, see Robert L. Birmingham, Breach of Contract, Damage Measures, and Economic Efficiency, 24 Rutgers L. Rev. 273 (1970); John H. Barton, The Economic Basis of Damages for Breach of Contract, 1 J. Legal Stud. 277 (1972). For a formal analysis, see Shavell, supra note 11.

contract law

the breaching party to internalize the costs her breach inflicts on the other party. The efficient breach theory, and especially its implication that expectation damages are preferable to specific performance, is controversial.105 Yet even economists who advocate specific performance in at least some circumstances share the view that there is nothing inherently immoral in breaching a contract as long as the breach is efficient. Another important factor is the extent of the promisee’s reliance on the contract. Often, the promisee can increase the value of the promisor’s performance by making investments and foregoing alternative courses of action. While such reliance increases the value of performance, it also increases losses in case of nonperformance. Whereas full expectation damages presumably generate optimal incentives for performance and breach by the promisor, creating optimal incentives for the promisee’s reliance requires denying her of any damages for breach or alternatively awarding her damages uncorrelated to either her expectation or reliance interests.106 In setting remedy rules, one should also consider the parties’ relative risk-aversion. At least theoretically, full expectation damages provide the promisee with a kind of insurance—an effect both parties would find desirable if the promisee is risk-averse and the promisor risk-neutral. However, if both parties are risk-averse and if purchasing commercial insurance is impracticable, sharing the costs of unintended breach by reducing damages below expectation damages, or sharing the prospect of extra profit from an alternative transaction by setting the damages above expectation, may be mutually beneficial. In that case, the exact measure of damages depends on the relative risk-aversion of the two parties. If the promisee is risk seeking, she would prefer a lower measure of damages coupled with price reduction. Finally, even a risk-averse promisee is unlikely to be willing to pay a premium for insuring against harms and injuries that would not affect her wealth.107

105. See Anthony T. Kronman, Specific Performance, 45 U. Chi. L. Rev. 351 (1978); Alan Schwartz, The Case for Specific Performance, 89 Yale L.J. 271 (1979); Thomas S. Ulen, The Efficiency of Specific Performance: Toward a Unified Theory of Contract Remedies, 83 Mich. L. Rev. 341 (1984); Daniel Friedmann, The Efficient Breach Fallacy, 18 J. Legal Stud. 1 (1989). 106. On the “Paradox of Compensation,” see generally Cooter & Ulen, supra note 11, at 261–67. 107. A. Mitchell Polinsky, Risk Sharing Through Breach of Contract Remedies, 12 J. Legal Stud. 427 (1983).

295

296


If renegotiation is feasible at zero cost, efficient performance or breach will occur regardless of remedy rules. If the remedy is too low to induce efficient performance, the promisee would “bribe” the promisor to perform; if the remedy is too high to induce efficient breach, the promisor would “bribe” the promisee to give up performance. While renegotiation costs are almost invariably positive (and even if they were zero, renegotiation entails redistribution of the contract surplus, which may lead to inefficient strategic behavior and may adversely affect the parties’ incentives regarding reliance), the possibility of renegotiation must be taken into account when establishing efficient remedy rules.108 Another important factor is the probability that the aggrieved party will actually obtain the legal remedy. Many breaches are unobservable or at least unverifiable. Even when a breach is observable and verifiable, the aggrieved party will not attain her legal remedies if the various costs involved are prohibitively high—which is very often the case. The lower the probability of obtaining the remedy, the lower the incentives for precautions and performance on the part of the promisor; which in turn induces suboptimal reliance on the part of the promisee. This concern arguably justifies multiplying the measure of damages by 1/p, where p is the probability that the damages will actually be paid.109 A countervailing concern has to do with nonlegal sanctions for breach, including the risk of retaliation, long-term damage to one’s reputation, and social disapproval. To induce optimal precautions and performance by the promisor (assuming no externalities), the totality of sanctions for breach— legal and nonlegal—should presumably equal the promisee’s expectation interest. The more powerful the nonlegal sanctions are, the greater the fear of overdeterrence generated by full expectation damages. Arguably, this concern justifies a decrease in the measure of damages.110

108. Richard Craswell, Contract Remedies, Renegotiation, and the Theory of Efficient Breach, 61 S. Cal. L. Rev. 629 (1988). 109. Daniel A. Farber, Reassessing the Economic Efficiency of Compensatory Damages for Breach of Contract, 66 Va. L. Rev. 1443 (1980); Alan Schwartz, The Myth that Promisees Prefer Supracompensatory Remedies: An Analysis of Contracting for Damage Measures, 100 Yale L.J. 369, 395–405 (1990); A. Mitchell Polinsky & Steven Shavell, Punitive Damages: An Economic Analysis, 111 Harv. L. Rev. 869, 936–38 (1998). 110. Robert Cooter & Ariel Porat, Should Courts Deduct Nonlegal Sanctions from Damages?, 30 J. Legal Stud. 401 (2001).

contract law

This list of incentive and allocative effects is far from exhaustive.111 Additional concerns include the incentives necessary to induce the promisee to share information about her expected losses from breach before and after contracting;112 difficulties in ascertaining the promisee’s subjective valuation of performance;113 incentivizing the promisee to mitigate her losses once a breach occurs, is imminent, or is merely probable;114 and the courts’ difficulties in obtaining and verifying the complex information necessary to consider all of the above factors. Not surprisingly, while the desirability of expectation damages remains the point of departure for any economic analysis of remedy rules, one could find efficiency arguments supporting almost any other remedy rule.115 On a more abstract level, economic analysis equates remedies for breach of contract to other doctrines of contract law, all of which aim at maximizing the “contractual pie” (assuming no externalities). Thus, damages are not seen as a means to rectify some wrong but merely as an execution price of the option to unilaterally nullify a party’s contractual obligations (“exit rules”). From the economic perspective, the enduring insistence to classify opting out of a contractual obligation as a “breach,” as well as the nonlegal sanctions that typically accompany it, is puzzling.

111. See generally Richard Craswell, Instrumental Theories of Compensation: A Survey, 40 San Diego L. Rev. 1135 (2003). 112. Ian Ayres & Robert Gertner, Filling Gaps in Incomplete Contracts: An Economic Theory of Default Rules, 99 Yale L.J. 87 (1989) (introducing the notion of penalty default rules and explaining the foreseeability rule as an incentive to efficiently share information at the contracting stage). 113. Timothy J. Muris, Cost of Completion or Diminution in Market Value: The Relevance of Subjective Value, 12 J. Legal Stud. 379 (1983). Cf. Alan Schwartz & Robert E. Scott, Market Damages, Efficient Contracting, and the Economic Waste Fallacy, 108 Colum. L. Rev. 1610 (2008). 114. Melvin A. Eisenberg, Actual and Virtual Specific Performance, the Theory of Efficient Breach, and the Indifference Principle in Contract Law, 93 Cal. L. Rev. 975, 1021–24 (2005); Alexander J. Triantis, Timing Problems in Contract Breach Decisions, 41 J. L. & Econ. 163 (1998). 115. See Richard Craswell, Against Fuller and Perdue, 67 U. Chi. L. Rev. 99, 111 (2000); Barak Medina, Renegotiation, ‘Efficient Breach’ and Adjustment: The Choice of Remedy for Breach of Contract as a Choice of a Contract-Modification Theory, in Comparative Remedies for Breach of Contract 51, 56–61 (Nili Cohen & Ewan McKendrick eds., 2005); Polinsky, supra note 15, at 69; Steven Shavell, Foundations of Economic Analysis of Law 360 (2004).

297

298


3. Deontology: Promises, Harms, and Contractual Obligations Most deontologists hold that breaching a contract infringes a constraint. One should perform one’s contractual obligation even if more good could be attained by breaching it. According to a view famously endorsed by Charles Fried, this constraint stems from the liberal ideal, from the notion that every person is an independent entity, autonomously deciding what to think and do while respecting the autonomy of other people. Following Kant, every person is an end in herself, and must not be used as mere means. Promises are the key to enjoying the skills, efforts, and property of other people without adversely affecting their liberty. When a person makes a promise, it enables the promisee to legitimately rely on the promise. The promise creates confidence and trust. Holding a person responsible for her past expression of will respects her as a continuing entity and promotes the mutual trust created by the promise.116 According to some deontological theories, the duty to keep a promise arises once a promise is made, even if the promisee has not yet relied on the promise and the promisor has not yet benefited in any way from making it, and even if no such reliance or benefit will ever occur.117 This position may rest on the notion that the constraint against promise breaking is independent of the constraint against harming people, or, alternatively, on the notion that any promise breaking ipso facto harms the autonomy or welfare of the promisee. Rival theories deny both the existence of such an independent constraint and the contention that any promise making and breaking necessarily harms the promisee. According to the latter theories, the scope of the constraint is narrower: only if the promisee has actually changed her position (either by acting in a certain way or by foregoing alternative courses of action), or the promisor has actually benefited from the promise, is the

116. Fried, supra note 4, at 7–17. The philosophical literature on promising is abundant. Recent contributions include: T.M. Scanlon, What We Owe to Each Other 295–327 (1998); Niko Kolodny & R. Jay Wallace, Promises and Practices Revisited, 31 Phil. & Pub. Aff. 119 (2003); David Owens, A Simple Theory of Promising, 115 Phil. Rev. 51 (2006). 117. Fried, supra note 4, at 9–12; Markovits, supra note 7, at 1442–46, 1491–514. For a critique of Fried’s theory, see, e.g., Anthony Kronman, A New Champion for the Will Theory (Book Review), 91 Yale L.J. 404 (1981); Patrick S. Atiyah, The Liberal Theory of Contract, in Essays on Contract 121 (1986). For a critique of Markovits’s theory, see Seana Valentine Shiffrin, Promising, Intimate Relationships, and Conventionalism, 117 Phil. Rev. 481, 493–96 (2008).

contract law

promisor morally obliged to keep her promise. These theories not only narrow the incidence of the constraint, but plausibly also limit the compensatory duty arising from violating the constraint to protecting the promisee’s reliance or restitution interest, rather than her expectation interest.118 A common thread of several theories of the moral duty to keep one’s promise is their reliance on the existence of a social practice or a convention of promising.119 Arguably, the very possibility of making promises depends on the existence of such practice or convention. Once such socially beneficial and fair practice/convention exists, a person who voluntarily participates in the practice and enjoys its benefits must not break her promise.120 Making and breaking a promise would both harm the promisee who (given the practice) relied on the promise and adversely affect the practice itself.121 Practice theories of promising are subject to various critiques, the most compelling being that the existence of a practice is either insufficient or unnecessary to justify the constraint against promise breaking. Even if the practice of promising is fair and socially beneficial, it is unclear why in a particular instance a promisor is bound by the practice. If the answer is that by the very act of promising, the promisor commits herself, or promises, to abide by the rules of the practice, then the argument becomes circular, as it assumes that a person is bound by her promises.122 If the answer is that the practice itself is justified, so that every moral person must abide by it because of its content, then the duty to keep promises does not stem from the existence of the practice but rather from the inherent moral justification for keeping promises. Finally, if one has to follow the practice because of the benefit one derives from the promise (ordinarily through inducing a reciprocal promise) or from the harm inflicted on the promisee who relied on the

118. See, e.g., Lon L. Fuller & William R. Perdue, Jr., The Reliance Interest in Contract Damages: 1, 46 Yale L.J. 52 (1936); Atiyah, supra note 6, at 177–215 (arguing that promises are mere admissions of pre-existing obligations resting on reliance or restitution). For a critique of Fuller and Perdue’s thesis, see, e.g., Craswell, supra note 115. For a critique of both theses, see Stephen A. Smith, Contract Theory 78–97 (2004). 119. David Hume, A Treatise on Human Nature: Vol. 1: Texts 331–37 (David Fate Norton & Mary J. Norton, eds., Oxford Univ. Press, 2007) (1740); John Rawls, Two Concepts of Rules, 64 Phil. Rev. 3 (1955); Fried, supra note 4, at 11–17. 120. John Rawls, A theory of Justice 301–08 (rev. ed. 1999); Fried, supra note 4, at 12–14. 121. This argument closely resembles the argument that lying both harms the deceived and violates a constitutive rule of language use. See supra pp. 275–76. 122. F.H. Buckley, Paradox Lost, 72 Minn. L. Rev. 775, 791–93 (1988).

299

300


promise, then once again these two factors would be sufficient to yield an obligation to keep promises—even absent the practice. In that last case, the scope of the constraint would possibly be narrower than under the view that reliance and benefit are unnecessary for the existence of the constraint (and yet, nothing would hinge on the existence of the practice).123 Whether there is an independent moral constraint against promise breaking per se or against harming other people by making and breaking promises, additional considerations impinge on the scope and outcomes of legally enforceable promises. Infringement of the constraint against promise breaking does not, in and of itself, warrant legal enforcement of a promise.124 Specifically, even those advocating an independent moral constraint against promise breaking may hold that harmless promise breaking should remain outside the law’s purview. In the same vein, even if morality views disrespect for the promisee’s autonomy or the mere disappointment experienced by the promisee as harms, the law may adopt a more restrictive definition and deny that such “harms” warrant legal treatment.125 This is a brief overview of some aspects of some deontological theories of promising and contractual liability. Alongside the theories focusing on the morality of promising, some theories build on notions of corrective justice126 and/or on a conception of contract as a transfer of rights or entitlements.127

123. Atiyah, supra note 6, at 112–14. For additional critiques of various aspects of different versions of practice/convention theories of promise, see Buckley, supra note 122; Scanlon, supra note 116, at 295–317; Shiffrin, supra note 117. 124. Joseph Raz, Promises in Morality and Law, 95 Harv. L. Rev. 916, 937 (1982); Smith, supra note 118, at 69–72; Shiffrin, supra note 23, at 709–19 (arguing that, while contract law need not enforce moral norms, it must not be incompatible with the conditions necessary for the flourishing of moral agency); Emmanuel Voyiakis, Contracts, Promises and the Demands of Moral Agency, in 10 Current Legal Issues: Law and Philosophy 288 (Michael D.A. Freeman & Ross Harrison eds., 2007) (advocating a division of labor between the morality of promising, which speaks to individuals, and contract law, which speaks to legal enforcement institutions). 125. Compare the analysis of harmless deception, supra pp. 278–79. 126. See Fuller & Perdue, supra note 118, at 56–57; Weinrib, supra note 13. While we propose to subject economic analysis to deontological constraints, Weinrib rejects any regard to consequences in formulating and applying contract law rules and private law rules in general. For this reason, his theory is incompatible with our proposed analytical framework. For a general critique of corrective justice theories of contract remedy rules, see Craswell, supra note 115, at 121–28. 127. See, e.g., Peter Benson, The Unity of Contract Law, in The Theory of Contract Law: New Essays 118 (Peter Benson ed., 2001); Daniel Friedmann, The Performance Interest in Contract Damages, 111 L.Q.R. 628 (1995); Weinrib, supra note 13. For a critical analysis of these theories, see Smith, supra note 118, at 97–103.

contract law

As we shall see in greater detail, there is a correlation between the basic characteristics of the various deontological theories of promise and contract (namely, whether they recognize an independent constraint against promise breaking, how they define harm, and whether they differentiate between morality and law); and their positions regarding the appropriate remedies for breach of contract. Yet this correlation is not particularly robust, as different conclusions about the appropriate remedies may be drawn from a single theory, and similar conclusions from dissimilar theories. Before discussing these competing conclusions and their implications for the feasibility of a deontologically constrained economic analysis of remedy rules, the next subsection argues that existing rules of contract remedies indeed reflect deontological morality.

4. Deontological Features of Contract Remedy Rules As indicated earlier and as will be further detailed, many rules and doctrines of contractual liability and remedies for breach are compatible with both standard economic analysis and deontological morality. There are, however, significant aspects of liability and remedy rules that are not reconcilable with the economic conception of breach, reflecting instead the notion that contract breach is an infringement of a deontological constraint. The very use of the term “breach” connotes a violation of a duty (rather than a legitimate choice), and the term “remedy” implies that something went wrong, that there is a difficulty or a problem that has to be corrected. The same is true of operational rules of contract law. Two simple examples will be used to substantiate this claim: liquidated damages versus bonuses and restitution exceeding expectation. Liquidated Damages vs. Bonuses Assume that a contractor undertakes to complete a project by a certain date X, at price P. The contract provides that for every day of delayed completion, the landowner will be entitled to $1000 in liquidated damages. Now assume a similar contract, where the agreed date of completion is X+100 and the price is P–$100,000. In the latter contract, there is a liquidated damages clause of $1000 for every day of delay but also a clause entitling the contractor to a $1000 bonus for every day of earlier completion, up to 100 days. The monetary consequences of completing the project at different dates seem to be identical under the two contracts and so are presumably the incentives created by the contracts.

301

302


This is not, however, the way contracting parties ordinarily perceive the two arrangements, and it is certainly not the way the law treats them. Assume, for example, that the project was completed on date X+50. Under the first contract, the contractor has breached her obligation to complete the contract by date X. The liquidated damages clause does not necessarily deprive the landowner of alternative or additional remedies for this breach.128 Even if the contract purports to exclude such alternative or additional remedies, this exclusion will be scrutinized; if the expected and actual loss due to the delay is significantly larger than the stipulated damages, the court may invalidate the exclusionary clause on the basis of legal doctrines such as unconscionability or pursuant to specific statutory authorization.129 At the same time, if the court views the liquidated damages as disproportionately high, it will void the clause in some legal systems, or reduce the liquidated damages to an acceptable level in others.130 Finally, if delay is due to unexpected and unavoidable circumstances which make completion on time impossible or impracticable, the contractor may not have to pay any damages at all, because the delay is not considered a breach under such circumstances.131 In contrast, under the second contract, completion on date X+50 does not constitute a breach but rather an expedited performance. Therefore, the landowner is not entitled to any remedy for breach of contract. Moreover, it is very unlikely that she could challenge the validity of the bonus clause on the ground that it is excessive, or that the contractor could challenge the bonus for being too low. Finally, the landowner could not avoid paying the bonus by claiming that the earlier completion was made possible thanks to unexpectedly favorable circumstances for the contractor. Each aspect of these very different legal effects of the two contractual arrangements may by justified—or criticized—on various grounds, including their incentive effects, the cognitive biases possibly affecting liquidated damages clauses, and the capacity of courts to accurately apply the pertinent norms.

128. Treitel, supra note 96, at 212–13, 214–19; Principles of European Contract Law, supra note 47, at 453–56; UCC § 2-719. 129. Treitel, supra note 96, at 216–17; French Civil Code, Art. 1152 par. 2; UCC § 2-718, Comment 3 (as amended in 2003) (“A liquidated damages term that provided for damages that are unreasonably small is likewise unenforceable”). See also UCC § 2-719. 130. Treitel, supra note 96, at 219–33. 131. See, e.g., UCC § 2-615.

contract law

What is more important for our purpose is, however, the undeniable fact that the legal treatment of the two arrangements around the world is radically different. While under the first contract completion at any date after X is treated as a breach (that is, as an infringement of a deontological constraint), under the second contract completion at any date prior to X+100 is treated as a legitimate realization of an option. For this reason, under the first contract—but not under the second—the landowner is entitled to additional remedies. For the same reason, undercompensatory liquidated damages may be viewed as an unconscionable exclusion of liability, whereas no complaint that the bonuses are too low (compared to the landowner’s additional gains from early completion) is likely to be accepted. Finally, whereas supervening circumstances may exempt the contractor from liability and hence from the payment of the liquidated damages, nothing of the sort may affect the entitlement of the contractor to the bonuses for early completion.132 Restitution Exceeding Expectation Suppose that a contractor is hired to do some work for a sum of $10,000, paid in advance. Before any work is done, the contractor repudiates, and the owner rightfully terminates the contract and sues for restitution. Suppose further that the contractor proves that the current market value of the same work has dropped to $8000, and perhaps the owner has even hired another contractor to do the work for this lower sum. In this hypothetical, standard economic analysis holds that the owner should only be entitled to a monetary relief of $8000 (assuming she has suffered no ancillary or consequential damage).133 The parties allocated to the contractor the risk that the costs and hence market price will increase, while the owner took upon herself the risk that costs and market price will decrease. Had market price increased, the owner would have rightfully been awarded the current, higher price, thereby putting her in the position she would have been in had the contract been performed. Restitution not capped by expectation puts the owner in a better position than she would have been in had the contract been performed. At least under ordinary circumstances, there is no reason to

132. See also infra note 137. 133. Andrew Kull, Restitution as a Remedy for Breach of Contract, 67 S. Cal. L. Rev. 1465 (1994).

