Micro-Simulation in Action, Volume 25: Policy Analysis in Europe using EUROMOD

LIST OF CONTRIBUTORS Olivier Bargain IZA (Institute for the Study of Labor), Bonn, Germany Tim Callan Economic and So...

Author: Olivier Bargain

27 downloads 552 Views 2MB 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

LIST OF CONTRIBUTORS Olivier Bargain

IZA (Institute for the Study of Labor), Bonn, Germany

Tim Callan

Economic and Social Research Institute, Dublin, Ireland

Manuela Coromaldi

Department of Economics, University of Rome ‘‘Tor Vergata’’, Italy

Carlos Farinha Rodrigues

Department of Economics, Technical University of Lisbon, Portugal

Herwig Immervoll

OECD Directorate of Employment, Labour and Social Affairs; Institute for the Study of Labor (IZA), Bonn; European Centre for Social Welfare Policy and Research, Vienna; Institute for Social and Economic Research (ISER), University of Essex, UK

Horacio Levy

Department of Applied Economics, Autonomous University of Barcelona, Spain; Institute for Social and Economic Research (ISER) University of Essex, UK

Christine Lietz

Institute for Advanced Studies, Vienna, Austria

Daniela Mantovani

Department of Economics,University of Modena e Reggio Emilia, Italy

Manos Matsaganis

Department of International and European Economic Studies, Athens University of Economics & Business, Athens, Greece vii

viii

LIST OF CONTRIBUTORS

Magda Mercader-Prats

Department of Applied Economics, Autonomous University of Barcelona, Spain

Cathal O’Donoghue

Department of Economics, National University of Ireland, Galway, Ireland; IZA, Bonn, Germany

Kristian Orsini

Department of Economics, University of Leuven, Belgium

Fotis Papadopoulos

Department of International and European Economic Studies, Athens University of Economics & Business, Athens, Greece

Amedeo Spadaro

Department of Economics, University of the Illes Balears, Palma de Mallorca, Spain

Holly Sutherland

Institute for Social and Economic Research (ISER),University of Essex, UK

Stefano Toso

Department of Economics, University of Bologna, Italy

Panos Tsakloglou

Department of International and European Economic Studies, Athens University of Economics & Business, Athens, Greece

Gerlinde Verbist

Centre for Social Policy, University of Antwerp, Belgium

John Walsh

Centre for Social Policy, University of Antwerp, Antwerp, Belgium

PREFACE This volume initiates a new collaboration between Research in Labor Economics (RLE) published by Elsevier Press and the Institute for the Study of Labor (IZA). Beginning 2006, the RLE series extends to two volumes per year. One volume will remain in the tradition of the series, consisting of empirical and theoretical contributions in labor economics, while the other volume will focus on specific policy questions. IZA has become one of the largest organizations of labor scholars worldwide while RLE is now a wellestablished publication containing labor economics research. We hope this new association will be a meaningful development for both IZA fellows and RLE readers. The first volume of this new editorial partnership focuses on a policyoriented research avenue concerning tax-benefit systems in Europe. The volume examines the tax-benefit micro-simulation model EUROMOD and its multiple uses. Micro-simulation models are computer-based tools used to simulate most of the direct tax and benefit instruments which affect household budget constraints. They are designed to substantially increase the level of realism in microeconomic models. Moreover, they are linked to household micro-datasets, representative of a national population, so that taxbenefit calculations can be repeated for all the sample households and aggregated. With these computations, it becomes possible to evaluate how existing tax and benefit policies alter inequality, poverty and the overall income distribution. In addition, the simulation of budget constraints and financial incentives considerably improves the econometric estimations of labor supply behavior, fertility choices and marital decisions, thereby allowing better prediction to policy changes. Although several micro-simulation calculators have been used for many years at a national level (e.g. TAXBEN and POLIMOD in the UK, SYSIFF, INES, MYRIAD and others in France, etc.), EUROMOD goes one step further by simulating tax-benefit systems of all EU-15 countries via an integrated and truly comparative methodology. This calculator has been designed to conduct international comparisons based on homogenized datasets and harmonized definitions of tax-benefit instruments, making it ix

x

PREFACE

possible to address the relevance of tax-benefit policies in a given institutional environment. In particular, EUROMOD allows one to assess the effect of a common reform in particular institutional settings as well as the transmission of a tax-benefit system from one country to another. This was the early motivation of the instrumental paper of Tony Atkinson, Franc- ois Bourguignon and Pierre-Andre´ Chiappori: ‘‘What Do We Learn About Tax Reform from International Comparisons? France and Britain’’ (European Economic Review, 1988, pp. 343–351). Following the line of research, the present volume provides interesting examples of counterfactual or hypothetical policy changes aimed to assist in designing actual policy and to answer such questions as: Could the British tax credit be fruitfully exported to continental Europe in order to encourage female employment? Are child benefits a necessary move to tackle child poverty in southern Europe? And, how do results differ if the same standard pension reform is implemented in different countries? While the first prototypes of comparative micro-simulation models were initiated in the early 1990s on a subset of countries (France, Italy and UK), the comprehensive EU-15 model has only been operational for a few years. Its construction took 10 years and two large EU-funded projects (the first one christened EUROMOD and the second known as MICRESA). This longterm effort has been necessary to overcome numerous technical difficulties and to produce a truly flexible and integrated structure. Along the way, this project has been an interesting example of a complex coordination process for 15 European countries and more than 18 different institutions. The model was officially released in 2004 and the number of papers making use of EUROMOD has flourished since then.1 Various contributions have explored several applications regarding tax policy analysis. They improved our understanding of fiscal and social policies, labor markets and labor market policies, wage setting mechanisms, income distribution and many other issues. This volume attempts to give an overview of these contributions and provide a selection of comparative policy analyses using EUROMOD. In the first chapter A Short Introduction to EUROMOD: An Integrated European Tax-Benefit Model, Christine Lietz and Daniela Mantovani establish a detailed description of the history of the EUROMOD project, its goals, achievements, limitations and future developments. This chapter also describes how EUROMOD may address various economic questions and policy issues. The authors illustrate how economists in labor, public and family economics can adopt the model and use it for their own research.2 With the anticipated decline in the ratio of workers to pensioners, policy makers have questioned the sustainability of Pay-As-You-Go pension

Preface

xi

systems in Europe. Whereas some promote private insurance schemes, other simply suggest modifying the existing arrangements, for example (i) to delay retirement, (ii) to lower replacement rates, (iii) to increase social insurance contributions rates and/or (iv) to introduce or extend minimum (solidarity) pensions. Relatively little is known about the likely distributional impact of these recommendations in the short-run, although it would provide useful insight into the political feasibility of proposed policy changes. In Pension Incomes in the European Union: Policy Reform Strategies in Comparative Perspective, Daniela Mantovani, Fotis Papadopoulos, Holly Sutherland and Panos Tsakloglou attempt to fill the gap by presenting original simulations of reforms of the type (ii), (iii) and (iv) along with various combinations. In each case, they identify potential winners/losers and the degree of redistribution across generations; they also evaluate the cost of the reform and its impact on relative poverty. They focus on four EU countries selected on the basis of the diversity of their pension systems (Denmark, Germany, Italy and the UK). The consequences of existing differences in institutional arrangements in particular societies are better understood thanks to the simulation of common reforms in the four countries. In particular, the introduction of a generous minimum pension considerably reduces old-age poverty in all countries but the appropriate means to finance it diverge. For instance, the large dispersion of income among German pensioners makes it possible to reach revenue neutrality in part by income redistribution within this population; in Denmark, the current flat pension system, and the concentration of pensioners’ incomes, leaves little room for intra-pensioner redistribution and requires increasing the tax burden of the non-elderly to finance the reform. This exercise helps define nationally appropriate directions for reform. While the previous chapter examines the effect of contributory pensions and non-contributory benefits (e.g. minimum pensions) on the relative position of pensioners, Gerlinde Verbist focuses on the tax side of the system in The Distribution Effects of Taxes on Pensions and Unemployment Benefits in the EU-15. Specifically, she investigates how the situation of pensioners and unemployed workers is affected by the specific tax treatment of replacement incomes. She observes that despite important heterogeneity across national systems, replacement incomes are less taxed than earnings in all EU-15 countries due to progressivity of tax schedules (and the fact that replacement incomes are lower on average), special provisions in personal income tax systems (e.g. tax allowances and tax credits) and lower levels of social insurance contributions. Consequently, the relative income position of both the elderly and the unemployed is better after payment of taxes and

xii

PREFACE

contributions, especially at the bottom of the distribution. Verbist discusses the implications in terms of redistribution within and between generations. In Family Transfers and Child Poverty in Greece, Italy, Spain and Portugal, Manos Matsaganis, Cathal O’Donoghue, Horacio Levy, Manuela Coromaldi, Magda Mercader-Prats, Carlos Farinha Rodrigues, Stefano Toso and Panos Tsakloglou examine the limited role of family policies in southern Europe and the potential need for more generous schemes to tackle child poverty. The chapter advances previous work on how the family rather than the state can support individuals in need. The authors also explore the scope for policy reform by simulating several types of universal child benefits, which entail a revenue-neutral reform and several schemes in force in the UK, Denmark and Sweden. On the one hand, they find that current expenditure on family transfers in southern Europe is too low to significantly reduce poverty through a simple reallocation of public spending alone. On the other, the Danish scheme would be the costliest, but it would also have the most substantial impact on child poverty in almost all examined countries. Nonetheless, results are mixed across countries; in the Italian case, the replacement of current relatively well-targeted policies by some of the suggested universal benefits could even increase child poverty. In Assessing the Impact of Tax-Transfer Policy Changes on Poverty: Methodological Issues and Some European Evidence, Tim Callan and John Walsh suggest a method to systematically assess the direct impact of tax and transfer policy changes on income distribution. The usual counterfactual – the situation where no change in tax, social security contributions and social benefits has taken place – is useful to keep track of the changes in government expenditures and tax revenues. Yet, it is not a satisfying benchmark against which to assess the distributional impact of policies actually implemented. Instead, the authors suggest a ‘distributionally neutral’ scenario approximated by indexation of tax and welfare parameters (i.e. tax allowances, tax credits, tax bands and welfare payment rates) in line with wage growth. This benchmark provides a more accurate picture of the impact of policy changes on inequality and relative income poverty. An initial application of the method suggests that over the 1998–2001 period, tax/transfer policy changes in Austria, Greece, Ireland, Portugal and the UK were particularly favorable to lower incomes. In Fiscal Drag: An Automatic Stabiliser?, Herwig Immervoll investigates the effect of taxes at the macro level and revisits the ‘automatic stabilizer’ argument when accounting for the true complexity of tax schemes and the heterogeneity of tax payers. Taxes are often thought of as automatic stabilizers. In this case, inflation may reduce the real value of nominally fixed

Preface

xiii

tax band limits, deductions and tax credits, thereby leading to higher real tax burdens, lower consumption and lower aggregate demand. On the other hand, higher tax burdens may lead to upwards pressure on wages and hence a cost-push upwards pressure on prices. Immervoll uses EUROMOD to derive distributions of inflation-induced changes of effective tax rates for representative samples of the population of five European countries. The effect of higher tax burdens on the cost of production is taken into account by simulating changes in the marginal and average tax burdens of employees and by using elasticities from the literature. The results suggest that inflation, combined with an un-indexed tax-benefit system, can produce a moderate upward pressure on wages. In Beans for Breakfast? How Exportable is the British Workfare Model?, Kristian Orsini and Olivier Bargain study the British working family tax credit (WFTC) and the potential effect of this policy when exported into other institutional settings. In the UK, the tax credit was introduced to redistribute income to working poor families, while at the same time to encourage labor force participation of inactive households. In continental Europe, such policy is the subject of numerous debates and is envisaged as a possible instrument to combat the disincentive effect of generous social assistance schemes. In effect, such in-work transfers would recreate a significant financial gap between inactivity and low-paid work, thereby encouraging participation of households living on welfare. To bring further insight into this question, Orsini and Bargain simulate the WFTC for France, Germany and Finland. Typically in these three countries, the generous social assistance schemes are suspected of creating inactivity traps. The authors analyze the impact of this policy measure on poverty rates, on the distribution of effective marginal tax rates (an indication of the potential effect on working hours) and on financial gains to work (an indication of the potential effect on participation). The redistributive impact toward lowearning families may be substantial. On the other hand, the reform is likely to discourage the employment of secondary earners due to the fact that the WFTC is means-tested on joint income of spouses. In particular, this instrument does not seem appropriate in countries like France, already characterized by high effective marginal tax rates on secondary earners due to joint income taxation. Finally, in Effective Marginal Tax Rates and Revealed Social Preferences in Spain, Amedeo Spadaro suggests an original use of micro-simulation models for the normative analysis of existing tax-benefit systems. EUROMOD allows simulating the distribution of effective marginal tax rates (EMTRs) by numerical derivation of the budget constraint of each

xiv

PREFACE

household at the observed levels of earnings. In the fiction of an optimizing social planner, these EMTRs can be thought of as revealing the national level of aversion to inequality, given a certain distribution of labor supply elasticities and wage rates. Then, under simplifying hypotheses and some regularity conditions, the author inverts Mirrlees’ optimal taxation model to retrieve social preferences consistent with actual EMTRs. In other words, it is possible to retrieve the level of social inequality aversion for which optimal tax rates coincide with actual ones, as derived by micro-simulation. The approach is applied to Spanish data and shows how social aversion to inequality has evolved with the change in political majority following the 1999 election. For insightful editorial advice in preparing this volume, I thank Margherita Borella, Andrea Brandolini, David Huffman, Leonor Vasconcelos Ferreira, Rachel Lloyd, Helene Perivier, Solomon Polachek, Andrew Shepard, Holly Sutherland, Justin Van de Ven, Klaas de Vos and Sally Wallace. I also warmly thank the authors and the whole EUROMOD team for their support as well as Tony Atkinson for kindly accepting to write the foreword. Authors who have prepared manuscripts that meet the stringent standards of RLE are encouraged to submit them to Solomon Polachek and Olivier Bargain via the IZA website (http://www.iza.org/index_html? long=en&mainframe=http%3A//www.iza.org/en/webcontent/index_html) for possible inclusion in a future volume.

NOTES 1. For more information on the project and its outcomes, see http://www.iser. essex.ac.uk/msu/emod/. 2. Two new EU-funded projects have started in 2005 and 2006 to extend the scope and capacity of EUROMOD. As well as laying the technical basis for a 25-country comparative research infrastructure, the Improving the Capacity and Usability of EUROMOD (I-CUE) project will begin to involve researchers, data providers and institutions from the 10 new member states in the long-standing EUROMOD collaboration. Together with design up-grades (e.g. a tool to guide the user through the model) and greater clarity in presenting the knowledge base embodied in the model (e.g. a menu of classification systems for taxes and benefits) these developments will prepare for a planned new phase of extending access to EUROMOD to the EU25 social science research community at large. The second project, Accurate Income Measurement for the Assessment of Public Policies (AIM-AP), is a program of three interlinked activities aiming to improve the comparability, scope and applicability of tools, methods and data for the measurement of income and the analysis of the effects of policies on inequality, poverty and social inclusion. It includes the

Preface

xv

distributional effects of non-cash incomes and the implementation of a more comprehensive income definition; the implications of (and methods to account for) errors in targeting social benefits, tax evasion and measurement error in income data; the incorporation of the effects of indirect, as well as direct taxes and social benefits in redistribution analysis. Each project will develop methodologies within a cross-national perspective and the resulting method enhancements will be made generally accessible and re-useable by implementing them within EUROMOD.

Olivier Bargain Editor

FOREWORD This volume is remarkable as the confluence of three important streams of research. The first is that of micro-simulation. Fifty years ago, Guy Orcutt had a vision of what could be achieved by the application of simulation methods to the economic and social behavior of individuals. At that time, progress was held back by the lack of computing power and by the paucity of micro-data. Both have been transformed in the last quarter of a century, allowing this branch of research to flourish. The development of tax benefit models, for example, has greatly increased the capacity of economists to advise about the implications of proposed reforms. I can remember in the early 1980s advising a Parliamentary Committee, where, each time a new tax benefit reform was proposed, the secretary would spend the weekend calculating the impact on a handful of stylized cases. Now, the ability to do just the arithmetic on a random sample of the population gives a new dimension to the policy debate, to say nothing of the modeling of incentive effects and behavioral responses. The second stream is that of comparative policy analysis. Economists have long sought to relate economic and social performance to differences in national policies, but the development of comparable models has allowed, at both micro- and macro-levels, greater depth to be achieved. This has involved in particular the modeling of institutions. It is not sufficient to refer to summary statistics such as replacement rates. As the work of the OECD and others has shown, one has to treat the institutional details, such as the entitlement conditions for unemployment insurance, that govern the impact on individual households of government programs. Modeling can never capture the entire flavor of national systems, but the papers in this volume show how one can learn a great deal from contrasting individual elements of the policy process. This has especial significance in the context of the European Union (EU), where policy-making has increasingly emphasized peer assessment and learning from other member states. This brings me to the third stream: pan-European research. Microsimulation analyses of policy are now widespread at national level, and much has been published on taxes and benefits by the OECD for its xvii

xviii

FOREWORD

members, but EUROMOD represents the first attempt to construct an integrated tax-benefit model of the EU. Christine Lietz and Daniela Mantovani describe very well the 10-year history of this project, in which some 18 institutions co-operated, under the leadership of Holly Sutherland. For those of us who took part, it has been a remarkable experience. Like many EU institutions, progress has been uneven, and at times frustratingly slow. Incorporation of the specificities of individual countries has been a challenge, and the same applies to the extension from EU-15 to EU-25. But the chapters in this volume demonstrate the great value of an integrated EUwide model. And there are many more potential uses. Suppose, for instance, that the EU President were to follow the example of the British Prime Minister and ask what it would take to abolish child poverty in Europe? EUROMOD provides an excellent starting point to begin to answer that question. Finally, and not least, EUROMOD has helped train a generation of researchers who will in the future contribute greatly to quantitative public policy research in Europe. A. B. Atkinson

A SHORT INTRODUCTION TO EUROMOD: AN INTEGRATED EUROPEAN TAX-BENEFIT MODEL Christine Lietz and Daniela Mantovani ABSTRACT By the mid-1990s the potential and usefulness of microsimulation models for researching tax-benefit systems had found widespread acceptance. Nevertheless, models were not widely available for independent or academic research in all countries of the European Union (EU). Even more important, carrying out consistent comparative tax-benefit microsimulation analysis was still an apparently impossible task. The time seemed ready for a European-Union-wide tax-benefit microsimulation model. Such a model, EUROMOD, is now available. This chapter is devoted to a short introduction to EUROMOD, including the reasons why it was built, its added value compared to existing models, the trade-offs faced by its builders and lessons that have been learnt from developing such an integrated model. Moreover, it aims to provide an insight into the wide range of possible applications of EUROMOD, underlined by summarizing some indicative findings of studies, which have used the model.

Micro-Simulation in Action: Policy Analysis in Europe using EUROMOD Research in Labor Economics, Volume 25, 1–26 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(06)25001-5

1

2

CHRISTINE LIETZ AND DANIELA MANTOVANI

INTRODUCTION EUROMOD is a tax-benefit microsimulation model covering all 15 preMay 2004 member states of the European Union (EU-15). It is a tool, which enables research on the effects of tax-benefit systems by allowing the assessment of their impact on incomes, poverty, inequality and social inclusion. In particular the model is designed to answer ‘‘What if’’ questions about different approaches to policy reform at European level. In general, tax-benefit models are computer programs based on household micro-data from representative sources. They calculate disposable income for each household in the dataset. This calculation is made up of elements of income taken from the survey data (e.g. employee earnings) combined with components that are simulated by the model (taxes and benefits). An evident advantage of such models is the possibility of evaluating the effects of hypothetical changes to tax- or benefit-rules. In a certain sense tax-benefit microsimulation models are tools, which allow ‘‘laboratory experiments’’ concerning tax-benefit systems, as by simulating reforms their potential effects can be studied before their actual implementation. While tax-benefit microsimulation techniques have been used extensively at the national level, EUROMOD is unique in being a research tool that is relevant both at the national level and as an integrated tool for European comparative social science research. EUROMOD does represent the first attempt at building this type of highly complex and ambitious multi-country infrastructure. The process of its design, construction and use has been based on ‘‘learning by doing’’, in many respects without prior or parallel experience to build or draw on. Owing to this approach unforeseen challenges have been encountered, and on the other hand unanticipated research applications for the model have been identified. The human effort and financial resources that have been invested in EUROMOD are considerable. The version of the model currently used is the result of almost 10 years of work. Some 50 individuals in 20 research institutions across the 15 member states are currently involved. Some 20 others have been in the past and most of these maintain contact. The chapter is structured as follows. The first section presents the background reasons for the construction of an integrated European tax-benefit microsimulation model and its added value compared to existing experiences. The second section reports on the experiences in designing and implementing EUROMOD and the trade-offs, challenges and limitations faced by its builders. The third section provides an insight into the wide range of

A Short Introduction to EUROMOD

3

possible applications of EUROMOD, underlined by some indicative findings of studies applying the model. The last section presents ongoing and future perspectives in using and extending the model.

1. BIRTH OF A EUROPEAN TAX-BENEFIT MICROSIMULATION MODEL The main impulse for building an integrated European tax-benefit microsimulation model came from research questions in public economics and more precisely those investigating the characteristics of tax-benefit systems and the comparative impact of common reforms across Europe. Exercises of ‘‘policy swaps’’ or ‘‘system swaps’’ have for instance consisted of analyzing the redistributive impact of replacing the French tax-benefit system by the UK one on a representative sample of French households (Atkinson, Bourguignon, & Chiappori, 1988).1Callan and Sutherland (1997) have also compared the effects of different types of fiscal and social policies on the welfare of households in certain EEC countries. These crucial policy questions lead to the need for a more consistent tool in order to perform cross-country comparisons and making use of tax-benefit microsimulation techniques applied at European level appeared to be a promising approach. Tax-benefit microsimulation models developed in parallel with the availability of computers. By the mid-1990s tax-benefit microsimulation models had spread out to many Western and some Eastern European countries.2 They ranged from simple models used for teaching purposes to multi-purpose models designed to support governments’ decision making. In the early 1990s, Merz (1991) surveyed more than 40 major national models across Europe (mainly Germany) and the US, some built in the 1960s and 1970s. A few years later, Sutherland (1995) described 19 static models already in use in five countries of the EU. Examples of such early developments of taxbenefit microsimulation models at the national level in Europe are TAXMOD model in the UK, developed around Atkinson, King and Sutherland’s work (Atkinson, King, & Sutherland, 1983), and the SYSIFF model in France (Bourguignon, Chiappori, & Sastre, 1988). Despite the growing availability of these tools and the widespread acceptance of their potential and usefulness the level of development and accessibility to researchers differed greatly among countries. Moreover, cross-country comparison remained a difficult task. The main obstacles in comparative analysis concerned data quality and consistency of data and definitions across countries.