303

304


assume that promisees would be willing to pay for this supracompensatory remedy, which is likely to reduce the contract surplus.134 Despite this sound economic reasoning, it is widely accepted that in such cases, the aggrieved party is entitled to full restitution.135 This rule suggests that contract law neither views contracts as mere means to efficiently allocating risks and prospects, nor perceives contractual obligations as options to perform or pay expectation damages. Rather, contracts create mutual trust and generate a moral duty to perform. Accordingly, the contractor cannot legitimately turn her back on her contractual obligations and at the same time rely on the contract to limit the owner’s remedy to her expectation interest. These examples demonstrate that existing contract law is more easily interpreted as reflective of a deontological conception of contractual liability than an economic or a consequentialist one. Additional examples are readily available.136 Other doctrines, such as impossibility and frustration of purpose, demonstrate that the law of contractual liability recognizes not only deontological constraints but also deontological options.137 As the next section

134. More generally, the option to choose ex post between different measures of relief in cases involving ex ante uncertainty places extra cost on the breacher and may thus result in incentives distortion. See Omri Ben-Shahar & Robert A. Mikos, The (Legal) Value of Chance: Distorted Measures of Recovery in Private Law, 7 Am. L. Econ. Rev. 484 (2005). 135. See supra note 100 and accompanying text. 136. For instance, no legal system has seriously considered the suggestion to reduce legal remedies for breach so as to avoid over-deterrence of breach due to the cumulative effect of legal and nonlegal sanctions for breach (see supra note 110 and accompanying text). Likewise, having to choose between over-reliance on the contract by the promisee due to the availability of full expectation damages, and between under-compliance by the promisor that would have resulted from denial of damages or decoupling damages from the promisee’s losses, all legal systems opt for the former. Alternative interpretations, such as the argument that the requirement of foreseeability in contract damages takes this consideration into account (Cooter & Ulen, supra note 11, at 264–66) are less compelling. It is questionable whether the foreseeability requirement is suited to attain this goal, and in any case it does not apply to specific performance, when available. Finally, no legal system awards negative damages, that is, letting the breacher recover the other party’s gains from the breach. This is so, despite the fact that such damages are at least sometimes necessary to induce an efficient breach. See Barry E. Adler, Efficient Breach Theory Through the Looking Glass, 83 NYU L. Rev. 1679 (2008). 137. These doctrines (and the related doctrine of impracticability) deal with unexpected and unavoidable supervening circumstances that make performance much more burdensome than initially expected. To date, it seems that all attempts to rest these doctrines on efficiency grounds have largely failed. See Posner, supra note 48, at 848–49; Yuval Procaccia, Contractual Accidents (unpublished manuscript, available at http://papers.ssrn. com/sol3/papers.cfm?abstract_id=1352358). Assume, for example, that the promisor’s

contract law

explains, we nevertheless have doubts regarding the feasibility of deontologically constrained economic analysis of contract remedies.

5. Challenges Facing Constrained Economic Analysis of Remedy Rules The most fundamental difference between consequentialist (including economic) and deontological conceptions of contract law is that only the latter recognize a constraint against contract breaching. This constraint is either a manifestation of the constraint against promise breaking, or an application of the general constraint against actively/intentionally harming other people— in this case, by making a contract and then breaking it. Conversely, standard economic analysis basically holds that contracts should be performed as long as performance is efficient and breached whenever breach is efficient. One could have assumed that this fundamental disagreement would yield different implications for remedies for breach of contract. As indicated, however, this is hardly the case. Various understandings of the meaning and scope of the constraint against contract breaching and various theories regarding the appropriate relationships between morality and law yield radically different conclusions about the appropriate goals of contract remedies. These differences are much greater than the gaps that separate some deontological theories of contract remedies and standard economic analysis. Before examining the implications of different deontological theories for contract remedies and hence for the integration of constraints into the economic analysis of remedy rules, we should address the claim that at least autonomy-based theories have no bearing whatsoever on the appropriate remedies. According to this argument, respect for individual autonomy requires that people be allowed to commit themselves to legally binding promises; yet it gives no reason to prefer any remedy over any other.138

expected cost of performance is 50, the agreed price is 60, the expected value of performance to the promisee is 700, and both parties are risk-neutral. Now assume that due to supervening circumstances, the cost of performance has unexpectedly tripled to 150. Performance is clearly more efficient than breach under these circumstances, and yet under appropriate circumstances practically all legal systems exempt the promisor from her liability. In this case, contract law seems to reflect a threshold option. 138. Fuller & Perdue, supra note 118, at 58; Richard Craswell, Contract Law, Default Rules, and the Philosophy of Promising, 88 Mich. L. Rev. 489, 517–20 (1989).

305

306


This claim proves too much. While autonomy-based theories do not necessarily entail expectation damages as the standard remedy for breach, they are not indifferent to remedies. If a breach of contract is regarded as intrinsically, morally wrong (and if this wrongness should be reflected in the law), then, all else being equal, one should prefer remedies providing the promisor with a stronger incentive to keep her promise. One should opt for remedies that more clearly express the inherent moral virtue of keeping one’s promise and condemnation for its breach. Reliance damages, for example, do not meet this criterion. Neither are other deontological theories unconcerned with contract remedies. If, for example, one holds that breaking a promise is not wrong in itself but only wrong to the extent that it harms the promisee, then the goal of remedies should be to redress this harm. In the same spirit, if the essence of promises and contracts lies in establishing relationships of recognition and respect, a kind of community, and if only remedies protecting the expectation interest “track the contracting parties’ transformation from strangers to collaborators,” then such remedies—rather than remedies protecting the reliance interest, which are available also to strangers—are warranted.139 Given these correlations between deontological theories of contract law and the appropriate remedies for breach, we can now examine how these implications impact the feasibility and fruitfulness of integrating deontological constraints with economic analysis of this sphere. Deontologists who hold that contracts are primarily a set of promises, that there is an independent moral constraint against promise breaking, and that the law should reflect this constraint, often advocate remedies aimed at protecting the aggrieved party’s expectation interest.140 The same is true of deontologists who perceive contracts as a kind of community or collaboration, and that conclude that only remedies protecting the expectation interest reflect the parties’ forward-looking commitment to affirmatively treat each other as ends in themselves.141 These theories thus converge with standard economic analysis which, at least as its starting point, favors remedies protecting the aggrieved party’s expectation interest. Furthermore, given that specific performance entails greater curtailment of the promisor’s liberty, whenever expectation damages adequately protect

139. Markovits, supra note 7, at 1503–14. 140. See, e.g., Fried, supra note 4, at 17–21. Cf. Weinrib, supra note 13, at 65–70. 141. Markovits, supra note 7, at 1503–14.

contract law

the promisee’s expectation interest, there is at least a prima facie deontological reason to prefer expectation damages over specific performance.142 This argument, too, accords with the common economic view that when expectation damages adequately protect the promisee’s expectation interest, they should be preferred over specific performance. Admittedly, there are many reasons to doubt that current doctrines of contract damages successfully put the aggrieved party in the same position she would have occupied but for the breach.143 However, this concern, too, is shared by standard economic analysis and by the deontological theories under consideration. Incorporating a constraint reflecting this brand of deontological theories into economic analysis of remedy rules is therefore unlikely to change the analysis. This (perhaps surprising) conclusion is even stronger when considering deontological theories which hold that contract remedies should only aim to protect the promisee’s reliance interest. These theories contend that contract breach is not a legitimate option open to the promisor but rather an infringement of a moral constraint against harming other people or against illegitimately enriching oneself at their expense. Yet they do not support harsher remedies than those regularly advocated by economists but lesser ones. Rather than reinforcing the immorality of contract breaching, incorporating a deontological constraint based on these theories may actually facilitate more breaches. Moreover, if one accepts that, but for its disregard for the immorality of contract breach, standard economic analysis adequately analyzes the incentive effects of different remedy rules, then one must reject this brand of deontological theories. If people are rational maximizers of their utility, contracts are made to promote the parties’ utility, and expectation—rather than reliance—damages incentivize the parties to act so as to maximize their utility, then the reliance interest must not be the standard measure for contract remedies.

142. Dori Kimel, From Promise to Contract: Towards a Liberal Theory of Contract 95–109 (2003). 143. The typical outcome of under-compensation is due to the difficulties of establishing the loss with sufficient certainty, the foreseeability requirement, the mitigation of loss rule, the reluctance to award emotional distress damages, the limited or no recovery of legal costs, and so forth. See generally Treitel, supra note 96, at 143–207; Eisenberg, supra note 114, at 989–97; John A. Sebert, Jr., Punitive and Nonpecuniary Damages in Actions Based upon Contract: Toward Achieving the Objective of Full Compensation, 33 UCLA L. Rev. 1565, 1566–71 (1986); Stewart Macaulay, The Reliance Interest and the World Outside the Law Schools’ Doors, 1991 Wis. L. Rev. 247, 249–53.

307

308


We are left with those deontological theories which maintain that remedies for breach of contract should be more stringent than expectation damages, to reflect the immorality of contract breaches. Such remedy rules may strive to protect the aggrieved party’s disgorgement, reliance, or restitution interests even when these interests surpass her expectation interest. They may regard specific performance as a standard remedy, minimize the promisee’s burden to mitigate her losses, more generously award damages for emotional harm, fully enforce liquidated damages clauses even when they seem to be unreasonably excessive, and allow for punitive damages for culpable breaches.144 Arguably, a fruitful analysis of the desirability of such supracompensatory remedies may be done through the integration of deontological constraints with economic analysis. But this argument moves too quickly. For one thing, there are powerful arguments—some of which are distinctively deontological and others that most deontologists would not wish to ignore—against such harsh remedies. For another, there are (as indicated above) significant efficiency arguments in favor of supracompensatory remedies. As regards the first type of arguments, remedies that considerably curtail the promisor’s liberty may infringe a constraint against harming the promisor.145 These would include specific performance, or monetary remedies that are so high as to leave the promisor with no practical alternative but to perform her obligations. In the same vein, repealing or minimizing the mitigation of losses rule would contradict the special moral obligation a contracting party owes her counterpart, even if the latter has breached her obligations. Plausibly, as part of the mutual trust, cooperation, and good faith expected between contracting parties, one should avoid losses that can be avoided “without undue risk, burden or humiliation.”146 In fact, doing away

144. Shiffrin, supra note 23, at 722–27. Shiffrin argues that the following features of American contract law fail to create the necessary conditions for the flourishing of moral agency: the primacy of damages rather than specific performance as remedy for breach, the denial of punitive damages for intentional breaches, the limitation of damages for consequential losses to those losses that have been reasonably foreseeable at the time of contracting, the mitigation of damages rule, and the unenforceability of punitive damages clauses. For the claim that a promise-based theory of contract should embrace specific performance as the standard remedy for breach, see Atiyah, supra note 117, at 124. See also Smith, supra note 118, at 418–20 (“[F]rom the traditional rights-based view of contract law, the refusal to punish deliberate breach is a genuine puzzle”). For a critique of these views, see Kraus, supra note 22. 145. See supra note 142 and accompanying text. 146. Restatement (Second) of Contracts § 350(1).

contract law

with the mitigation of losses rule may be injurious to the conception of contract as a cooperative relationship in quite the same way as is the efficient breach theory.147 With respect to the suggestion to indiscriminately enforce liquidated damages clauses even when they seem to set unreasonably large damages, it seems to disregard the prevailing cognitive biases and other problems due to which contracting parties often agree to unreasonably and inefficiently high liquidated damages.148 Comparable arguments may militate against other proposals to impose stringent remedies for breach of contract. More generally, economic analysis teaches us that in the absence of costless renegotiation, supracompensatory remedies are likely to distort the incentives for efficient behavior by both parties, thus decreasing the expected joint surplus of the contract and harming both the promisor and the promisee.149 While moderate deontology may certainly endorse rules that constrain the attainment of best outcomes, setting inefficient, supracompensatory remedy rules is arguably undesirable for two related reasons. First, to the extent that the immorality of breach lies, in part, in the harm it inflicts on the promisee, measures that may deter breach yet concomitantly reduce the expected joint surplus of the contract, and hence, the expected gain of the promisee, may seem counterproductive. Second, to the extent that supracompensatory remedies decrease the joint surplus of the contract, one may expect that parties acting as rational maximizers would try to contract around such remedies, thus neutralizing their effect. A multitude of more pragmatic reasons also militate against harsher remedies for breach of contract. For instance, specific performance is often much more costly to administer than a monetary remedy. Awarding punitive damages may also encumber the judicial process, for it requires fact finding and normative deliberations beyond those generally required for awarding compensatory damages for pecuniary losses. For all of these reasons, even a deontologist who insists that there is an independent constraint against promise breaking (or that every promise

147. Cf. Roy Kreitner, Frameworks of Cooperation: Competing, Conflicting, and Joined Interests in Contract and Its Surroundings, 6 Theoretical Inq. L. 59, 80–81 (2005). 148. Melvin Eisenberg, The Limits of Cognition and the Limits of Contract, 47 Stan. L. Rev. 211, 225–36 (1995). But see Robert A. Hillman, The Limits of Behavioral Decision Theory in Legal Analysis: The Case of Liquidated Damages, 85 Cornell L. Rev. 717 (2000). On cognitive biases and legal paternalism, see infra pp. 313–47. 149. Schwartz, supra note 109, at 372–83.

309

310


breaking ipso facto harms the promisee) and who believes that this constraint should equally apply in the legal sphere must be very cautious before advocating supracompensatory or punitive remedies for breach of contract. Deontologically constrained economic analysis of contract remedies may also be too complicated for reasons having to do with economic analysis itself. The above discussion largely assumed that economic analysis supports expectation damages as the standard remedy for breach of contract. Indeed, if one is forced to choose a single remedy that is most likely to be efficient, it may well be expectation damages. However, as demonstrated, current economic analysis is complex and rather indeterminate with regard to remedy rules. Alongside economic arguments in favor of damages exceeding expectation, such as the typical underenforcement of remedies for breach and the characteristic difficulty of ascertaining high subjective valuations, and arguments for blanket enforcement of liquidated damages rules,150 there are arguments for less-than-expectation damages, including the concern of overreliance by the promisee, the existence of nonlegal sanctions, and risk sharing. To date, the bottom line of economic analysis of remedy rules is that no bottom line exists. Integrating deontological constraints with the economic analysis of remedy rules may thus be too challenging a task at this time.151 Hopefully, once our understanding of the incentive effects of different remedy rules will be enhanced (possibly through empirical and experimental studies), such integration would become more feasible.

•

F. Conclusion

This chapter sought to establish that certain deontological constraints apply to contracting behavior and that incorporating deontological constraints with economic analysis of contract law may be fruitful. With respect to the

150. Schwartz, supra note 109. 151. Recognizing a deontological constraint against promise breaking and contract breaching may however have a more limited, expressive role (on the expressive role of law, see generally supra pp. 117–22). As noted in p. 292 above, while some legal systems consider enforced performance as the primary remedy for breach of contract and others view damages as the standard remedy, all modern systems largely converge in terms of the remedies actually sought for by aggrieved parties and awarded by courts. The former systems, however, more forcefully express the moral notion that promises and contracts should be performed, and are therefore more compatible with deontological morality.

contract law

formation stage, we exposed the limitations of standard economic analysis of deception and precontractual disclosure duties. We explained how recognizing a constraint against precontractual deception may contribute to a normative and interpretative analysis of the pertinent issues. We analyzed the scope and main features of such a constraint, submitting that the integration of the constraint with economic analysis facilitates a structured consideration of germane moral factors that standard economic analysis fails to consider. As regards the performance and breach stage, we demonstrated that current legal norms are incompatible with a perception of breach as a legitimate option for the promisor, reflecting instead the notion that there is a moral constraint against contract breaching. At the same time, we conceded that due to the inconclusiveness of both deontological and economic analyses of contract remedies and the multiplicity of pertinent policy considerations, a deontologically constrained economic analysis of remedies for breach of contract may be too challenging at this developmental phase of contract theory. The proposed analytical framework is likely to contribute to the analysis of other aspects of contract law. Inter alia, deontologically constrained CBA may facilitate better understanding of such legal doctrines as duress, unenforceability on grounds of illegality and public policy, and the duty to negotiate and to perform contracts in good faith. In addition, the normative analysis of disclosure duties and of such doctrines as impossibility and impracticability may be advanced by incorporating threshold options into their economic analysis.

311


•

ten

Legal Paternalism

•

A. General

paternalism is an intervention in a person’s freedom aimed at furthering

her own good.1 Paternalistic interventions prevail in both private and public life, yet their legitimacy and desirability are hotly debated. The central justification for paternalism is welfarist: the paternalistic intervention aims at promoting the well-being of the person whose freedom is curtailed. The most powerful objection to paternalism is that it infringes upon the autonomy and freedom of people. This conflict between welfare and selfdetermination echoes the fundamental conflict between consequentialism and deontology. While deontology prioritizes liberty and autonomy over promotion of the good, consequentialism values freedom only as a component of well-being. Hence, one could have assumed that deontological morality would reject paternalism, whereas consequentialist theories would embrace it. In fact, numerous attempts have been made to justify paternalism according to deontological theories, and many consequentialists— including, so it seems, most welfare economists—strongly object to it. Until recently, economists (including legal economists) have mostly eschewed the issue of paternalism, assuming as a matter of course that efficiency and paternalism are irreconcilable. Explicit and detailed analyses of paternalism from an economic point of view have been conducted only in

1. For additional meanings and definitions of the term “paternalism,” see Gerald Dworkin, Paternalism, 56 The Monist 64 (1972) [hereinafter Dworkin, Paternalism]; Gerald Dworkin, Paternalism: Some Second Thoughts, in Paternalism 105, 105–07 (Rolf Sartorius ed., 1983) [hereinafter Sartorius, Paternalism]; Donald VanDeVeer, Paternalistic Intervention: The Moral Bounds of Benevolence 16–40 (1986); John Kleinig, Paternalism 3–17 (1984); John Kultgen, Autonomy and Intervention: Paternalism in the Caring Life 60–73 (1995); Thaddeus Mason Pope, Counting the Dragon’s Teeth and Claws: The Definition of Hard Paternalism, 20 Ga. St. U. L. Rev. 659 (2004).

313

314


the past 15 years or so.2 In recent years, the powerful empirical assault on the standard assumption of economic rationality have sparked a lively discussion of paternalism among economists.3 Yet the issue is anything but settled. The goal of this chapter is twofold. First, it critically analyzes the consequentialist—and particularly the economic—position regarding paternalism. Contrary to the prevailing view among economists, we will demonstrate that principled antipaternalism is incompatible with consequentialism, including economic efficiency. Economists’ attempts to counter this conclusion (i.e., to establish that paternalism is necessarily and invariably inefficient) follow some of the attempts to defend consequentialism against the deontological critique, discussed in chapter 1, and are as problematic. We will propose a simple model for assessing the efficiency of any paternalistic legal rule, a model that strives to avoid ideological biases which sometimes distort the economic discussion of this issue. The second goal of this chapter is to point out the incompleteness of such standard CBA of paternalism and to demonstrate how it can be improved by incorporating deontological constraints into the analysis. The incorporation will follow the general discussion in chapter 4. The integration of deontological constraints into the economic analysis will provide a firmer basis for opposing or restricting otherwise efficient paternalistic interventions, without relying on dubious consequentialist arguments.

2. Paul Burrows, Patronizing Paternalism, 45 Oxford Econ. Papers 542 (1993) [hereinafter Burrows, Patronizing Paternalism]; Paul Burrows, Analyzing Legal Paternalism, 15 Int’l Rev. Law & Econ. 489 (1995); Eyal Zamir, The Efficiency of Paternalism, 84 Va. L. Rev. 229 (1998). Some earlier studies, which examined the implications of empirically studied phenomena of bounded rationality, arrived at conclusions that are tantamount to legitimizing paternalism, but most of them did not explicitly address the issue. See, e.g., George A. Akerlof, Procrastination and Obedience, 81 Am. Econ. Rev. (Papers & Proc.) 1 (1991); George A. Akerlof & William T. Dickens, The Economic Consequences of Cognitive Dissonance, 72 Am. Econ. Rev. 307 (1982); Richard J. Arnould & Henry Grabowski, Auto Safety Regulation: An Analysis of Market Failure, 12 Bell J. Econ. 27 (1981); Martin Feldstein, The Optimal Level of Social Security Benefits, 100 Q.J. Econ. 303 (1985). 3. See, e.g., Ted O’Donoghue & Matthew Rabin, Studying Optimal Paternalism, Illustrated by a Model of Sin Taxes, 93(2) Am. Econ. Rev. (Papers & Proc.) 186 (2003); Edward L. Glaeser, Paternalism and Psychology, 73 U. Chi. L. Rev. 133 (2006); George Loewenstein & Ted O’Donoghue, “We Can Do This the Easy Way or the Hard Way”: Negative Emotions, SelfRegulation, and the Law, 73 U. Chi. L. Rev. 183, 191–92 (2006); Jonathan Klick & Gregory Mitchell, Government Regulation of Irrationality: Moral and Cognitive Hazards, 90 Minn. L. Rev. 1620 (2006); Douglas Glen Whitman & Mario J. Rizzo, Paternalist Slopes, 2 N.Y.U. J. L. & Liberty 411 (2007); George Loewenstein & Emily Haisley, The Economist as Therapist: Methodological Ramification of ‘Light’ Paternalism, in The Foundations of Positive and Normative Economics: A Handbook 210 (Andrew Caplin & Andrew Schotter eds., 2008).

legal paternalism

This chapter proceeds as follows: section B surveys the different forms and degrees of paternalistic interventions and demonstrates the prevalence of paternalism in Western legal systems, thus laying the ground for the ensuing discussion. Section C discusses paternalism from a consequentialist, and particularly economic, perspective, demonstrates that paternalism is sometimes efficient, and addresses objections to this conclusion. It then provides a simple model for the assessment of the efficiency of legal rules. Finally, section D argues that a fruitful discussion of paternalism must take into account the intrinsic value of autonomy and freedom, discusses deontological perspectives on paternalism which recognize such intrinsic value, and integrates deontological constraints into the economic analysis of paternalism.

•

B. Paternalism: Classifications and Prevalence

To grasp the vast practical significance of legal paternalism, it is useful to describe the variety of paternalistic interventions in private and public life. There are numerous forms and degrees of paternalistic interventions. A central variable is the mental condition and the intellectual capabilities of the person whose freedom is curtailed. Accordingly, a distinction is commonly drawn between paternalism toward minors and the mentally disabled on the one hand and paternalism toward competent adults on the other hand.4 Another factor is the extent to which the frustrated choice is voluntary and informed. Arguably, an intervention in a choice based on partial information or induced by deceit is not paternalistic because a voluntary choice presupposes full information and free exercise of discretion. A broader definition of paternalism would, however, encompass such cases as well.5 Paternalistic interventions also vary with regard to the rationality of the frustrated choice6

4. See, e.g., John Stuart Mill, On Liberty (1859), reprinted in On Liberty and Other Essays 14 (Oxford Univ. Press, 1991); Anthony T. Kronman, Paternalism and the Law of Contracts, 92 Yale L.J. 763, 786–97 (1983). For the argument that competence is necessarily decision-relative, see Allen E. Buchanan & Dan W. Brock, Deciding for Others: The Ethics of Surrogate Decision Making 18–20 (1989); Duncan Kennedy, Distributive and Paternalist Motives in Contract and Tort Law, with Special Reference to Compulsory Terms and Unequal Bargaining Power, 41 Maryland L. Rev. 563, 642–46 (1982). 5. See, e.g., Joel Feinberg, Legal Paternalism, 1 Can. J. Phil. 105 (1971). 6. On different meanings of “rationality” and their implications for the justification of paternalism, see Danny Scoccia, Paternalism and Respect for Autonomy, 100 Ethics 318, 319–27 (1990); David Luban, Paternalism and the Legal Profession, 1981 Wis. L. Rev. 454, 461–86.

315

316


and the kind and magnitude of the injury or loss that may ensue in the absence of intervention. The probability of such injury or loss7 and the expected probability of their avoidance by virtue of the intervention should also be considered. Additionally, paternalistic interventions differ from one another regarding the degree of closeness between the paternalist and the other person. Even people who object to any and all governmental paternalism more willingly accept instances of paternalism among family members.8 The intrusiveness of paternalistic interventions, ranging from mere refusal to assist a person in carrying out her plans, to physical restriction of her movements, is similarly crucial. Thus, the total elimination of one’s physical or legal capacity to do something is considered an extreme intervention. Criminal sanctions for disobeying the paternalist’s instructions constitute a severe, if not extreme, intervention. Providing assistance in-kind rather than in money is a less severe form of intervention, as are “sin taxes” on such products as cigarettes and alcohol. On the opposite end of the scale, the use of rational arguments to persuade a person not to do something is not considered paternalistic by most accounts, as there is no paternalism without intervention in a person’s freedom. Setting default rules that people are (more or less) free to opt out or deviate from, is a very mild form of paternalism.9 One may further classify paternalistic interventions according to whether the intervention prevents an action or requires one;10 the existence or absence of other bases for intervention, such as protection of third persons and redistribution of power and wealth;11 and the extent to which the

7. One may also distinguish between actions people do in order to harm themselves and actions involving varying degrees of risk taking. See Feinberg, supra note 5, at 109–10. 8. See Kennedy, supra note 4, at 638–41, 646–49; Kultgen, supra note 1, at 161–68; M. Todd Henderson, The Nanny Corporation and the Market for Paternalism, U. Chi. L. Rev. (forthcoming, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1348235) (comparing governmental and corporate paternalism). 9. On this parameter, see also Mill, supra note 4, at 111–13; Dworkin, Paternalism, supra note 1, at 66–67; James F. Childress, Practical Reasoning in Bioethics 124, 125 (1997) (distinguishing between active and passive paternalism); infra note 14 and accompanying text. 10. Joel Feinberg, The Moral Limits of the Criminal Law: Harm to Self 8 (1986); Pope, supra note 1, at 688–90. 11. See, e.g., Kronman, supra note 4. Cf. Kennedy, supra note 4, at 624–26; Feinberg, supra note 10, at 8, 16–17; Richard J. Arneson, Mill versus Paternalism, 90 Ethics 470, 471–72 (1980); Lawrence O. Gostin & Kieran G. Gostin, A Broader Liberty: J.S. Mill, Paternalism, and the Public’s Health, 123 Pub. Health 214, 216, 218 (2009); Henderson, supra note 8.

legal paternalism

paternalistic intervention restricts the freedom of people other than the person protected.12 These parameters and distinctions reflect the ongoing, extensive debates regarding the legitimacy and desirability of various paternalistic measures in both the private and public spheres. All of these distinctions are relevant to legal paternalism. Notwithstanding the prevailing hostility to state paternalism in the public discourse in Western, liberal democracies,13 all Western legal systems include numerous paternalistic norms. Limitations on the legal capacity of minors, the mentally disabled, and intoxicated people are obvious examples, as is the compelled use of various safety measures while driving or working in high-risk environments. Paternalism is also the primary grounds for norms prohibiting the use of some drugs and gambling and for rules forbidding swimming in the absence of a lifeguard. Additional cases include the exclusion of victim consent as a defense to certain criminal offenses; inalienability of certain basic liberties; compulsory social security and pension arrangements; compulsory elementary education; and “sin taxes” on tobacco and other unhealthy products. In the market domain, paternalism underlies coolingoff periods in door-to-door sales and other types of transactions, certain applications of the undue influence and unconscionability doctrines, and the limited enforceability of some contract terms, such as forfeiture clauses and liquidated damages. Moreover, since the distinction between default rules (applying unless people effectively depart from, or contract around, them) and mandatory rules (applying notwithstanding attempts to depart from, or contract around, them) is blurred for a variety of reasons, even default rules may sometimes direct people’s choices and may therefore be used as a (very mild) form of legal paternalism.14

12. See Dworkin, Paternalism, supra note 1, at 67–68. See also Feinberg, supra note 10, at 9–10; Pope, supra note 1, at 686–88; Loewenstein & O’Donoghue, supra note 3, at 191–92. 13. See, e.g., VanDeVeer, supra note 1, at 16–20, 25–26; David L. Shapiro, Courts, Legislatures, and Paternalism, 74 Va. L. Rev. 519, 529–45 (1988). 14. See, e.g., Eyal Zamir, The Inverted Hierarchy of Contract Interpretation and Supplementation, 97 Colum. L. Rev. 1710, 1785–88 (1997); Colin F. Camerer et al., Regulation for Conservatives: Behavioral Economics and the Case for “Asymmetric Paternalism,” 151 U. Pa. L. Rev. 1211 (2003); Cass R. Sunstein & Richard H. Thaler, Libertarian Paternalism Is Not an Oxymoron, 70 U. Chi. L. Rev. 1159 (2003); Richard H. Thaler & Cass R. Sunstein, Nudge: Improving Decisions About Health, Wealth, and Happiness (2008). On objective and subjective obstacles to contracting around seemingly default rules, which make them to some extent mandatory, see Zamir, id. at 1738–53, 1755–68; Russell Korobkin, The Status Quo Bias and Contract Default Rules, 83 Cornell L. Rev. 608 (1998).

317

318


Can all or some of these manifestations of paternalism be justified? We will first examine this question according to consequentialist normative theories, focusing on economic efficiency.

•

C. Normative Economic Analysis of Paternalism

This section argues that paternalism is compatible with economic efficiency (subsection 1) and addresses possible objections to this claim (subsection 2). It then proposes a mathematical formulation for determining the efficiency of paternalistic rules (subsection 3).

1. The Compatibility of Efficiency and Paternalism The conflict between freedom and welfare lies at the heart of the debate about paternalism.15 Attributing intrinsic value to freedom and autonomy militates against paternalism, while promoting people’s well-being is the strongest argument for paternalism. Within normative economics, freedom does not have an inherent value but rather plays a secondary, instrumental role. First, the freedom to choose among different alternatives is a component of well-being. It is one object of people’s preferences, something they actually and rationally prefer to have. Second, some degree of freedom is also a necessary means to achieve and safeguard other components of one’s well-being. Third, private will, as reflected in people’s willingness to pay for entitlements, provides economic analysis with a yardstick for measuring well-being. However, since respecting individuals’ preferences is not an end unto itself from an economic standpoint, it contains no constraints on the maximization of total well-being—including no constraint against curtailing people’s freedom.16

15. Ian M.D. Little, Ethics, Economics, and Politics 151 (2002); Buchanan & Brock, supra note 4, at 40–47; Donald VanDeVeer, Deciding for Others: The Ethics of Surrogate Decision Making by Allen E. Buchanan & Dan W. Brock (Book Review), 52 Phil. & Phenomenological Res. 232, 233–34 (1992); Tom Law Beauchamp & James F. Childress, Principles of Biomedical Ethics 176–87 (5th ed. 2001). 16. Kleinig, supra note 1, at 48–51. On the role of freedom within normative economics and other consequentialist moral theories, see generally Richard A. Posner, The Economics of Justice 92–99 (1983); Amartya Sen, Freedom of Choice, 32 Eur. Econ. Rev. 269 (1988); Michael D. Bayles, Legally Enforceable Commitments, 4 Law & Phil. 311, 321–22 (1985);

legal paternalism

One could have thus assumed that consequentialist theories, including normative economics, would embrace paternalism. Since consequentialism places no constraints on promoting the good, it should endorse paternalism whenever it brings about the best outcomes, all things considered. In fact, many consequentialists—including, so it seems, most welfare economists— object to paternalism. We maintain that this objection is unsound and that principled antipaternalism is incompatible with consequentialism, including normative economics. In what follows, we substantiate this claim with regard to several theories of well-being, focusing on the preferences theories underlying normative economics. Mental-State Theories Paternalism is clearly compatible with consequentialism resting on mentalstate theories of human well-being.17 These theories maintain that human well-being is determined by the presence of pleasure and the absence of pain. A person’s belief that a certain course of action will yield the greatest happiness for her may be erroneous, and, in such a case, obstructing the action may increase her happiness despite the displeasure involved in having her choices frustrated.18 The epistemological objection—that every person knows best what would maximize her happiness—is unpersuasive.19 Both common experience and scientific studies attesting to human fallibility and bounded rationality refute the claim that people invariably know best what would make them happy.20

Lawrence Haworth, Autonomy and Utility, in The Inner Citadel: Essays on Individual Autonomy 155 (John Christman ed., 1989). 17. For general surveys of the competing theories of well-being, see Derek Parfit, Reasons and Persons 493–502 (1984); Shelly Kagan, Normative Ethics 25–41 (1998). 18. Donald H. Regan, Paternalism, Freedom, Identity, and Commitment, in Sartorius, Paternalism, supra note 1, at 113, 113–14; Colin F. Camerer, Wanting, Liking, and Learning: Neuroscience and Paternalism, 73 U. Chi. L. Rev. 87, 106–09 (2006). 19. On this objection, see Mill, supra note 4, at 84–85, 92–93; Arneson, supra note 11; Dworkin, Paternalism, supra note 1, at 70–76; Feinberg, supra note 5, at 107–09; Douglas N. Husak, Paternalism and Autonomy, 10 Phil. & Pub. Aff. 27, 27–28 (1981); Rolf E. Sartorius, The Enforcement of Morality, 81 Yale L.J. 891, 900–10 (1972). 20. Cognitive psychology studies attesting to people’s bounded rationality are abundant. For clear, book-long surveys of the literature on various phenomena of bounded rationality, see Robin M. Hogarth, Judgement and Choice: The Psychology of Decision (2d ed. 1987); Scott Plous, The Psychology of Judgment and Decision Making (1993). For shorter surveys, see Ward Edwards & Detlof von Winterfeldt, Cognitive Illusions and Their Implications for the Law, 59 S. Cal. L. Rev. 225 (1986); Howard Rachlin, Judgment,

319

320


Objective-List Theories Objective theories of well-being posit that well-being consists of having certain things (such as health, autonomy, and accomplishment) that are intrinsically good. According to these theories, whenever a person’s action would detract from her objective good, and an intervention would, all things considered, enhance her well-being, paternalism is justified. Indeed, personal freedom is an important element of anyone’s well-being, and hence not just any choice that detrimentally affects a person’s health or excellence in life should be paternalistically obstructed.21 However, to be a principled antipaternalist within the objective goods framework, one must hold that the total well-being of a person can never increase if her liberty is decreased. This position is no longer consequentialist, but rather deontological: it introduces a constraint that personal freedom or autonomy should not be infringed upon even for the sake of increasing the total amount of well-being.22 Ideal Preferences Theories Ideal preferences theories of well-being maintain that people’s well-being is enhanced to the extent that their ideal desires are fulfilled. The pertinent desires are those a person would have were she to calmly and rationally consider the issue, paying heed to all the relevant information and without being subject to any external pressure or prejudice. Ideal preferences theory may endorse paternalistic interventions when a person’s actual preferences are not the ideal ones. This is not to say that an ideal preferences theory gives carte blanche to paternalism. Having relatively broad freedom to make one’s own decisions, including wrong ones, is quite high on most people’s list of

Decision, and Choice 43–72, 95–112 (1989). Useful collections of studies of bounded rationality are Judgment under Uncertainty: Heuristics and Biases (Daniel Kahneman, Paul Slovic & Amos Tversky eds., 1982); Judgment and Decision Making: An Interdisciplinary Reader (Terry Connolly, Hal R. Arkes & Kenneth R. Hammond eds., 2d ed. 2000); Richard H. Thaler, Quasi Rational Economics (1991); Choices, values, and frames (Daniel Kahneman & Amos Tversky eds., 2000); Heuristics and Biases: The Psychology of Intuitive Judgment (Thomas Gilovich, Dale Griffin & Daniel Kahneman eds., 2002); The Construction of Preferences (Sarah Lichtenstein & Paul Slovic eds., 2006). See also infra p. 326. 21. Buchanan & Brock, supra note 4, at 33–36. 22. Following the distinction between factoral and foundational normative theories introduced in pp. 26–27 above, the described position is deontological on the factoral level.

legal paternalism

ideal preferences.23 It does, however, open the door to paternalism under appropriate circumstances. Actual Preferences Theory An actual preferences theory of well-being—measuring one’s well-being by the extent to which a person’s actual desires are fulfilled—presents the greatest difficulty for justifying paternalism. If promoting one’s well-being consists exclusively in satisfying one’s actual preferences, then it seems that intervention aimed at enhancing one’s well-being cannot be defended under any circumstances. For many economists, this alleged antipaternalistic implication of the actual preferences theory is one of its most compelling features.24 In what follows, we will argue, first, that even consequentialism which rests on actual preferences theory of well-being is compatible with paternalism; and second, that normative economics does not rest on pure actual preferences theory but rather on a theory that is closer to ideal preferences. Consequentialism based on an actual preferences theory of well-being is not incompatible with paternalism for at least three reasons. First, people have preferences not only regarding different bundles of goods and services or different courses of action, but also regarding their own preferences.25 For example, a person may regularly eat junk food and at the same time wish her eating preferences were different. Whenever a person’s first-order preferences are not in harmony with her second-order ones, her actual choices and behavior are likely to reflect only the former. Yet the second-order preferences are as actual as the first-order ones.26 In the typical case of dissonance between first- and second-order preferences, the first-order ones are conceived by the person as injurious to herself. In such cases, she may approve

23. See Allan Gibbard, Interpersonal Comparisons: Preferences, Good, and the Intrinsic Reward of a Life, in Foundations of Social Choice Theory 165, 170–71 ( Jon Elster & Aanund Hylland eds., 1986). 24. See, e.g., Daniel M. Hausman & Michael S. McPherson, Economic Analysis and Moral Philosophy 72–73, 90–93 (1996); Robin West, Rationality, Hedonism, and the Case for Paternalistic Intervention, 3 Legal Theory 125, 126 (1997). 25. See generally Harry G. Frankfurt, Freedom of the Will and the Concept of a Person, 68 J. Phil. 5 (1971); Amartya K. Sen, Rational Fools: A Critique of the Behavioral Foundations of Economic Theory, 6 Phil. & Pub. Aff. 317 (1977); Timothy J. Brennan, A Methodological Assessment of Multiple Utility Frameworks, 5 Econ. & Phil. 189 (1989). 26. A famous example is Ulysses and the Sirens. See Dworkin, Paternalism, supra note 1, at 76–84; Jon Elster, Ulysses and the Sirens: Studies in Rationality and Irrationality (rev. ed. 1984).