4


The only attempt to compare tax-benefit systems was for many years a periodical research on the impact of tax and benefit systems on household income of workers earning an average salary (average production worker), for different family types, carried out by the OECD.3 Though this research contained a great deal of cross-country information on tax-benefit systems, this method did not allow the investigation of inequality, poverty and other indicators that need to be computed on the whole population (or a representative sample).4 A study presenting consistent evidence for income distribution in a wide range of OECD countries, commissioned by the OECD in the early 1990s and released in 1995, represented a significant breakthrough (Atkinson, Rainwater, & Smeeding, 1995). It was based on a large and consistent multi-country database made available by Luxembourg Income Study (LIS).5 This database is derived from national micro-data provided by single institutions in the respective countries. Though national micro-data are not uniform in terms of source (some from administrative sources, some from sample surveys), quality, objectives and definitions, the common LIS requirement for highly detailed data allows for a reclassification of income components, in order to obtain consistent income definitions across countries. Though there is still room for improvements – in fact, full crosscountry consistency is not an achievable goal when starting from different sources and in this sense comparability is still a matter of degree – with the LIS database a reasonable level of comparability across OECD countries was reached for the first time. In this period the European Community Household Panel (ECHP) was also established. This survey covered a wide range of topics including income, housing, demographics and employment characteristics, and was conducted on yearly basis from 1994 to 2001. The ECHP covered a more limited number of countries than LIS, basically all (then) EU countries (at different levels of participation duration), but it was designed with comparability in mind. It was co-ordinated by Eurostat and was based on a standardized design and common technical and implementation procedures.6 After LIS and ECHP proved that the derivation of a consistent multicountry database suitable for cross-country comparisons is possible, the use of tax-benefit microsimulation techniques seemed a natural step ahead. The microsimulation approach has two main advantages compared to solely applying a multi-country database. The most obvious is that it is not restricted to analyses of existing tax-benefit systems, but allows the study of a wide range of reforms and hypothetical tax-benefit approaches. In addition,


5

microsimulation techniques may enhance comparability and the measurement of redistribution performed by tax-benefit systems. In fact, measuring redistribution effects of tax-benefit systems with ECHP and with some LIS data is not possible, given that all ECHP data and a number of LIS national databases do not provide information about taxes, a number benefits and gross incomes.7 Even if available, micro-data on taxes and benefits are affected by response errors to an extent that depends on data quality and that is bound to vary across national databases. Thus it might be argued that the consistent application of policy rules ensured by a multi-country model makes tax and benefits comparison less vulnerable to the different quality of data across countries. Different approaches to comparative tax-benefit microsimulation have been discussed in the literature, two of them are the comparison of published results from national simulations and the use of national models in parallel. However, published results reflect a vast range of different national priorities, classification systems and practices, and comparisons of originally independent studies at least raise some doubts on their consistency. Some experiments on using national models in parallel have been conducted. In general they suggest that, if the number of countries that are to be considered together becomes large, differences in modeling and in assumptions are too difficult to accommodate. If the range of possible approaches is also large, then difficulties in making the models comparable are clearly prohibitive. Callan and Sutherland (1997) used national tax-benefit microsimulation models for Ireland (SWITCH) and the UK (POLIMOD) to study the differential impact of comparable Basic Income Schemes. They found that data and model constraints limited the schemes that could be modeled and the ways in which the results could be presented considerably. They also found (Callan & Sutherland, 1996) that designing a realistic basic income scheme that could be modeled and simulated consistently with only five national models (Belgium, France, Ireland, Italy and the UK) was in practice impossible. Furthermore, the level of development of national models that could potentially be used in the mid-1990s differed greatly: in some countries no model had been developed yet, in others all this activity took place within government, in others developments were in their infancy. As this seemed to suggest that the only way of carrying out consistent comparative tax-benefit microsimulation analysis was an integrated EU-wide model – with the side effect and strong motivation for countries without an own model that it could be used for national policy analyses as well – the time was right for the challenge of building EUROMOD.

6


EUROMOD is the result of almost 10 years of work, starting with a preparatory study funded by the European Commission’s Targeted SocioEconomic Research (TSER) program and carried out between March 1996 and May 1997. This study was aimed to assess the technical feasibility of constructing an integrated European tax-benefit microsimulation model.8 It included an exercise to construct a prototype model for three countries: France, UK and Italy, based on the previous experience of EUROMOD reseachers gained on national models. This first model was used to simulate the impact of common reforms on the populations of those three countries (Bourguignon, O’Donoghue, Sastre-Descals, Spadaro, & Utili, 1997). As all tax-benefit microsimulation models, EUROMOD is the object of the endless work of maintenance, revision and improvement, so that defining a date when it has been completed is to some extent arbitrary. The first fully working version of the model was available in 2001 and it has been updated and improved considerably since then. One might even argue that the model is not yet finalized, and never will be, as there is always room for improvement. The first full EUROMOD model was constructed within a project also funded by the TSER program called EUROMOD: An integrated European benefit-tax model. It was carried out in three years, from January 1998 to December 2000 (see Sutherland, 2001a). Between 2001 and 2004 a new project offered the chance to improve the model and use the full potential of EUROMOD: the Micro-Level Analysis of the European Social Agenda (MICRESA) project, funded by the European Commission’s improving human potential program. This project studied the impacts of social and fiscal policies, and reforms to these policies, on poverty reduction by applying EUROMOD. In fact, most of EUROMOD’s outcomes, including several of the contributions to this book, have been made possible by this pioneering research project.9

2. BUILDING EUROMOD: CHALLENGES AND LIMITATIONS Building EUROMOD involved a great deal of expertise and – unlike national models – a great number of experts spread across countries, so that the process required a clear planning of the procedures and steps necessary to complete the work and a clear division of labor among participants. The process of construction was organized in four main, in general successive, but frequently overlapping tasks: (1) finding expert respondents and


7

dealing with a large-scale European coordination; (2) finding representative and comparable micro-databases for each country; (3) designing the model framework and implementing it (collection, coding and parameterization of policy rules for 15 tax-benefit systems) and (4) testing, validating and documenting the model. 2.1. A European Cooperation A first step consisted in finding expert respondents in each EU country, this meant involving some 40 researchers with 12 different mother languages from 18 institutions in 15 countries. EUROMOD’s construction was then carried out following a flexible division of labor among national teams. Various issues and unexpected problems often had to be dealt with on a case-by-case basis. Given the number of participants and the different locations of teams, the project involved a great deal of coordination, achieved by Holly Sutherland and her team at the University of Cambridge (now at the University of Essex). They were also responsible for programming the core part of the model and designing and documenting common features. Each national team was responsible for providing data and information on tax and benefit rules necessary to model their national system. In collaboration with the model-building team they also carried out the testing and validation of the model for their country. What is unusual, compared to other research projects, is the truly European diversity of the group and the longevity of the project. To be able to work together EUROMOD members had to go beyond national traditions and priorities to which they were accustomed and they had to establish new common practices. One important, though unintended, achievement of the EUROMOD project was the creation of an international community able to communicate in a homogeneous (scientific) language with relative success. 2.2. Data Comparability The EUROMOD preparatory study (Sutherland, 1997) surveyed existing micro-data sources in terms of their general quality and suitability for being the basis for EUROMOD. The characteristics of data provided at the European level by Eurostat (ECHP) as well as national sources were discussed, with the conclusion that national teams selected which micro-data, among those available for their country were the most suitable for tax-benefit modeling. Therefore EUROMOD national datasets differ in origin and type

8


(see Table 1) according to the different data availability and needs in each country.10 The ECHP was chosen for reasons of comparability and homogeneity in several countries,11 but has the problem of small sample sizes. Other national datasets of good quality and larger size were chosen when available, especially if they proved to be reliable enough for tax-benefit microsimulation since they had been used in national models. Twelve separate data contracts with distinct conditions and requirements were signed. Table 1. Country

Austria Belgium Denmark Finland France

Sources of Micro-Date for EUROMOD.

Base Dataset

European Community Household Panel Panel Survey on Belgian Households (PSBH) European Community Household Panel Income distribution survey (IDS) Budget de Famille (BdF)

Germany

German Socio-Economic Panel (GSOEP)

Greece

European Community Household Panel Living in Ireland Survey (LIS) Survey of Households Income and Wealth (SHIW) PSELL-2

Ireland Italy

Luxembourg Netherlands Portugal Spain Sweden UK

Sociaal-economisch panelonderzoek (SEP) European Community Household Panel European Community Household Panel Income Distribution Survey (IDS) Family Expenditure Survey (FES)

Type

Year of Collection (Sample Size: Individuals)

ECHP

1998+1999(7,386)

National panel

1997(7,057), 1999(9,089) 1995(7,044)

ECHP Register+survey Household budget survey National panel

1998(25,010), 2001(28,303) 1994/1995(29,158)

ECHP

1998(18,227), 2001(16,874), 2002(26,249) 1995(15,062)

National panel

1994(14,585)

Income survey

1996(23,924)

National panel

1999(6,566), 2001(6,226) 1996(11,035), 2000(10,344) 1996(14,468), 2001(13,092) 1996(18,991), 2000(14787) 1997 (38,756), 2001 (33,223) 1995/1996(16,586), 2000/2001(15,914)

National panel ECHP ECHP Register+survey Household budget survey


9

Many of the general complications associated with collecting information on tax-benefit rules and using micro-data became particularly acute because of the multi-country nature of EUROMOD. Solutions adopted in single countries were put into question when confronted with non-comparable approaches from other countries. A typical case is the level of detail of variables: this varies greatly across national data. Some countries have very detailed information that allows the modeling of taxes and benefits with precision, including those details less relevant when it is comparison that matters (for example taxes and duties at municipal level).12 Some others can only access much less detailed data and they are often accustomed to a more schematic modeling of their tax-benefit systems. This problem was treated by defining as many ‘‘common variables’’ (variables available in all 15 datasets and needed for the simulation of many national systems) as possible, whereas a minimum possible number of ‘‘country-specific variables’’ were added to the databases, if they were necessary for the simulation of the national taxbenefit system, but were not available or needed for many other countries. In general, national teams were responsible for dealing with data quality issues, such as exclusion from the sample, non-responses and under reporting. However, during the course of the project a number of technical problems had to be dealt with in a coordinated and common manner. Three issues emerged as being particularly important. First, the reference time period was an issue. The original data used for deriving the EUROMOD database refer to different time periods in different countries. Income is most commonly available on an annual basis, but in some countries for some variables the reference period is as short as a week. The month was chosen as the common reference period for the model output. Second, imputation of gross incomes was necessary since many of the micro-datasets used by EUROMOD originally contained only incomes net of taxes and social insurance contributions.13 Finally, updating the databases to allow simulations for different countries at the same point in time is an obvious requirement for comparability, but not all national sources are made available every year so that the available years among some countries did not correspond. Therefore, a program for static upsdating was developed to adjust datasets to the same year and to later simulation periods.14 2.3. Design Issues: Ease of Use versus Flexibility A multi-country model like EUROMOD must permit all the tax-benefit systems that it covers to be simulated in a consistent way. This, together with the fact that the key issue is comparability, makes such a model

10


necessarily complex. Moreover, while most national tax-benefit microsimulation models limit the range of reforms that can be simulated to national requirements to minimize complexity for the model developer and user. Such a limitation is much more difficult in a multi-country framework, as there is no established tradition suggesting reasonable restrictions.15 This means that there are trade-offs between flexibility and user friendliness and partly between flexibility and comparability. EUROMOD is in general a highly flexible tool and able to address a wide range of policy questions, though several limitations have been accepted at least temporarily, to keep the development of the model a manageable task (see below).16 2.4. Test, Validation and Documentation Currently, EUROMOD contains tax-benefit rules for EU-15 for the years 1998 and 2001, and for the majority of them also for 2003. To verify the quality of EUROMOD’s outcomes country experts, in close collaboration with the model-building team, carried out testing and validation exercises. These exercises comprise mainly three sorts of checks. The first is case-by-case checking, which consists of comparing EUROMOD micro-results for some observations with corresponding micro-results obtained in a completely independent way. The second, aggregate validation, involves the comparison and reconciliation of aggregate indicators and statistics – such as the Gini coefficient for household incomes, poverty rates or aggregate budgetary amounts – produced by EUROMOD with corresponding statistics provided by external sources. Where national models were accessible to EUROMOD members, EUROMOD results were compared with national model results as well. The aggregate validation and comparisons against national models are documented in Country Reports, which also contain comprehensive documentation on the manner in which each tax-benefit system is implemented in EUROMOD.17 In addition, two aggregate validation exercises, encompassing all 15 countries, were carried out by comparing EUROMOD results with ECHP statistics, with reassuring results on reliability (see Mantovani & Sutherland, 2003; Lietz & Sutherland, 2005). 2.5. Limitations and New Developments As mentioned above, several limitations have been accepted, at least temporary, to limit complexity and to keep model building a manageable task.


11

The most important are: non take-up of benefits is considered only to a very limited extent and tax evasion is not accounted for; EUROMOD is a static tax-benefit microsimulation model, therefore long-term effects of policy reforms are not accounted for since behavioral responses to policy reforms, such as the changing of labor supply behavior, are generally not considered. Moreover, not all tax-benefit instruments can be simulated. Generally speaking, the following instruments are simulated in all countries: income taxes (national and local); social insurance contributions (paid by employees, employers and the self-employed); family benefits; housing benefits; social assistance and other income-related benefits. The following instruments are generally not simulated: capital and property taxes; real estate taxes; disability benefits and finally – due to absence of information of previous work history in most datasets – contributory benefits, most importantly pension/survivor benefits and unemployment benefits. These instruments that are not simulated are nevertheless taken into account in EUROMOD’s output concept of disposable income, as they can be derived from the original data.18 However, EUROMOD is an ongoing project and there are several new developments, which promise to tackle some of the limitations discussed above. One planned project will investigate possibilities to take account of non-cash incomes, tax evasion and non-take-up of benefits and methods for imputation of expenditure data for analyzing indirect taxes.19 Great effort has been and is put in making EUROMOD a suitable tool for research on the impact of tax-benefit policies on work incentives and labor supply (see Section 3.4 for more details). To improve user friendliness a prototype user interface has been developed. It provides a graphical interface aimed to help the user to understand and apply EUROMOD and its concepts. The intention of this prototype was to facilitate the use of EUROMOD while not limiting any of its capabilities. As a result it mainly addresses the needs of users with some experience. One main focus of a recently started new project is to further enhance ease of use.20 In fact, user friendliness is a fundamentally important issue if the model is to be widely used. Therefore a user group, composed of frequent EUROMOD users as well as less experienced users has been constituted, which will meet on regular basis to discuss model improvements. Finally, EUROMOD is frequently revised on the basis of feedback provided by model users. Since EUROMOD started to be used, the group of users has increased, including new people who were not involved in the construction of the model, all of them making a valuable contribution in testing the model and in validating results.

12


3. THE SCOPE OF A EUROPEAN TAX-BENEFIT MICROSIMULATION MODEL EUROMOD has been conceived to be a tool that enables a wide range of questions about the impact of social and fiscal policy on the population of the EU to be answered, and allows a large variety of conceptual frameworks and assumptions to be adopted. The numerous research projects that have been carried out on several different subjects and applying different approaches show that this aim has been achieved to a large extent. EUROMOD offers the classic features of a tax-benefit microsimulation model like (i) case study analysis (e.g. drawing budget curves for any family type); (ii) information at the individual or aggregate level (total tax revenue, total cost of an instrument, a reform, etc.); (iii) analysis of the redistributive potential of a system (inequality and poverty indices) or the contribution of a given instrument to inequality and poverty; (iv) analysis of the incentive potential of a system (financial gains to work, effective marginal tax rates) or the contribution of a given instrument (e.g. an inwork transfer) and (v) analysis of the incidence of a given instrument or reform (gainers/losers, etc.). These features allow consistent analyses for all 15 countries included and some of them can be applied on a hypothetical ‘‘Europeland’’. Beyond this, as an integrated multi-country model, EUROMOD offers specific possibilities like: (i) cross-country comparisons of certain instruments (e.g. the progressivity/equalizing role of income tax throughout EU-15); (ii) policy swapping (i.e. exporting one instrument from country A to country B and maybe vice versa) and system swapping (e.g. Spadaro, 2004); (iii) analysis of the impact of common changes across countries and (iv) analysis of the impact of country-specific changes at the European level with common objectives. In fact, as several applications have shown, EUROMOD is flexible and adaptable enough to have uses far beyond its builders’ original plans. Therefore, to give a complete list of all potential applications of the model is an impossible task. Instead, in the following we intend to provide some insight into research areas that EUROMOD is most suited to address and in the techniques that it offers to answer specific questions. This will draw to a large extent on research already carried out with EUROMOD, as examples show best how the model’s capabilities can be used. Moreover, it allows us to report some indicative findings and to give account of work in addition to that included in this book.


13

3.1. Inequality, Poverty and Redistribution A noticeable feature of EUROMOD is its standardized calculation of household disposable income and its components (original income, benefits, taxes and social insurance contributions21) for all countries included. This allows doing a range of calculation in a consistent way, among others it provides a common base for computing measures of income distribution, redistribution and poverty. A basic but most indicative output of EUROMOD is the provision of an articulated picture of the effects of actual European tax-benefit systems. Many EUROMOD applications are based on this feature, including several contributions in this book. A standard set of statistics on taxes, benefits and income distribution in the 15 member states has been produced.22 They show income components and household characteristics by decile group for each country and for EU-15 as a whole. Fig. 1, which is derived from these statistics, is an illustration on how this EUROMOD output can be used to derive interesting findings.23 It shows the incidence of benefits and taxes (including social insurance contributions) on disposable income of poor households. Benefits are, in turn, defined as ‘‘gross’’ and ‘‘net’’ benefits, where net benefits are gross benefits minus taxes. The picture drawn by using the concept of net benefits is rather different from that obtained by using gross benefits alone. In several countries, the poor pay a significant amount in taxes (and/or contributions). These taxes may be levied on other sources of income but in some cases on the benefits themselves. This illustrates that to establish the impact of social transfers on the poor across countries, account needs to be taken of the effect of taxes and contributions as well as cash transfers. It also shows that higher poverty rates are to be found mostly in southern countries (Portugal, Italy, Greece and Spain) where family and social benefit schemes are not as developed as in the rest of Europe. A more sophisticated analysis of tax-benefit systems – however still based on EUROMOD basic output (i.e. decomposed household disposable income) – was recently accomplished by Immervoll et al. (2005a). The authors investigate the direct effect of tax-benefit systems on cross-sectional inequality. Unlike many other international studies, the analysis encompasses both direct taxes and cash benefits. They find that non-means-tested benefits and taxes play an essential role in reducing inequality in countries with comparably more redistributive tax-benefit systems, such as the Scandinavian and most of the countries in central Europe. UK and Ireland, where

14


90% 80%

Taxes and contributions Benefits Net benefits Poverty rate (rh axis)

70%

25%

20%

60% % gross income

50% 15%

40% 30% 20%

10%

10% 0% 5% -10% -20% 0%

-30% DK AT LU FI

BE NL SW GE FR UK SP EU15 GR IT PT IR

Fig. 1. The Effect of Taxes and Benefits on Incomes in Poor Households, 2001. Note: ‘‘Poor’’ is Defined as Having Equivalized Household Income Below 60% of the National Median Household Incomes are Equivalized Using the Modified OECD Equivalence Scale (1/0.5/0.3). Source: EUROMOD; MICRESA Final Report at http://www.iser.essex.ac.uk/msu/emod/micresa-final.php.

means-tested benefits are relatively more important, have an intermediate position concerning reduction of inequality. Finally, southern European countries show the lowest degree of redistribution, with the exception of Spain if pensions are included in benefits. Though EUROMOD basic output is already able to provide some interesting insights, the typical use of EUROMOD as a tax-benefit microsimulation model consists in analyzing the effects of planned or illustrative reforms of tax-benefit policies. The standard definition of household disposable income is very useful for such analyses as well, as measurements of redistribution, changes in risk of poverty and budgetary effects can be computed on a comparable base. This is especially useful for studies which compare cross-country effects of reforms implemented at European level. This is the reason why applications of this sort appear as the earliest studies carried out with EUROMOD. Examples for such studies are Atkinson, Bourguignon, O’Donoghue, Sutherland, and Utili (1999), which explores a


15

universal minimum pension at European level and Atkinson (2000) which considers ‘‘poverty benchmarking’’ among countries and calculates a basic measure for the trade-off between poverty reduction and social spending. Two more recent examples of such methodology, referring to pensioners’ well-being and tax-transfer policies, are presented in both the contribution of Mantovani et al. and that of Callan in the present book.

3.2. Example: Child Poverty One key common social objective within the EU is the reduction of poverty and social exclusion concerning children and many member states are paying particular attention to the position of children living in households at risk of poverty. EUROMOD has the potential for enriching the knowledge about how to protect children from social exclusion by allowing the assessment and comparison of the impact of different European tax-benefit systems on child poverty. A recent study by Corak, Lietz, and Sutherland (2005) analyses the impact of fiscal policy on the economic resources available to children, and on the child poverty rate. The study is not only interesting because of its significant findings, but also because it applies a novel technique to measure those resources available to children. As the study illustrates and is briefly summarized below, the way in which ‘‘child benefits’’ are defined is important in assessing the impact of public policy. The definition used in the study and referred to as child-contingent incomes corresponds to the extra income the household receives from the state because of the presence of children and is assessed by the re-calculation of taxes and transfers as though there were no children. The difference between taxes and transfers in this ‘‘no children’’ scenario and the original taxes and transfers form the child-contingent incomes. This calculation is not the same as simply counting up the value of child and family benefits. Social assistance benefits may ‘‘fill the gap’’ left by family benefits; alternative housing benefit schemes may exist for parents and non-parents; some child-related components may be taxable and in this case their absence would result in a reduction in tax liability (though in general the removal of tax concessions for children will result in taxes rising). EUROMOD re-calculates liabilities and entitlements and thus measures the net effect of child-contingent tax-benefit components. Fig. 2 illustrates that the definition of resources available to households with children is of significant importance, by showing child poverty measured with different income bases: (1) household disposable income,

16

CHRISTINE LIETZ AND DANIELA MANTOVANI 40%

40% 35%

Disposable income Without child-contingent incomes Without all transfers Without all transfers+taxes

35%

30%

30%

25%

25%

20%

20%

15%

15%

10%

10%

5%

5%

0%

0% SW DK FI

AT BE LU GE NL FR GR UK PT IT

SP IR

Fig. 2. The Child Poverty Rate in EU-15 (2001), with and without Child-Contingent Incomes, all Transfers, and all Transfers and Taxes. Note: Child Poverty is Defined as the Proportion of Households with Children Below the National Poverty Line (50% of the National Median of Equivalized Household Incomes, using the Square Root Equivalence Scale). Source: EUROMOD; Corak et al. (2005).

(2) household income without child-contingent incomes; (3) household income without all transfers, and (4) household income without all transfers and before taxes. The contrast between the darkly shaded bars and the height of the circle-topped lines illustrates that child poverty rates are strongly affected by public resources not specifically addressed to children. The contrast is largest in the Scandinavian countries, because of public support to working parents. It is also important in the southern countries, but in this case because of the relatively more important role of intrahousehold transfers from other co-resident adults. It also has proved instructive to compare the ‘‘before transfers’’ child poverty rates (a measure often used to assess the effect of government policies on poverty) with those that would apply before both transfers and taxes. As the figure illustrates, taking away the transfers without giving back the taxes paid on them overestimates the impact of the transfer system on poverty reduction, the effect being very significant in some countries. Of course some of the taxes correspond to taxes and contributions on original income, not on transfers. The point, in this context, is to demonstrate the relevance of taxes. Another fruitful source of investigation is the comparisons of countries’ various family policies and their success in combating child poverty. In particular, EUROMOD’s standardized structure facilitates a special approach: to transfer family benefits between countries, by implementing one


17

country’s benefits in another and vice versa. This kind of application fully exploits the most distinctive characteristics of EUROMOD: comparability and homogeneity. Examples of such ‘‘benefit swapping’’ exercises are given in Immervoll, Sutherland, and de Vos (2000), Sutherland (2001b), Levy (2003) and Levy, Lietz, and Sutherland (2005). The contribution of Matsaganis et al. in the present book also analyze the impact of importing Northern child benefit policies to Southern Europe in order to alleviate child poverty. All these studies find that there is clearly some scope for improvements in looking beyond national borders.