321

322


of external intervention that would impede the fulfillment of her first-order, revealed preferences.27 We do not argue that people who dislike their firstorder preferences necessarily wish them to be paternalistically frustrated. The claim is that in at least some cases, a second-order approval of interference with first-order preferences is plausible. Therefore, actual preferences theory does not entail principled antipaternalism. Arguably, indirect evidence for the prevalence of such second-order preferences may be found in the fact that democratically elected governments throughout the world widely engage in paternalistic regulation.28 A second reason why paternalism is not incompatible with an actual preferences theory of well-being is that norms influence and shape people’s preferences in various ways.29 Paternalism toward children not only protects them from immediate mishaps but also, and no less importantly, shapes their preferences so that they avoid such mishaps in the future. Interventions in people’s choices and behavior frequently result in an adaptation of their preferences to conform with the rules.30 These people may in retrospect be grateful for the initial paternalistic treatment. To the extent that this is true, there may be no consequentialist reason to give greater weight to the ex ante actual preferences than to the ex post ones. As stated, this argument

27. Cass R. Sunstein, Legal Interference with Private Preferences, 53 U. Chi. L. Rev. 1129 (1986) [hereinafter Sunstein, Legal Interference]; Cass R. Sunstein, Disrupting Voluntary Transactions, in Nomos XXXI: Markets and Justice 279 (John W. Chapman & J. Ronald Pennock eds., 1989) [hereinafter Sunstein, Disrupting Voluntary Transactions]; Dworkin, Paternalism, supra note 1, at 76–84; Husak, supra note 19; Robert E. Goodin, Laundering Preferences, in Foundations of Social Choice Theory, supra note 23, at 75, 82–85. 28. Examples of such regulation were provided in p. 317 above. An important objection to this justification of legal paternalism is that the paternalistic rule applies not only to people who actually hold these second-order preferences but also to people who do not (see Sunstein, Legal Interference, supra note 27, at 1141–42; Jonathan Schonsheck, Deconstructing Community Self-Paternalism, 10 Law & Phil. 29 (1991); Rosemary Carter, Justifying Paternalism, 7 Can. J. Phil. 133, 144–45 (1977)). While this objection may be compelling from a deontological perspective, it is less powerful within a consequentialist, cost-benefit analysis. 29. Sunstein, Legal Interference, supra note 27, at 1145–58; Robin West, Taking Preferences Seriously, 64 Tulane L. Rev. 659, 670–75 (1990); Tyler Cowen, The Scope and Limits of Preference Sovereignty, 9 Econ. & Phil. 253, 254–58 (1993); Robert E. Goodin, Political Theory and Public Policy 19–56 (1982); G. Peter Penz, Consumer Sovereignty and Human Interests 87–119 (1986). 30. See, e.g., Burrows, Patronizing Paternalism, supra note 2, at 563–64 (reporting that resistance to seat belt and safety helmet legislation in the UK declined sharply after its implementation). See also Gostin & Gostin, supra note 11, at 217 (arguing that bans on smoking in public places have affected social norms about tobacco).

legal paternalism

is, of course, incomplete because it does not answer the questions when and why the ex post preferences should take precedence over the ex ante ones. Yet it suffices to show that in principle, paternalism is not incompatible with a consequentialist theory resting on an actual preferences theory of well-being. Finally, even if people’s actual preferences are not affected by legal norms, they do change over time. For example, people may regret their past decisions regarding pension savings or higher education. According to an actual preferences theory of well-being, the well-being of a person is enhanced to the extent that one’s actual desires are satisfied over the course of one’s life. Hence, not only current desires matter but also future ones. There is, however, no reason to assume that people can foresee their future desires precisely. Even if they could, their current preferences need not correspond to the overall balance of their lifetime desires. The gap between present actual preferences and the overall balance of lifetime actual preferences leaves room for paternalistic intervention. Once again, this argument does not warrant any specific paternalistic intervention. It only establishes that paternalism is not necessarily inconsistent with an actual preferences theory of well-being. Having established that the actual preferences theory of well-being does not rule out paternalism, we now turn to an alternative (or additional) argument, focusing on normative economics. Despite claims to the contrary, standard economic analysis does not rest on a pure actual preferences theory. The crux of standard economic theory is that, given certain assumptions, the sum of human well-being (“social welfare”) is maximized if everyone acts so as to maximize one’s own well-being.31 Market forces lead to an optimal allocation of resources and entitlements, and consequently to the maximization of total well-being. For this happy result to ensue, two general sets of assumptions must hold true. The first set concerns the conditions of perfect competition, such as numerous actors, full information, and no restraints of trade. The second refers to attributes of the participants’ rationality. This second set of assumptions, concerning economic rationality, consists of both cognitive and motivational elements. A thin, cognitive rationality requires only formal elements regarding the structure of a person’s set of preferences and her strategy of decision-making, such as transitivity and

31. Adam Smith, The Wealth of Nations 423 (Edwin Cannan ed., 1937) (1776).

323

324


completeness of preferences, and ignoring irrelevant information.32 A thick definition of rationality adds to the above structural elements an assumption about human motivation. It assumes that people always choose the option that yields the greatest benefit to them (they are “rational maximizers”). This assumption excludes both true altruism (choices and actions aimed solely at furthering the well-being of others) and idealism or commitment (acting out of a sense of duty even contrary to one’s self-interest and sympathetic preferences).33 It is often said that economic analysis is not committed to a thick definition of rationality and that it requires only a few very basic elements of thin rationality.34 Most conventional economic studies, however, including economic analyses of law, are based on both thin and thick notions of rationality.35 All standard concepts of market failure make sense only under the rationalmaximizer assumption. If economic agents were not self-interested, but rather devoted altruists or idealists, then the prisoners in the prisoner’s dilemma scenario would face no dilemma. Accordingly, there would be no problems with the provision of public goods (commonly explained as an instance of prisoner’s dilemma), and information (characteristically a public good) would be produced and supplied like any private good. The basic prediction regarding the efficiency of the competitive market would not hold if a considerable portion of the participants preferred unprofitable transactions to profitable ones. Widespread infringement of the assumptions of thin rationality would have the same destructive results for standard economic analysis. Decisions based on irrelevant information, disregard of

32. See also supra pp. 11–12. 33. Sen, supra note 25; Hausman & McPherson, supra note 24, at 51–65. 34. See, e.g., Gary S. Becker, The Economic Way of Looking at Life, in Accounting for Tastes 139 (1996); Donald P. Green & Ian Shapiro, Pathologies of Rational Choice Theory: A Critique of Applications in Political Science 18 (1994). 35. Elster, supra note 26, at 141–46, 153–56; Hausman & McPherson, supra note 24, at 38–65; Green & Shapiro, supra note 34, at 19 (regarding economic analysis of law); Brennan, supra note 25, at 201; David Goetze & Peter Galderisi, Explaining Collective Action with Rational Models, 62 Pub. Choice 25, 25–26 (1989); Russel Hardin, Autonomy, Identity, and Welfare, in The Inner Citadel, supra note 16, at 189, 190–91; Amartya Sen, Rationality and Social Choice, 85 Am. Econ. Rev. 1, 15 (1995) (on the tendency of the public choice tradition). Some economic models, especially those relating to the behavior of firms, assume tremendous capacity to gather and process information. See Green & Shapiro, supra note 34, at 19; Thomas S. Ulen, Cognitive Imperfections and the Economic Analysis of Law, 12 Hamline L. Rev. 385, 385–86 (1989).

legal paternalism

risks, shortsightedness, or computational errors are unlikely to accomplish one’s aims, whatever they may be, or to maximize aggregate social utility. Standard economic analysis thus does not embrace a pure actual preferences criterion of well-being. Between a pure actual preferences theory of well-being and a pure ideal preferences theory, there is a whole spectrum of compromised, complex theories. The more demanding the definition of rationality introduced into a theory, the closer it comes to an ideal preferences theory.36 Arguably, economic analysis does not rest on the normative claim that rational preferences are a superior criterion for human well-being than actual ones. It merely rests on the factual claim that people’s actual preferences are rational. However, to the extent that standard economic analysis is built on the assumption that people are rational maximizers, its normative implications are the same as those of a rational preferences theory of well-being. Moreover, some economists explicitly maintain that in cases of bounded rationality, such as inability to process information or wishful thinking, maximization of social welfare requires the consideration of preferences people would have in the absence of such limitations and biases.37 The claim that economic analysis, and economic analysis of law in particular, ordinarily assumes both thin and thick rationality is descriptive, rather than normative or logical. Nothing inherent in economic analysis precludes inquiry into the effect of various phenomena of bounded rationality on maximization of human well-being.38 Just as economic analysis readily explores the implications of deviations from the perfectly competitive market (market failures), it can, should, and in recent years, increasingly does, investigate the implications of deviations from the assumptions of (thin and thick) rationality.39

36. Dan Brock, Paternalism and Promoting the Good, in Sartorius, Paternalism, supra note 1, at 237, 251–52. Cf. John C. Harsanyi, Morality and the Theory of Rational Behaviour, in Utilitarianism and Beyond 39, 54–56 (Amartya Sen & Bernard Williams eds., 1982). 37. See Louis Kaplow & Steven Shavell, Fairness Versus Welfare 23, 410–13 (2002). 38. Kenneth J. Arrow, Rationality of Self and Others in an Economic System, in Rational Choice, The Contrast between Economics and Psychology 201 (Robin M. Hogarth & Melvin W. Reder eds., 1987). 39. See, e.g., Akerlof & Dickens, supra note 2; Akerlof, supra note 2; Arnould & Grabowski, supra note 2; Symposium on Views and Comments on Bounded Rationality as Applied to Modern Institutional Economics (Rudolf Richter ed.), 146 J. Inst. & Theor. Econ. 648–748 (1990); Rational Choice, The Contrast between Economics and Psychology (Robin M. Hogarth & Melvin W. Reder eds., 1987); Burrows, Patronizing Paternalism, supra

325

326


Empirical data evidencing widespread deviations from the assumptions of rationality is abundant. Research indicates that people’s perceptions, analyses, and decisions are affected by systematic biases. Common errors include misjudgment of probability, particularly the probability that a future risk will be realized; overoptimism; inability to make a correct cost-benefit analysis when the data is complex; dependence of decisions on the manner in which data is presented (so-called framing of decisions); overestimation of the frequency of outstanding events due to the tendency to remember them better (known as the availability problem); consideration of irrelevant factors; ignorance of relevant information; undervaluation of future benefits and costs in comparison to present ones (myopia); ignorance of the incompleteness of data and the limitations of judgmental skills leading to overconfidence in evaluations and predictions; and failure to rationally analyze relevant information or examine alternatives.40 Research also reveals deviations from the assumption of self-interested maximization. For example, many people prefer to share the proceeds of a transaction fairly rather than to monopolize them.41 Experiments also indicate deviations toward altruism and toward maximization of the difference between one’s own utility and that of others.42 Any attempt to understand human behavior, and in particular to predict what rules and institutions will maximize aggregate social utility, should take these deviations into account, though it is more difficult to model deviations from the thick rationality assumption than from the thin one. Once the prevalence of systematic deviations from the rationalmaximizer model is acknowledged, principled antipaternalism is no longer a tenable position of economic analysis.

note 2; Robert Sugden, The Opportunity Criterion: Consumer Sovereignty without the Assumption of Coherent Preferences, 94 Am. Econ. Rev. 1014 (2004). This trend is reflected in the law-and-economics literature as well. See generally Robert C. Ellickson, Bringing Culture and Human Frailty to Rational Actors: A Critique of Classical Law and Economics, 65 Chi.-Kent L. Rev. 23 (1989); Behavioral Law and Economics (Cass R. Sunstein ed., 2000); Russell B. Korobkin & Thomas S. Ulen, Law and Behavioral Science: Removing the Rationality Assumption from Law and Economics, 88 Cal. L. Rev. 1051 (2000). 40. See sources cited supra note 20. 41. See, e.g., Elizabeth Hoffman & Matthew L. Spitzer, The Coase Theorem: Some Experimental Tests, 25 J. L. & Econ. 73 (1982). For a brief review of experimental studies using the Ultimatum Game, see Daphna Lewinsohn-Zamir, The Choice Between Property Rules and Liability Rules Revisited: Critical Observations from Behavioral Studies, 80 Tex. L. Rev. 219, 227–31 (2001). 42. See, e.g., Norman Frohlich & Joe Oppenheimer, Beyond Economic Man: Altruism, Egalitarianism, and Difference Maximization, 28 J. Conflict Resol. 3 (1984).

legal paternalism

We may conclude that paternalistic intervention in people’s freedom is not necessarily inefficient. Not only does normative economics not require principled antipaternalism, but the latter is in fact incompatible with the former. Once second-order, ex post, and overall lifetime preferences are recognized and respected, paternalism is sometimes efficient even under the prevailing, though doubtful, assumption that normative economics rests on an actual preferences theory of well-being. Paternalism can certainly be efficient once it is understood that the theory of well-being underlying normative economics is, in fact, closer to rational preferences theory.

2. Possible Objections The claim that economic efficiency and paternalism are not incompatible troubles many economists. In particular, it is troubling for free marketers who distrust the government and oppose any sort of regulation. The antipaternalist sentiments of some economists are so strong, in fact, that they object to even the mildest forms of paternalism, such as setting default rules which the decision-maker believes would benefit the interests of the people to whom they apply (rather than simply trying to imitate their choices).43 Some antipaternalists have questioned the validity of the empirical studies substantiating the prevalence of bounded rationality, claiming that seemingly mistaken judgments are not mistakes, that they would disappear with monetary incentives, or that the market would drive out bad judgment.44 However, as Jeffery Rachlinski nicely pointed out, these claims fail both empirically and conceptually. “Empirically, the cognitive phenomena on which the case for paternalism rests are well understood and have support from hundreds of studies that have used a wide variety of subjects, contexts, and incentives. . . . Conceptually, . . . [a]rguments that incentives or real-world

43. See, e.g., Glaeser, supra note 3; Klick & Mitchell, supra note 3; Whitman & Rizzo, supra note 3. 44. See, e.g., Richard A. Posner, Rational Choice, Behavioral Economics, and the Law, 50 Stan. L. Rev. 1551 (1998); Gregory Mitchell, Why Law and Economics’ Perfect Rationality Should Not Be Traded for Behavioral Law and Economics’ Equal Incompetence, 91 Geo. L.J. 67 (2002); Steven Winter, A Clearing in the Forest: Law, Life, and Mind 92–96 (2001); Joshua D. Wright, Behavioral Law and Economics, Paternalism, and Consumer Contracts: An Empirical Perspective, 2 N.Y.U. J.L. & Liberty 470 (2007).

327

328


circumstances cure people of their reliance on misleading information fail to identify exactly how this would occur. The extra motivation that comes with monetary incentives and real decisions cannot just ‘work by magic.’”45 Other attempts to reconcile the conclusion that paternalism may be efficient with antipaternalist sentiments follow the familiar lines of argument aimed at defending consequentialism against the deontological critique.46 One such line of argument stresses the long-term and indirect adverse effects of paternalism on people’s motivation to act deliberately and carefully, on their “educational investments,” and more generally, on the “development of effective decision-making skills and strategies.”47 While such possible effects should be taken into account, they are often overstated. They ignore the fact that paternalistic interventions may actually serve as a means to further learning.48 By having a certain entitlement (such as a safe working environment or social security rights), people learn to appreciate its true worth and develop a stronger preference for it.49 Those arguments also disregard the fact that some decisions (e.g., those involved in buying a house) are made very infrequently, and thus learning through trial and error may be very costly and ineffective in these cases.50 Some choices (such as riding a motorcycle without a helmet) may have fatal consequences, and the outcomes of other choices (such as smoking, not saving, and not acquiring an education) are spread out over a very long time. In all of these cases, learning from one’s

45. Jeffery J. Rachlinski, The Uncertain Psychological Case for Paternalism, 97 Nw. U. L. Rev. 1165, 1167–68 (2003) (citations omitted) [hereinafter Rachlinski, Paternalism]. See also Amos Tversky & Daniel Kahneman, Rational Choice and the Framing of Decisions, 59 J. Bus. S251, S274 (1986); Daniel A. Farber, Toward a New Legal Realism, 68 U. Chi. L. Rev. 279, 290–94 (2001); Jeffrey J. Rachlinski, The “New” Law and Psychology: A Reply to Critics, Skeptics, and Cautious Supporters, 85 Cornell L. Rev. 739 (2000); Colin F. Camerer & Robin M. Hogarth, The Effects of Financial Incentives in Experiments: A Review and Capital Labor Production Framework, 19 J. Risk & Uncertainty 7, 33 (1999); David Hirshleifer, Investor Psychology and Asset Pricing, 56 J. Fin. 1533 (2001) (describing the survival of cognitive biases in securities markets); Eldar Shafir & Robyn A. LeBoeuf, Rationality, 53 Ann. Rev. Psychol. 491, 500–09 (2002). 46. See supra pp. 21–33. 47. Klick & Mitchell, supra note 3, at 1626. For similar arguments, see, e.g., Mill, supra note 4, at 62–82; Regan, supra note 18, at 115–16; Whitman & Rizzo, supra note 3, at 24–26. Cf. supra pp. 22–24. 48. On various ways in which paternalistic interventions facilitate learning that is unlikely to occur otherwise, see Camerer, supra note 18, at 96–97, 99–102, 104–06. 49. See supra p. 322. 50. Rachlinski, Paternalism, supra note 45, at 1223.

legal paternalism

mistakes may be impossible because recognition of the mistake comes too late.51 Moreover, many cognitive biases are so deeply rooted, even in people of high intelligence and with rich life experience, that they are unlikely to disappear as a result of letting individuals make more mistakes. Acknowledging that the case for paternalism depends upon demonstrating that the costs of learning exceed the costs of paternalistic intervention does not imply that such a case cannot be made. Impediments to learning include people’s profound overconfidence in the correctness of their decisions, the reluctance to admit that one has made a mistake, the fact that it is often impossible to know with much certainty what would have happened had a different choice been made, and a lack of feedback.52 Indeed, numerous studies have shown that cognitive biases are prevalent among experienced professionals, who presumably have ample opportunity to learn from their mistakes.53 In addition, models of learning from experience (compatible with experimental data) indicate that learning leads to decidedly greater risk aversion in the domain of gains (where expected returns of alternatives are positive) than in the domain of losses (where different choices result in different harms).54 Yet the latter is the area typically relevant to legal paternalism. Presumably, the antipaternalistic development argument is particularly relevant to minors, who are in their formative years. Nevertheless, there is practically a consensus that legal paternalism toward children and adolescents is justifiable. Extreme, comprehensive state paternalism may indeed lead to undesirable results in terms of people’s development. Such totalitarian measures are, however, unacceptable under any normative theory anyway. Another antipaternalist argument often used by consequentialists follows the familiar move from act- to rule-consequentialism:55 even if people sometimes know what is best for others, the risk of error is high enough to

51. Camerer, supra note 18, at 99–100; Rachlinski, Paternalism, supra note 45, at 1219, 1223. 52. David Dunning et al., The Overconfidence Effect in Social Prediction, 58 J. Personality & Soc. Psychol. 568 (1990); Robert P. Vallone et al., Overconfident Prediction of Future Actions and Outcomes by Self and Others, 58 J. Personality & Soc. Psychol. 582 (1990); Rachlinski, Paternalism, supra note 45, at 1220–22. 53. See, e.g., Chris Guthrie, Jeffrey Rachlinski & Andrew J. Wistrich, Inside the Judicial Mind, 86 Cornell L. Rev. 777, 782–83 (2001); Jeffrey J. Rachlinski, Cognitive Errors, Individual Differences, and Paternalism, 73 U. Chi. L. Rev. 207, 219–21 (2006). 54. James G. March, Learning to Be Risk Averse, 103 Psychol. Rev. 309 (1996). 55. See supra pp. 24–27.

329

330


warrant a general ban on paternalism.56 Such a ban would also avoid the potential for abuse of power, especially when paternalism is practiced by state officials. Antipaternalists also resort in this context to slippery slope arguments.57 While these considerations must not be overlooked, they are clearly exaggerated.58 In setting paternalistic rules that aim to protect people from their shortsightedness, computation limitations, overoptimism regarding low-probability risks, wishful thinking, and other common manifestations of bounded rationality, policy-makers enjoy a clear advantage, as they can rely on professional, technical, and statistical knowledge that is often unavailable to individuals.59 Even more importantly, policy-makers weigh the alternatives in a more detached and calm manner, based on objective, general statistics. For example, while a policy-maker may suffer from an availability problem (which may be further manipulated by interest groups), her decision as to what is best for drivers, borrowers, or would-be smokers is unlikely to be affected by emotional stress, peers’ social pressure, wishful thinking, overoptimism, disregard of low-probability risks, “illusion of control,” and so forth.60 It may be reasonably assumed that the various inputs to the

56. For discussions of this argument in the philosophical literature, see the sources cited in supra note 19. In the recent economic literature, see, e.g., Whitman & Rizzo, supra note 3. 57. See Whitman & Rizzo, supra note 3. 58. For a critique of the slippery slope argument against paternalism, see Thaddeus Mason Pope, A Definition and Defense of Hard Paternalism: A Conceptual and Normative Analysis of the Restriction of Substantially Autonomous SelfRegarding Conduct 405–24 (2003, unpublished Ph.D. dissertation, Georgetown University, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1087383). For a critique of the simplistic, one-dimensional depiction of political and regulatory actors as motivated solely by self-interest and greed, offered by public choice theory, see Jerry L. Mashaw, Greed, Chaos, and Governance: Using Public Choice to Improve Public Law (1997). For a balanced analysis of the pros and cons of private and public ordering, see also Cass R. Sunstein, After the Rights Revolution: Reconceiving the Regulatory State (1990). 59. Cass R. Sunstein, Behavioral Analysis of Law, 64 U. Chi. L. Rev. 1175, 1190 (1997) (pointing out the advantages of statistical, as opposed to intuitionistic, judgments); David B. Spence & Frank Cross, A Public Choice Case for the Administrative State, 89 Geo. L.J. 97 (2000) (administrative decision-makers are likely to select the public policy the median voter would prefer if she had full information and the opportunity to deliberate); Robert C. Clark, Contracts, Elites, and Traditions in the Making of Corporate Law, 89 Colum. L. Rev. 1703, 1718–19 (1989) (criticizing the “odd strategy” of antipaternalists who claim that regulators do not know better than the regulated what is in the latter’s best interest, a claim that is “desperately implausible in many real-world contexts”). 60. See, e.g., Neil D. Weinstein, Unrealistic Optimism About Susceptibility to Health Problems: Conclusions from a Community-Wide Sample, 10 J. Behav. Medicine 481 (1987) (demonstrating that people consistently believe that they are less prone to suffer harm than their

legal paternalism

legislative, administrative, and judicial processes, along with the wisdom and cumulative experience of the relevant decision-makers, are likely to result in a sensible assessment of the relevant factors. Moreover, judicial review, and more generally policy-makers’ accountability, are likely to mitigate decision-makers’—including collective decision-makers’—cognitive errors and biases.61 As for the risk of abuse, this was indeed a real concern two or three centuries ago and may still be a concern in some societies. However, in the current public atmosphere in Western liberal democracies, policy-makers very rarely conceal their ulterior motives behind paternalistic rhetoric. Rather than justifying a policy on the ground that the policy-maker knows better than her constituency what is good for them, she would ordinarily justify it on distributive or efficiency grounds. Telling people that they are stupid is seldom a winning strategy in a democracy. There is a greater risk of abuse in letting policy-makers rely on justifications such as “unequal bargaining power” or market failure, than on paternalistic arguments. The risks of policy-makers’ errors and abuse should be taken into account in identifying the matters in which state paternalism may bring about the best outcomes and in choosing among different forms of intervention; but they cannot serve as a general, principled argument against paternalism. Finally, economists may argue that disinterested people would experience displeasure knowing that other people are treated paternalistically.62 While such disinterested preferences may well exist, so may the opposite preferences of people who resent the notion that the state should do nothing

peers—a bias that is likely to affect one’s self-regarding decisions but not the decisions made regarding other people); Shelley E. Taylor, Positive Illusions: Creative SelfDeception and the Healthy Mind 3–45 (1989) (providing ample evidence of the various ways in which “the self is self-serving”); Daniel Kahneman & Dan Lovallo, Timid Choices and Bold Forecasts: A Cognitive Perspective on Risk Taking, 39 Management Science 17, 24–27 (1993) (demonstrating that “inside view” of risk taking, focusing on the unique features of a problem and reflecting the “illusion of control” tends to produce over-optimistic assessments, while an “outside view,” focusing on statistical aggregations, is likely to produce much more accurate predictions and decisions). While the efficacy of health and safety (and other) regulation is controversial, some studies clearly demonstrate the positive effect of paternalistic regulation. See, e.g., Alma Cohen & Liran Einav, The Effects of Mandatory Seat Belt Laws on Driving Behavior and Traffic Fatalities, 85 Rev. Econ. & Stat. 828 (2003); Thomas S. Dee & William N. Evans, Behavioral Policies and Teen Traffic Safety, 91(2) Am. Econ. Rev. (Papers & Proc.) 91 (2001). 61. Mark Seidenfeld, Cognitive Loafing, Social Conformity, and Judicial Review of Agency Rulemaking, 87 Cornell L. Rev. 486 (2002). 62. Compare the “preferences for constraints” argument discussed in pp. 27–29 above.

331

332


to protect people from losses and suffering people bring upon themselves. More fundamentally, normative judgments are not made by aggregating preferences but rather by assessing the strength of the competing arguments.63 In conclusion, we maintain that being a consequentialist normative theory, welfare economics does not rule out paternalism. In fact, welfare economics endorses paternalism whenever its benefits even slightly exceed its costs. The next subsection provides a general model to evaluate the economic efficiency of paternalistic rules.

3. A Simple Model Within a consequentialist framework, legal paternalism is justified whenever its expected benefits exceed the expected costs. The goal of the following formulation is to provide a framework for determining when a paternalistic rule is efficient. Due to the difficulty of modeling nonselfish behavior, the model does not deal with the issue of altruistic or idealistic behavior (which may detract from a person’s well-being), but with suboptimal decisions due to cognitive limitations only.64 To keep the formula simple and avoid issues of standard market failures (such as monopolies and externalities) and distributive concerns—both of which often accompany manifestations of bounded rationality—we assume a perfectly competitive market and disregard issues of distribution. The model assumes that neither the agent’s relevant choice nor the paternalistic rule affects third persons. It further assumes that people are self-interested maximizers, who sometimes make suboptimal decisions due to various cognitive imperfections. The policy-maker’s sole motivation is presumed to be the maximization of aggregate social utility. The model deals with legal rules that apply to many people and takes into account their heterogeneity. The paternalist is assumed to attain perfect compliance with the paternalistic rule and to bear the costs involved in attaining such compliance. In the model, agents face a binary choice between alternatives X and Y, and choose the one they think is the correct one, that is, the one with the

63. See supra pp. 27–29. 64. We leave open the question of whether it may be efficient to paternalistically interfere in people’s altruistic or idealistic behavior when such behavior has a detrimental effect on the actor. For further discussion of this point, see Zamir, supra note 2, at 255–56. See also infra pp. 340–41.

legal paternalism

greater expected utility for each one of them. Absent the paternalistic rule, agents may choose either X or Y, and in either case, their choice may be correct or incorrect in the above sense. The paternalistic rule limits the agents’ choice to X, possibly by eliminating or prohibiting Y. N = {1, 2, . . . n} is the set of instances in which any of the agents to whom the paternalistic rule applies faces a choice between X and Y (or would have faced such a choice absent the rule).65 Pli is the probability that in the i-th case, absent the paternalistic rule the agent would have incorrectly chosen Y and as a result would have suffered a loss Li. Pgi is the probability that in the i-th instance, absent the paternalistic rule, the agent would have correctly chosen Y and consequently gained Gi (in comparison to her utility from choosing X).66 As defined, both Li and Gi are nonnegative. The expected disutility of letting the agent choose Y in instance i is a function of Pli, Pgi, Li, and Gi. Therefore, the expected benefit from the paternalistic rule in instance i equals: PliLi –PgiGi. In words, it is the loss from incorrectly choosing Y, multiplied by the probability of such a choice, minus the gain from correctly choosing Y, multiplied by the probability of such a choice (as defined, this expected “benefit” may be negative). Whenever the agent would have chosen X, the rule does not alter the choice’s outcomes. Such cases, therefore, do not affect the assessment of the rule discussed thus far.67 To see whether the paternalistic rule is efficient, one should weigh its expected benefits against its expected costs. The model assumes that there

65. N may be much larger than the number of agents to whom the rule applies (e.g., all instances in which any driver or passenger decides whether to use a seat belt), but also much smaller (e.g., all instances in which a person decides whether to sign a voluntary enslavement agreement). 66. The following is a formal presentation of these definitions: Pli = Pi(Y is chosen ∩ X is correct) Pgi = Pi (Y is chosen ∩ Y is correct) Li = Ui (X|Y is incorrectly chosen) – Ui (Y|Y is incorrectly chosen) Gi = Ui (Y|Y is correctly chosen) – Ui (X|Y is correctly chosen) 67. To illustrate, assume that in 60 percent of the instances, a particular driver uses a seat belt and in 40 percent she does not (i.e., the probability of choosing Y in instance i is 0.4). Assume further that the probability that her decision not to use a seat belt is correct is 0.3. The expected loss from not using a seat belt in this instance, assuming it is the incorrect choice, is 10 (the expected magnitude of various avoidable injuries, weighted by their probabilities, less the inconvenience involved in wearing the seat belt). The expected gain from not using a seat belt where this is the correct choice is 5 (avoiding the inconvenience involved in using the seat belt, less the expected, weighted magnitude of the various avoidable injuries). In this example, the expected net benefit from forcing the agent to use a seat belt in instance i is: 0.4 × (0.7 × 10 – 0.3 × 5) = 2.2.

333

334


are only two types of costs: Fi, the frustration costs, reflects the agent’s disutility from being treated paternalistically in each instance i; and C, the paternalist’s total costs, which include the legislative, administrative, and judicial costs involved in establishing and implementing the paternalistic rule. An unconventional feature of the model is that it measures the losses to the agent from an incorrect choice of Y, Li, and her gains from a correct choice of Y, Gi, according to the preferences she would have had if her thin rationality were perfect, whereas her disutility from being treated paternalistically, Fi, is measured according to her actual preferences. Hence, a paternalistic rule that limits agents’ choice to X (whenever they face a binary choice between X and Y) is efficient if and only if its expected net benefit, B, is positive. B is calculated as follows: n

(1) B = ∑ ( Pli Li − Pg i Gi − Fi ) − C i =1

At this point, it should be stressed that the fact that the decision-maker believes that B is positive does not, in and of itself, guarantee the rule’s efficiency. This is because the decision-maker is not omniscient. Her judgments may be erroneous for two main reasons: her own bounded rationality and her lack of familiarity with each agent’s personal values, beliefs, and needs. The legal paternalist may be wrong in assessing any of the above variables. Pli , Pgi , Li , Gi and Fi largely depend on agent-specific characteristics; they may therefore be misconceived by a decision-maker who is not intimately familiar with these characteristics. The paternalist may also miscalculate C, because it may be borne in the future by various branches of government (e.g., a paternalistic statute may impose future costs on the courts). The paternalist’s error may be in either direction and of any magnitude. However, just as any agent must make a decision whenever she is faced with a choice between X and Y, the policy-maker has to choose between introducing the paternalistic rule and not introducing it. The paternalist’s estimation of B is similar to the cost-benefit analysis required for any type of regulation under conditions of uncertainty. An attempt should be made to aggregate the various possible over- and underestimations of each of B’s variables, weighted by their estimated probabilities. A detailed analysis of the model, the interrelations between its variables, and possible relaxations of its assumptions—including the assumptions that paternalism has no long-term effects, that neither the agent’s choice nor the paternalistic rule affect third persons, that there are no other market imperfections, that there is no risk of misuse by the paternalist, that the

legal paternalism

paternalist attains perfect compliance, and that no distributive issues are involved—has been provided elsewhere.68 For our purposes, the central point is that the formula, while facilitating a structured, rational assessment of the efficiency of any existing or proposed paternalistic rule, ascribes no intrinsic value to personal freedom and sets no constraint against its infringement. Indeed, people’s possible aversion to paternalism is likely to affect the function’s variables in two indirect ways. First, the greater one’s aversion to paternalism, the more subjective disutility Fi , one is likely to experience whenever treated paternalistically; Second, the larger the aversion to paternalism, the higher the expected costs of enforcing the paternalistic policy, C.69 A consequentialist may thus use (or rather misuse) the model to limit paternalistic interventions in various ways. Feelings of displeasure and frustration are hard to measure and monetize, and are therefore susceptible to manipulation. The antipaternalist may thus overestimate the total disutility to agents resulting from being treated paternalistically, F = ∑ Fi. However, if one believes that legal paternalism is morally objectionable on liberal or libertarian grounds, and should therefore be limited to cases in which nonintervention would result in large or even huge losses, the right way to do so is to add a constraint to the model. Manipulating other variables only obscures the real dilemma inherent in paternalism, the dilemma between freedom and welfare. The following section examines deontological perspectives on paternalism and demonstrates how deontological constraints may be incorporated into the formula. Such incorporation provides a better and more inclusive framework for a balanced evaluation of existing and proposed paternalistic rules, based on the pertinent normative factors.

•

D. Incorporating Deontological Constraints 1. Deontological Perspectives on Paternalism

Deontology views principles of the right as prior to, and largely independent of, conceptions of the good. Deontological moral theories commonly attribute

68. Zamir, supra note 2, at 254–83. 69. The model attributes no weight, however, to disinterested, “moral preferences” about paternalism. Cf. supra pp. 27–29.

335

336


intrinsic value to people’s autonomy and freedom, prioritizing them over the promotion of good outcomes. Autonomy and freedom are complex concepts. Generally speaking, individual autonomy refers to the capacity to be one’s own person, to live one’s life according to reasons and motives that are one’s own, rather than the product of external forces.70 Accordingly, individuals should be respected as independent agents, having a right to set their own goals and make their own decisions even if such goals and decisions are not in their best interest according to some external criteria. At least an active/ intentional (and not insignificant) interference in a person’s autonomy or freedom is thus an infringement of the constraint against interfering with people’s autonomy or freedom, or more generally, the constraint against harming people. Even under these theories, one’s freedom may be restricted for the purpose of preventing a person from interfering with the autonomy and freedom of other people or otherwise harming them. This basic tenet of liberalism is known as “the harm principle.”71 By definition, paternalism entails intervention in people’s freedom, and sometimes also infringements of their privacy and control over their body, in situations where such intervention cannot be grounded in the harm principle.72 One may thus conceive of three basic deontological stances toward paternalism. The deontologist may (1) rule out paternalism under any circumstances; (2) justify seemingly paternalistic measures whenever it can be

70. On the complex notion of autonomy, see generally Gerald Dworkin, The Theory and Practice of Autonomy (1988); John Christman, Autonomy in Moral and Political Philosophy, in Stanford Encyclopedia of Philosophy (2003), http://plato.stanford. edu/entries/autonomy-moral. 71. Mill, supra note 4, at 13–14. 72. Indeed, self-harming behavior often involves actual or potential harm to others as well. Thus, for example, when a person is badly injured due to her carelessness, such injury not only harms that person but also imposes additional costs on the health care system and adversely affects the welfare of her dependents and associates. Such external effects are relevant factors in assessing the justification for any interference in people’s freedom, both from a consequentialist perspective (because CBA takes into account everyone’s welfare) and from a deontological one (because such interference may be justified by the harm principle). An analysis of “mixed-motives” interventions exceeds the scope of the present discussion (see generally Zamir, supra note 2, at 277–80, and the sources cited in supra note 11), which for simplicity’s sake focuses on harm to self. For a critique of the tendency of both consequentialists and deontologists to overstate external effects in order to legitimize interventions without resorting to “infamous” paternalistic arguments, see Burrows, patronizing paternalism, supra note 2, at 563; Dworkin, paternalism, supra note 1, at 65.

legal paternalism

shown that such measures do not, in fact, infringe any deontological constraint; or (3) justify paternalism whenever enough good (or bad) is at stake to override the constraint. The first, absolutist, position—ruling out paternalism altogether—is internally coherent but suffers from all the drawbacks of absolutist deontology. It is irreconcilable with commonsense morality that legitimizes a great deal of paternalism in both private and public spheres. Hardly anybody supports such an extreme position nowadays.73 In any case, an absolutist position obviates the need for threshold functions because it denies that there is any amount of good that may justify paternalism. The second position—justifying certain instances of paternalism without relying on its contribution to good outcomes—is quite common in the philosophical literature. Several types of arguments are made in this context. Some arguments rely on the agent’s consent, be it hypothetical, tacit, prior, anticipated, or subsequent.74 Arguably, when such consent exists, the paternalistic intervention does not infringe upon the agent’s self-determination. Another argument is that when a person’s choice is uninformed or involuntary, proscribing, or frustrating, it does not violate her autonomy because the frustrated choice is not truly autonomous.75 The same may be said of people who are incompetent to make (certain) decisions due to their very young age, mental retardation, dementia, or comparable causes.76 Finally, paternalistic interventions that aim to secure a person’s future liberty—such as

73. Even Mill did not object to paternalism towards “children, . . . young persons below the age which the law may fix as that of manhood or womanhood,” and “those backward states of society in which the race itself may be considered as in its nonage.” (Mill, supra note 4, at 14). But see Gregory Mitchell, Libertarian Paternalism is an Oxymoron, 99 Nw. U. L. Rev. 1245 (2005) (criticizing Sunstein & Thaler’s argument, supra note 14, that some, very mild, forms of paternalism are compatible with libertarianism); Mark D. White, Behavioral Law and Economics: The Assault on Consent, Will, and Dignity, in New Essays on Philosophy, Politics, and Economics: Integration and Common Research Projects (Gerald Gaus, Christi Favor & Julian Lamont eds., forthcoming) (available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1274444). 74. See, e.g., Kleinig, supra note 1, at 55–67; VanDeVeer, supra note 1, at 45–94; John D. Hodson, The Principle of Paternalism, 14 Am. Phil. Q. 61 (1977). 75. Mill, supra note 4, at 106–07; Feinberg, supra note 5, at 111–16; Kronman, supra note 4, at 786–97; Arneson, supra note 11, at 482–89. One may extend the scope of justified paternalism by using a restrictive definition of what are to be regarded as “autonomous” preferences or decisions. See Scoccia, supra note 6, at 327–34. 76. See, e.g., Kronman, supra note 4, at 786–97.