3.3. Example: Links Between Macro-and Microeconomic Variables Research carried out with tax-benefit microsimulation models is usually not expected to deal with macro economic issues, such as growth, inflation and unemployment. However, this has been precisely the subject of a number of studies using EUROMOD. The contribution of Immervoll in the present book provides an important example of this kind of application. The author analyses the stabilization effect of fiscal drag in presence of inflation. One other example of research on fiscal stabilizers in Europe is presented by Mabbett (2004). Using EUROMOD, the author produces estimates of the automatic stabilizers, which focus on the stabilization of household income rather than the budgetary effects of cyclical changes in taxes and benefits. Mabbett’s work confirms previous results that suggested that stabilization and redistribution are not separate functions, because welfare states provide insurance against excessive fluctuations of households’ income due to the economic cycle. Another study (Immervoll, Levy, Lietz, Mantovani, & Sutherland, 2005b) explores the sensitivity of (micro-based) indicators for social inclusion on ‘‘macro’’ changes for EU-15. Late in 2001 the EU adopted a set of commonly agreed indicators for social inclusion (see Atkinson, 2005). These so-called Laeken indicators – among them some income based poverty measures – are meant to be part of a toolbox to assess member states’ success in reducing poverty and social exclusion. However, the positive effects of policy initiatives may be mitigated by other, independent changes in the economy or society. The study uses EUROMOD’s capability to emulate such ‘‘macro’’ changes. Poverty measures can be computed based on original household disposable income before and after the simulated change. This allows an exploration of the sensitivity of indicators to such changes, i.e. if they are affected by the change, and if so, whether in the way that

18


might be expected. The changes considered are (a) an increase in unemployment, (b) real income growth and (c) an increase in earnings inequality. The results suggest that indicators are sensitive to such types of macro changes and that the extent of sensitivity does differ across countries. The most important result consists of demonstrating the dangers of relying on single indicators instead of maintaining a portfolio. In fact, unemployment, as simulated has an ambiguous effect on the poverty rate. Also the simulation of an increase in earnings inequality – an example of increasing the income of the rich at the expense of the poor – results in some countries in a decrease in measured poverty heads count, as the poverty line moves down.

3.4. Example: Work Incentives and Behavioral Responses The basic version of EUROMOD is a static model, designed to calculate the immediate, ‘‘morning after’’ effect of policy changes. As such, it does not take into account the effects of behavioral changes nor the long-term effect of change. However, great effort has been put in making EUROMOD a suitable tool for research assessing the impact of tax-benefit policies on work incentives and labor supply. The starting point was the observation that EUROMOD is an ideal tool to compute the different budget constraints faced by workers under different tax-benefit systems. A next step was the implementation of facilities to calculate implicit tax rates (often referred to as Effective Marginal Tax Rates, EMTR) to give an insight of the potential effect of a given tax-benefit system or reform on labor supply behaviors at the intensive margin (i.e. in terms of the number of working hours). This was motivated by the fact that macro-based effective tax rates do not capture differences among population sub-groups nor do they provide distribution of EMTRs by income levels. In this respect, EUROMOD not only allows the comparison of effective tax burdens over EU-15 in a consistent way, but can also show how many and which types of individuals face different EMTR levels. Moreover, results can also be broken down to isolate the influence of income taxes, social contributions and various types of social benefits. Immervoll (2004) recently covered the broad analysis of average and marginal tax rates for 14 of the EU countries. In addition, EUROMOD allows the computation of replacement rates, namely the theoretical variation in household incomes due to a change in employment status of one or more of its members. An application of this feature is found in Immervoll and O’Donoghue (2004b). The contribution of


19

Bargain and Orsini in the present volume also illustrate how to characterize the (dis)incentive potential of a given tax-benefit system or a reform (the UK working family tax credit in their case). In particular, financial gains to work (or replacement rates computed when switching from inactivity to employment) give a clear idea of the potential effects of reforms on labor market participation of different sub-groups of the population. Finally, computations of EMTRs or budget constraints at the individual level can feed in labor supply models for richer comparative analyses involving behavioral responses. Applications of EUROMOD in this promising direction are starting to develop (see Bourguignon & Spadaro, 2000a, b; Spadaro, 2004; Bargain & Orsini, 2004; Immervoll, Kleven, Kreiner, & Saez, 2004a).

4. CONCLUSION AND FUTURE ISSUES The long process of building EUROMOD has resulted in a success to a large extent; EUROMOD has been used widely for comparative studies by academic researchers. It has been and is used for a number of high-scale projects by international institutions as well. Some recent examples are the contribution to the UNICEF report on child poverty in rich countries (UNICEF, 2005), the OECD research using typical household calculations (Immervoll, Marianna, & Mira D’Ercole, 2004b) and the contribution to the Social Situation Report 2005 of the European Commission (DG-EMPL, 2006). Nevertheless EUROMOD is now facing the challenge of keeping up with new knowledge needs of social scientists and policy makers due to the enlargement. Several new issues have to be addressed for the model to continue to be useful and to realize its full potential. Some developments seem of particular importance: (i) Ways to update EUROMOD, i.e. implement tax-benefit rules for further policy years and incorporate recent data, must be found. This involves further improvement and standardization of procedures to facilitate these tasks. (ii) The extension of EUROMOD to cover the 10 new member states of the EU seems to be another evident task. Considering the fact that this means extending EUROMOD for tax-benefit systems, which are likely to involve some features that are quite different form those already encountered in EU-15, this task is certainly not trivial and might also involve a reassessment of the way EU-15 is treated. The EUROMOD team has recently started a new project that, among others, investigates the inclusion of the new member states into EUROMOD.24

20


(iii) The community of users needs to be expanded to fully exploit the research possibilities the model offers. It is important in this context to improve EUROMOD’s user-friendliness to facilitate the application of the model.

NOTES 1. De Lathouwer (1996) also simulates the effects of taxation of the unemployment benefit system, enforced in the Netherlands, on a sample of Belgian households, thus reflecting the importance of the socio-demographic characteristics of the population on the resulting effects. 2. Bourguignon and Spadaro (2005) give a general overview of what can be achieved using microsimulation techniques. 3. This kind of analysis is still published by the OECD as ‘‘Benefits and Wages – OECD indicators’’, though the series has changed contents and title. 4. Analysis based on hypothetical family-type methodology can be replicated with microsimulation models. An example of such an exercise carried out with EUROMOD is Berger et al. (2001). 5. The LIS project began in 1993 and currently includes data for 25 countries. For more information see: http://www.lisproject.org/. 6. For more information about ECHP see: http://www.forum.europa.eu.int/irc/ dsis/echpanel/info/data/information.html. 7. The use of imputation techniques, directly derived from microsimulation, is a possibility to obtain missing variables, see for example Immervoll and O’Donoghue (2001b). 8. Main findings and recommendations of the preparatory study are presented in a final report (see Sutherland, 1997). 9. For further details see Sutherland (2005). 10. A ‘‘Database Robustness Assessment Exercise (DRAE)’’ was carried out in order to assess the quality of data. Country respondents were asked to answer a questionnaire about the databases and the responses were analyzed and compared. 11. For some countries a more comprehensive national version is used. 12. Other examples are variables necessary to simulate certain country-specific benefits, that cannot be simulated for other countries (where such variables are not collected). 13. Net-to-gross imputation had to be performed in nine countries using a numerical inversion routine (see Immervoll & O’Donoghue, 2001b). 14. The possibility of applying static aging techniques was also explored in Immervoll, Lindstro¨m, Mustonen, Riihela¨, & Viitama¨ki (2005c). 15. In many countries changes involve reprogramming part of national models; there are exceptions, such as the Spanish national model, EspaSim (available at: http://selene.uab.es/espasim) and the Irish national model, SWITCH (Callan, O’Donoghue, & O’Neil, 1996). 16. The appendix provides insight of how these issues are dealt with in EUROMOD.


21

17. See http://www.iser.essex.ac.uk/msu/emod/countries/. 18. This income concept is however based on monetary income and does not include non-cash benefits, such as free health care, child care provision and public housing. Also indirect taxes, most importantly consumption taxes, cannot be analyzed with the current version of EUROMOD. 19. FP6-specific targeted research project on ‘‘Accurate Income Measurement for the Assessment of Public Policies’’ (AIM-AP). 20. FP6 research infrastructures design study on ‘‘improving the capacity and usability of EUROMOD’’ (I-CUE). 21. EUROMOD is unique in allowing such a decomposition of disposable income, as it provides information on taxes and benefits for all 15 countries, while for some of them other micro-data sources contain only disposable incomes. 22. See http://www.iser.essex.ac.uk/msu/emod/emodstats/. 23. The figure is taken from the MICRESA final report (Sutherland, 2005). 24. FP6 research infrastructures design study on ‘‘improving the capacity and usability of EUROMOD’’ (I-CUE). 25. A more detailed description of EUROMOD’s programming can be found in Immervoll, O’Donoghue and Sutherland (1999) and in Immervoll and O’Donoghue (2001a).

ACKNOWLEDGMENT This chapter is based on the work of all the researches who contributed to EUROMOD’s construction and who carried out research with EUROMOD. We have benefited in particular from previous studies carried out by our former colleagues at Microsimulation Unit, Herwig Immervoll and Cathal O’Donoghue, and by the team coordinator Holly Sutherland. We wish to thank Olivier Bargain and Holly Sutherland for their comments and suggestions during the various stages of the writing process. We are also grateful to Herwig Immervoll for his comments on an earlier version of the chapter. The views expressed are those of the authors. We are the one responsible for any errors as well.

REFERENCES Atkinson, A. B. (2000). A European social agenda: Poverty benchmarking and social transfer. EUROMOD Working Paper, EM3/00. Atkinson, A. B. (2005). EUROMOD and the development of EU social policy. DIW Discussion Papers, 467. Atkinson, A., Bourguignon, F., & Chiappori, P. A. (1988). What do we learn about tax reforms from international comparisons? France and Britain. European Economic Review, 32.

22


Atkinson, A. B., Bourguignon, F., O’Donoghue, C., Sutherland, H., & Utili, F. (1999). Microsimulation and the formulation of policy: A case study of targeting in the European Union. EUROMOD Working Paper, EM2/99. Atkinson, A. B., King, M. A., & Sutherland, H. (1983). The analysis of personal taxation and social security. National Institute Economic Review, 103, 63–74. Atkinson, A. B., Rainwater, L., & Smeeding, T. M. (1995). Income distribution in OECD countries. OECD, Paris. Bargain, O., & Orsini, K. (2004). In-work policies in Europe: Killing two birds with one stone? EUROMOD Working Paper, EM4/04. Berger, F., Borsenberger, M., Immervoll, H., Lumen, J., Scholtus, B., & De Vos, K. (2001). The impact of tax-benefit systems on low income households in the Benelux countries. A simulation approach using synthetic datasets. EUROMOD Working Paper, EM3/01. Bourguignon, F., Chiappori, P.A., & Sastre, J. (1988). SYSIFF: A simulation program of the French tax-benefit system. In: A. B. Atkinson, & H. Sutherland (Eds), Tax benefit models. London: STICERD. Bourguignon, F., O’Donoghue, C., Sastre-Descals, J., Spadaro, A., & Utili, F. (1997). Eur3: A prototype European tax-benefit model. Microsimulation Unit Discussion Paper, MU9703. Bourguignon, F., & Spadaro, A. (2000a). Redistribution and labour supply incentives: An application of the optimal tax theory. Paper presented at the 56th Congress of the IIPF, Seville, 28–31 August. Bourguignon, F., & Spadaro, A. (2000b). Social preferences revealed through effective marginal tax rates. Working Paper DELTA, No 29-2000. Bourguignon, F., & Spadaro, A. (2005). Microsimulation as a tool for evaluating redistribution policies. ECINEQ Working Paper, 2006-20. Callan, T., O’Donoghue, C., & O’Neil, C. (1996). Simulating welfare and income tax changes: The ESRI tax-benefit model. Dublin: ESRI. Callan, T., & Sutherland, H. (1996). Comparative analysis of basic income proposals: Prospects for the use of national tax-benefit models in five European countries. Microsimulation Unit Discussion Paper, MU/RN/21. Callan, T., & Sutherland, H. (1997). The impact of comparable policies in European countries: Microsimulation approaches. European Economic Review, 41, 627–633. Corak, M., Lietz, C., & Sutherland, H. (2005). The impact of tax and transfer systems on children in the European Union. EUROMOD Working Paper, EM4/05. De Lathouwer, L. (1996). A case study of unemployment scheme for Belgium and the Netherlands. In: A. Harding (Ed.), Microsimulation and public policy. Amsterdam, North Holland: Contributions to Economic Analysis 232. DG-EMPL (2006). Network on social inclusion and income distribution. Final Report, European Observatory on the Social Situation (SSO). Immervoll, H. (2004). Average and marginal effective tax rates facing workers in the EU. A micro-level analysis of levels, distributions and driving factors. EUROMOD Working Paper, EM6/04. Immervoll, H., Kleven, H. J., Kreiner, C. T., & Saez, E. (2004a). Welfare reform in Europe: A micro-simulation analysis. CEPR Discussion Paper, 4324. Immervoll, H., Levy, H., Lietz, C., Mantovani, D., O’Donoughue, C., Sutherland, H., & Verbist, G. (2005a). Household incomes and redistribution in the European Union: Quantifying the equalising properties of taxes and benefits. EUROMOD Working Paper, EM9/05.


23

Immervoll, H., Levy, H., Lietz, C., Mantovani, D., & Sutherland, H. (2005b). The sensitivity of poverty rates to macro level changes in the European Union. Cambridge Journal of Economics, 30, 181–199. Immervoll, H., Lindstro¨m, K., Mustonen, E., Riihela¨, M., & Viitama¨ki, H. (2005c). Static data ‘‘ageing’’ techniques. Accounting for population changes in tax-benefit microsimulation. EUROMOD Working Paper, EM7/05. Immervoll, H., Marianna, P., & Mira D’Ercole, M. (2004b). Benefit coverage rates and household typologies: Scope and limitations of tax-benefit indicators. OECD Social, Employment and Migration Working Paper No. 20, OECD, Paris. Immervoll, H., & O’Donoghue, C. (2001a). Towards a multi-purpose framework for tax-benefit microsimulation. EUROMOD Working Paper, EM1/01. Immervoll, H., & O’Donoghue, C. (2001b). Imputation of gross amounts from net incomes household surveys, an application using EUROMOD. EUROMOD Working Paper, EM2/01. Immervoll, H., & O’Donoghue, C. (2004). What difference does a job make? The income consequences of joblessness in Europe. In: D. Gallie (Ed.), Resisting marginalization: Unemployment experience and social policy in the European Union. Oxford: Oxford University Press. Immervoll, H., O’Donoghue, & C., Sutherland, H. (1999). An introduction to Euromod. EUROMOD Working Paper, EM0/99. Immervoll, H., Sutherland, H., & de Vos, K. (2000). Child poverty and child benefits in the European Union. EUROMOD Working Paper, EM1/00. Levy, H. (2003). Child-targeted tax-benefit reform in Spain in a European context: A microsimulation analysis using EUROMOD. EUROMOD Working Paper, EM2/03. Levy, H., Lietz, C., & Sutherland, H. (2005). Alternative tax-benefit strategies to support children in the European Union: Recent reforms in Austria, Spain and the United Kingdom. EUROMOD Working Paper, EM10/05. Lietz, C., & Sutherland, H. (2005). Social indicators and other income statistics using EUROMOD: An assessment of the 2001 baseline and changes 1998–2001. EUROMOD Working Paper, EM06/05. Mabbett, D. (2004). Fiscal stabilizers in Europe: The macroeconomic impact of tax and benefit systems. EUROMOD Working Paper, EM7/04. Mantovani, D., & Sutherland, H. (2003). Social indicators and other income statistics using EUROMOD baseline: A comparison with Eurostat and national statistics. EUROMOD Working Paper, EM1/03. Merz, J. (1991). Microsimulation – a survey of principles, developments and applications. International Journal of Forecasting, 7, 77–104. Spadaro, A. (2004). Microsimulation and normative policy evaluation: An application to some EU tax-benefits system. DELTA Working Paper, 2003–23. Sutherland, H. (1995). Static microsimulation models in Europe: A survey. Microsimulation Unit Discussion Paper, MU9503. Sutherland, H. (1997). The EUROMOD preparatory study: A summary report. Microsimulation Unit Discussion Paper, MU9705. Sutherland, H. (2001a). EUROMOD: An integrated European benefit-tax model-final report. EUROMOD Working Paper, EM9/01. Sutherland, H. (2001b). Reducing child poverty in Europe: What can static microsimulation models tell us? EUROMOD Working Paper, EM5/01.

24


Sutherland, H. (Ed.) (2005). Micro-level analysis of the European social agenda: combating poverty and social exclusion through changes in social and fiscal policy-final report. EUROMOD Working Paper, EM8/05. UNICEF. (2005). Child poverty in rich countries, 2005. Innocenti Report Card No.6, UNICEF Innocenti Research Centre, Florence.

APPENDIX. ARCHITECTURE OF THE MODEL The programming language used for EUROMOD is C/C++, with input data stored in Microsoft Access databases and all other input parameters stored as spreadsheet tables that can be read and manipulated with any spreadsheet software.25 Any country’s tax-benefit system might be seen as a collection of complex policy instruments, such as income tax and social assistance. In turn, policy instruments are made up of more elementary blocks, e.g. application of rates to tax-base brackets and tax credits. Fig. A1 shows EUROMOD’s hierarchical structure. The figure also introduces the terminology used by EUROMOD builders and users. Each national tax-benefit System is made up of individual Policies. The order in which single policies are simulated is not ‘‘hard-wired’’ in the

Fig. A1.

EUROMOD Hierarchical Structure.


25

program, but might be chosen by editing an input file, called Policy Spine, that controls the flow of the computational operations taking place within the model. In practice, the Policy Spine is a list of policies indicating the sequence by which they are computed in the tax-benefit system. For example, if social insurance contributions are tax deductible, then the corresponding policy would have to appear before the policy that computes income tax because the model requires the amount of contributions as a prerequisite for calculating income tax; similarly, if social assistance depends on after tax income, then the social assistance policy would have to appear after the income tax policy. For the implementation of a policy elementary building blocks, called Modules, are used. For instance, a child benefit policy may contain a module for means-testing, a module for general eligibility conditions (e.g. test on the age of children) and a last module for computation of the entitled amount. Technically a module is a piece of code (a C-function), but its behavior is not ‘‘hard-wired’’ as it requires parameters to specify what it does. If all inputs of a module are parameterized it is possible to implement changes without any reprogramming. The concept of Modules as distinct elementary building blocks has special advantages. If Modules are designed in a sufficiently general manner one can build up a large ‘‘library’’ of algorithms and can use these for different policies and different tax-benefit systems. An example for such a Common Module is the so-called ‘‘benefit calculator’’, which allows the implementation of a wide range of different types of benefits. EUROMOD’s Common Modules provide a general structure, which can be seen as using a standardized language to describe policy instruments of their sub-components. Once EUROMOD users are accustomed to this language, their understanding of other (foreign) countries’ benefits and taxes improves considerably. Moreover, the use of a common structure facilitates the detection of similarities and differences between (maybe only seemingly very different) tax and benefit instruments. EUROMOD’s Modules have three types of parameters. Tax-benefit parameters are those related to tax-benefit system amounts, such as rates, income brackets and means-testing thresholds. Fiscal units, define the group of person(s) on which the tax-benefit rules are to be performed; for example the persons over whom taxable incomes are to be aggregated in order to determine total taxable income for a joint tax system, or the definition of family relevant for a certain benefit. Users may not only choose among a standard set of already defined fiscal units, they are able to create their own new definitions as well. This is made possible by two EUROMOD’s characteristics. First, sufficiently disaggregated and detailed micro-data (each

26


individual is identified separately and some relationships, such as parent/ child, partner or living in same household, between these individuals are recorded), second, there is an algorithm within the model that can aggregate individuals according to criteria specified by the user. Finally, another type of parameter refers to Income Concepts, used as both, input to the taxbenefit algorithms (e.g. taxable income, ‘‘means’’, etc.) and as output of the model (e.g. disposable income). Income concepts can be defined in terms of all monetary variables (whether contained in the micro-data or simulated by the tax-benefit model) available in the model.

PENSION INCOMES IN THE EUROPEAN UNION: POLICY REFORM STRATEGIES IN COMPARATIVE PERSPECTIVE Daniela Mantovani, Fotis Papadopoulos, Holly Sutherland and Panos Tsakloglou ABSTRACT This paper considers the effects on current pensioner incomes of reforms designed to improve the long-term sustainability of public pension systems in the European Union. We use EUROMOD to simulate a set of common illustrative reforms for four countries selected on the basis of their diverse pension systems and patterns of poverty among the elderly: Denmark, Germany, Italy and the UK. The variations in fiscal and distributive effects on the one hand suggest that different paths for reform are necessary in order to achieve common objectives across countries, and on the other provide indications of the appropriate directions for reform in each case.


27

28

DANIELA MANTOVANI ET AL.