337

338


prohibitions on enslavement contracts and on the use of drugs—are justified as enhancing autonomy and liberty.77 All of these arguments have been subjected to powerful critiques, and in any event justify only a fraction of the prevailing paternalistic practices. For instance, actual contemporaneous consent eliminates the paternalistic nature of the “intervention,” and hence cannot justify truly paternalistic measures. Subsequent approval may perhaps reduce the blameworthiness of an otherwise immoral paternalistic intervention, but cannot retroactively justify it because there was no actual consent at the time of the intervention.78 Neither is a hypothetical, rational consent an actual one. Relying on hypothetical consent is more often than not a rhetorical devise aimed at presenting a welfarist concern—what the agent should have rationally consented to for her own sake—under the guise of self-determination. It is therefore prone to legitimizing too much paternalism.79 Prior authorization to restrict an agent’s freedom concerning a specified behavior in well-defined circumstances may indeed justify a paternalistic intervention, but the practical significance of this justification is very limited, particularly in the public sphere where people rarely express specific consent to being treated paternalistically (general authorization, such as by participation in democratic elections, is insufficient under most accounts of this justifications). Furthermore, this justification does not legitimize legal paternalism toward people who have not expressed their prior consent (even if most of the people to whom a rule applies have consented to it).80 The notion of incompetence is not very helpful either, unless it is accompanied by welfarist considerations. This is because competence is (almost) always decision-related, and the question of which decisions an agent is competent to make depends on the expected effect of any decision on her well-being.81 Finally, the argument that some paternalistic measures are justified as a means

77. See, e.g., Mill, supra note 4, at 113–15; Joseph Raz, The Morality of Freedom 422–23, 425 (1986); Feinberg, supra note 10, at 62–81; Kronman, supra note 4, at 774–86. See also Gostin & Gostin, supra note 11, at 215–16 (arguing that paternalistic policies aiming at reducing the use of damaging narcotics and other unhealthy activities enhance people’s autonomy). 78. Feinberg, supra note 10, at 187; Kleinig, supra note 1, at 62–63; Thaddeus Mason Pope, Monstrous Impersonation: A Critique of Consent-Based Justifications for Hard Paternalism, 73 UMKC L. Rev. 681, 693–98 (2005). 79. Beauchamp & Childress, supra note 15, at 184–85; Pope, supra note 78, at 698–703. 80. Pope, supra note 78, at 688–91. Cf. supra note 28 and accompanying text. 81. Buchanan & Brock, supra note 4, at 18–20, 29–47.

legal paternalism

to maximize future autonomy is available to consequentialists (assuming plausibly that autonomy is a component of human well-being) and to moderate deontologists (who justify an infringement if enough good is at stake). However, this argument cannot stand on its own, regardless of the goodness of outcomes, because deontology is not a maximization theory.82 It turns out that arguments attempting to demonstrate that some paternalistic interventions do not infringe a deontological constraint—and may accordingly be justified regardless of their effect on the well-being of the person whose freedom is curtailed—are very problematic. At most, they can justify only a small fraction of the paternalistic policies currently implemented in liberal democracies. In any event, there is no need to elaborate on these issues: for once it is established that a certain paternalistic act or rule does not infringe any deontological constraint, such a constraint need not be incorporated into a cost-benefit analysis of such an act or rule.83 We then come to the third position open to deontologists, namely justifying a paternalistic infringement of one’s autonomy for the sake of securing enough good (or, more often, eliminating enough bad) for that person. Indeed, as indicated earlier, we believe that this is the soundest and most fruitful depiction of the moral dilemma inherent in paternalism. It is, therefore, important to set a threshold that has to be met for the paternalistic intervention to be deontologically permissible (K in the notation proposed in chapter 4) and to resolve what types of costs and benefits should be deemed relevant in calculating the net benefit of the intervention (B in our notation). Finally, even if a certain paternalistic measure, standing on its own, is permissible, it may be judged impermissible if one can implement alternative measures with a less adverse effect on people’s autonomy and freedom. We discuss these three issues in turn.

2. Size of the Threshold In section C above, we proposed a simple model to determine when a paternalistic rule is efficient. To incorporate a constraint against limiting

82. See, e.g., Regan, supra note 18, at 116–17; Feinberg, supra note 10, at 75–79. In addition, a person who adversely affects her own future autonomy is not necessarily infringing a deontological constraint, and thus preventing her from taking an action that would have limited her future autonomy does not necessarily decrease the total infringements of deontological constraints. 83. Cf. supra p. 281.

339

340


people’s freedom, a threshold function may deviate from the original model in two respects. One modification would be to require that the net benefit should not only be positive (as required in the simple efficiency function (1)), but greater than some value K, representing the inherent value of personal freedom. Accordingly, function (1) would be modified as follows: n

(2) T = ∑ ( Pli Li − Pg i Gi − Fi ) − C − K i =1

The size of the constraint K may vary considerably from one case to another. It depends, among other things, on the extent to which the frustrated choice is regarded as an important element in the agent’s life plan. To illustrate, the threshold that has to be met to permit a rule forcing people to use seat belts is plausibly higher than the threshold for a rule invalidating an unconscionable clause in a standard-form, consumer contract, assuming it rests on paternalistic grounds. At the same time, the threshold for the seat belts rule should be considerably lower than the threshold for a paternalistic rule forbidding the engagement in self-harming religious practices.84 The centrality of the frustrated choice in the agent’s life plan is closely connected to the notion of autonomy and to our model’s incidence. The reason for limiting the basic efficiency model to deviations from thin, cognitive rationality was primarily methodological: it rested on the difficulty to model deviations from thick, motivational rationality. Once we set a constraint against interfering in people’s autonomy, the distinction between choices reflecting cognitive limitations and choices based on nonselfish motivations (altruism, commitment, etc.) takes on a different significance. Interference resting on cognitive errors does not necessarily infringe upon one’s autonomy, or at least less significantly so. It basically assists people in accomplishing their own goals, unaffected (or less affected) by their biases and cognitive errors. In contrast, frustrating choices on the alleged basis of the agent’s “motivational irrationality” is a much more blatant assault on people’s autonomy, as it refers to ends and not merely to means.85 It should

84. See Dan W. Brock, Paternalism and Autonomy, 98 Ethics 550, 551 (1988); Regan, supra note 18, at 119; Robert N. Van Wyk, Children and Community: A Reply to Jonathan Schonsheck’s “Deconstructing Community Self-Paternalism,” 15 Law & Phil. 75, 77–78 (1996). 85. Raz, supra note 77, at 422–23; Beauchamp & Childress, supra note 15, at 187–88; Pope, supra note 58, at 333–67. Cf. Ronald Dworkin, Equality and the Good Life, in Sovereign Virtue: The Theory and Practice of Equality 237, 242–76 (2000). Distinguishing

legal paternalism

be conceded, however, that the very distinction between ends and means (or, for that matter, between cognitive and motivational rationality), is sometimes blurred. Consider a person who is about to rush into a burning building, where she is likely to get killed. She may do so because she does not understand the risk involved, out of carelessness or adventurousness, or in a heroic attempt to rescue people trapped inside. Paternalistically preventing this person from entering the building is much more justifiable in the first case, where she acts out of cognitive deficiency, than in the third case, where she acts out of deeply held moral conviction (the second case is somewhere in between). However, it may be very difficult to distinguish between the cases, as even the agent herself may not be able to disentangle the cognitive and motivational causes of her action. The size of the threshold should also depend on the rule’s intrusiveness. A rule concerning private and intimate aspects of one’s life is likely to entail a higher threshold than a rule that does not concern such aspects (possibly because the former involves the infringement of an additional constraint: the constraint against interfering in one’s privacy). The size of the threshold may also reflect the degree to which the rule restricts one’s freedom. For instance, the threshold that has to be met when permitting involuntary civil commitment is much higher than the one applicable when invalidating a freely made enslavement agreement. While such invalidation interferes with the agent’s freedom of contract, on the whole it does not diminish her autonomy. In the same vein, the harsher the sanctions used to ensure compliance with the paternalistic rule, the higher the threshold that has to be met to justify the rule. The same rule may be enforced by different means, only some of which meet the threshold. Thus, for example, to encourage people to save for retirement, the paternalist may, at one extreme, impose criminal sanctions on anybody who does not join a pension plan, or, at the other extreme, merely set the participation in a pension plan as the default.

between volitional interests (having or achieving something a person wants) and critical interests (having or achieving those things that would make a person’s life better), Dworkin argues against state paternalism aimed at interfering with the critical interests people have and endorse. Apparently, Dworkin does not object to volitional paternalism, forcing people to avoid harms they presumably want to avoid anyway. On Dworkin’s position, see Christopher Wolfe, Liberalism and Paternalism: A Critique of Ronald Dworkin, 56 Rev. Pol. 615 (1994); T.M. Wilkinson, Dworkin on Paternalism and Well-Being, 16 Oxford J. Legal Stud. 433 (1995). See also Childress, supra note 9, at 124.

341

342


Arguably, sometimes infringing upon one’s right for her own good calls for a lower threshold compared to the case in which a constraint is infringed in order to promote the interests of another person. Breaking into someone’s property to save it from damage (assuming it is impossible to get the owner’s consent on time) is more easily permitted than breaking into the same property to save someone else’s property from damage.86 The former is, however, hardly a case of paternalistic intervention, because it involves infringement of property (and possibly privacy) rights but not a direct infringement of the agent’s freedom. A higher threshold should plausibly be set in truly paternalistic cases, i.e., when the actor ignores the agent’s choice. At times (for instance, when saving one’s life requires amputating her leg against her will), an action simultaneously infringes the constraint against curtailing a person’s autonomy and other constraints. Plausibly, a greater amount of good is necessary to justify the cumulative infringement of two constraints.87 Thus far, our analysis reflected the judgment that different paternalistic regulations harm the agent’s autonomy to different extents by setting different additive thresholds for different paternalistic measures, depending on the extent to which each one of them curtails people’s autonomy. An alternative way to formalize this notion, which would enable one to use the same threshold function to assess the permissibility of different paternalistic measures, is to opt for a multiplier, or a combined—additive and multiplier— threshold.88 Such a function would mean that the more harmful a measure is to one’s autonomy and freedom, the greater the expected net benefit of the regulation necessary to render it permissible. It would require adding to the formula not only a multiplier threshold K ′, but also a notation for the adverse effect of the assessed regulation on the agent’s freedom.

3. Relevant Types of Benefits and Costs As explained in chapter 4, threshold functions often deviate from standard CBA not only in setting a threshold that the net benefit of the rule must meet

86. Samantha Brennan, Paternalism and Rights, 24 Can. J. Phil. 419 (1994). 87. See supra p. 97. The last proposition assumes that cutting off a person’s leg to save her life constitutes harming her, but one may argue that such a case involves no harm at all, because the agent is actually benefited on balance. On this and related questions, see Kagan, supra note 17, at 86–90; Brennan, supra note 86. 88. See supra pp. 93–96.

legal paternalism

to render it deontologically permissible but also in limiting the type of benefits and costs taken into account. Following the discussion in chapter 4,89 some or all of the following “excluders” may be incorporated into the threshold function (and others may be considered as well). Small Benefits One may hold that respect for autonomy permits frustrating an agent’s choice only if the expected loss to her from incorrectly choosing Y (absent the rule) surpasses a certain threshold, m.90 This requirement may apply to the expected loss from choosing X when Y is incorrectly chosen in the i-th case (that is adding a cumulative limitation, Li > m, to function (2)), or alternatively to the expected loss from choosing X when Y is incorrectly chosen in the i-th case, multiplied by the probability of such a choice (thus adding the limitation: PliLi > m).91 In practical terms, to meet this constraint, the applicability of a paternalistic rule may have to be confined to instances in which the expected loss from incorrectly choosing Y surpasses a certain threshold. For example, a central function of the statute of fraud is to paternalistically caution people against acting hastily.92 Assuming that there is a correlation between the scope and importance of a transaction and the risks involved in making it hastily, limiting the statute’s application to important and relatively large transactions may thus rest on the present conception.93 In a similar fashion, laws regarding the legal capacity of minors and the mentally disabled may impose varying restrictions on the validity of different transactions according to their potential adverse effect on the protected person’s well-being.94

89. See supra pp. 86–93. 90. In support of this proposition, see, e.g., Jeffrie G. Murphy, Incompetence and Paternalism, 60 Archiv für Rechts- und Sozialphilosophie 465, 479, 483 (1974); Childress, supra note 9, at 125. For a critique of this position, see Hodson, supra note 74, at 64, 65. 91. For the latter proposition, see Eric D. Johnson, Sounds of Silence for the Walkman Generation: Rock Concerts and Noise-Induced Hearing Loss, 68 Ind. L.J. 1011, 1026–27 (1993). 92. On the cautionary function of formal requirements, see, e.g., Lon L. Fuller, Consideration and Form, 41 Colum. L. Rev. 799, 800 (1941). 93. On the statute’s scope of applicability, see generally Restatement (Second) of Contracts § 110. 94. See, e.g., § 786 of the Civil Code of Puerto Rico (31 L.P.R.A. § 786 (2004)) (requiring court authorization for certain acts by guardians of minors and incapacitated persons, including the alienation of their real property, making of contracts requiring recording, the

343

344


Even under the basic function (1), to be efficient (B > 0), the applicability of some paternalistic rules may have to be restricted to instances in which Li (or PliLi) exceeds a certain value. For example, given a very high Fi and C, involuntary civil commitment would be efficient only if the harm the patient might inflict upon herself is quite serious. However, whereas under function (1), this conclusion depends on the values of the other variables (such as the subjective disutility experienced by the patient and the implementation costs of the rule), the deontological constraint—embodying the inherent value of freedom and autonomy—is independent of such other variables. The Existential Constraint Instead of (or in addition to) excluding any expected loss from incorrectly choosing Y (Li) that is smaller than m, it may be required that in at least one instance, the loss Li would exceed a (higher) threshold h.95 For example, a risky activity may be prohibited if in at least one instance it may result in serious physical harm to the actor, even if under most circumstances its expected harm is not serious. The lesser harms, however, would still be taken into account. Chronologically Remote and Probabilistic Costs and Benefits A threshold function may plausibly exclude some chronologically remote or low-probability costs or benefits of the paternalistic rule. For example, respect for personal freedom may militate against prohibiting activities involving risks with a very low probability. Similarly, one may wish to exclude any paternalistic intervention whose probability of success in preventing the loss or injury to the agent is too low.96 At the same time, the argument that paternalism adversely affects the development of judgment skills may similarly be excluded in at least some contexts as too remote. Indeed, such effects were excluded from the original model.97 However, while the exclusion from the original model rested on methodological grounds (the

alienation of personal property the value of which exceeds one thousand dollars, and the execution of lease contracts for a period longer than six years); sections 7, 20, and 47 of the Israeli Capacity and Guardianship Law, 1962 (16 L.S.I. 106 (1961–62)) (laying down comparable requirements). 95. On the existential constraint, see generally supra p. 88. 96. Beauchamp & Childress, supra note 15, at 186–87; Pope, supra note 58, at 378–83. 97. See supra pp. 332–35.

legal paternalism

difficulty of verifying and quantifying such remote and manipulable effects), within the deontological framework such exclusion may rest on the normative judgment that such remote and uncertain effects should not bear on the permissibility of paternalism. Promoting the Good vs. Eliminating the Bad Following the prevailing economic convention, the original model equates losses with foregone gains and gains with foregone losses. In contrast, there seems to be a prevalent notion that limiting people’s freedom for their own sake is much more justifiable when done to prevent loss and suffering than when it is done to ensure greater gains and enjoyment.98 For example, forcing a patient to undergo a medical procedure that would save her from premature death may be justifiable under certain circumstances, while forcing her to undergo the same procedure in order to prolong her life beyond her normal life expectancy may not. Similarly, prohibiting the manufacturing, dissemination, and use of dangerous food seems much more justifiable than compelling the manufacturing, dissemination, and consumption of particularly healthy food.99 A threshold function may take this consideration into account by discounting or even disregarding the benefits of a paternalistic rule belonging to the latter type. It should be noted, however, that the very distinction between promoting the good and eliminating the bad is often unclear, and that in some contexts— particularly, but not exclusively, with regard to children—paternalism aimed at promoting the good is considered permissible. Thus, mandatory elementary and high school education may be seen as paternalism aimed at promoting the career prospects of children and adolescents, or alternatively as preventing the hardships facing uneducated people in societies where most people have at least high school education. The same ambiguity exists with regard to mandatory or semi-mandatory pension arrangements. The appropriateness of any of the above excluders, their content and magnitude, plausibly depend on the type and scope of the paternalistic intervention. If the regulator merely sets default rules that would enhance

98. Kleinig, supra note 1, at 13–14; Pope, supra note 58, at 312–20. On the distinction between promoting the good and eliminating the bad in the context of deontological constraints, see generally supra pp. 91–93. See also supra pp. 287–88. 99. This example, however, involves not only the promotion of the good/elimination of the bad distinction, but also the preventing/forcing distinction (discussed in supra note 10 and accompanying text), which also affects our moral intuitions.

345

346


the agent’s welfare, the necessary constraint should be minuscule because the harm to people’s autonomy is trivial, and the scope of benefits and costs taken into account may be unlimited. As the practical obstacles to deviating from the default get higher, the constraint may be increased and the types of benefits considered in justifying it restricted correspondingly.100 At the other extreme, forced hospitalization is a harsh form of paternalism. In function (1), this fact was likely to affect the frustration costs experienced by the hospitalized person, Fi. From a deontological perspective, it justifies the incorporation of an independent constraint (of the type Li > m, or PliLi > m), and the exclusion of many types of expected benefits of the hospitalization. Importantly, such a constraint and such exclusion are justified irrespective of the subjective frustration experienced by the patient.

4. Marginal Net Benefit and Alternative Measures Thus far, we proposed threshold functions aimed at determining the permissibility of any paternalistic intervention examined on its own. In real life, however, there is often more than one way to protect individuals from loss or injury, varying in their effectiveness and their harm to people’s autonomy and freedom. For instance, one may try to reduce smoking hazards by providing people with information about those hazards, by restricting cigarette advertisement, by imposing “sin taxes” on tobacco products, or by imposing an absolute prohibition of smoking backed up by criminal sanctions. The first measure does not infringe the constraint against curtailing people’s autonomy and freedom. Arguably, it actually expands freedom of choice by enabling people to rationally consider the pros and cons of smoking and make an informed decision. The last measure is the most disrespectful of people’s autonomy, leaving no room (or very little room) for independent, free choice. The second and third measures lie somewhere in between these two extremes. As we argued in other contexts,101 in choosing among these measures (or between any combination thereof), a moderate deontologist should compare each measure not only to inaction, but also to other courses of action that

100. On various obstacles to deviations from default rules, and their undermining effect on the very distinction between default and mandatory rules, see sources cited in supra note 14. 101. See supra pp. 149–50, 155–56, 170, and 222–23.

legal paternalism

either do not infringe any constraint, or whose net benefit is large enough to render the infringement permissible.102 The comparison to alternative, deontologically permissible courses of action may be formalized by redefining the relevant variables of the threshold function according to their marginal values. Specifically, the measure’s relevant benefit is only the difference between the expected reduction of smoking hazards if an alternative course of action is pursued, and the magnitude of this reduction under the current one. Similarly, the relevant cost is only the added harm that the infringement inflicts on people’s autonomy, in comparison to the alternative. Thus, the inquiry is whether the more harmful infringement is justified given its marginal net benefit. This qualification may well result in a change to the value of the threshold K. The comparison to an alternative should not be restricted to an assessment of whether the action under consideration is the least harmful among all possible infringements that are equally effective in thwarting the risk. In the above example, presumably none of the less intrusive antismoking measures is as effective as a complete ban on smoking; and yet the marginal net benefit of adopting this measure may be too small to justify it. The evaluation of an infringement must therefore include a comparison to less harmful alternatives that are less effective in protecting people.

•

E. Conclusion

Many discussions of paternalism, including legal paternalism, focus either exclusively on welfare or exclusively on liberty. Such discussions miss the fundamental tension inherent in paternalism—between welfare and liberty. In this chapter, we argued that the best way to grasp the dilemma and to satisfactorily resolve it in legal contexts is by integrating a deontological constraint against restricting people’s freedom into a welfarist cost-benefit analysis. This analysis rejects both the notion that economic efficiency entails antipaternalism, and the belief that a sound justification for paternalism may disregard its effects on well-being. As demonstrated, the combination of threshold deontology with economic modeling techniques provides a framework for a balanced, comprehensive, and reasoned analysis of paternalism.