1. INTRODUCTION According to many commentators, one of the most important problems facing the majority of the European Union (EU) member states in the coming years is related to the rapid ageing of their populations and the concomitant difficulties in the ability of their social protection systems to maintain the living standards of their senior citizens. In most EU countries ‘‘defined benefit’’ pay-as-you-go pension systems are in operation. With the anticipated decline in the ratio of workers to pensioners, serious doubts have been expressed regarding their sustainability. Broadly speaking, two types of policies have been recommended in the public discourse in order to avoid the collapse of the system: (i) policies aiming to move the pensions systems closer to ‘‘defined contributions’’ rather than ‘‘defined benefits’’ and/or promote private insurance schemes and (ii) policies aiming to modify the existing arrangements, without altering their main characteristics. The most important of the policy recommendations of the second type come under the headings of (a) delayed retirement, (b) lower replacement rates, (c) increased social insurance contributions rates and (d) introduction or extension of systems of minimum (solidarity) pensions in order to avoid situations of extreme poverty and deprivation among the senior citizens. Our paper focuses on proposals aiming to modify existing arrangements, rather than altering the very nature of the pension systems. While studies examining the fiscal consequences of these types of proposal can be found in the literature,1 relatively little detailed attention has been paid to the likely distributional impact of these recommendations, especially in the short run. This is an important gap, since, as the experience of many EU countries suggests, the short-term political acceptability and longer term sustainability of the reform proposals may depend primarily on distributional rather than fiscal considerations. Our paper attempts to fill this gap by providing estimates of the likely short-term effects of policy reforms of the type (b)–(d), if such reforms were to be introduced in the existing social protection systems of four European countries representing different ‘‘welfare-state regimes’’ (Esping-Andersen, 1990; Ferrera, 1996): Denmark (‘‘social-democratic’’), Germany (‘‘corporatist’’), Italy (‘‘southern’’) and the UK (‘‘liberal’’).2 For the purposes of our analysis we use EUROMOD, a detailed static microsimulation model covering all 15 (pre-May 2004) EU member states. As well as the fiscal implications of the reforms, emphasis is placed on their likely impact on the gap between the elderly and non-elderly and on the poverty status of the two groups. Detailed comparisons are

Pension Incomes in the European Union

29

performed under alternative scenarios and similarities and differences across countries are identified. Pension reform is a long-term process and analysis using dynamic microsimulation models studying various aspects of pension reform can be found in the literature.3 However useful, these models have a number of important limitations, especially regarding the huge number of assumptions required (regarding demographic trends, labour force participation and unemployment dynamics as well as interest rate dynamics) and their focus on distributional issues is typically related primarily to aspects of equity across generations. In addition to the obvious difficulties in predicting future social and economic conditions, there are also uncertainties surrounding the extent of political consensus behind reforms.4 So, while it is clear that a dynamic analytical approach of some variety is necessary to evaluate the long-term consequences of pension reform, a short-term static approach is also informative, particularly in a comparative perspective. Furthermore, by using a static microsimulation model we may ignore some, potentially important, behavioural aspects. Nevertheless, we can identify potential winners and losers under alternative reform scenarios and gain a richer understanding of the consequences of existing differences in institutional arrangements in particular societies. Moreover, using such models we can examine aspects of equity across generations as well as across income groups at a particular point in time. As a result, we obtain a clearer picture of the specific challenges facing each country, of the appropriateness of particular reform strategies and, perhaps, get useful insights regarding the political feasibility of proposed policy changes. In addition, in this paper we focus on a somewhat different notion of ‘‘adequacy’’ than that highlighted in recent official European discussions. Rather than considering the ways to encourage greater saving and private pension provision, as a route to the achievement of adequate retirement incomes in the future, we consider how to protect and improve the level of pension income now, and through the coming period of reform, thus achieving adequate retirement incomes in comparison with the living standards of the society under examination. If political consensus for pension reform is to be achieved, and other EU goals in relation to social inclusion are to be met, then one important factor will be the extent to which the solidarity component of the evolving pension systems protects the incomes of current pensioners (World Bank, 1994; Gillion, Turner, Bailey, & Latulippe, 2000; ILO, 2001). Of course, our meaning of ‘‘adequacy’’ may be debatable and, perhaps, costly to finance, while the very existence of a

30


guaranteed adequate level could reduce individual incentives to save for retirement. However, to make judgements about the appropriate balance between improving incentives and ensuring income security requires that one understands the costs and benefits of each strategy. This paper is intended as a contribution to such an understanding. The remainder of the paper is organised as follows. Section 2 provides a detailed description of the current situation of elderly and non-elderly in EU countries using data from the European Community Household Panel (ECHP) and estimates derived using EUROMOD. Section 3 analyses the separate effects of three policy reforms: the introduction (or enhancement) of systems of minimum pensions, lower replacement rates and increased social insurance contributions rates. Section 4 describes the impact of combinations of the above reforms. Section 5 concludes the paper.

2. INCOMES AND POVERTY OF ELDERLY PERSONS IN THE EU One of the main objectives of most pension systems is to provide elderly citizens with sufficiently high incomes, so that their living standards do not lag seriously behind those of the rest of the population, and so that they do not live in poverty. Therefore, we start by examining the current situation in the EU in these respects. Table 1 provides a snapshot picture of the income position of the elderly (persons aged 65+) in the 15 (pre-May 2004) EU member states using the information from the 2000 wave (with incomes for 1999), of the ECHP.5 The first row shows that the share of elderly persons living in private households varies between 11.1% in Ireland and 19.3% in Greece. This variation depends both on the demographic structure of the population and on institutional arrangements for care of the elderly.6 The second line reports the ratio of the mean equivalent disposable income of the elderly to the mean equivalent disposable income of the non-elderly.7 In most countries the percentage is between 80% and 100%. It is lowest in Denmark with a ratio of 78% while in the Netherlands the average income of the elderly is 4% higher than that of the non-elderly. Although in most countries the elderly lag behind the non-elderly, in most of them the differences are not as large as the differences observed 10 or 15 years ago (ISSAS, 1990; Hagenaars, de Vos, & Zaidi, 1994; Tsakloglou, 1996a, 1996b).8 While in aggregate the incomes of the elderly are reasonably close to those of the working aged in all countries, there is much more variation in the extent to which poverty occurs among the elderly, compared with the rest of

Relative Income Position of Elderly People in EU Countries.


Table 1.

31

32


the population. In four countries (the Netherlands, Luxembourg, Italy and Sweden) the poverty rate of the elderly is lower than that of the non-elderly – by a very large margin in the case of the Netherlands – while in Germany there is no difference in the poverty rates of the two groups. On the contrary, in Austria, Belgium, Finland, Ireland and, especially, Denmark, the poverty rates of the elderly are more than twice as high as those of the nonelderly. As a result, the share of the elderly in aggregate poverty varies enormously; from 3.5% in the Netherlands to 40.2% in Denmark.9 However, since in many countries a considerable proportion of the elderly receive a pension that is slightly lower than the poverty line, the elderly may be poor but, on average, they may be less likely than the non-elderly poor to be located far below the poverty line. Table 1 provides estimates of the poverty index of Foster, Greer, and Thorbecke (1984).10 In seven countries (Spain, Finland, Italy, Luxembourg, the Netherlands, Sweden and the UK) the estimates for the elderly are lower – sometimes substantially so – than the estimates for the non-elderly. In three countries the estimates for the two groups are very close (Germany, France and Portugal) and only in five countries (Austria, Belgium, Denmark, Greece and Ireland) the estimates for the elderly are clearly higher than those for the non-elderly. Once again, the contribution of the elderly to the aggregate poverty index varies considerably across countries; from around 2% in the Netherlands and Sweden to almost 40% in Denmark. Clearly, an important part of the explanation for poverty among the elderly lies in the pension systems. The final row of Table 1 shows the share of pensions in the incomes of the elderly, at the individual level. The aggregate picture is one of relative uniformity with the share of pensions in the incomes of those aged over 64 varying between 81% and 91% in all countries apart from the UK (76%). Thus, to explain the widely diverging patterns of poverty among the elderly we must investigate what lies behind these averages and the nature of the pension systems themselves. For example, high levels of poverty intensity indicate that pension coverage may be incomplete, or that the levels of some pensions are very low. High rates of poverty with low intensity suggest that the general level of pension is too low. Devising reform strategies involves taking existing inadequacies of level, structure and coverage of existing systems into account. In the next step we focus on the four countries of interest, using EUROMOD, a tax-benefit model for the EU (see Immervoll, O’Donoghue, & Sutherland (1999) for a general description). Tax-benefit models calculate disposable income for each household in a representative set of micro-data. The datasets used as the basis for this paper are listed in Appendix A. They


33

were chosen on the grounds that they provide the best quality input for a tax-benefit model and are at the same time available and accessible to an international scientific project. Although they include data collected at various points around the mid-1990s, they have all been adjusted to 1998 prices and incomes and, where necessary, gross incomes have been imputed from net (Immervoll & O’Donoghue, 2001). The calculation of household disposable income is made up of elements of gross income taken (or imputed) from the survey data combined with elements of income – taxes and benefits – that are simulated by the model. For the purposes of the paper, the calculations are performed once for the 1998 tax-benefit system, and again for each alternative scenario. The first round effect of the simulated change is the arithmetic difference in the ‘‘before’’ and ‘‘after’’ calculations. Table 2 provides some background information about the 1998 pension systems. The statutory retirement age is usually 65, except for women in UK and Italy (60), and for all retirees in Denmark (67). In the main part of our analysis we consider people aged 65 and over (the ‘‘elderly’’), which in the Danish case includes some people not yet eligible. In all countries, some of the elderly do not receive a pension. In addition, some pension recipients are not past retirement age. However, we do not consider early retirement in this paper, although this is a key issue for the sustainability of some pension systems: notably that of Italy among the four countries that we consider.11 Three countries have effective minimum levels of state pensions. These, as well as existing minimum income levels, are below the respective national poverty lines (calculated for a single person). Thus, it is not surprising that, as shown in Table 3, there are considerable proportions of elderly living below the poverty line in all four countries. These indicators of poverty before any reforms, as estimated using EUROMOD, provide a ‘‘baseline’’ picture against which to measure the effect of reforms.12 They confirm the picture provided by the ECHP data in Table 1 that while the population poverty headcount is lower in Denmark and Germany than in the other two countries, it is highest for the elderly in Denmark, followed by the UK and then Italy. On the other hand, intensity of poverty is lower for the elderly in the UK and Denmark, but higher in Germany. In fact, it is striking how the ranking of countries almost completely reverses when one compares the estimates of the poverty rate, which only takes into account the extent of poverty, and the FGT(2) which also takes into account the intensity of poverty (as well as the degree of inequality in the distribution of income among the poor). Table 2 shows that the coverage rates of pensions are much lower for the poor than for the population of elderly as a whole in Germany and Italy, suggesting that lack

34

Table 2.

State Pension Incomes in 1998 in Four EU Countries: Some Key Figures.

Pension agea Maximum public pension for a single personb Minimum public pension for a single personc Minimum income level, single-person age 70 Poverty line (single person)d % Elderly with public pension incomee % Elderly below the poverty line with public pension incomee % Elderly with private pension income % Employed people agedo65 paying state pension contributions Income ratio

Denmark

Germany

Italy

UK

67 (65) 14,726 6,276 No minimum income 10,263 88 89 n/a 100 0.67

65 No maximum No minimum 3,829 (average) 8,346 90 74 21 77 0.92

65/60 No maximum 4,684 1,216 6,275 81 76 1 99 1.00

65/60 13,666 4,784/1,198 5,088 7,747 98 98 54 87 0.81


Note: Except where otherwise stated money amounts are in 1998 euro per year using the following exchange rates per euro (31 December, 1998) DK,:7.4587; DE, 1.9558; IT, 1936.3; UK, 0.7032. These exchange rates have not been adjusted for purchasing power differences. Money amounts in this table should be compared within columns, not between columns. However, even information within columns has been drawn from a variety of sources for a range of years and is not necessarily fully comparable. Source: EUROMOD, unless otherwise specified; n/a, not available. a Economic Policy Committee (2002) Table 3.2. Men/women; (after 1998 reforms). b For DK this includes maximum old-age pension plus maximum ATP; UK includes basic pension plus maximum SERPS. c ‘‘Minimum pension’’ has different meanings in different contexts. For UK this is the full basic state pension/minimum basic state pension for those with minimum contributions leading to a pension. For DK it is the full ‘‘basic amount’’ of the old-age pension for someone meeting the residency requirement in full. A pensioner with no other income would be entitled to a means-tested addition. For Italy it is the minimum payable to a person with a contribution record. d 60% median equivalised income in 1998 euro per year. Calculated using EUROMOD (see Mantovani & Sutherland, 2003; Appendix B). e As defined in Section 3.1.


Table 3.

35

Poverty Indicators for 1998 in Four EU Countries: Some Key Baseline Figures from EUROMOD.

Poverty rate (all) Poverty rate (65+) FGT (a ¼ 2) (all) FGT (a ¼ 2) (65+)

Denmark

Germany

Italy

UK

11.4 31.0 0.973 0.483

9.9 13.9 0.673 0.978

20.1 18.0 3.131 1.189

19.9 22.6 1.107 0.730

Note: Poverty is defined as income below 60% of median equivalised household disposable income, using the modified OECD equivalence scale. Source: EUROMOD.

of coverage is at least a contributory factor to elderly poverty in these countries. On the other hand, in Denmark and the UK coverage is just as good for the poor as for the elderly as a whole, suggesting that inadequate pensions levels are to blame for poverty among the elderly. It appears that pension and benefit systems are successful at either keeping the substantial majority of elderly people out of poverty altogether while allowing a minority to fall substantially below the poverty line or preventing most from falling far below the poverty line through near-comprehensive minimum pension or minimum income schemes, even though a considerable proportion of them are located below – but not far from – the poverty line. None of the systems considered here either completely fail or completely succeed on both grounds.

3. PENSION SYSTEMS AND REFORM SCENARIOS The four countries we consider represent different types of pension systems facing different challenges. The German and Italian systems have at their heart an earnings-related contributory pay-as-you-go pension. The Italian system is supplemented by minimum pension schemes that are designed to make sure that pensioners with low entitlements under the earnings-related scheme receive at least a minimum level of pension. There is also a meanstested social pension for elderly people without contributory pensions. The Danish system is based on a non-contributory residence-based tax-financed pension. There is a means-tested addition and also a, now compulsory, contributory pension that is not earnings-related (but depends on hours worked). The UK system has a contributory basic pension plus an earningsrelated contributory state pension that may be replaced by private provision. In all four countries there are regulated or incentivised private

36


schemes. With the major exception of the UK, and to a smaller extent Germany, these schemes do not yet constitute a significant part of pension incomes for current pensioners. See Appendix B for more detail about current (1998) state pension schemes and contribution arrangements. In the analysis that follows we focus on public pensions and their associated contribution systems. To some extent we ignore the parts of the state system that are effectively contracted out to private occupational schemes or to personal pension savings schemes, even where these are compulsory and regulated by the state, or where they are subsidised through tax incentives. The reason for this is that it is impossible to draw a common line for all EU systems that divides ‘‘public’’ pensions from ‘‘private.’’ Any choice is to some extent arbitrary and here we make a choice based on pragmatic considerations. Using the data available to us we are unable to distinguish between all forms of quasi-public and private pensions and do not have sufficient information to calculate liability for private pension contributions. However, in the case of the UK ignoring state-subsidised private pensions would make that the remaining state pension provision appear to be very meagre, when compared with the other three systems. For many current pensioners, a significant source of pension income is from an occupational defined benefit scheme (in many respects similar to continental social insurance schemes and a major source of inequalities between elderly men and women in the UK, according to Bardasi & Jenkins, 2004). These pensioners (and their employers) have contracted out of the state earnings-related scheme and paid lower contributions. So in what follows these pensions are included along with state pensions for the UK, but not for other countries (where they are much less significant). Thus, conclusions that we draw about the UK system relative to the other three should bear this departure from comparability in mind. We consider three illustrative reforms, first individually and then in combination. Reform 1 is the introduction of a common minimum pension scheme in all four countries. The purpose of this is to strengthen the ‘‘solidarity’’ element of current systems with the aim of protecting pension incomes at some acceptable level, while other reforms take place. We consider two alternative, stylised methods of paying for this minimum, placing the burden on the working and the retired population respectively. Reform 2 is an increase in the rate of the main tax or contribution paid by employees and self-employed which finances public pensions. Reform 3 is a proportional reduction in pensions in payment. In each case we wish to consider the budgetary cost of the reform as well as the distributional consequences. In order to be able to compare across


37

countries, we measure the cost of the reform in terms of the proportional change in aggregate household disposable income. This is not equivalent to the proportional change in the public budget. When comparing different schemes within countries, one can interpret the change in aggregate household income as being equivalent to the fiscal cost. However, when comparing across countries account must be taken of differing aggregate levels of income and different relative sizes of the tax and benefit systems. Changes in poverty are assessed against the poverty line that corresponds to 60% of median equivalised disposable income in the baseline (pre-reform) scenario. The poverty line does not shift with changes in median incomes.

3.1. Reform 1: A Minimum Pension In recent years, several European governments have introduced or strengthened existing minimum pension schemes. These schemes may be particularly useful in periods of pension reform, since they strengthen social solidarity and make the reforms more easily acceptable from a political point of view.13 The evidence of Fig. 1, where the elderly are grouped by the decile of the equivalised income distribution to which they belong, shows that in all four countries persons aged above 64 are overrepresented in the bottom half of the income distribution – more so in the case of Denmark and, to a lesser extent, the UK. Since, according to the evidence of Tables 1 and 2, the elderly rely primarily on pensions, the introduction or the strengthening of existing minimum pension schemes like those simulated below are likely to be quite effective in reducing elderly poverty. In this reform scenario, the design of the scheme is as follows. The minimum is set at a common proportion of current national average earnings (20%, 30%, 40% and 50%). In the central scenario that will be analysed in detail in order to understand the mechanics of the reform, the proportion is set at 40%. It is designed as an individual pension entitlement, guaranteeing that everyone aged 65 or more who has existing public pension rights will receive at least this level of pension income. Any resulting increase in pension is treated in the same way by the rest of the tax and benefit system as existing contributory pensions. Thus, the gain from the addition of the minimum pension may be withdrawn to some extent due to income taxation, contribution payments or reduction of income-related benefits such as housing benefits. It is intended to be financed out of social contributions and so is only available to those who have made their own contributions. It does not guarantee a pension to all citizens over the age of 65 and does not

38

DANIELA MANTOVANI ET AL. 40% 35% Denmark Germany Italy UK

percentage of elderly

30% 25% 20% 15% 10% 5% 0% 1

2

3

4

5

6

7

8

9

10

equivalised household income decile

Fig. 1.

Position of People Aged 65+ (‘‘Elderly’’) in the Distribution of Household Equivalised Disposable Income. Source: EUROMOD.

improve the situation of those with no contributory rights.14 The amount of the payment is independent of the marital status of the pensioner and of the existence of any dependants. It is not income-tested, except against other pension income. This is defined as follows, for those aged 65+: Denmark: all non-contributory and ATP pension income. Germany: all contributory pensions and survivors’ benefits. Italy: all earnings-related pensions and survivors’ benefits and all existing contributory minimum pension payments. UK: basic state pension, widow’s pension and SERPs and occupational ‘‘defined benefit’’ pensions. As explained above, occupational (private) pensions are included for the UK because for many people these replace rather than supplement the state earnings-related scheme. Omitting them from the eligibility test for the minimum pension, combined with the low level of the basic state pension, would result in very large proportions of pensioners – including some of those with the highest pension incomes – benefiting from the minimum.


39

The proportions of individuals aged 65+ in receipt of their own state pension income varies across the four countries (see Table 2). The treatment of married women without pension rights based on their own contributions also differs. In most countries, pension payments are independent of marital status (although in all contributory systems widows and widowers inherit some part of their spouse’s contributory pension; in Denmark a widow/er has her/his own old age pension rights based on citizenship). In the UK, couples receive a higher basic pension than single people even if the pension is only based on one set of contributions. It is paid partly to the ‘‘dependent’’ spouse. This makes it difficult to determine which individuals should have rights to the minimum under this scenario. To aid comparability across countries we assume in a second variant that UK women in receipt of a state pension at only the dependant’s rate do not qualify for their own minimum pension.15 In all four countries, the great majority of the elderly live either alone or with their spouses. It should be noted that the implicit equivalence scale used here in the case of couples with two low-pension spouses is higher than the equivalence scale used for the purposes of poverty analysis. Thus, ceteris paribus, our scheme is likely to be more effective in fighting poverty among elderly (two-pension) couples than among single elderly persons. However, it is less likely to be effective for couples with only one pension entitlement between them. Table 4 shows the level of the minimum pension corresponding to 40% of national average earnings before tax or employee contributions, and the aggregate effects of implementing it. First, in terms of nominal 1998 euro it is clear that the level of the minimum varies greatly. Adjusting for purchasing power differences narrows the range but does not eliminate it. Under variant 1 incomes as a whole rise by 4.2% in the UK compared with 2.9% in Germany and 1.4% in both Italy and Denmark. The second UK variant restricts eligibility for the minimum to those receiving a state pension based on their own contributions or those of a deceased spouse. This makes a large difference to the cost, reducing it to a 2.2% increase in disposable income, lower than the aggregate change in income for Germany. Further discussion concentrates on this more restrictive version of the scheme (although it should be noted that this treatment is not strictly comparable with that in the other three countries). In particular the inclusion of non-state pension income in other countries would reduce the cost to some extent. This is likely to be significant only in Germany where 21% of elderly receive some amount of this form of income (Table 2). However, it is interesting to note that there is very little difference in the impact on poverty

40


Table 4. Effects of a Minimum Pension Set at 40% Average Gross Earnings.

Minimum pension level: h/year PPS/yearb

Denmark

Germany

Italy

UKa

13,208 10,678

11,668 10,978

7,056 8,005

10,675 9,909

% Change Household disposable income Benefits and state pensions Social contributions Taxes Poverty rate (age 65+) % Affected (aged 65+)

Variant 1

Variant 2

1.4 7.5 0.0 1.8 63.5

2.9 9.9 1.6 0.1 69.1

1.4 5.9 0.0 1.7 54.4

4.2 22.2 0.0 3.2 96.5

2.2 11.7 0.0 2.0 93.8

82.9

47.5

42.3

79.9

58.1

Source: EUROMOD. a Variant 1 allows married women to qualify on their own account based on pension income arising from their husbands’ contributions; variant 2 removes eligibility on the basis of ‘‘dependent’’ pensions. b PPP conversion factors calculated by Eurostat for 1998 and provided in Dennis and Guio (2003).

in the UK using either of the variants, in spite of the large difference in cost. In fact, both variants are sufficient to virtually eliminate elderly poverty in the UK. The main effect on incomes is directly through the increase in pension income although in each country taxes (and in Germany, contributions) are levied on pensions making the net aggregate effect lower. In each country income from minimum income schemes is replaced to some extent by the minimum pension. The effects are small in Denmark, Germany and Italy where the cost of the gross minimum pension is 2–4% higher before taking account of withdrawal of social assistance, housing benefits and other minimum income schemes. In the UK the effect is much larger: 29% of the gross cost is compensated by a reduction in the cost of pensioner means-tested benefits. The impact on poverty rates for the elderly is large with reductions of at least half in all countries. This leaves just 4.3% of elderly in poverty in Germany and 8.2% in Italy. The reduction of 64% (or 19.7 percentage points) in Denmark from a high starting point of 31% (Table 3) leaves 11.4% of Danish elderly below the poverty line. The UK case is the most


41

dramatic – all but 1.4% of elderly are taken out of poverty by the minimum pension under variant 2, a reduction of 94% (under variant 1 the poverty rate falls from 97% to 0.8%). These effects depend partly on the size of the gains: how far existing pension incomes are below the new minimum level, and partly on the position of those with low pensions in the household income distribution. These are considered in turn. Fig. 2 shows that the distribution of the size of the net benefit varies considerably across countries. In Denmark a very large proportion of the population aged 65+ are net gainers (83%). However, as many as 34% are gaining less than 10% of their (current) income. About the same proportion in the UK as in Denmark (47% and 49%) qualify for amounts that lead to gains of more than this. The share of those aged over 64 who experience a very large increase in their incomes (over 60%) is considerably higher in the UK than in Denmark – 13% compared with 3%. Although the cost of this reform scenario is of a comparable relative magnitude in Italy and Denmark, in Italy the beneficiaries of the reform among those aged over 64 are in the minority with 37% gaining between 100% 90% 80%

% of elderly

70% 60% 50% 40% 30% 20% 10% 0% Denmark

Germany

Italy

United Kingdom

gain 10-30% gain > 60% gain < 10% gain 30-60% no change

Fig. 2. Increases in Household Disposable Income of People Aged 65+ Due to a Minimum Pension of 40% of Average Earnings. Source: EUROMOD.

42


10% and 60% of their income (with few gaining more than this) and 4% less than 10%. In Denmark, more people gain smaller amounts (34% benefit by less than 10%) and a small minority (3%) gain very large amounts. The German beneficiaries are divided quite evenly between those gaining a lot and those gaining relatively little. Germany has the highest proportion of those who gain more than 60% (18%). Fig. 3 shows the average proportional gain in disposable income among all households across the income distribution. The distributional effect is partly a consequence of the position in the income distribution of pension recipients. For example, as shown in Fig. 1, in Denmark pensioners are likely to be concentrated at the bottom of the income distribution. Therefore, the concentration of gains from the minimum pension at these levels of income shown in Fig. 3 is not surprising. In the other three countries pensioners are spread more evenly by income level: UK pensioners are to be found disproportionately in the lower–middle parts of distribution (deciles 3 and 4) and the number of Italian pensioners in the bottom decile group is disproportionately small.16

percentage change in household income

14.0% Denmark Germany Italy UK

12.0%

10.0%

8.0%

6.0%

4.0%

2.0%

0.0% 1

2

3

4

5

6

7

8

9

10


Fig. 3.