102. Cf. Beauchamp & Childress, supra note 15, at 186; Pope, supra note 58, at 368–77, 389–400.

347


Conclusion

economic analysis of law, and of human behavior in general, is a powerful

analytical methodology. At the same time, standard cost-benefit analysis (CBA), the common measure of efficiency employed by legal economists, is normatively objectionable on various grounds. This book focused on the consequentialist aspect of welfare economics, primarily on the lack of constraints on promoting the best outcomes, and secondarily on the lack of options not to promote such outcomes. Moderate deontology is normatively more defensible and is more in accord with commonsense morality and existing legal doctrines. Often, however, it is less determinate and rigorous than standard CBA. We argued that integrating threshold constraints (and options) with CBA would overcome both deficiencies. Furthermore, such integration is compatible with, and perhaps even mandated by, the more sophisticated versions of foundational consequentialism. The proposed integration raises principled and methodological concerns but we believe that these concerns can all be suitably addressed. Adding threshold constraints (and options) to economic analysis makes it not only normatively more acceptable, but also descriptively more valid, without significantly compromising its methodological rigor. The advantages of constrained CBA are particularly conspicuous in the analysis of nonmarket spheres, such as constitutional, criminal, tort, and family law, where deontological constraints loom large, but they are significant in market contexts as well, as we have demonstrated in the contexts of antidiscrimination laws and precontractual deception. Hopefully, such integration would contribute to bridging the increasing gulf between economic analysis and other approaches to law and legal theory. Putting aside other contested aspects of standard CBA (such as its underlying theory of human welfare and its stance regarding welfare distribution), deontologically constrained CBA holds that the legal system should strive to maximize social welfare subject to threshold constraints. In evaluating an act (or a rule), the first question to be asked is whether it infringes upon a 349

350


deontological constraint, that is, primarily, whether it involves an active/ intentional infliction of harm. If the answer is in the negative, then one may proceed using standard CBA. If, however, the act does infringe a constraint, one must ask whether the net benefit of such an infringement is great enough to justify the infringement. This question should be answered through the construction and employment of a threshold function. A threshold function determines the amount of net social benefit that is required to render the infringement permissible (the threshold), and the type of effects (costs and benefits) that are taken into account in calculating the net benefit. Based on normative judgments, some effects of the infringing act may be excluded altogether and others may be discounted. The pertinent threshold may or may not depend on the magnitude of the harm caused by the infringing act. Depending on the variety of costs and benefits that are deemed relevant in calculating the net benefit and setting the threshold, a threshold function may face greater or lesser challenges of commensuration and quantification. In some contexts, explicit mathematical formalization with concrete numbers may be fruitful. In others, the threshold function only serves as a heuristic to highlight the pertinent moral factors and their interrelations. In yet other contexts, expressive and other concerns may militate against the use of mathematical functions, and we must content ourselves instead with verbal formulations embodying moderate deontology. Having determined that a certain act is morally permissible because its net benefit is large enough to override the threshold constraint, one often faces another task, namely choosing among alternatives that either do not infringe any constraint or produce enough net benefit to override the constraint. This stage may require lexical ordering of different constraint infringements. It may also be combined with the previous stage by constructing a threshold function that focuses on the marginal net benefit of a certain infringing act, compared to acts infringing lesser constraints. The first part of the book presented our proposal in relatively general terms. We indicated the major distinctions between different types of costs and benefits that should be considered when formulating threshold functions and proposed ways to formalize the shape and size of the threshold within such functions. It is clear, however, that specific choices among the different possibilities have to be made and defended in any specific context. The second part of the book thus illustrated the deficiencies of standard economic analysis in several prominent legal issues and critiqued attempts to overcome these deficiencies within a purely consequentialist analytical framework. We then described in some detail our alternative method,

conclusion

integrating deontological constraints (and options) with CBA of those issues. Some of the illustrations included the construction of mathematical threshold functions. We sought to demonstrate that deontologically constrained CBA facilitates a better grasp of, and more fruitful grapple with, the relevant normative issues. The proposed integration can by no means serve as a substitute for moral, policy, and legal deliberation. It merely highlights the crucial factors that must be taken into account and elucidates their interrelations. Deontologically constrained CBA is more complex than standard CBA, but this normative complexity exists anyway. One may ignore this complexity or try to cope with it by manipulating other elements of the analysis. However, the first option seriously decreases the fertility and relevance of the analysis and the second blurs the real issues instead of clarifying them. Integrating deontological constraints with CBA is a more rigorous and sophisticated solution.

351


Index

Act/omission. See Doing/allowing Actual preferences. See Theory of the good Adler, Jonathan 285 Adler, Matthew 2–3, 16n17, 28n66, 28n67, 107n8, 108n10 Agent-relativity (See also Deontology) 20, 31, 33n86, 36–37, 41–43, 49, 57–59, 71, 172–74, 289–91 Affirmative action (See also Antidiscrimination laws; Discrimination) 228–29 Aggregation of harms (See also Taurek, John M.; Deontology, Threshold deontology) 54–55 Aggressors (See also Self defense; Culpability) 136–38, 141–45, 151– 60, 164–65, 167–68, 171–74 Alexander, Larry 20n34, 49n29, 53n46, 53n47, 55–56, 59n6, 67n36, 89n16, 89n17, 94n33, 102n54, 113n25, 114n27, 179n4, 181n14, 187n50, 190, 193n82, 197n94, 204n118, 204n119, 214n152, 217, 245n74, 245n79, 248n92, 249n95, 255n115, 274n46, 276n57, 279n62, 280n66, 281n70 Americans with Disabilities Act of 1990, 225, 228n7, 229n9, 230–31, 236n40, 236n42, 246, 248n90, 252n105, 253–54 Antidiscrimination laws (See also Affirmative action; Discrimination) 66, 225–56 autonomy 226–27, 240–46, 249–50 accommodation 226, 228–31, 235–37, 239, 246–48, 250n100, 251–55 Americans with Disabilities Act of 1990 225, 228n7, 229n9, 230–31, 236n40,

236n42, 246, 248n90, 252n105, 253–54 Civil Rights Act of 1964, Title VII 225–26, 228n7, 229n8, 229n11, 230, 236n40, 237, 239, 245n74, 246n83, 246n84, 252n104, 252n105, 253 conflicting constraints 229–30, 244–46 constrained cost-benefit analysis 251–55 disparate impact 228–31, 246–55 deontological constraints 240–44 doctrinal background 225–26, 228–31 economic analysis 231–40 incentives, effect on 235–37 long-term effects 235–40 market failures and 235–40 preferences and 237–40 racial discrimination 225, 230, 234, 241–42, 251–53 redistribution 246–51, 255 statistical discrimination 234, 247–49 Arrow’s impossibility theorem 14 Autonomy (See also Paternalism; Contract law) 20, 41, 59–63, 104, 163, 187, 190–94, 200, 202–06, 212, 216, 218–22, 226–27, 240–46, 249–50, 262, 275, 278–79, 298–300, 305–06, 313–17, 320, 335–39, 340, 341, 344, 346–47 contract law 260–62, 275–76, 278–79, 298–300, 305–06 discrimination 240–46 free speech 187, 190–94, 200, 202–06, 212, 216, 218–22 paternalism 313–17, 320, 335–39, 340, 341, 344, 346–47 torture 163–64

353

354

index Balance of Interests 127–28, 130–34, 180, 185–86, 214–16 Becker, Gary 231–33, 237, 324n34 Behavioral law and economics. See Bounded Rationality; Cognitive psychology Bounded rationality (See also Cognitive psychology; Rationality) 15–16, 17n24, 76, 90, 107, 110n14, 152, 200n103, 309, 317n14, 319–20, 325–32, 334, 340 Brandenburg v. Ohio 183, 188 Brennan, Samantha 21n40, 26n57, 29n70, 46n21, 48n24, 78n64, 87n11, 88, 94n33, 95n36, 97n41, 97n42, 342n86, 342n87 Canadian Charter of Rights and Freedoms 150n89, 179, 184 CBA. See Cost-benefit analysis Civil Rights Act of 1964, Title VII 225–26, 228n7, 229n8, 229n11, 230, 236n40, 237, 239, 245n74, 246n83, 246n84, 252n104, 252n105, 253 Cognitive psychology 16, 17n24, 43n8, 76, 90, 97n41, 107, 110n14, 152, 200n103, 309, 317n14, 319–20, 325–32, 334, 340 Combatants vs. civilians (See also Terrorism) 157–60, 171–75 Commodification 111 Commonsense morality 2, 23, 24, 33–36, 48, 56, 77–78, 84, 90–92, 99, 117, 189, 211, 213, 224, 240, 263–67, 286, 337, 349 Consequentialism (See also Deontology; Public morality) 1–4, 12, 18–40, 57–78, 91–92, 95, 100, 106, 113–14, 117, 130–36, 167–68, 171, 189–91, 193–95, 214–15, 235–37, 250–51, 259, 262–67, 313–14, 318–27 definitions 18 demandingness objection 20–21, 33–40, 289–301 collective, rule-consequentialism 37–38 extremism 35–36

long-term and indirect effects 34–35, 117 “preferences for options” 36 satisficing 38–39 theories of the good 36–37 lack-of-constraints critique 19–33, 104, 131–34, 194, 215, 238–40, 262–67, 271, 274, 320–21 “consequentialist vacuum cleaner” 31 feelings of virtue and remorse 29–30, 237–38 ideal preferences 30–32, 110, 131, 189n61, 194, 201, 215, 238–40, 320–21 theory of the good 30–32, 104, 110, 131, 189n61, 194, 201, 215, 238–40, 320–21 long-term and indirect effects 22–24, 196–201, 205–08, 232–37, 328 “preferences for constraints” 27–29, 30, 207, 237, 239–40, 331–32 rule-consequentialism 24–27, 48, 62–63, 131–32, 151–52, 163, 188–89, 193, 196–98, 200, 207–08, 237, 329–31 public morality 57–78 rule-consequentialism 24–27, 37–38, 48, 49, 62–63, 131–32, 151–52, 163, 188–89, 193, 196–98, 200, 207–08, 237, 329–31 “Consequentialist vacuum cleaner” 31, 36–37 Constitution, United States 131, 177–84, 188, 198, 213, 216, 219, 229n9 Constraints (See also Deontology; Options; Threshold deontology; Threshold functions) 19–33, 41–56, 60–70, 81, 106, 135–39, 189–95, 240–51, 260–61, 274–91, 298–310, 335–47 conflicting constraints 81, 137, 203–04, 242–44 deception, prohibition on 192, 203–04, 274–77 discrimination, prohibition on 157–60, 240–51

index harming people 19–33, 42–46, 50–51, 135–38, 152–53, 161–63, 166–68, 171–74, 240–44, 260–61, 274–77, 298–301, 305–10, 335–47 lying, prohibition on 42, 46, 52, 73–74, 83, 101n50, 95, 102, 117, 192, 203–04, 260, 262, 274–88, 291 multiple/cumulative 97, 242–44, 341, 342 options and 47, 60–63, 98, 255 “preferences for constraints” 27–29, 207, 237, 239–40, 331–32 quantification and monetization 28, 93–94, 108–16 size of threshold 59, 93–96, 116–17, 150– 61, 165, 169–75, 213–22, 251–55, 286–87, 339–42 Contingent valuation methods (CV/CVM) 28 Contract law (See also Deception; Promises) 114–15, 257–311, 317, 343 allocation of risks, 258, 269, 270n33, 303–04, 310 deception 267–91 doctrinal background 267–69 economic analysis 269–74 questions, in response to 283–86 silence, by 268, 275, 278, 283–86 spontaneous 283–86 deontological constraints 260–61, 274–91, 298–310 disclosure duties 267–74, 278, 280–81, 286, 289–91 disgorgement 293, 308 economic analysis 258–59, 262–67, 269–74, 294–97 efficient breach 95n35, 114–15, 264–67, 294–98 expectation interest 292–97, 301–10 liquidated damages 301–03 mistake and misrepresentation 267–91 doctrinal background, 267–69 economic analysis 269–74 reservation value 280–81 constrained cost-benefit analysis 277–91

nondisclosure 267–74, 278, 280–81, 286, 289–91 precontractual liability 258, 267–74 promises 260–62, 264, 266–67, 298–301, 305–06, 308n144, 310n151 reliance interest 259–63, 279, 285, 292–93, 295, 299–300, 306–08, 310 remedies for breach of contract 262–63, 292–310 constrained cost-benefit analysis 301–10 deontological constraints 261, 298–301 doctrinal background 292–94 economic analysis 294–97 restitution 262, 293, 299, 301, 303–05, 308 will theory 260–64, 298 Contractarianism 48, 62, 192–93 Convention on the Prevention and Punishment of the Crime of Genocide 200–01 Cost-benefit analysis (See also Economic analysis; Normative economics) 2–5, 15–18, 26, 27, 29–32, 35–37, 39–40, 47, 79–108, 111–16, 122–23, 129–34, 161, 184–87, 332–35 normative neutrality 105–08 theory of the good 15–18, 27, 30–32, 59n7, 105–06, 131, 189n61, 194, 201, 215, 238–40, 263, 321–25 Culpability (See also Self defense) 19, 136–38, 141–45, 151–60, 164–65, 167–68, 171–74 Deception (See also Contract law) 42, 46, 52, 73–74, 83, 101n50, 95, 102, 117, 192, 203–04, 260, 262, 267–91 deontology 42, 46, 52, 73–74, 83, 101n50, 95, 102, 117, 192, 203–04, 260, 262, 274–77 doctrinal background 267–69 economic analysis 269–74 harmless 102, 275–77, 278–80 intentionality 281–83 lying 42, 46, 52, 73–74, 83, 101n50, 95, 102, 117, 192, 203–04, 260, 262, 274–88, 291

355

356

index Deception (Cont.) questions, in response to 283–86 reservation value 280–81 silence, by 268, 275, 278, 283–86 spontaneous 283–86 Democracy 62–63, 66–67, 73, 121, 190, 199, 214, 331, 338 Deontology (See also Autonomy; Constraints; Deception; Doing/ allowing; Intending/foreseeing; Options; Threshold deontology) 1–4, 19–21, 25, 41–78, 79–104, 116–17, 130, 141–44, 150, 161–66, 167, 172–74, 264–67, 274–77, 282, 289–91, 313, 335–48 absolutist 46, 49–52, 55, 116, 130, 141–44, 161–66, 167, 264–67, 275–77, 282, 336–37 agent-relativity 20, 31, 33n86, 36–37, 41–43, 49, 57–59, 71, 172–74, 289–91 critique of deontology 49–56, 116–18, 121, 130–34, 150 deception 42, 46, 52, 73–74, 83, 101n50, 95, 102, 117, 192, 203–04, 274–77 doing/allowing 21, 43, 45–46, 50, 57, 60–63, 76, 92, 132, 134, 135, 153, 243–44 exclusion of costs and benefits 47, 59, 85–93, 98, 140–45, 147–49, 195–211, 237–40, 254–55, 287–88, 342–46 combining excluders 147–49, 210–11 chronologically-remote costs and benefits 89, 147–48, 153–56, 182–86, 188, 194, 196–200, 203, 210–02, 222, 254, 344–45 existential constraint 88, 344 lexical priority 47, 82–83, 86–87, 148, 222–23, 287, 350 low-probability costs and benefits 89–91, 95, 100, 147–48, 151–57, 195–96, 199–201, 254, 287, 344–45 promoting the good vs. eliminating the bad 47, 77, 91–93, 100, 287–88, 345–46 small costs and benefits 87–88, 201–02, 208, 210, 343–44

lying 42, 46, 52, 73, 83, 101n50, 102, 117, 192, 260, 262, 274–88, 291 options 20, 33–40, 41, 43, 46–47, 51, 53, 59–63, 83–84, 98–103, 255, 289–91 “paradox of deontology” 49–50 promise 20, 29, 41–42, 46, 47, 52, 73, 94–97, 102, 114–15, 117, 260–62, 264, 266–67, 276, 298–301, 305–06, 308n144, 310n151 public morality 57–78 threshold deontology 2, 25, 46–56, 52, 116–17, 140–75, 195–224, 251–55, 277–91, 305–10, 336–37, 339–47 Deterrence 22, 57, 68n38, 77, 120, 128–29, 136–37, 140–43, 145, 147, 150–51, 170, 190, 214, 296, 304n136, 309 Dignity 20, 41–42, 61, 73, 104, 134, 135–39, 141–44, 148, 152, 161–64, 167, 184, 187, 190–94, 200, 202–06, 208, 212, 216, 219–21, 226–27, 242, 275, 278–79 Disclosure duties. See Contract law; Deception Discrimination (See also Antidiscrimination laws) 66, 73–74, 225–56 anti-terrorist measures and 157–60 constraint against 157–60, 240–51 economic analysis 231–40 contestable markets 232–33 equal respect in curtailing speech 192, 219 incitement to 182–84, 205 motivation for 66, 231–40 “rational” 233–34 statistical 234, 247–49 Disgorgement interest. See Contract law Distributive justice (See also Redistribution) 13–14, 17–19, 80–81, 104, 105, 246–51, 255, 316, 331, 332 Doing/allowing 21, 43, 45–46, 50, 57, 60–63, 76, 92, 132, 134, 135, 153, 243–44, 274–75 intending/foreseeing and 45–46, 243–44 Doomed people 167, 170–71

index Economic analysis. See Cost-benefit analysis; Normative economics; Positive economic analysis; Rational choice theory Edmond v. Goldsmith 157n120 Eisenberg, Melvin A. 273n41, 273n43, 291n95, 297n114, 307n143, 309n148 Efficiency. See Normative economics; Kaldor-Hicks efficiency; Pareto Efficient breach. See Contract law Employment. See Antidiscrimination Laws; Discrimination Epstein, Richard A. 189n59, 226, 232n24, 233n31, 234n33, 235n35, 238n49, 239n53, 245n80 Errors, types of 152–53, 186n45, 213–15, 334 Existential constraint (See also Threshold deontology; Threshold functions) 88, 344 Expectation interest (See also Contract law) 292–97, 301–10 Expressive role of law 53, 74, 117–22, 214, 237–40, 266–67, 301–03, 306, 310n151 Factoral vs. foundational moral theories (See also Consequentialism; Deontology) 26–27, 48, 49, 56, 62, 71n47, 114, 117, 191–95, 198, 215, 251, 320n22 Fairness 17, 27, 42, 47, 97, 138, 160, 240–41, 280, 299, 326 First amendment. See Freedom of speech Freedom of speech 66–67, 177–224 categories of speech 178–80, 182–87, 190–91, 203, 218–21 choice among permissible courses of action 222–23 chronologically remote harms 182–83, 185–86, 188, 194, 196–200, 203, 210–02, 222 commercial speech 179, 206, 218–20 consequentialist justifications 189–91, 194, 196–98, 200–01, 207–08, 218 constrained cost-benefit analysis 187–88, 195–224

content-based restrictions 66–67, 77, 180–87, 213–15 content-neutral restrictions 66–67, 77, 180–81, 215–17 deontological justifications 187–95, 198–201, 208 doctrinal background 178–84 economic analysis 133, 184–88 evaluative neutrality 187, 218–21 First Amendment 67, 178–80, 183 genocide, advocacy of 200–01 hate speech 178–84, 197, 200–01 imminent lawless action 182–83, 185–86, 188, 194, 196–200, 203, 210–12, 222 incitement to violence 133, 182–84, 196–200, 205–07, 210–11, 219, 222–23 low-probability harms 195–96, 199–201 “marketplace of ideas” 184, 186, 197–98, 202–04 offensiveness 206–10 persuasion principle 192–94, 202–04 political speech 218–19 “protected” vs. “unprotected” speech 178–80, 182–87, 190–91, 203, 218–21 rational persuasion 192–94, 202–04 pornography 178, 183, 185, 206–09, 218–21 small harms 201–02 “track one” and “track two” restrictions 66, 77, 180–87, 213–17 violence, incitement to 133, 182–84, 196–200, 205–07, 210–11, 219, 222–23 Fried, Charles 52n43, 116n32, 169n158, 260n4, 263n13, 275n53, 276n54, 276n55, 276n58, 284n76, 298, 299n119, 299n120, 306n140 Genocide, advocacy of (See also Freedom of speech) 200–01 German Constitutional Court 118, 136–37, 139, 167, 170, 182, 184 Good, theories of the. See Theory of the good