Proportional Changes in Income by Decile Due to a Minimum Pension Equal to 40% of Average Earnings. Source: EUROMOD.


43

However, the average gains for households located close to the bottom of the income distribution are even larger in relative terms in Germany than in Denmark. This may be an indication that even though many Danish pensioners are concentrated at the bottom decile, their pensions are not extremely low, at least in comparison with their German counterparts at this position in the income distribution. The picture in Italy is less pro-poor since the beneficiaries of the reform are more evenly spread across the income distribution, in comparison with the other three countries. In order to explore further the relationship between cost and effectiveness in poverty reduction among pensioners, other levels of minimum pension are also simulated. Table 5 shows the level in terms of proportions of average earnings (20%, 30%, 40% and 50%) in euro per year, not adjusted for purchasing power, our measure of net cost that can be compared across countries (percentage increase in household disposable income) and the proportion of elderly people in households which gain. The effect of a low level of minimum pension (20% of average earnings) is very small in Italy and Denmark. Existing minimum pensions are higher in value than this minimum (see Table 2) and the only beneficiaries are those not eligible for these existing Table 5. Effects of a Minimum Pension Scheme, by Level of the Minimum. % of Average Earnings 20 Minimum pension level (h/year) Denmark 6,604 Germany 5,834 Italy 3,528 UK 5,337 % Increase in household disposable income Denmark 0.02 Germany 0.45 Italy 0.06 UK 0.08 % People aged 65+ that gain Denmark 3.7 Germany 14.3 Italy 1.5 UK 12.7 Source: EUROMOD.

30

40

50

9,906 8,751 5,292 8,006

13,208 11,668 7,056 10,675

16,511 14,586 8,820 13,344

0.35 1.30 0.31 0.82

1.40 2.87 1.38 2.16

2.98 5.24 2.36 3.94

49.2 30.2 31.4 43.3

82.9 47.4 42.3 58.0

84.9 66.3 47.7 63.9

44


minima (but in receipt of pension income). There is a more substantial effect in Germany and the UK: 14% and 13% respectively of elderly are beneficiaries. Generally, the relationship between the level of the minimum and its cost is very similar in Denmark and Italy: the cost is relatively low at low levels of the minimum because few people qualify for small amounts. In the Danish case large numbers of pensioners become entitled to relatively small amounts as the level of the minimum rises. Proportionately fewer pensioners benefit in Italy, but those who do, on average, gain more. In Germany, implementation of the minimum pension is more costly than in the other countries at all levels of the minimum; the cost rises in line with the level of the minimum, due to the (mainly) proportional nature of German contributory pensions. In the UK, the numbers entitled rise steeply between the 20% and 30% levels due to the concentration of pensioners on similar levels of income. This is rather like in Denmark, but in the UK case the numbers entitled at high levels of the minimum are not so large, due to more widespread and developed private occupational pensions. Thus at high levels of the minimum the proportions entitled are more like the proportions in Germany, but the cost is lower. Table 6 provides detailed estimates of the effects of the minimum pension on elderly poverty. Clearly, in all countries the larger the minimum pension the larger the reduction in the elderly poverty rate. However, the relationships are not equally strong in all countries. In Denmark, the minimum pension is very effective at reducing elderly poverty from its high starting rate. At the 40% level nearly all elderly people who could benefit are in Table 6.

Effects of the Introduction of Minimum Pension on Poverty Among the Elderly.

Index

Scenario

Poverty rate (65+)

Baseline 20% 30% 40% 50%

FGT (a ¼ 2) (65+)

Baseline 20% 30% 40% 50%

Source: EUROMOD.

Denmark 31.0 31.0 23.1 11.4 1.7 0.483 0.481 0.324 0.177 0.121

Germany 13.9 12.9 9.6 4.3 3.7 0.978 0.656 0.369 0.344 0.334

Italy 18.0 17.9 16.9 8.2 7.6 1.189 1.181 1.003 0.624 0.491

UK 22.6 22.1 14.6 1.4 1.2 0.730 0.675 0.125 0.065 0.062


45

receipt and two thirds of those below the poverty line before the reform are now above it. But due to the relatively high level of the Danish poverty line, it takes a minimum pension of 50% of earnings to raise (nearly all) the final third above the poverty line. In Italy, only higher levels of minimum pension are effective at reducing elderly poverty. There are two reasons for this. First, a relatively low proportion of Italian elderly qualify, leaving 8% below the poverty line even when the minimum pension is at the highest level considered. Second, it is only at 40% of earnings that the minimum pension is above the Italian poverty line. In Germany and the UK the minimum pension level is above the poverty line (for a single person) at the 30% level. In Germany the returns, in terms of poverty rate reduction, are greater for higher levels of minimum pension up to the 40% level. There is little improvement in poverty rate from increasing the level of the minimum pension to 50%, even though the proportion gaining continues to rise. Similar ‘‘diminishing returns’’ set in for the poverty rate in the UK at 50%, and the numbers benefiting level off too. The objective of minimum pension schemes is not only to reduce the poverty rate of the elderly but, also, to improve the living standards of those who remain below the poverty line after the transfer. Nevertheless, this type of exercise is sometimes used in order to calculate the ‘‘cost per person taken out of poverty’’ at each level of the minimum pension. Note that since resources are shared at the household level, some of the persons taken out of poverty as a result of the minimum pension would be non-elderly, although here we focus on the elderly alone. In all countries the relationship between cost per elderly person taken out of poverty and the level of minimum pension is U-shaped. This is because at low levels of minimum pension very few people are moving above the poverty line, while at high levels many recipients are already above the poverty line. For example, in Denmark, the annual cost per elderly person taken out of poverty when the minimum pension is set at 20% of average earnings is 28,716 euro. It declines sharply to 4,202 euro when the minimum pension is set at 30% of average earnings and, then, rises to 6,801 and 9,694 euro as the minimum pension is successively set at 40% and 50% of average earnings.17 Minimum pensions, especially when set at levels equal to, or higher than, 40% of average earnings, appear to reduce the elderly poverty rate substantially, particularly in Denmark and the UK, but at significant cost per elderly person taken out of poverty, especially in Germany. On the other hand, the intensity of poverty using FGT(a ¼ 2) is reduced significantly in Germany with quite low levels of minimum (at 20% of average earnings it is

46


% increase in household disposable income

6

5 Germany 4

3

40% Denmark

40%

2 Italy

40%

40

50

40%

UK

1

0 0

10

20

30

60

70

80

90

100

% reduction in poverty (FGT(a=2))

Fig. 4. Reduction in Poverty Among the Elderly by Levels of Generosity of the Minimum Pension. Source: EUROMOD. Note that Successive points on each line indicate minimum pension levels equal to 20%, 30%, 40% and 50% of average earnings, respectively.

reduced by nearly one-third). Fig. 4 shows the returns to higher levels of minimum pension generosity in terms of reductions in poverty among the elderly as measured by the FGT(a ¼ 2); that is, taking into account the impact of the simulated minimum pension schemes on the intensity, as well as extent of poverty among the elderly. This clearly shows that in Germany and the UK there are levels of minimum pension above which there is rather little improvement. For Germany this corresponds to a minimum pension of about 30% of average earnings and it appears that this policy cannot reduce poverty as measured by FGT(a ¼ 2) to less than about a third of its baseline level. This residual must be explained by elderly not covered by pension entitlements and hence, not entitled to the minimum. Table 6 shows that higher levels of minimum pension do result in reduced poverty rates, indicating that a minimum pension of around 40% is needed to raise some remaining pensioners from just below to just above the poverty line. A similar picture is apparent in the UK, although a higher level of minimum (40%) is necessary to secure large reductions in poverty intensity. Poverty is then almost eradicated, indicating that while the level of existing pension is low for many UK elderly, coverage is relatively comprehensive.


47

The small residual poverty rate in Denmark at the 50% level of minimum pension corresponds to the low-income elderly aged 65–66 who were not qualifying for pension in 1998. The value of the FGT(a ¼ 2) remains at 25% of its baseline level, even with the high minimum pension in place is a result of, on the one hand, the low level of baseline intensity of poverty in Denmark and, on the other, the fact that those not qualifying for pension due to being aged under 67 may include a minority with very low incomes. Fig. 4 also shows how, in all four countries, the fiscal cost of the introduction of minimum pensions appears to rise at an increasing rate with the generosity of the minimum pension. The following two sections consider two mechanisms by which this may be financed; the first by increasing the fiscal burden on those currently working, and the second by redistributing incomes within the pensioner populations. 3.2. Reform 2: Increasing Contributions We explore the impact of increasing contribution rates on the amount of revenue raised and on poverty by simulating increases of 1, 2 and 3 percentage points in all rates of pension contribution.18 In Germany this is straightforwardly done: in first case the employee’s rate is increased from 10.15% to 11.15% in both East and West Germany. In Denmark the rate of tax that finances the old age pension is increased from 8% to 9%. In the UK, each of the Class 1 rates of contribution (for those contracted out and not contracted out of the state earnings-related pension) are increased by 1 percentage point (from 8.4% to 9.4% and from 10% to 11%, respectively); the rate on earnings below the lower threshold in 1998 is also increased (from 2% to 3%); and the rate on self-employed profits is increased from 6% to 7%. In Italy, there are many rates for people in different forms of employment and self-employment: these are all increased by 1 percentage point. We also take account of the fact that in contributions paid on a proportional basis (within limits) in some contributory systems, there also are fixed or flat contributions. In these cases, we increase the flat payment in proportion to the change effected in the main simulation. Such increases also apply in Denmark and the UK. Table 7 shows the aggregate effect of the three levels of increase in contribution rates, in terms of the proportional change in household disposable income. Within countries, larger increases in marginal rates result in aggregate effects that are proportionately larger. The effects differ across countries with the net effect of a 1-percentage point increase in contribution varying from a 1.2% drop in income in Denmark to 0.5% in Italy.19

48


Table 7.

Aggregate Effects of Increased Contribution Rates.

Index

Scenario

% Change in household disposable income SIC+1% SIC+2% SIC+3%

Denmark Germany 1.22 2.44 3.67

0.73 1.45 2.18

Italy

UK

0.47 0.59 0.94 1.17 1.40 1.76

Source: EUROMOD.

This can be explained by cross-country differences in the structure of the systems of pension contribution: a common percentage point increase in rate is of differing significance across systems. In Germany a 1-percentage point increase in pension contribution represents a relatively small proportional increase in contribution income as a whole (4.4%) as there are other parallel systems of contribution for other purposes (unemployment, etc). In Italy, the proportional increase in contribution income is twice as high as in Germany (8.8%), reflecting the greater relative importance of pensions in the contributory system and employee contribution rates that are on average lower than in Germany (8–9% in the most common scheme). The proportional increase is larger in the UK (10.7%) because the starting average rate is relatively low. In Denmark, most of the pension system is financed by part of the income tax system. A 1- percentage point increase in contributions represents a 35% proportional increase in existing contributions, due to the very small starting rate. The effects of the increase in contributions according to position in the distribution of household income are shown in Fig. 5. Since, as shown in Table 7, the aggregate effect of the change is not equivalent across countries, the proportional change in income is shown as the ratio of the change in each decile group to that of the population as a whole. This normalisation allows us to focus on differences in distributional effects, rather than differences in the overall size of the impact across countries. These effects are quite similar in Germany and the UK with the proportional reduction in household income increasing fairly steeply with income up to the top one (UK) or two (Germany) decile groups. Ceilings on contributions in both systems, as well as in Italy, make the contribution system regressive at higher income levels. This, combined with tax relief on contributions in Germany, and the fact that some high earning Germans opt out of the system altogether, make the proportional reduction in income less for the top 10% than for the rest of the top 50% (Germany) or 40% (UK). The Danish system is only slightly regressive at the very top, but less progressive over the bottom half of the distribution than the German or UK


49

ratio of % fall in income to mean % fall in income

1.4 1.2 1.0 0.8 0.6 Denmark Germany Italy UK

0.4 0.2 0.0 1

2

3

4

5

6

7

8

9

10


Fig. 5. Proportional Reduction in Income by Decile Group Relative to the Mean Following an Increase in Social Insurance Contribution Rates by 1-Percentage Point. Source: EUROMOD. Note that for each decile group the proportional reduction in income is shown as the ratio to the mean proportional reduction.

systems. The Italian system is much less differentiated by income level, although this may have as much to do with household composition effects as with the contribution burden by levels of gross earnings.20 In common with Germany and the UK the proportional effect is less in the top decile group in Italy than in the rest of the top half of the distribution. Given that the increases in contribution rates affect almost exclusively the non-elderly, they are likely to leave the poverty situation of the elderly largely unchanged since in our analysis the poverty line is kept fixed. At the same time, to the extent that at least some contributors are in households near or below the poverty line, it might be expected that increasing contributions might exacerbate poverty in general. Table 8 shows the percentage point increase in poverty rates for the national populations following each of the three increases in contribution rate. While the effects are generally small, they are largest in Italy, where (as shown in Table 2, above) nearly all employed persons pay pension contributions. Only 2% of Italian employees in the bottom quintile group by disposable income pay no

50


Table 8.

Effects on Poverty Rates of Increasing Contribution Rates.

Index

Scenario

Denmark

Germany

Italy

UK

Poverty rate (all)

Baseline

11.4

9.9

20.1

19.9

Increase in poverty rate (percentage points) SIC +1% 0.16 SIC +2% 0.24 SIC +3% 0.32

0.04 0.22 0.32

0.23 0.45 0.61

0.03 0.12 0.22

Source: EUROMOD.

contribution. In Germany and the UK low earnings are exempt from contributions: 29% and 39% respectively of employees in the bottom quintile group pay no contributions. We can conclude that raising revenue to finance pensions through higher contributions is most problematic from a distributional point of view in the Italian system. Italian contributors are more likely to have low incomes and a significant share of the extra burden is borne by them.

3.3. Reform 3: Reducing the Replacement Rate in Contributory Pensions in Payment This reform scenario focuses exclusively on the incomes of the pensioners. The purpose is to gauge the size of savings and the distributional effects on the elderly of a proportional reduction of pensions in payment (by 5%, 7% and 10%). Although in a short-term framework it is very unlikely that pensions can be cut in real terms like this – unless there is a severe financial crisis – the common practice of many governments has been to let the pensions lag behind incomes from other sources by linking pensions to the retail price index rather than average earnings.21 In this scenario, we reduce the earnings-related and flat rate contributory components of state pension systems. The main flat rate components are the basic pension in the UK and the old age citizen’s pension in Denmark. These are near-universal in coverage although the Danish scheme, operating from the age of 67 in 1998, excludes some 12% of the elderly because they were aged 65 and 66. The state earnings-related schemes cover 90% of those aged over 64 in Germany and 87% in Italy. Table 9 shows the aggregate effects of a 5% cut. The importance of the earnings-related components, and in particular the lack of a ceiling on payments (see Table 2) in Germany and Italy is shown by


Table 9.

51

Effects of Decreasing Contributory State Pension Payments (by 5%).

% Change Household disposable income Benefits and state pensions Social contributions Taxes

Denmark

Germany

Italy

UK

0.27 1.44 0.00 0.34

0.92 3.25 0.40 0.27

0.79 3.78 0.00 1.52

0.23 1.24 0.00 0.18

Source: EUROMOD.

the large aggregate reduction in benefits as a whole – more than 3% in both countries. The effect is mitigated by the income tax system, to a greater extent in Italy than Germany. In the former case, pensions are taxed as earnings, whereas in Germany they are taxed much less heavily.22 On the other hand, in Germany pension income itself attracts contributions, which fall along with pension income. In the UK and Denmark, state pensions make up a relatively small component of total benefit expenditure. While pensions are lowered by 5% this represents only a 1.2% and 1.4% decrease in total benefits respectively in the two countries. In the UK this is partly because the fall in contributory state pension is compensated by an increase in social assistance or housing benefits for many low-income pensioners. Increases in these benefits make up for about 30% of the fall in contributory pension. There is a similar effect in the other three countries but it is much smaller: 4% in Denmark and Germany and 3% in Italy. Fig. 6 shows the distributional effects in terms of the proportional reduction in disposable income across the household income distribution. In Denmark, the concentration of losses from the pension reduction at low levels of income corresponds with the location of pensioners at the bottom of the income distribution. In the UK, the rather flat distributional effect is due to the compensation of low-income pensioners by an increase in social assistance benefits. Another feature is the low level of UK state pensions in relation to other incomes: a proportional reduction is smaller in absolute size than the same reduction in Italy or Germany. However, it is clear that in all four countries a cut in the state contributory pension is a regressive measure due to the position of pensioners in the income distribution as a whole: proportional reductions in pensions affect disproportionately the lower decile groups. Table 10 reports the effects on poverty among the elderly of the proportional cuts in pensions. The poverty rate rises in all countries – most in

52

DANIELA MANTOVANI ET AL. 0.0%


-0.2% -0.4% -0.6% -0.8% -1.0% Denmark Germany Italy UK

-1.2% -1.4% -1.6% 1

2

3

4

5

6

7

8

9

10


Fig. 6. Proportional Changes in Income by Decile following a Proportional Decrease in Contributory Pension Payments by 5%. Source: EUROMOD.

Denmark and, to a lesser extent, Germany. Apparently, in these countries a considerable proportion of the elderly are located just above the poverty line. Poverty intensity increases too: many of the elderly who are below the poverty line and are not eligible for social assistance face a decline in their social transfers. Hence, the increase in the FGT(a ¼ 2). In fact, after the pension cuts, in all countries apart from Germany the proportional increases of the FGT(a ¼ 2) are larger than those of the poverty rate. While the revenues (as measured by the proportional reduction in disposable incomes) resulting from the cuts are quite small in the case of Denmark, the impact on the poverty measures is the largest, at each level of the cut. The proportional income reductions are larger for Germany and Italy, and the impact on poverty indictors is less, reflecting the earningsrelated nature of their pension systems. In the UK, both the aggregate income and the poverty effects are dampened by the role of income-tested benefits. It seems there is most scope for redistribution among pensioners in Germany and Italy but very little in Denmark.


53

Poverty and Aggregate Income Effects of Proportional Reductions in Pensions.

Table 10. Index

Scenario

Poverty rate (65+)

Baseline 5% 7% 10%

FGT (a ¼ 2) (65+)

Baseline 5% 7% 10%

0.483 0.653 0.739 0.894

0.978 1.065 1.107 1.186

1.189 1.334 1.396 1.491

0.730 0.855 0.906 0.985

5% 7% 10%

0.27% 0.38% 0.54%

0.92% 1.28% 1.83%

0.79% 1.11% 1.59%

0.23% 0.32% 0.46%

% Change in total disposable income

Denmark 31.0 36.8 39.2 42.2

Germany 13.9 16.7 17.0 18.6

Italy 18.0 18.9 19.7 20.7

UK 22.6 23.8 24.6 25.4

Source: EUROMOD.

4. REFORM PACKAGES We can combine the three elements of reform into revenue-neutral packages.23 Examination of the individual components above has shown that some types of reform are more cost-effective in some countries than in others. There are many dimensions to consider and in this first attempt we select, at the national level, combinations of components that maximise the reduction in the elderly poverty rate while being mindful of the tax burden of the working age population. Focus on alternatives such as the FGT(a ¼ 2) indicator for the elderly, gender-specific poverty indicators, poverty indicators for the whole population or on the relative incomes of elderly and nonelderly (or many other important outcomes) might lead us in different directions. The main purpose at this stage it to demonstrate the existence of a range of options and of the need to take national circumstance into account. The packages we consider all involve a minimum pension set at 40% of average earnings, together with: Denmark:

Percentage point increase in pension contributions (and earmarked taxes): 1.2 Change to existing pensions: none

54


The rationale for the Danish package is that there is little point in reducing existing pensions if they are not earnings related. The cost of the 40% minimum is borne by a large increase in taxes/contributions paid by people in employment. Percentage point increase in pension contributions: 3.0 Change to existing pensions: 5% reduction The German package combines a significant increase in pension contributions with a modest decline in existing pensions. The increase in contributions is large because the base is relatively small (23% of the employed non-elderly do not pay state pensions contributions), while, as shown below, inequality among the pensioners is relatively high, thus leaving room for a moderate redistribution among them. Germany:

Change in pension contributions: none Change to existing pensions: 11% reduction We have seen that increasing Italian contributions has some impact on poverty. There is already significant income inequality among pensioners. So the package involves redistribution within pensioner incomes – a large proportional cut in existing pensions. Italy:

UK:

Percentage point increase in pension contributions: 3.3 Change to existing pensions: 10% reduction

As in the case of Germany, the UK package combines all elements. Both the cut in state pensions and the increase in contribution rate are relatively large because in each case the base is relatively small. Under these packages, the cost of the reform is borne exclusively by the working age population in Denmark and by pensioners themselves in Italy. In Germany and the UK the cost is shared between workers and pensioners. Pensioners bear 24% of the total cost of the reform package in Germany, while the corresponding figure in the UK is 10%. The distributional effects of the four packages are shown in Table 11. First, and not surprisingly, the ratio of household income of the elderly rises when compared with incomes of other people. This effect is smallest in Italy (the effect is only due to the way redistribution occurs across households containing elderly people). The effect is particularly large in Germany, in spite of the large cut in earnings-related pensions. Poverty among the elderly falls considerably in all countries and the net effect of the packages on poverty in general is to clearly reduce it. The progressive effect across the distribution of all household incomes is illustrated in Fig. 7. In UK and


Table 11.

55

Distributional Effects of the Reform Packages.

Index

Scenario

Denmark

Income ratio

Baseline Reform Change

Poverty rate (all)


11.4 8.8 2.6

9.9 8.6 1.3

20.1 18.5 1.7

19.9 16.8 3.1

Poverty rate (65+)


31.0 11.4 19.6

13.9 4.8 9.1

18.0 8.6 9.4

22.6 1.5 21.1

FGT (a ¼ 2) (all)


0.973 0.941 0.032

0.673 0.565 0.108

3.131 3.006 0.125

1.107 1.012 0.095

FGT (a ¼ 2) (65+)


0.483 0.178 0.305

0.978 0.345 0.633

1.189 0.658 0.531

0.730 0.070 0.660

Gini coefficient (65+)


0.207 0.160 0.047

0.340 0.230 0.110

0.439 0.396 0.043

0.330 0.290 0.040

0.67 0.76 0.09

Germany

Italy

0.92 1.03 0.11

1.00 1.03 0.02

UK 0.81 0.90 0.09

Source: EUROMOD.

Germany decile groups 1–5 are all net gainers. In Denmark this is the case for decile groups 1–3, and for Italy groups 1–4. Another way to look at the distributional impact of the simulated changes but focusing exclusively on the elderly is through the examination of the frequency distributions of their incomes before and after the reforms. The evidence is reported in Fig. 8. Naturally, in all countries the distribution shifts to the right, fewer elderly persons are left below the poverty line and in all countries (with the partial exception of Italy) a considerable proportion of the elderly are concentrated around the new minimum pension. However, cross-country differences are striking. In Denmark, which has the most concentrated income distribution among the elderly, after the reform a twin-peak distribution emerges. The elderly making up the left peak are persons aged over 64 but below state retirement age (67). The elderly around the right peak are elderly persons who qualify for the new minimum pension simulated in our reform, which is more generous than the minimum pension of the baseline scenario. Italy has the least concentrated income

56

DANIELA MANTOVANI ET AL. 14.0%


12.0% Denmark Germany Italy UK

10.0% 8.0% 6.0% 4.0% 2.0% 0.0% −2.0% −4.0% 1

2

3

4

5

6

7

8

9

10


Fig. 7.

Proportional Changes in Household Income by Decile due to Reform Packages. Source: EUROMOD.

distribution of the elderly among the countries examined here; a feature hardly affected after the simulation of our reforms. Compared with the other three countries considered a relatively large number of the Italian elderly do not qualify for a pension, and at the same time a relatively high proportion share households with non-elderly – see below for indirect evidence – the effect of the simulated reform in Italy is not as effective in reducing elderly poverty as in the other countries. In the UK, the frequency distribution of the elderly has two peaks in the baseline scenario. In fact, the first of these peaks lies below the poverty line. After the reform, poverty in Britain is almost eradicated with a single peak above the poverty line, indicating the high proportion of British concentrated around the level of the simulated minimum pension. Perhaps the most interesting post-reform picture emerges in Germany. Few elderly are left below the poverty line after the reform, while the reform produces a twin-peak distribution of the elderly, with both peaks clearly lying to the right of the poverty line. This is due to the fact that both members of many German low-income couples benefit from the minimum pension – these make up the peak on the right. Couples

Baseline Reform

poverty line

20

25 % of elderly

% of elderly

25

Germany 30

15 10

20 15 10 5

5

0

0

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 Proportion of median household disposable income

Proportion of median household disposable income

Italy 30

poverty line

UK 30

Baseline Reform

20

25 % of elderly

% of elderly

25

Baseline Reform

poverty line

15 10 5

poverty line


Denmark 30

Baseline Reform

20 15 10 5

0

0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 Proportion of median household disposable income

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 Proportion of median household disposable income

57

Fig. 8. Distribution of the Elderly by Level of Household Income Before and After the Reform Packages. Source: EUROMOD. Note that incomes are measured in relation to the pre-reform median. The upper tails of the distributions have been truncated at three times median income. The proportions of the elderly not shown in each country are 0.05% (Denmark), 0.39% (Germany), 3.98% (Italy) and 1.45% (UK).

58

DANIELA MANTOVANI ET AL. 100% 90% 80%

% of people

70% 60% 50% 40% 30% 20% 10% 0% DK 65+ DE:65+ IT:65+ UK:65+ gain > 60% gain 30-60%

Fig. 9.

gain 10-30% gain < 10%

DK:all DE:all IT:all UK:all no change ( 10%

Gainers and Losers from Pension Reform Packages. Source: EUROMOD.

with one beneficiary and single beneficiaries are concentrated in the lowerincome peak. Fig. 9 shows the proportions gaining and losing different amounts, first for the elderly and then, on the right hand of the chart, for the populations as a whole. Denmark shows large proportions of elderly gaining both large and small amounts, paid for exclusively by relatively small proportional reductions in income among the working age population. In contrast, in the other three countries there are some quite large reductions in income. These are mainly confined to the elderly. As shown in Table 11, poverty among the elderly is all but eradicated in the UK. The German poverty rate for the elderly also falls to a very low level, and the FGT(2) indicator declines by around the same proportion. While still substantial, the fall in poverty rate among the elderly in Italy is less dramatic than in the other three countries. This is partly because a residual group of Italian low-income elderly do not qualify for public pensions and are not affected by the reform package. As shown in Table 2 above, the pension


59

Table 12. Proportion of People Aged Under 50 whose Household Income Comprises At Least Half Pension Income From an Elderly Person. % All In bottom quintile group

Denmark

Germany

Italy

UK

0.2 0.3

1.6 3.3

8.0 12.1

1.6 2.0

Source: EUROMOD.

coverage rate is lowest in Italy among the four countries considered (81% compared with 88%, 90% and 98% in Denmark, Germany and UK, respectively). This is part of the explanation for the persistence of poverty both in terms of rate and severity in spite of the generous minimum pension. Another part of the explanation may lie in the extra burden of other household members on pension income. Table 12 shows how pension income is the major source of household income for a much larger proportion of younger Italians (aged under 50), than it is for the same age group in other countries. The difference is particularly marked in the bottom quintile groups where 12% of Italians depend on pension income of elderly persons in the same household, compared with 3% in Germany and 2% in the UK. It is simply that larger increases in pension income are needed to push all Italian households containing the elderly across the poverty line.24 Finally, we see from the reduction in the Gini coefficient – shown in Table 11 above – that the reforms bring about a considerable reduction in inequality among the elderly. Again, this is to be expected and is especially large in the German case. Fig. 9 shows particularly large average gains at the bottom of the German income distribution. About 18% of German elderly increase their incomes by more than 60%, whereas the corresponding proportions are 12% in the UK and only 3% and 1% in Denmark and Italy, respectively.

5. CONCLUSIONS We have compared four pension systems and analysed their characteristics by discussing a number of illustrative EUROMOD simulations with the aim of highlighting cross-country differences and similarities and suggesting nationally appropriate directions for reform. In doing so, our approach has been entirely normative and takes little account of political economy considerations or indeed, pragmatic political realities. Reform may be

60


constrained in ways that we have not considered, although the numbers of gainers and losers shown in Fig. 9 do indicate something of the relevant ‘‘political arithmetic.’’ Whatever reform strategy is adopted in practice may well be sub-optimal in terms of the normative framework that has been adopted in this paper. While our analysis probably cannot be used as a blueprint for realistic reform proposals, it can inform their development. As anticipated, simply applying proportional cuts to state contributory pensions is a regressive measure due to the position of pensioners in the income distribution in all four countries: proportional reductions in pensions affect disproportionately the lower deciles of the distribution. As a result, additional measures are needed to ensure that reducing the level of pensions does not result in increasing pensioner poverty. So we have placed particular emphasis on the consideration of a minimum pension scheme as a central component of a reform strategy. The relatively generous level of minimum pension set at 40% of average earnings makes a big difference to low-income pensioners in Germany. Given the relatively high inequality among pensioners in this country, we found that the scheme could – within our normative framework – be financed at least in part by redistribution within the pensioner population. The minimum is of benefit to the vast majority of pensioners in Denmark where the numbers just below the poverty line are very large. At the core of the Danish pension system there is a residence-based flat amount, which makes the system virtually comprehensive (for those aged 67+) but the pension is set at a low level, just around the poverty line. The minimum pension scheme would bring a great many pensioners above it, reducing the elderly poverty rate by almost 20-percentage points. The concentration of pensions around the basic amount leaves very little room for intra-pensioner redistribution. So the scheme must be financed by increasing the tax burden of the non-elderly. The situation in the UK is similar, where the provision of a minimum pension of 40% of earnings virtually eradicates poverty in old age. However, in this case there is some scope for financing this by reducing existing UK earnings-related pensions. The effectiveness of the minimum in reducing poverty in Denmark and the UK is due to the already near-comprehensive nature of the basic pension schemes in both countries. In Italy there are significant gaps in public pension provision, which are duplicated in the coverage of the minimum pension that we have simulated and make it, in some ways, less effective in this country than in the others. The latter might be exacerbated by the fact that in Italy a relatively high proportion of non-elderly persons live with, and depend on, the pensions received by elderly household members.


61

Breaking the link with past contributions, or finding ways of crediting non-contributors, would improve the distributional effects of the minimum in Italy. This may also have a beneficial effect in countries such as the UK where – under the minimum scheme as we have simulated it – many couples only receive one entitlement to a minimum payment because they have only one contributory pension between them. If each member of the couple had their own minimum entitlement it would be possible for the level to be lower while still protecting all elderly – singles and couples – from falling below the poverty line. This issue is relevant, to some extent, in each country examined and is particularly evident in the German case, as shown in Fig. 8. More generally, there is an important aspect of pensions and pension reform that is missing from this paper: an analysis by gender. It is well known that women tend to have lower pension entitlements than men because of (on average) lower earnings and less of their lives spent in paid work, and that they are more likely to spend longer in poverty in old age.25 – It is likely that the types of reform packages that we have considered would be to the net benefit of women, at least in the three countries that depend on contributory systems. Nevertheless, further analysis – which takes account of dependency and survivorship – is needed to establish the extent to which this is so. Our analysis refers to 1998. Since then, the pension systems in a number of EU member states have been reformed to varying degrees – in an effort to meet the Laeken 2001 objectives – sometimes along the lines outlined in our simulations. Of the four countries examined in the paper, this is the case for Germany and, to a lesser extent, Italy. In most cases, the legislated or announced reforms will take a rather long time until they mature and will have more affect on the living standards of future rather than current pensioners. The results of the paper illustrate the likely distributional consequences of such reforms. The major challenge identified by most commentators is the growing pressure on the public finances due to pensions. So a critical issue is the cost of the reform in relation to its beneficial effects. We have explored packages that revenue neutral in the short run as a way of illustrating the different trade-offs across systems and countries. Naturally, if in the future the dependency ratio is higher or private pension income has grown, the conditions will be different and revenue neutrality may not hold. The approach we have taken – exploring reform components individually and in combination – allows us to anticipate the effects of packages that are designed to raise revenue or release resources. For example, we can see from Table 5 that setting the minimum pension at 30% rather than 40% of average

62


earnings would release revenue corresponding to about 1% of average disposable income (more in Germany and the UK). Poverty reduction measured using FGT(2) would still be considerable and the fall in the poverty rate for the elderly still significant. More generally, we have seen that the cost is relatively low if (a) the shortfall between the existing pension and the minimum level is small (Denmark, UK); (b) the numbers qualifying are small (Italy, Germany) or (c) the minimum pension replaces existing meanstested pensioner benefits (UK). However, in assessing the balance between the fiscal costs and anti-poverty impacts it is important to acknowledge that our comparisons do not capture some important effects that may differ in significance across countries. For example, when increasing contribution rates we take no account of any effects on employment of increasing non-wage labour costs in this way. Likewise, we assumed that the proposed reforms do not affect the savings behaviour of the population. More generally, we have assumed that there is no ‘‘leakage’’ of revenue from increases in contributions through evasion or the development of new avoidance mechanisms (Gillion et al., 2000; Manchester, 1999). Similarly, we have also assumed that the minimum pension could be administered in such a way as to achieve 100% take-up. This may be possible where existing systems are not fragmented and where a single authority possesses all the information necessary to determine entitlement automatically, but perhaps not in other cases. Administrative functions and capacities differ across countries, as do cultural norms in relation to benefit claiming and tax paying. These aspects matter in practice and could be important factors in influencing the relative impact of actual reforms in different countries. The main conclusion that can be drawn from this preliminary exercise is that the appropriate direction of reform is not the same across EU member states. Even if the long-term goals are common, the most effective pathways are different.

NOTES 1. See, for example, Feldstein and Siebert (2002) and the references cited there. 2. Although widely used, this categorisation is not always relevant in the case of pension systems. For similarities and differences in the pension systems of the four countries under consideration, see Appendix B. 3. See, for example, Kennell and Sheils (1990), Galler (1996), Morrison (2000), Emerson et al. (2004) and the special issues of Revue Economique (2000) and Economie et Prevision (2003) – particularly the paper of Dupont, Hagnere, and Touze


63

(2003). See also Nelissen (1995, 2000) who focuses on the effects of pension arrangements on the distribution of lifetime income. 4. In fact, some authors dispute the urgency of pension reform in OECD countries on economic grounds and claim that the reform agenda that is currently promoted by international organisations and implemented in a number of countries is primarily motivated by political considerations (Weller, 2004). 5. The ECHP is an ambitious effort at collecting information on the living standards of the households of the EU member-states using common definitions, information collection methods and editing procedures. It contains detailed information on incomes, socio-economic characteristics, housing amenities, consumer durables, social relations, employment conditions, health status, subjective evaluation of well-being, etc. For a detailed description see Eurostat (1996); for a critical appraisal see Peracchi (2002). 6. ECHP excludes people living in institutions such as old age homes but the extent to which this occurs among the elderly is not the same in all countries. 7. The ‘‘income ratio’’ compares equivalised household incomes across individuals. Following Eurostat’s methodology, throughout the paper we use the ‘‘modified OECD’’ equivalence scale, which assigns a weight of 1 to the household head, weights of 0.5 to every other adult in the household and 0.3 to each child (person aged below 14) in the household. 8. The estimates reported in Table 1 are not strictly comparable with the estimates of these studies, primarily because the latter’s concept of resources includes private incomes in-kind. 9. It is worth noting that alternative sources of micro-data may show other patterns. For example, the 2001 Danish Law Model database (European Commission, 2003; Table 2) gives an at risk of poverty rate of 8% for non-elderly adults – which compares well with the ECHP estimate for all non-elderly in Table 1. However the corresponding estimate for the elderly is 24%, much lower than the ECHP figure of 31%. 10. We set the poverty aversion parameter of the index at a ¼ 2 (FGT(2)). For this value, the index is sensitive to both the distance of the average poor person from the poverty line and the extent of inequality among the poor. Since, in absolute terms, the estimates of the FGT(2) are very small, for expositional purposes in the tables of the paper they are multiplied by 100. 11. Although increasing early retirement age is one of the most popular elements in recent pension reform efforts in EU countries, using EUROMOD it is difficult to model delayed retirement since its database does not containthe information necessary to model the counterfactual of later retirement To put things in perspective, there is no early retirement on state pensions in the UK, and in Denmark the great majority of the early retirees receive unemployment rather than pension benefits (see Appendix B). According to EUROMOD the proportion of pension recipients below official retirement ages (see Table 2) in Germany and Italy was 27% in both cases. However, this includes recipients of survivor benefits as well as early retirees and the share of the latter is likely to be below this figure. Our definition of the elderly as being 65 and over in this analysis also excludes a proportion of female pensioners who have an official retirement age of 60 in Italy and the UK. These proportions are 11% and 12% respectively of all pension recipients aged over 60.

64


12. Naturally, due to different reference years, data sources and the fact that taxes and benefits in EUROMOD are simulated, a number of figures in Table 3 differ from the corresponding estimates of Table 1. See Mantovani and Sutherland (2003) for a discussion. One important source of difference is the treatment of non-take-up of means-tested benefits and pensions by pensioners. For simplicity, EUROMOD estimates assume complete take-up whereas, at least for the UK, there is evidence of considerable non-take-up (Hancock & Barker, 2005). While under 1998 policies this would not make much difference to the poverty rate as measured using the proportion in households below 60% of the median, poverty rates using lower poverty lines, or measures that take account of intensity such as FGT(2) would be expected to be underestimated using EUROMOD. 13. For a description of the interaction between existing minimum income schemes with general social assistance schemes or other such schemes aimed specifically at the elderly in the countries under examination, see Appendix B. 14. See Atkinson, Bourguignon, O’Donoghue, Sutherland, and Utili (2002), Atkinson and Sutherland (1998) and Sutherland (1998) for illustrations of the effects of comprehensive minimum pension schemes that aim to cover all the elderly. 15. This is in fact somewhat arbitrary because some of these women might have made their own contributions but are better off claiming a pension as their husband’s dependant than in their own right. So the scheme most comparable to that in the other three countries is somewhere between the two variants. 16. The position of the elderly is similar to that of pensioners although, with the exception of Denmark, there is a tendency for the elderly to be more concentrated at lower income levels than those with pension income. This effect is not observed for Denmark because many of the 65–66 group (non-pensioner elderly) are still economically active. 17. The corresponding sets of figures for Germany, Italy and the UK are: 29,223, 19,431, 19,441 and 33,436; 19,481, 13,756, 6,997 and 11,250; and 11,735, 7,336, 7,276 and 13,106 euro, respectively. These estimates would be lower if account were also taken from the non-elderly. 18. In practice it is not unusual for an increase in pension contribution rates to be accompanied by an increase in ceilings, where such ceilings are in operation (all countries apart from Denmark in our case – see Appendix B). In order to keep things simple and focus on cross-country similarities and differences, we do not simulate changes in these ceilings along with the change in rates. 19. The effect is net of small changes in current benefit expenditure in all four countries (because some income tested benefits are based on income after contributions) and in income tax (except in the UK where contributions are not tax deductible). 20. The net effect of increasing contributions is the average of the effects for each individual within the household. 21. Our illustrative reduction in pension payments is also indicative of the effects of other short-to-medium-term strategies for reducing pension spending available to, or practised by, governments. Replacement rates can be reduced explicitly (e.g. as proposed in Germany) or implicitly (by changing the definition of reference earnings). 22. See Keenay and Whitehouse (2003) for a cross-country comparison of the taxation of incomes in retirement and the paper of Verbist (2006) in this volume.


65

23. It should be noted that the condition of revenue neutrality refers to the reference year. If, in the future, demographic and/or economic conditions change, the corresponding packages may not continue to be revenue neutral. 24. Taking into account that Italy is one of the very few EU countries where a generalised Minimum Income Guarantee scheme does not exist, it can be argued that it might be preferable to support this segment of the population directly through social assistance policies rather than through generous across-the-board increases in existing minimum pension schemes. 25. The need to address gender inequalities in pension incomes is explicitly recognised in EU-level policy statements as part of the pensions ‘‘modernisation’’ agenda. See for example European Council (2003, p. 103). 26. See Hansen (2001). 27. See Grabka (2001) and MISSOC for 1998 [http://europa.eu.int/comm/employment_social/soc-prot/missoc98/english/07/index.htm]. 28. See Atella, Coromaldi, & Mastrofrancesco (2001). 29. See Sutherland (2001).

ACKNOWLEDGEMENTS This paper was written as part of the Micro Analysis of the European Social Agenda (MICRESA) project, financed by the Improving Human Potential programme of the European Commission (SERD-2001-00099). We are indebted to our present and former colleagues Herwig Immervoll, Christine Lietz and Cathal O’Donoghue for their invaluable contributions to the construction of the EUROMOD. Thanks are due to participants of the International Microsimulation Conference on Population Ageing and Health, Canberra, December 2003; the MICRESA project meeting, Lisbon, May 2004; a seminar at the University of Sheffield; and to Olivier Bargain, Herwig Immervoll, Manos Matsaganis and an anonymous referee, for helpful comments on earlier versions of the paper. The views expressed in this paper, as well as any errors, are the responsibilities of the authors. Part of the research was carried out while Panos Tsakloglou was a visiting Senior Research Fellow at the Hellenic Observatory of the LSE. He wishes to thank the members of the Observatory for their hospitality.

REFERENCES Atella, V., Coromaldi, M., & Mastrofrancesco, L. (2001). EUROMOD Country Report: Italy. http://www.econ.cam.ac.uk/dae/mu/emodcty.htm. Atkinson, A. B., Bourguignon, F., O’Donoghue, C., Sutherland, H., & Utili, F. (2002). Microsimulation of social policy in the European Union: Case study of a European minimum pension. Economica, 69, 229–243.

66


Atkinson, A. B., & Sutherland, H. (1998). Microsimulation and policy debate: A case study of the minimum pension guarantee in Britain. In: H. P. Galler & G. G. Wagner (Eds), Empirische forschung und wirtschaftspolitische Beratung-Festschrift fu¨r Hans-Ju¨rgen Krupp zum 65 Geburtstag (pp. 79–98). Frankfurt: Campus. Bardasi, E., & Jenkins, S. (2004). The gender gap in private pensions. ISER Working Paper No 2004-29. University of Essex, UK. Dennis, I., & Guio, A-C. (2003). Poverty and social exclusion in the EU after Laeken – part 1. Statistics in Focus, Theme 3 8/2003, Eurostat. Dupont, G., Hagnere, C., & Touze, V. (2003). Les modeles de microsimulation dynamique dans l’analyse des reformes des systemes de retraites: Une tentative de bilan. Economie et Prevision, 160–161, 167–191. Economic Policy Committee. (2002). Reform challenges facing public pension systems: the impact of certain parametric reforms on pension expenditure, EPC/ECFIN/237/02. Economie et Prevision. (2003). 160–161 (special issue on pensions microsimulation). Emerson C., Reed, H., & Shephard, A. (2004). An assessment of PenSim2. IFS Working Paper No W04/21. Esping-Andersen, G. (1990). The three worlds of welfare capitalism. Cambridge: Polity Press. European Commission. (2003). Draft joint inclusion Report: Statistical Annex, [COM(2003)773final] http://europa.eu.int/comm/employment_social/news/2001/jun/ napincl_03_da_en.pdf. European Council (EC). (2003). Joint report on adequate and sustainable pensions http:// europa.eu.int/comm/employment_social/soc-prot/pensions/2003jpr_en.pdf. Eurostat. (1996). The European Community Household Panel (ECHP): Survey methodology and implementation, Theme 3, Series E, Eurostat, Luxembourg. Feldstein, M., & Siebert, H. (Eds) (2002). Social security pension reform in Europe. Chicago: University of Chicago Press. Ferrera, M. (1996). The ‘southern model’ of welfare in social Europe. Journal of European Social Policy, 6, 17–37. Foster, J. E., Greer, J., & Thorbecke, E. (1984). A class of decomposable poverty measures. Econometrica, 52, 761–766. Galler, H. P. (1996). Microsimulation of pension reform proposals: Modelling the earnings of couples. In: A. Harding (Ed.), Microsimulation and public policy. Contributions to Economic Analysis (Vol. 232, pp. 293–312). Amsterdam: Elsevier. Gillion, C., Turner, J., Bailey, C., & Latulippe, D. (2000). Social security pensions: Development and reform. Geneva: ILO. Grabka M. (2001). EUROMOD country report: Germany, http://www.econ.cam.ac.uk/dae/ mu/emodcty.htm Hagenaars, A.J.M., de Vos, K., & Zaidi, M.A. (1994). Poverty statistics in the late 1980s: Research based on micro-data, Theme 3, Series C, Eurostat, Luxembourg. Hancock, R. M., & Barker, G. (2005). The quality of social security benefit data in the British family resources survey: implications for investigating income support take-up by pensioners. Journal of the Royal Statistical Society, Series A 168(Part 1), 63–82. Hansen H. (2001). EUROMOD country report: Denmark, http://www.econ.cam.ac.uk/dae/ mu/emodcty.htm Immervoll, H., & O’Donoghue, C. (2001). Imputation of gross amounts from net incomes in household surveys: An application using EUROMOD. EUROMOD Working Paper EM1/01. University of Cambridge, Cambridge.