357

358

index Greenawalt, Kent 178n3, 183n22, 190, 193n80, 193n81, 201n108, 207n125, 210n137, 210n139 Harm. See Constraints; Deontology; Doing/allowing; Intending/ foreseeing; Threshold deontology; Threshold functions Harm principle 249n97, 250, 336 Hate speech. See Freedom of speech Holmes, Oliver Wendell, Jr. 182, 210n138, 266n22 Human welfare. See Theory of the good Ideal preferences. See Theory of the good Imminent lawless action. See Freedom of speech Incommensurability 17, 110, 112–17, 149, 186 Incomparability 53, 112–13, 115 Innumerability argument. See Aggregation of harms; Taurek, John M. Intending/foreseeing (See also Deontology; Doing/allowing) 21, 44–46, 50, 57, 61n13, 63–70, 135–36, 166–70, 174–75, 180–81, 189, 194, 213–18, 243–44 antidiscrimination laws 243–44 anti-terrorist measures 67–68, 70, 135–36, 166–70, 174–75 causality test 45, 135–36 deception 281–83 doing/allowing and 45–46, 243–44, 281–83 free speech 180–81, 215–17 means/side-effect 44–46, 63–70, 135–36, 166–70, 174–75, 180–81, 189, 194, 213–18 public morality 57, 63–70 trolley problem 44–46, 282n72 Israeli Supreme Court 118, 142–44, 147, 158n126, 158n127, 174n171, 288n88 Judicial review 63–70, 76, 118–19, 122n57, 133, 142, 178–84, 187, 197, 213–15, 219, 331 Kagan, Shelly 19n32, 25–27, 35n92, 43n8, 49n30, 49n31, 50, 53n48, 89n19,

193n81, 260n3, 260n5, 261n8, 265n20, 275n50, 282n72, 289n91, 342n87 Kaldor-Hicks efficiency (See also Normative economics) 14–15, 17–18, 27, 257 Kalven, Harry 217 Kamm, Frances Myrna 41n1, 47n22, 50n36, 50n37, 54n51, 92n26, 96n38, 96n39, 96n40, 99n45, 170n160 Kant, Immanuel 48, 52, 62, 121, 137n44, 141, 263n13, 275–76, 298 Kaplow, Louis 16n16, 16n17, 26n57, 27, 30n76, 50–51, 78n65, 106n3, 132n28, 325n27 Kelman, Mark 229n9, 241n57, 245, 247–48, 253 Kraus, Jody S. 2n8, 3n10, 266n22, 266n23, 308n144 Kronman, Anthony T. 269n32, 271, 272n39, 286n82, 295n105, 298n117, 315n4, 316n11, 337n75, 337n76, 338n77 Law vs. morality 75, 77–78, 80, 98, 102, 103, 117–22, 141–42, 161–62, 278–79, 300–04 Legislation. See Public morality Legal paternalism. See Paternalism Lexical priority (See also Deontology) 47, 82–83, 86–87, 148, 222–23, 287, 350 Lies. See Deception Locke, John 138 Low-probability harms. See Threshold functions Lying. See Deception Marginal net benefit (See also Threshold functions) 149–50, 155–56, 170, 223, 346–47 Marginal utility 18, 101 McNaughton, David 31 Means/side-effect. See Intending/foreseeing Mental state. See Theory of the good Mill, John Stuart 190n63, 203, 277n59, 315n4, 316n9, 319n19, 328n47, 336–38 Mistake and misrepresentation. See Contract law; Deception

index Moderate deontology. See Threshold deontology Monetization 15, 17, 28, 82, 93–94, 108–16, 335 Multiplier threshold function. See Threshold functions “Negative utilitarianism” 47, 77, 91–93, 100, 287–88, 345–46 Net benefit. See Threshold deontology; Threshold functions Normative Economics (See also Costbenefit analysis) 11–19, 27, 31–32, 37, 39–40, 47, 79–108, 111–16, 122–23, 129–34, 140–76, 184–89, 195–224, 234–40, 251–54, 258–59, 262–67, 269–74, 277–91, 294–97, 305–10, 318–35, 339–47 Kaldor-Hicks efficiency 14–15, 17–18, 27, 257 liberty and 130–34, 187–89, 318 Pareto efficiency 13, 42, 50–51, 106, 133, 257, 262 normative neutrality 105–08, 132, 150 theory of the good 15–18, 27, 30–32, 36–37, 59n7, 105–06, 131, 189n61, 194, 201, 215, 238–40, 263, 321–25 Nozick, Robert 32n83, 116n32, 153n101 Objective-list theories of the good. See Theory of the good Offensiveness (See also Freedom of speech) 183–86, 202, 206–10 Omission bias 76 Omission/commission. See Doing/ allowing Options (See also Consequentialism; Constraints; Deontology; Threshold deontology) 20, 33–40, 41, 43, 46–47, 51, 53, 59–63, 83–84, 98–103, 289–91 conferring a benefit smaller than the agent’s sacrifice 98–99 constraints and 47, 60–63, 98, 255 disclosure duties and 289–91 maximal sacrifice 102–03, 255

“preferences for options” 36 proximity 20, 33–35, 99–100 thresholds 53, 98–103 Pareto 13, 42, 50–51, 106, 133, 257, 262 Paternalism (See also Autonomy; Harm principle) 313–47 adaptive preferences 322–23 antipaternalism 317, 319–22, 327–32, 347 abuse of power 329–31 learning from mistakes 328–29, 344 choice among permissible courses of action 346–47 classifications 315–18 consent 337–38 consequentialism and 313–14, 318–27 constrained cost-benefit analysis 335–47 types of costs and benefits 342–46 threshold size 339–42 criminal law 316–17, 341, 346 default rules 316–17, 327, 341, 345–46 definition 313 deontology 313–14, 318, 335–47 economic analysis 313–14, 318, 323–35, 344–45 efficiency and 313–14, 318–27, 332–35 frustration costs 333–35, 344, 346 incompetence 315, 337–39, 343 indirect paternalism 316–17 injury from the frustrated choice 316, 344 probability of 316, 344 type and magnitude of 316 information problems and 315, 337, 346 intrusiveness of intervention 316, 341 legal 315–17, 322–23, 327, 329–35, 338 “libertarian paternalism” 316, 327, 345–46 rationality 314–15, 319, 323–28, 332, 334, 340–41 redistribution and 316, 331, 332 second-order preferences 321–22, 338 “sin taxes” 316–17, 346 speech and 193 theory of the good and 319–25, 334 Persuasion principle. See Freedom of speech

359

360

index Pigou, Arthur 12 Policy-making. See Public morality Positive economic analysis 11–12, 128–29, 231–34 Posner, Eric A. 2–3, 16n17, 26n55, 28n66, 28n67, 62n16, 107n8, 120n51, 122n57, 129, 130n13, 131–34, 150, 162, 274n48, 304n137 Posner, Richard A. 1n2, 3n9, 16n18, 23n42, 109, 118n38, 119n45, 127–33, 157n120, 162, 184–89, 191, 196–97, 207n127, 208n130, 213, 214n149, 214n151, 215n153, 217n159, 218n164, 235–36, 251, 253–54, 286, 290, 304n137, 318n16, 327n44 Precontractual deception. See Contract law; Deception Preemptive measures. See Terrorism Preferences. See Consequentialism; Paternalism; Theory of the good Preferences vs. judgments 27–29, 30, 36, 109–10, 331–32 Prerogatives. See Options Private morality. See Public morality Private/public distinction. See Public morality Probabilistic costs and benefits (See also Threshold functions) 89–91, 95, 100, 147–48, 151–57, 195–96, 199–201, 254, 287, 316, 344–45 Promises (See also Contract law) 20, 29, 41–42, 46, 47, 52, 73, 94–97, 102, 114–15, 117, 260–62, 264, 266–67, 276, 298–301, 305–06, 308n144, 310n151 practice/convention theories 299–300 Public morality 57–78, 117–22 collective decision-making 57–59, 63–78 doing/allowing 60–63 intending/foreseeing 57–58, 63–70 legislation 53, 57–58, 63–75, 117–22, 214 private/public distinction 58–59, 180 Punishment 137–38, 141–42, 145 capital 76–77, 143

Rachlinski, Jeffery 327–28, 329n51, 329n52, 329n53 Rational choice theory (See also Cognitive psychology) 11–12, 15–16, 34, 39, 231–34, 258–59, 323–27, 332, 334, 340–41 Rationality (See also Bounded rationality; Cognitive psychology; Economic analysis) 11–12, 15–16, 31–32, 39, 231–34, 258–59, 323–25, 332, 340 cognitive 11–12, 15–16, 31–32, 258–59, 323–25, 332, 334, 340–41 motivational 12, 15–16, 31–32, 39, 231–34, 258–59, 324–25, 332, 340–41 Rawling, Piers 31, 50n33 Rawls, John 14n8, 41n1, 42n6, 48n27, 78n63, 105, 299n119, 299n120 Raz, Joseph 31n78, 110n16, 112n22, 112n23, 179n4, 214n50, 242, 250, 300n124, 338n77, 340n85 Redistribution (See also Distributive justice) 246–51, 255, 296, 316, 331–32 antidiscrimination laws 246–51, 255 contract law 296 paternalism and 316, 331–32 Reliance interest (See also Contract law) 259–63, 279, 285, 292–93, 295, 299–300, 306–08, 310 Restitution interest (See also Contract law) 262, 293, 299, 301, 303–05, 308 Retribution (See also Terrorism) 137–38, 141–42, 145 Risk allocation 258, 269, 270n33, 303–04, 310 Risk aversion 91n22, 265n21, 295, 329 Rule-consequentialism (See also Consequentialism) 24–27, 37–38, 48, 49, 62–63, 131–32, 151–52, 163, 188–89, 193, 196–98, 200, 207–08, 237, 329–30 Security. See Terrorism Self defense (See also Culpability) 138–39, 153–56, 164 imminence of danger 153–56 forfeiture argument 136–39

index Sen, Amartya 12n2, 16n15, 17n25, 24n47, 24n49, 31, 36n97, 37n99, 51n40, 106n3, 318n16, 321n25, 324n33, 324n35 Shavell, Steven 16n16, 16n17, 26n57, 27, 29, 30n76, 50–51, 78n65, 95n35, 106n3,132n26, 132n28, 262n11, 263–67, 271n37, 294n104, 296n109, 297n115, 325n27 Siege (See also Terrorism) 144 Slippery slope 119–20, 163, 330 Social welfare function (See also Normative economics) 13–14, 79, 103–04, 250 Speech. See Freedom of speech Sunstein, Cass R. 1n1, 57n2, 60–62, 63n18, 76–77, 107n8, 129, 130n13, 131–34, 143n66, 179n4, 191n70, 220n171, 252n103, 317n14, 322n27, 322n28, 322n29, 330n58, 330n59 Taurek, John M. 54–55, 97n43 Terrorism (See also Self defense; Torture) 127–76 choice among permissible courses of action 149–50, 155–56 combatants vs. civilians 157–60, 171–75 chronologically-remote costs and benefits 147–48, 153–56 culpability 136–38, 141–45, 151–60, 164–65, 167–68, 171–74 definition 128 deterrence 129, 137, 141–43, 145, 147, 150–51, 170 economic analysis 129–34 Geneva Convention relative to the Protection of Civilian Persons in Time of War 134, 144n68, 144n71, 157n122, 158, 168 Intending/foreseeing 67–68, 70, 135–36, 166–70, 174–75 low-probability harms 147–48, 151–57 nationality of victims 172–74 preemption 136–45, 149–61 pressure 143–44 probability of attack 147–48, 151–57 retribution 137–38, 141–42, 145 self defense 138–39, 153–56, 164

side-effects of anti-terrorist measures 135–36, 166–75 siege 144 shooting down an aircraft 118, 136–37, 139, 167, 170 targeted killing 135–61, 166–75 torture 141, 143, 152, 161–66 Theory of the good 12, 15–19, 21, 24, 27, 30–32, 36–37, 105–06, 131, 189n61, 194, 201, 215, 238–40, 263, 318–25 autonomy 131, 194, 201, 215, 275, 320, 338–39 actual preferences 15–16, 19, 21, 24, 27, 31–32, 275, 321–25, 334 “Consequentialist vacuum cleaner” 31, 36–37 cost-benefit analysis 15–18, 27, 30–32, 59n7, 105–06, 131, 189n61, 194, 201, 215, 238-40, 263, 321–25 ideal preferences 16, 19, 31–32, 110, 131, 189n61, 194, 201, 215, 238–40, 320–21, 334 mental state 19, 275, 319 objective list 19, 320 preferences 12, 15–16, 19, 21, 27–29, 31–32, 110, 131, 189n61, 194, 201, 215, 237–40, 275, 320–25, 334 Threshold deontology (See also Constraints; Deontology; Incommensurability; Options; Threshold functions) 2, 25, 46–56, 59, 79–104, 116–17, 140–75, 195–224, 251–55, 277–91, 305–10, 336–37, 339–47 choice among permissible courses of action 81–83, 87, 91, 96, 149–50, 155–56, 222–23, 346–47 critique 49–56, 116–18, 121, 130–34, 150 arbitrariness 53, 132, 150 disrespectfulness 52–53, 117–18, 121 incoherence 52, 132, 150 setting constraints too low 116–17 puzzles 55–56 Threshold functions (See also Threshold deontology) 79–123, 143–60, 166–75, 195–221, 251–55, 339–47 additive 84, 93–96, 101, 146, 169, 213, 253, 339–42

361

362

index Threshold functions (cont.) choice among permissible courses of action 81–83, 87, 91, 97, 149–50, 155–56, 222–23, 346–47 chronologically-remote costs and benefits 89, 147–48, 153–56, 182–83, 185–86, 188, 194, 196–200, 203, 210–12, 222, 254, 344–45 combining excluders 146, 211–12 combined thresholds 84, 93–96, 101–02, 146, 169, 174–75, 211–12, 342 commodification 111 critique 105–23, 132, 150 exclusion of costs and benefits 47, 59, 85–93, 98, 140–45, 147–49, 195–211, 237–40, 254-55, 342–46 combining excluders 147–49, 210–11 chronologically-remote costs and benefits 89, 147–48, 153–56, 182–83, 185–86, 188, 194, 196–200, 203, 210–12, 222, 254, 344–45 existential constraint 88, 344 lexical priority 47, 82–83, 86–87, 148, 222–23, 287 low-probability costs and benefits 89–91, 95, 100, 147–48, 151–57, 195–96, 199–201, 254, 287, 344–45 promoting the good vs. eliminating the bad 47, 77, 91–93, 100, 287–88, 345–46 small costs and benefits 87–88, 96, 201–02, 208, 210, 343–44 marginal net benefit 83, 149–50, 155–56, 170, 223, 346–47 multiplier 84–85, 93–96, 101–03, 146, 155, 160–61, 211–12, 253, 286–87, 342 net benefit 86–93, 109, 147–50, 147–50, 195–211, 254–55, 342–46 object of analysis 80 options 59, 83–84, 98–103, 289–91 probabilistic costs and benefits 89–91, 100, 147–48, 151–57, 195–96, 199– 201, 254, 287, 344–45

promoting the good vs. eliminating the bad 47, 77, 91–93, 100, 287–88, 345–46 quantification and monetization of constraints 28, 93–94, 108–16 shape of the threshold 84–86, 93–96, 98, 101–02, 145–46, 155, 169–71, 174–75, 211–13, 253, 286–87, 342 size of threshold 59, 93–96, 116–17, 150–61, 165, 169–75, 213–22, 251–55, 286–87, 339–42 small costs and benefits 87–88, 96, 201–02, 208, 210, 343–44 structure 84–86, 93–96, 98, 101–02, 145–46, 155, 169–71, 174–75, 211–13, 253, 286–87, 342 terrorism 140–75 Threshold options. See Options; Threshold deontology; Threshold functions Torture (See also Terrorism) 49, 52, 55–56, 62, 74, 83, 86–87, 88, 90–91, 92–93, 117–18, 119n46, 120, 141, 143, 152, 161–66 act vs. practice 166 definition 161–62 killing, compared to 162–64 legitimate purposes 141–45, 164 Tradeoff 115, 116, 130–31, 134 Tribe, Laurence H. 66n26, 66n28, 66n32, 177n2, 181, 183n26, 207n126, 211n140, 212n142, 214n151, 216n156, 217n158, 217n160, 219n166 Trolley problem (See also Intending/ foreseeing) 44–46, 282n72 United Nations Covenant on Civil and Political Rights 182 Universalizability (See also Kant, Immanuel) 48, 62 Utilitarianism 12, 14–15, 18–19, 106n2 Vermeule, Adrian 57n2, 60–62, 63n18, 76–77, 129, 131–34, 143n66, 150, 162

index War on terror. See Terrorism Welfare. See Social welfare function; Theory of the good Welfare economics. See Normative Economics Welfarism (See also Theory of the good) 12, 16–17, 37 Well-being. See Theory of the good

Williams, Bernard 23 Willingness to accept (WTA) 17, 93, 108–10, 113 Willingness to pay (WTP) 15, 17, 21, 93, 108–10, 113, 238–39, 275, 318 Zerbe, Richard 27–29

363

Law, Economics, and Morality

Law, Liberty, and Morality

Morality, Politics, and Law

Tort Law and Social Morality

Responsibility in Law and Morality

Tort Law and Economics (Encyclopedia of Law and Economics)

Morality and Economics: De Moribus est Disputandum

Law and Economics in Civil Law Countries

Natural Law, Liberalism, and Morality: Contemporary Essays

Monopsony in Law and Economics

Law, Morality, and the Relations of States

Principles methods law and economics

Law and Economics, 6th Edition

Law and Economics (3rd Edition)

Economics, Law and Individual Rights

Law, Politics, and Morality in Judaism

Conflicts of Law and Morality (Clarendon Law Series)

Law economics marriage divorce

Civil Law and Economics (Encyclopedia of Law and Economics , Vol 2)

Antitrust Law and Economics, 2nd edition (Encyclopedia of Law and Economics)

Antitrust Law and Economics, Volume 21

The Law and Economics of Cybersecurity

Odious Debt: Law and Economics Perspectives

Economic Efficiency in Law and Economics

Islamic Finance: Law, Economics, and Practice

Economics of Public and Tax Law (Encyclopedia of Law and Economics , Vol 4)

Research in Law and Economics, Volume 19 (Research in Law and Economics)

Morality

Law, Economics, and Morality

Law, Liberty, and Morality

Morality, Politics, and Law

Tort Law and Social Morality

Responsibility in Law and Morality

Tort Law and Economics (Encyclopedia of Law and Economics)

Morality and Economics: De Moribus est Disputandum

Law and Economics in Civil Law Countries

Natural Law, Liberalism, and Morality: Contemporary Essays

Monopsony in Law and Economics

Law, Morality, and the Relations of States

Principles methods law and economics

Law and Economics, 6th Edition

Law and Economics (3rd Edition)

Economics, Law and Individual Rights

Law, Politics, and Morality in Judaism

Conflicts of Law and Morality (Clarendon Law Series)

Law economics marriage divorce

Civil Law and Economics (Encyclopedia of Law and Economics , Vol 2)

Antitrust Law and Economics, 2nd edition (Encyclopedia of Law and Economics)

Antitrust Law and Economics, Volume 21

The Law and Economics of Cybersecurity

Odious Debt: Law and Economics Perspectives

Economic Efficiency in Law and Economics

Islamic Finance: Law, Economics, and Practice

Economics of Public and Tax Law (Encyclopedia of Law and Economics , Vol 4)

Research in Law and Economics, Volume 19 (Research in Law and Economics)

Morality

Recommend Documents