67

Immervoll H., O’Donoghue, C., & Sutherland, H. (1999). An introduction to EUROMOD. EUROMOD Working Paper EM0/99. University of Cambridge, Cambridge. International Labour Office (ILO). (2001). Social security: A new consensus. Geneva: ILO. ISSAS. (1990). Poverty in figures: Europe in the early 1980s, Theme 3, Series C, Eurostat, Luxembourg. Keenay, G., & Whitehouse, E. (2003). Financial resources and retirement in nine OECD countries: The role of the tax system. OECD Social, Employment and Migration Working Papers No. 8. OECD, Paris. Kennell, D. L., & Sheils, J. F. (1990). PRISM: Dynamic simulation of pension and retirement income. In: G. H. Lewis & R. C. Michel (Eds), Microsimulation techniques for tax and transfer analysis (pp. 137–172). Washington, DC: Urban Institute Press. Manchester, J. (1999). Compliance in social security systems around the world. In: O. S. Mitchell, R. J. Myers & H. Young (Eds), Prospects for social security reform. Philadelphia: University of Pennsylvania Press. Mantovani, D., & Sutherland, H. (2003). Social indicators and other income statistics using the EUROMOD baseline: A comparison with Eurostat and national statistics. EUROMOD Working Paper EM1/03. University of Cambridge, Cambridge. Morrison, R. J. (2000). DYNACAN, the Canada pension plan policy model: Demographics and earnings components. In: A. Gupta, & V. Kapur (Eds), Microsimulation in government policy and forecasting. Contributions to Economic Analysis (Vol. 247, pp. 341–360). Amsterdam: Elsevier. Nelissen, J. H. M. (1995). Lifetime income redistribution by the old-age state pension in the Netherlands. Journal of Public Economics, 58, 429–451. Nelissen, J. H. M. (2000). The lifetime redistributive impact of capital-funded Dutch supplementary pensions. Geneva Papers on Risk and Insurance, 25, 407–423. Pensions Policy Institute (PPI). (2003). State Pension Models, PPI: London (http://www. pensionspolicyinstitute.org.uk). Peracchi, F. (2002). The European community household panel: A review. Empirical Economics, 27, 63–90. Revue Economique, 2000, Vol. 51 (special issue on pensions microsimulation). Sutherland, H. (1998). A citizen’s pension. Microsimulation Unit Discussion Paper MU9804. Department of Applied Economics, University of Cambridge, Cambridge. Sutherland, H. (2001). EUROMOD country report: United Kingdom, available from http:// www.econ.cam.ac.uk/dae/mu/emodcty.htm Tsakloglou, P. (1996a). Elderly and non-elderly in the European Union: A comparison of living standards. Review of Income and Wealth, 42, 271–291. Tsakloglou, P. (1996b). On the relative welfare position of older persons in the member-states of the European Union. In: C. Paraskevopoulos, R. Grinspun & T. Georgakopoulos (Eds), Economic integration and public policy: NAFTA, EU and beyond (pp. 154–168). Cheltenham: Edward Elgar. Verbist G. (2006). The distributional effects of taxes on pensions and unemployment benefits in the EU-15 (in this volume). Weller, C. E. (2004). The future of public pensions in the OECD. Cambridge Journal of Economics, 28, 489–504. World Bank. (1994). Averting the old age crisis: Policies to protect the old and promote growth. Oxford: Oxford University Press.

68


APPENDIX A. EUROMOD BASE DATASETS Country

Base Dataset for EUROMOD

Denmark

European Community Household Panel (W2) German Socio-Economic Panel (W15) Survey of Households Income and Wealth Family Expenditure Survey

Germany Italy UK

Type

Date of Collection

Reference Time Period for Incomes

ECHP

1995

Annual 1994

National panel

1998

Annual 1997

Income survey

1996

Annual 1995

1995/1996

Month in 1995/1996

Household budget survey

APPENDIX B. A SHORT DESCRIPTION OF THE PENSION SYSTEMS OF DENMARK, GERMANY, ITALY AND THE UK Denmark26 The old age pension is tax financed and based on length of residence and citizenship rather than former income or contribution record. The maximum pension is paid on the basis of residence from the age of 15 until retirement at the age of 67 years (reduced to 65 in 2004). For residence of less than 40 years the pension is paid on a pro-rata basis. Danish early retirement schemes that depend on voluntary contributions to the unemployment insurance have no influence on the old-age pension. This cannot be claimed before the age of 67 and there is no premium for delaying retirement. In 1998, the residence-based old-age pension consisted of three flat rate benefits. 1. A ‘‘basic amount’’ of 46,812 DKK per year is means-tested against current earnings. The full basic amount is received by 95% (PPI, 2003). 2. A ‘‘supplement’’ which is 39,576 DKK for singles and 20,568 DKK for each member of a couple; this is means-tested against income from the earnings and capital of both partners with a threshold of 43,900 DKK for a single and 88,000 DKK for a couple. The taper is 30% on income above


69

the thresholds (15% each in the case of old-age couples). In 2002, 64% received the supplement in full (EC, 2003). 3. A ‘‘special supplement’’ of 6,828 DKK for singles (not means-tested). The ‘‘contributions’’ are direct components of the income tax scheme. They are levied at a combined rate of 9% on all gross wages, taxable fringe benefits and the part of the business income that is categorised as personal income. In addition, there is a social insurance supplementary pension (ATP), established in 1964. Contributions are compulsory for employees working at least nine hours a week. Pensions are paid depending on the number of contributions and hours of work: they are not earnings related. The maximum payable in 1998 would be 16,620 DKK (for someone working full-time since 1964). Some 68% of pensioners receive some ATP (PPI, 2003). The employee contribution in 1998 was 894 DKK per year for 27 or more hours work per week, 2/3 of that for 18–27 hours and 1/3 for 9–18 hours. All ATP contributions are deductible from income for personal income tax purposes. Private pension schemes cover 80% of current employees but are not yet widespread as pensions in payment. Germany27 Germany operates a contributory pay-as-you-go earnings-related state pension. Contributions are paid half each by employers and employees at a combined rate of 20.3% of salary up to a ceiling. The ceiling and the average pension payments are different in the former West and East Germany. The contribution assessment limit in 1998 was DM 8,400 a month in western Germany and DM 7,000 in eastern Germany. Pension payments are also partially subsidised to allow the system to cover people who have been carers. Pensions can be taken early or deferred beyond the pension age of 65, with a corresponding adjustment to benefit. The self-employed can make voluntary contributions. There are no additions for dependants. There is no minimum pension. Social assistance is paid to 1.4% of pensioners (PPI, 2003). This acts as a top-up to other incomes. In addition, voluntary company schemes cover about half of men (fewer women) and personal pensions (with tax incentives) cover 21% of male and 9% of female pensioners (PPI, 2003). Italy28 The Italian state pension system is in the process of transformation. In 1998, the main contributory pension (made up of many similar but distinct

70


schemes) was earnings-related based on contributions up to a ceiling. Reference earnings were the last 10 years and the proportion of pension that is assessed in this way is being reduced as the reforms are phased in, replaced by whole working lives. Retirement could be taken early, and also deferred, but only to 65. Where entitlement to this pension is below a certain limit (minimum pension) there is entitlement to a means tested supplement up to the difference between that limit and the actual entitlement. A full supplement is paid if personal taxable income is lower than minimum pension and couple income is lower than four times the minimum pension. A taper of 100% is applied to higher incomes. This affects about 39% of people aged 65+. The Social Pension is a means-tested minimum income scheme, which covers about 6% of elderly who do not have contributory rights. There are also some occupational-defined contribution schemes and some personal tax incentivised schemes that are not very widespread. EUROMOD estimates show that just 1% of current elderly receive such income (Table 2). United Kingdom29 The state system was in 1998 made up of (a) a flat rate contributory pension (the ‘‘basic pension’’), (b) a state earnings-related component, (c) occupational and personal pensions that are regulated and may substitute for the state earnings-related pension and, (d) for pensioners whose income falls below a certain level, a minimum income scheme. If they meet the contribution conditions, people over state pension age get a flat rate basic pension. If conditions are only partly met, a reduced pension of at least 25% of the basic can be paid. Spouses who do not meet the conditions may receive a lower pension based on their partner’s contributions. At age 80 contribution conditions are removed. There is no early retirement; extra pension increments can be earned if retirement is delayed. Additions are paid for dependent spouses under pension age (subject to a means test) and dependent children. The basic pension is taxable. For pensioners who contributed to the State Earnings Related Pension Scheme (SERPS) an additional earnings-related pension is payable. This is taxable and there are no additions for dependants. Income support (IS) is the main social assistance benefit for people whose family incomes are lower than a specified level The level of payment of the basic pension is less than the income level offered by IS to the elderly. Unless they or their partners have other sources of income a basic pensioner will be entitled to a top-up from IS.


71

Employees pay contributions on earnings between a lower and upper limit. Employees who are contracted out of SERPS pay a lower rate of contribution. People with self-employment income are liable for contributions, which only bring entitlement to the basic retirement pension, not to any earnings-related benefit. Contributions are not tax deductible.

THE DISTRIBUTION EFFECTS OF TAXES ON PENSIONS AND UNEMPLOYMENT BENEFITS IN THE EU-15$ Gerlinde Verbist ABSTRACT Research has shown that the tax treatment of replacement incomes differs considerably among countries. Consequently, the ranking of countries by expenditure level is different for gross and net social expenditures. On a micro level this is translated into a gap between gross and net benefits; this gap varies among countries. In this chapter, we use EUROMOD for an international comparison of the difference between gross and net benefits at the micro level. We investigate the distribution effects of the income tax treatment of replacement benefits, focusing on old-age $

This study was carried out during a research visit at the Microsimulation Unit of the Department of Applied Economics, University of Cambridge (August–November 2003). The visit was supported by funding of the Flemish Fund for Scientific Research. I also like to thank the members of the EUROMOD-team and network for making this research possible. More in particular, I am very grateful to Holly Sutherland, Christine Lietz, Daniela Mantovani, Herwig Immervoll, Fre´de´ric Berger, Tim Callan, Klaas de Vos, Bengt Eklind, Horacio Levy, Panos Tsakloglou and Heikki Viitamaki for support and many useful remarks in the process of writing this text. Remaining errors are of course all mine.


73

74

GERLINDE VERBIST

pensions and unemployment benefits. We present a summary overview of the different ways of levying taxes on benefits in the pre-2004 EU-15 countries. We then try to answer the question how the tax treatment of social security benefits affects the distribution of these benefits and how progressive taxes on benefits are compared to taxes on earnings.

1. INTRODUCTION There are important links between the tax system and social benefits. Both play an important role in the income-generating process of households. Taxes and benefits affect disposable income of households, as well as the distribution and redistribution of income. Both systems are also technically linked: both social contributions and social benefits can be part of taxable income, and there are special provisions in the personal income tax system with a social goal (the so-called ‘social-fiscal’ measures). Up until now the relationship between personal income taxation and the social security system has received relatively little attention in research. An exception is the field of family policy, where e.g. the joint effect of child benefits and tax concessions for children has been analysed (e.g. Bradshaw & Finch, 2002; O’Donoghue & Sutherland, 1999). If we want to gain a more thorough understanding of the way in which the tax system influences social policy, and thus plays a role in the shaping of the welfare state, we need to include the other sectors of social security in the analysis, and more specifically the allowances that replace income from work that is lost due to old age, sickness, invalidity, unemployment, etc. Research has shown that the tax treatment of replacement incomes differs widely among countries (Adema et al., 1996; Adema, 1999). Consequently, the ranking of countries by expenditure level is different for gross and net social expenditures. On a micro level this is translated into a gap between gross and net benefits, which varies among countries. An international comparison of the difference between gross and net benefits at the micro level is the goal of this chapter. As EUROMOD covers the major part of the different national personal income tax and social benefit systems and runs on a representative set of micro-data (see Chapter 1 of this book), it is the most appropriate tool for this kind of analysis. We investigate empirically how taxes affect the distribution of replacement incomes in the EU-15 (the 15 countries that formed the European Union before 2004, see Table A1 for an overview of the EUROMOD datasets). We limit ourselves to the tax treatment of old-age pensions and unemployment benefits. This means that we look at only part of the redistributive process,

The Distribution Effects of Taxes

75

namely the specific interaction between taxes and social security benefits. We do not investigate the redistributive effect of taxes and benefits as a whole; for this we refer to Immervoll et al. (2006). We start this chapter with the relative income position of the elderly and the unemployed, pointing at the difference between gross and net incomes. One of the reasons for this difference is found in the special provisions in the various tax systems for old age (pensions) and unemployment, which are discussed in Section 3. This differential tax treatment is an important factor in explaining the difference in tax burden between earners and benefit recipients. In Section 4 we present the distribution characteristics of taxes on replacement incomes: what is the tax burden on replacement incomes at the bottom and the top end of the income distribution? Is there a difference in tax progressivity between the various income categories? Section 5 concludes.

2. RELATIVE INCOME POSITION OF THE ELDERLY AND THE UNEMPLOYED We first define our three categories of income recipients, namely workers, the elderly and the unemployed. Workers are either civil servants or employees and have a strictly positive value for income from employment.1 Only individuals who are 18 and older and have not yet reached the age of 60 are classified as workers. Old-age individuals are all individuals older than 65, living in private households.2 Unemployed are those individuals who either have a strictly positive value for unemployment benefits or have ‘Unemployed’ as their employment status. The same age limits apply as for workers. For Luxembourg the absolute number of unemployed in the sample is too small to have reliable results; therefore this country is not included in this part of the analysis. 2.1. Relative Income Position of Old-Age Individuals As an indicator of the relative income position of old-age individuals we express their individual average income as a percentage of that of workers and of that of the entire population. A relative income ratio of 100% or more indicates that the elderly are on average better off than the entire population, or than the group of workers. What strikes immediately from Table 1 is that in all countries the net relative income ratio is higher than the gross relative income ratio. This is irrespective of whether the denominator

76

GERLINDE VERBIST

Table 1. Average Income of Old-Age Individuals as a Percent of Average Income of all Individuals and Average Income of Workers. Country

As % of Income of all Individuals Gross

Austria (AT) Belgium (BE) Denmark (DK) Finland (FI) France (FR) Germany (GE) Greece (GR) Ireland (IR) Italy (IT) Luxembourg (LU) Netherlands (NL) Portugal (PT) Spain (SP) Sweden (SW) UK (UK)

96.3 91.2 69.3 88.9 110.5 88.5 79.3 81.2 115.3 110.7 104.1 87.5 97.7 96.1 87.8

Net (6) (8) (15) (9) (3) (10) (14) (13) (1) (2) (4) (12) (5) (7) (11)

109.7 105.7 82.8 97.8 120.6 112.2 88.2 93.3 127.6 119.4 119.9 101.1 109.1 105.5 98.6

As % of Income of Worker Gross

(6) (8) (15) (12) (2) (5) (14) (13) (1) (4) (3) (10) (7) (9) (11)

55.6 43.3 40.7 44.7 58.1 46.2 35.3 31.2 59.5 58.8 55.2 44.5 35.1 55.1 46.6

(4) (11) (12) (9) (3) (8) (13) (15) (1) (2) (5) (10) (14) (6) (7)

Net 65.8 54.7 50.6 52.6 66.6 66.5 40.3 38.9 68.8 64.9 66.0 54.5 40.5 65.3 55.4

(5) (9) (12) (11) (2) (3) (14) (15) (1) (7) (4) (10) (13) (6) (8)

Note: The figures in brackets indicate the ranking of countries for the relative income position of old-age individuals (lower ranks have better positions). Source: EUROMOD.

is income of all individuals or income of workers only. The difference between gross and net rates is biggest for Germany (more than 20%), and this according to both measures. The difference is much smaller in Finland, Greece and Luxembourg. This difference between gross and net relative income ratios indicates that the tax system has a positive impact on the relative welfare position of the elderly. If we compare the average income position of old-age individuals with the entire population, then old-age individuals have a gross relative income ratio below 100% in all countries, except France, Italy, Luxembourg and the Netherlands. But in net terms, the elderly are almost everywhere better off than the average for the entire population; the exceptions are Denmark, Greece and Ireland and to a lesser extent Finland and the UK. Compared to workers, the average relative income ratio is in all countries below 100% (both in gross and in net terms). Gross relative income ratios vary between 31% (Greece) and almost 60% (Italy), whereas net relative income ratios range from 40% (Greece and Spain) to 69% (Italy). The ranking of countries according to the relative income position of the elderly is different in


77

3500 3000 2500 2000 1500 1000 500 0 GR

PT

IR

SP

Gross Income - 65+

FI

DK

BE

Net Income - 65+

UK

SW

GE

IT

Gross Income - Workers

AT

FR

NL

LU

Net Income - Workers

Fig. 1. Average Gross and Net Individual Income of Workers and Old-Age Individuals in the EU-15, 1998 (Monthly Amounts in Euro PPP). Source: EUROMOD.

gross and in net terms: Luxembourg, Spain, Sweden and Finland fall back two or three positions (ranking of comparison with all individuals), whereas Germany and Portugal make a considerable leap forward. Fig. 1 compares the average amounts of income of workers and pensioners per month; all amounts are converted to Euro purchasing power parities (see Table A2). Countries are ranked according to the average net income level of old-age individuals. Luxembourg stands out as the country with the highest average income for both workers and old-age individuals, followed by the Netherlands. Portugal and Greece have the lowest average income for the two categories of individuals. For the countries in between these two extremes, the ranking changes in the transition from gross to net average income of workers. In general, we see that average net income of workers is less dispersed among countries (between 1320 Euro in Sweden and 1510 Euro in Austria) than gross income. The ranking of countries based on income of the elderly, on the contrary, is roughly the same for gross and net income. The figure also shows quite clearly that the gap between gross and net income is different for workers and elderly, which was already expressed by the differences in relative income ratios. 2.2. Relative Income Position of the Unemployed The pattern of relative income ratios of the unemployed is quite similar to that of old-age individuals (Table 2). Also for the unemployed the net

78

GERLINDE VERBIST

Table 2. Average Income of Unemployed Individuals as a Percent of Average Income of all Individuals and Average Income of Workers. Country

Income of all Individuals Gross

Austria Belgium Denmark Finland France Germany Greece Ireland Italy Netherlands Portugal Spain Sweden UK

93.1 95.5 119.8 90.0 76.6 80.5 62.2 77.1 21.9 69.4 64.5 67.5 104.9 52.1

Income of Worker

Net (4) (3) (1) (5) (8) (6) (12) (7) (14) (9) (11) (10) (2) (13)

106.5 108.0 122.3 98.4 84.1 94.9 69.1 90.1 24.4 72.4 74.1 73.3 107.6 61.5

Gross (4) (2) (1) (5) (8) (6) (12) (7) (14) (11) (9) (10) (3) (13)

53.7 45.4 70.4 45.2 40.3 42.0 27.7 29.7 11.3 36.8 32.8 24.3 60.2 27.6

(3) (4) (1) (5) (7) (6) (11) (10) (14) (8) (9) (13) (2) (12)

Net 63.9 55.9 74.7 52.9 46.5 56.3 31.6 37.6 13.1 39.9 39.9 27.2 66.6 34.5

(3) (4) (1) (6) (7) (5) (12) (10) (14) (8) (9) (13) (2) (11)

Note: The figures in brackets indicate the ranking of countries for the relative income position of the unemployed (lower ranks have better positions). Source: EUROMOD.

relative income ratio is higher than the gross relative income ratio in all countries, irrespective of the denominator. The difference between gross and net rates is again largest for Germany (around 14%), and this according to both measures. The difference is much smaller in Denmark, Italy, the Netherlands and Sweden. The difference between gross and net relative income ratios for unemployed is however considerably smaller than for oldage individuals, and also the ranking of countries is less affected. Fig. 2 compares the average amounts of income from workers and unemployed per month in Euro purchasing power parities. Southern Europe, Ireland and the UK have the lowest average income of unemployed. There is a middle group including France, the Netherlands, Finland, Belgium and Germany. Income of the unemployed is on average relatively high in Austria, Sweden and especially Denmark. 2.3. Income from Work, Pensions and Unemployment Benefits We have seen that both for old-age and unemployed individuals net relative income ratios are considerably higher than gross relative income ratios. Also


79

3000 2500 2000 1500 1000 500 0 IT

PT

GR

SP

UK

IR

NL

Gross Income - Unemployed Gross Income - Workers

FI

FR

GE

BE

SW

AT

DK

Net Income - Unemployed Net Income - Workers

Fig. 2. Average Gross and Net Individual Income of Workers and Unemployed in the EU, 1998 (Monthly Amounts in Euro PPP). Source: EUROMOD.

the ranking of countries with respect to the relative income positions is different for gross and for net income, especially for old-age individuals. This is mainly due to the role of the tax system, which will be investigated in the rest of this chapter. In order to highlight the effect of tax differentials between workers, the elderly and unemployed individuals, we focus on their specific individual income components namely income from work for workers, pension income for the elderly and unemployment benefits for the unemployed. Income from work (or earnings) equals gross income from employment and includes also 13th/14th monthly salaries, as well as other employment incomes, such as bonuses, extra holiday pay, occasional pay, etc. Pensions include both public and private pensions, as we are interested in the effects of taxes on income that is old age specific. Unemployment benefits are those benefits that are aimed specifically at unemployed and include both insurance-related and social assistance allowances. Gross income components are taken directly from the datasets used in EUROMOD (see Table A1) or, where necessary, are imputed from net income (see Immervoll & O’Donoghue, 2001; for more details on the composition of gross income see Verbist, 2004). To arrive at net income we subtract personal taxes (PT) and social insurance contributions (SIC) from gross income. PT include taxes both at the national and the local level. SIC do not include employer contributions, following customary

80

GERLINDE VERBIST

Table 3. Earnings, Pensions and Unemployment Benefits of Workers, Old-Age Individuals and Unemployed as a Percent of Their Gross Individual Income. Workers Gross Earnings

Old-Age Individuals Gross Pension

Unemployed Gross Unemployment Benefits

96.5 93.7 96.4 92.8 93.1 96.4 96.4 96.9 92.9 93.8 95.6 97.3 98.9 91.0 95.8

91.3 84.4 65.7 83.7 79.1 88.7 74.5 66.6 67.4 80.5 90.9 71.3 85.2 78.7 69.1

23.1 38.4 39.6 26.1 29.5 33.2 4.0 62.9 21.2 32.6 27.9 51.4 31.6 34.0 6.5

Austria Belgium Denmark Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain Sweden UK Source: EUROMOD.

practice in most distributional studies. Policy measures used here refer to 1998. For workers almost their entire gross income consists of income from work (more than 90%, Table 3), and this applies for all the EU-15 countries. This is not the case for the elderly and unemployed. On average, gross pensions make up between 66% and 91% of gross income of old-age individuals (the highest proportions are found in Austria and the Netherlands, which also have the highest proportion of private supplementary pension provisions). This variation is mainly due to differences in level and coverage of the public pension system, what old-age individuals can and cannot earn as income from work, and regulations of possible combinations with other benefits. The variety is even bigger for the unemployed (between 4% in Greece and 63% in Ireland). This is also mainly due to variations in regulation of the benefits (level and coverage of the benefits, possible combinations with other income sources, etc.).3


81

3. TAXES AND SOCIAL CONTRIBUTIONS ON REPLACEMENT INCOMES 3.1. Measuring Taxes on Replacement Incomes in EUROMOD As personal income taxes are often levied on the total of taxable income, it is not obvious how to calculate the exact amount of taxes on separate income components such as replacement incomes. There are two possible methods to calculate the tax burden on benefits: the marginal method and the proportional method (Verbist, 2002; Ministry of Health and Social Affairs, 2003). With the marginal method we compare the present system with a hypothetical one in which replacement incomes are tax exempt. The difference between both systems is a measure of the tax burden on benefits. This method seems to be the most appropriate. Yet there is a problem for households or individuals with other sources of income: when replacement incomes are taken out, then the tax burden that concerns other remaining incomes is computed at lower marginal tax rates due to progressive taxation. The tax burden on replacement incomes is then overstated, or corresponds to an upper bound. A lower bound of the tax burden is given by taxes on replacement incomes alone (all other incomes assumed to be tax exempt). The proportional method provides an intermediary case: we attribute taxes to income components according to their respective shares in taxable income before application of allowances and deductions (for the specification of allowances and deductions, see Verbist, 2004). We use here the proportional method, so we have to specify for each country: (a) taxable replacement incomes (RY), (b) taxable income before applications of tax deductions and tax allowances (Y), (c) taxes to be apportioned (in most cases only personal income taxes, for the Scandinavian countries these include also local taxes) (T). Taxes on replacement income (TRY) are then TRY ¼ T RY/Y. When replacement incomes are in EUROMOD an undistinguishable part of income liable for SIC, the same procedure is applied for these contributions.4

82

GERLINDE VERBIST

3.2. Differential Tax Treatment of Replacement Incomes In most countries pensions and unemployment allowances are subject to PT. But quite often the tax treatment is distinct from that of income from work, because of special provisions in the personal income tax system for old age, pensions and unemployment benefits.5 This kind of provisions have been called ‘social fiscal measures’ by Adema et al. (1996), and are defined as ‘‘those reductions, exemptions, deductions or postponements of taxes, which (a) perform the same policy function as cash transfers which, were they to exist, would be classified as social expenditures; or (b) are aimed at stimulating private provisions of benefits.’’ Such a social-fiscal measure can be applied in different forms. The OECD (1984, 1996a) distinguishes the following different categories of tax reliefs: (1) A tax allowance, which is the deduction of a fixed amount of taxable income. As most tax systems are progressive, the amount of the advantage increases with income, and as such it favours the higher income groups. (2) An income-related tax deduction occurs when the amount subtracted from taxable income is not fixed but depends on the level of the income (3) A tax credit is a (fixed) amount that is subtracted from tax liability. Sometimes the tax credit exceeds tax liability, in which case two possibilities arise: either the amount of the excess of the tax credit over the tax liability is paid to the taxpayer, in which case it is a ‘non-wastable tax credit’, or this does not happen, and then the tax credit is ‘wastable’, providing no or a minor advantage to the lowest-income groups. (4) A tax exemption is part of income or specific sources of income that are tax exempt. If exemptions are applied at the bottom of the tax scale, then the effect is analogous to that of a tax credit. If a certain source of income is not included in taxable income, then the effect is similar to that of a tax allowance. (5) A preferential tax rate: some incomes or sources of income are taxed at a lower rate than others are. This can for instance be done by dividing taxable income in different parts, such that each part is taxed at a lower marginal rate. Table 4 presents an overview of the special provisions in the personal income tax system for old age, pensions and unemployment benefits. Pensions are in almost all countries part of taxable income, though often provisions for old age or for pension income have been made (see also Fenge & Werding, 2004; Keenay & Whitehouse, 2003a, 2003b). Part of pension

Austria Belgium Denmark Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain Sweden UK

Special Provisions in the Personal Income Tax System for Old Age and Unemployment in the EU15, EUROMOD, 1998. Concessions for Old Age or Pension Income

Concessions for Unemployment Benefits

Tax credit for pensioners Tax credit for pension incomes Guaranteed minimum income for old persons is tax exempt – Deduction for pensions Minimum pension and social benefit for dependent elderly are tax exempt Deduction for pensions Deduction for old age, for pensions – Tax allowance for old age Social pension is tax exempt Tax credit for pensioners Deduction for pensioners Tax allowance for old age Deduction for pension income Tax credit for elderly inactive Part of pensions are deducted from taxable income, and thus tax exempt Tax allowance for old age and for married old-age couples

Unemployment benefits are tax exempt Tax credit for unemployment benefits


Table 4.

– – – Unemployment benefits are tax exempt – Unemployment benefits are tax exempt – – – Unemployment benefits are tax exempt – – –

Source: Based on the country reports of EUROMOD.

83

84

GERLINDE VERBIST

income is tax exempt in Belgium, Finland, France, Germany, Italy, Luxembourg, Portugal and Sweden. In Belgium, France and Italy pensions that can be characterised as a guaranteed social minimum are not taxable. In Finland, Sweden and Luxembourg older people are entitled to a special deduction that depends on the level of pension income. Germany has made the most extensive favourable tax treatment for pension income: the share of public pensions that is taxable depends on age (e.g. 38% of public pensions are taxable when the individual is aged 55, 27% is taxable at age 65 and 21% at age 70). Ireland, the Netherlands and the UK have a tax allowance for old age, which means that for individuals older than 65 a fixed amount can be deducted from taxable income. A tax credit, which is the subtraction from tax liability, is granted in Austria, Belgium and Spain; in Austria and Belgium the tax credit depends on the level of pension income, whereas in Spain the tax advantage is granted to inactive persons older than 65. In four countries unemployment benefits are partly or entirely tax exempt, namely in Austria, Germany, Ireland and Portugal. Belgium provides a tax credit for unemployment benefits, which is designed in such a way that the households whose income consists only of these benefits do not pay taxes. These differences in tax treatment are reflected in the relative tax burden on the income components. In Table 5 we present PT and social contributions (SIC) on earnings, pensions and unemployment benefits. We have used the proportional method and express these taxes as a percentage of gross earnings, gross pensions and gross unemployment benefits respectively. In all countries income from work is taxed more heavily than either pensions or unemployment benefits. There are two reasons for this. First, gross pensions and unemployment benefits are on average lower in all countries. As most personal income tax systems are progressive, this will lead to a higher tax burden for workers. A second reason is found in the special provisions in the personal income tax system for old age, for pension incomes or – to a minor extent – for unemployment benefits. This becomes apparent in a large gap between PT on earnings and on pensions in some countries, namely Belgium, Finland, Germany, Ireland, Portugal and the UK. The difference between earners and the unemployed is quite big in Austria, Belgium, Germany, Ireland, Portugal, Spain and the UK. In most of these countries, unemployment benefits are tax exempt or can benefit from a special tax relief. Taxes on earnings are mainly PT, except for France, Greece and the Netherlands, where SIC have the biggest weight. Also for pensions personal income taxes are the most important component for all countries but France, Germany and Greece. In almost half of the countries no SIC are payable on


85

Table 5. PT and SIC on Earnings as a Percent of Gross Earnings (Only Earners), on Pensions as a Percent of Gross Pensions (Only Pensioners) and on Unemployment Benefits as a Percent of Gross Unemployment Benefits (Only Unemployed). Earnings


Pensions

Unemployment Benefits

PT

SIC

PT

SIC

PT

SIC

15.0 20.9 31.3 27.9 4.5 18.8 9.8 20.5 19.0 13.9 11.2 13.0 16.7 31.3 16.0

13.8 11.6 10.5 7.6 19.7 17.7 11.5 4.0 8.9 10.0 19.4 11.0 3.7 6.2 6.7

13.7 9.7 25.3 18.8 3.6 1.7 3.6 3.3 14.4 9.8 9.0 2.6 6.9 25.9 6.3

3.7 2.5 0.0 3.0 5.2 6.4 4.0 0.0 0.0 2.6 8.2 0.0 0.0 0.0 0.0

0.0 5.8 28.8 20.1 0.9 0.0 4.9 0.0 10.3 – 7.0 0.0 3.1 26.0 0.7

0.0 0.0 2.5 1.4 1.5 0.0 0.0 0.0 0.0 – 20.3 0.0 2.5 6.9 0.0

Source: EUROMOD.

pension income; in the other countries the level of SIC on pensions is rather small due to the fact that it is levied at a reduced rate (for more details see Verbist, 2004). This is so because for certain risks pensioners do not have to pay any more insurance contributions (e.g. for unemployment) (see also Disney et al., 1999; Keenay et al., 2003a, b). Social contributions on unemployment benefits are negligible, except in the Netherlands and Sweden.

4. DISTRIBUTION CHARACTERISTICS OF TAXES ON REPLACEMENT INCOMES 4.1. The Bottom and the Top End of the Income Distribution As many of the special provisions for replacement incomes are targeted at lower income recipients, the difference between gross and net income components is probably much smaller at the lower end of the income distribution than at the top. Figs. 3–5 give on the one hand the tax rates on the

86

GERLINDE VERBIST 35.0 30.0 25.0

%

20.0 15.0 10.0 5.0 0.0

IR

UK

PT

SP

BE

FR

IT

LU

GR

PT

SIC

AT

GE

FI

NL

SW

DK

Fig. 3a. Taxes (PT) and Social Contributions (SIC) Paid on Pensions and Earnings (Left and Right Bar Respectively for Each Country) as a Percent of the Individual Gross Income Component in the Bottom Quintile, EUROMOD 1998. Countries are Ranked in Ascending Order of Taxes and Contributions on Pensions. Source: EUROMOD.

respective income components for the bottom and the top income quintiles,6 split into PT and SIC; on the other , we also present average gross and net income components for our three categories of income recipients (see Appendix B for the numbers). Let us first look at the comparison of pensions with earnings at the bottom end of the income distribution. In Fig. 3a countries are ranked from low to high tax burden on pensions (total of PT and SIC). For each country the left bar present the tax burden on pensions and the right bar the tax burden on earnings. A general observation is that in the bottom quintile, the total of PT and SIC paid on pensions is considerably lower than taxes paid on earnings. One of the reasons is that pensions are in general lower than earnings (see Fig. 3b). The tax burden on pensions tends to increase with the tax and contribution rate on earnings, indicating that, despite the existence of special provisions for pensions, the general structure and weight of the tax system is still similar for earnings and pensions. In Fig. 3b we compare gross and net pensions (left bar for each country) with earnings (right bar); countries are ranked in ascending order of their net pension level. The ranking of countries in Figs. 3a and b is not the same: countries with higher pensions tend to have higher taxes on their pensions, and vice versa, but the


87

1600 1400 1200

Eur

1000 800 600 400 200 0 GR

PT

SP

IR

BE

UK

IT

GE

Gross

SW

FI

DK

FR

AT

NL

LU

Net

Fig. 3b. Pensions and Earnings (Left and Right Bar Respectively for Each Country) in the Bottom Quintile, Euro PPP, EUROMOD 1998. Countries are Ranked in Ascending Order of Net Pensions. Source: EUROMOD.

relationship is not a very strict one. This is illustrated by the Scandinavian countries who rank among the countries with the highest tax burden, but who are somewhere in the middle of the ranking of pension levels. There is quite some diversity among countries with respect to taxes and contributions paid on pensions at the bottom of the income distribution. PT and contributions are zero in Ireland, the UK and Portugal. In these countries no SIC are due on pensions. In Ireland and the UK low-income pensioners hardly pay income taxes due to the tax allowances, of which the allowance for old age is an important component. In Portugal the tax deduction for pension income is especially important for the lower incomes. PT and social contributions are very low in Spain, Belgium and France (less than 1%) and in Italy, Luxembourg and Greece (less than 5%). In Belgium social security contributions on pensions are only liable from a certain threshold onwards, and low-income pensioners hardly pay PT due to the tax credit for pension incomes. In Italy and Spain the low average tax rate for pensioners is mainly due to the fact that no SIC are due on pension income, other special provisions for older people are too small to make a difference. In France, Luxembourg and Greece the tax burden on pensions consists mainly of SIC. This is also the case in Austria and Germany, though at a

88

GERLINDE VERBIST

somewhat higher rate. The SIC rate, however, is considerably lower than that on earnings, because pensioners only contribute for specific sectors in social security. The tax burden on low pensions is highest in the Netherlands, Finland, Sweden and especially Denmark. In the Scandinavian countries this follows mainly from taxes (especially local taxes), whereas for the Netherlands, social contributions are far more important. It is also in these countries that the gap between average gross and net pensions for low incomes is highest (Fig. 3b). The situation is quite different for the higher income groups. For most countries the average tax liabilities on pensions in the upper quintile are much closer to those on earnings (Fig. 4a). SIC paid on pensions are still remarkably lower than those paid on earnings, but for PT this does not apply any more. In some countries high income pensioners pay on average even relatively more in PT than workers (Austria, Belgium, Denmark, Finland, France, Luxembourg and the Netherlands). This is mainly due to the fact that any special provisions for old age or pensions are outweighed by the tax advantages granted to (high) earnings (e.g. deduction for professional expenses, see also Verbist, 2004 for more details on the various tax

50.0 45.0 40.0 35.0

%

30.0 25.0 20.0 15.0 10.0 5.0 0.0 PT

GE

GR

FR

IR

SP

UK

IT PT

LU

NL

AT

BE

SW

FI

DK

SIC

Fig. 4a. Taxes (PT) and Social Contributions (SIC) Paid on Pensions and Earnings (Left and Right Bar Respectively for Each Country) as a Percent of the Individual Gross Income Component in the Top Quintile, EUROMOD 1998. Countries are Ranked in Ascending Order of Taxes and Contributions on Pensions. Source: EUROMOD.


89

5000 4500 4000 3500

Eur

3000 2500 2000 1500 1000 500 0 DK

IR

GR

PT

SP

FI

SW

IT

Gross

UK

BE

GE

FR

AT

LU

NL

Net

Fig. 4b. Pensions and Earnings (Left and Right Bar Respectively for Each Country) in the Top Quintile, Euro PPP, EUROMOD 1998. Countries are Ranked in Ascending Order of Net Pensions. Source: EUROMOD.

allowances, deductions and credits). The most notable exceptions are Germany and Portugal: the tax liability of high income pensions is still remarkably lower than that of earnings, due to the considerable tax deduction for pensions. It is also remarkable that the countries with the highest pensions (the Netherlands, Luxembourg, Austria, France and Germany) do not have the highest liability on these pensions. The highest tax burden is found again in the Scandinavian countries and in Belgium, who have a pension level in the middle of the country ranking (Fig. 4b). Exception is Denmark, which has the lowest net pension of all countries, mainly because its pension system is not yet fully ‘mature’. For unemployment benefits, we only present here the results for the bottom quintile (see Figs. 5a and b), as the number of unemployed at the top of the income distribution is rather small. The highest tax burden on unemployment benefits is found in Scandinavia (mainly PT) and the Netherlands (mainly SIC). These are also the countries with the largest difference between gross and net unemployment benefits. In all other countries, gross and net unemployment benefits are (almost) the same in the bottom quintile.

90

GERLINDE VERBIST 35.0 30.0 25.0

%

20.0 15.0 10.0 5.0 0.0 UK

IR

PT

AT

GE

BE

FR

GR

SP

IT

FI

NL

DK

-5.0 PT

SW

SIC

Fig. 5a. Taxes (PT) and Social Contributions (SIC) Paid on Unemployment Benefits and Earnings (Left and Right Bar Respectively for Each Country) as a Percent of the Individual Gross Income Component in the Bottom Quintile, EUROMOD 1998. Countries are Ranked in Ascending Order of Taxes and Contributions on Unemployment Benefits. Source: EUROMOD.

1600 1400 1200

Eur

1000 800 600 400 200 0 GR

UK

IT

PT

SP

NL

FI

FR

Gross

AT

DK

BE

SW

GE

IR

Net

Fig. 5b. Gross and Net Unemployment Benefits and Earnings (Left and Right Bar Respectively for Each Country) in the Bottom Quintile, Euro PPP, EUROMOD 1998. Countries are Ranked in Ascending Order of Taxes and Contributions on Net Unemployment Benefits. Source: EUROMOD.


91

4.2. Progressivity of Taxes on Pensions, Unemployment Benefits and Earnings The differences in tax liabilities between the bottom and top of the income distribution can be summarised by the degree of progressivity. We compare progressivity of taxes on replacement incomes for the various countries with progressivity of taxes on earnings. In the literature progressivity is measured as the deviation of a tax system from proportionality. We use the Kakwani index PK T to measure progressivity (Kakwani, 1977, 1984). Kakwani indices are calculated here with respect to the respective income components: income from work, pensions and unemployment benefits. This means that progressivity of taxes on income component i (Ti on Yi) is measured as the difference between the Gini coefficient G of Yi (which equals the concentration coefficient of a proportional tax) minus the concentration coefficient C of Ti: PK T i ¼ G Yi C Ti PT on earnings are most progressive in France (Kakwani of 0.344) and least progressive, rather proportional in Denmark (Kakwani of 0.053) (Table 6). This conforms to the general progressivity indices calculated for all individuals (see Verbist, 2004). SIC on earnings are more or less proportional in all countries. Only Spain has a more pronounced regressive effect, mainly due to its upper bound for SIC liability. In almost all countries progressivity is much smaller for PT on earnings than on pensions. The only exceptions are Denmark and Greece, where taxes on earnings have higher progressivity index. PT on pensions are most progressive in Germany, and they are far more progressive than taxes on work, mainly because of the extensive tax relief granted to pensioners. Also for old-age individuals, taxes are least progressive, i.e. proportional in Denmark. SIC on pensions are progressive in Belgium, Finland and France, whereas in the Netherlands they are regressive. In most countries where taxes are levied on unemployment benefits, these taxes are proportional. Only in France there is a stronger inclination towards progressivity and in Belgium towards regressivity. As France has a highly progressive personal income tax system it is not surprising that taxes on unemployment benefits also exhibit a more progressive pattern. The negative sign of the Kakwani index of taxes on unemployment benefits in Belgium is somewhat surprising, as there is a tax credit for low unemployment benefits. This tax credit, however, decreases when the benefit is combined with

92

GERLINDE VERBIST

Table 6. Kakwani Index of Personal Taxes (PT) and Social Contributions (SIC) on Earnings, Pensions and Unemployment Benefits, EU-15, 1998. Earnings (W)


Pensions (P)

Unemployment Benefits (UB)

PT on W

SIC on W

PT on P

SIC on P

PT on UB

SIC on UB

0.215 0.150 0.053 0.071 0.344 0.127 0.296 0.153 0.102 0.301 0.261 0.272 0.187 0.064 0.111

0.058 0.000 0.043 0.006 0.030 0.043 0.072 0.032 0.066 0.064 0.048 0.000 0.167 0.067 0.042

0.313 0.437 0.051 0.288 0.408 0.562 0.254 0.465 0.260 0.319 0.433 0.437 0.402 0.144 0.450

0.006 0.335 – 0.223 0.208 0.069 0.000 – – 0.005 0.144 – – – –

– 0.122 0.009 0.012 0.160 – 0.084 – 0.017

– – 0.071 0.018 0.236 – – – –

0.060 – 0.036 0.011 0.032

0.009 – 0.079 0.003 –

Source: EUROMOD.

earnings. As the lowest unemployment benefits are often combined with income from work (from the partner in the household), the tax credit does not or hardly apply for these groups, thus probably explaining the more regressive pattern of PT. In most countries no SIC on unemployment benefits are levied. In those countries where they are, the distribution is proportional. Only in France, these contributions are progressive, and even much more than SIC on earnings, due to the progressive rate structure of the ‘Cotisation Sociale Geńe´raliseé’ on unemployment benefits.

5. CONCLUSION In this chapter we have studied the interaction between the tax system and social benefits. We are well aware that this is only a part of the redistributive process through taxes and benefits. But it is a part that is often neglected; undeservedly, as we have tried to show.


93

As we have seen, there are important differences in the way PT and SIC are levied on the income components we have studied here (earnings, pensions and unemployment benefits). There is also a wide variety among countries in the EU-15. On average old-age pensions and unemployment benefits are taxed at a lower rate than income from work in all the EU-15 countries. There are three reasons for this. First, gross income of old age and unemployed individuals is on average lower, and as personal income tax systems are progressive this will lead to a higher tax burden for workers. Second, some countries have special provisions in their personal income tax system for old age and for pension and unemployment income, the so-called social fiscal measures. Third, pensioners and unemployed persons pay less in SIC than workers. This has consequences for the relative income position of replacement income recipients: the relative income position of both the elderly and the unemployed is better in net terms than in gross terms. These factors are especially important at the bottom of the income distribution, e.g. many of the social-fiscal measures are targeted at lower incomes. This is apparent in the large differences between taxes on pensions and taxes on earnings in the bottom quintile. The same applies for taxes on unemployment benefits. The differences are smaller in the top quintile, especially for personal income taxes. It is also remarkable that the countries with the highest pension levels are not the ones that charge the heaviest liabilities. In the Scandinavian countries we find the highest taxes levied on unemployment benefits and pensions, even though these income categories are not the highest in the EU. These three countries have an overall high tax level and an almost proportional structure, also for replacement incomes; the impact of social-fiscal measures is rather limited. In other countries, these special provisions have a considerable impact on tax liabilities of benefit recipients: the old age tax allowance in Ireland and the UK, the tax credit in Belgium, the substantial tax deduction in Germany are some important examples of measures that make pension incomes in the lower end of the income distribution as good as tax exempt. As the relative income position of the elderly is enhanced by the income tax system, these taxes also play a role in the process of intergenerational solidarity. This is also relevant for international comparisons, as the ranking of countries according to the relative income positions of old-age individuals is different for gross and net incomes. Taxes on pensions are in almost all EU countries (with Denmark and Greece as the only exceptions) more progressive than taxes on earnings; this indicates that PT also reduce inequality within the old-age generation. Thus, if one wants to investigate the

94

GERLINDE VERBIST

redistribution between and within generations, it is not sufficient to analyse only the pension system. One also needs to incorporate the personal income tax system.

NOTES 1. The self-employed have been excluded, together with those individuals with an income that does not originate from work, i.e. unemployment benefits, pensions and invalidity allowances. 2. The age of 65 is the most common state pension age in EU countries; recent increases in pension age indicate a convergence on this level in the future (Disney & Whitehouse, 1999). The samples used in EUROMOD all refer to private households; this means that the elderly living in an institution are not represented in the study. The proportion of old-age individuals in institutional households varies widely among countries (for figures see OECD, 1996b). 3. In Greece and the UK the proportions of unemployment benefits are very low. In Greece, unemployment benefits are low, virtually flat, provided for a relatively limited period of time and with rather stringent conditions. Moreover, in the European Community Household Panel (ECHP) respondents are asked about their income during the previous year (so this applies not only for the Greek data, but also for the other countries that use ECHP in EUROMOD, see Appendix A). The combination of these two factors explains why the gross income of the unemployed consists for almost 75% of income from work, and that for most unemployed the unemployment benefit is rather a small supplement to their income. In the UK gross income of the unemployed includes mainly income support (35%), which is a general social assistance benefit; as it is not aimed specifically at the unemployed, we have not classified it as an unemployment benefit. Other important income components of the unemployed in the UK are income from work (15%) and housing benefits (17%). 4. For apportioning social contributions to replacement incomes, this formula had to be used for Finland, Luxembourg, the Netherlands and Sweden (see Verbist, 2005 for the exact specification). 5. These provisions are part of what Titmuss (1969) called ‘fiscal welfare’: ‘‘Allowances and reliefs from income tax, though providing similar benefits and expressing a similar social purpose in the recognition of dependencies, are not, however, treated as social service expenditures. While one is ‘a cash transaction’, the other is ‘an accounting convenience’. Despite this difference in administrative method, the tax saving that accrues to the individual is, in effect, a transfer’’. Titmuss wanted to make clear that in many countries the tax system can to some extent serve as an alternative for the social security system. This was translated by Adema et al. (1996) into the concept of ‘social-fiscal measures’. Kvist and Sinfield (1997) use the concept of ‘tax benefits’, analogous to social benefits. 6. The quintiles are constructed on the basis of equivalised household income at the personal level. Income (components) are corrected for differences in household size and composition with the modified OECD scale. According to this equivalence scale the first adult has a weight 1, every other adult 0.5 and each child


95

0.3. We assume that household income is equally shared among all household members.

REFERENCES Adema, W. (1999). Net social expenditure. Labour Market and Social Policy Occasional Papers nr. 39, OECD, Paris. Adema, W., Einerhand, M., Eklind, B., Lotz, J., & Pearson, M. (1996). Net public social expenditure. Labour Market and Social Policy Occasional Papers nr. 19, OECD, Paris. Bradshaw, J., & Finch, N. (2002). A comparison of child benefit packages in 22 countries. Research Report 74, Leeds, Department for Work and Pensions. Disney, R., & Whitehouse, E. (1999). OECD pension systems in crisis: What are the reform options? Pension Reform Primer Series, Social Protection Discussion Paper no. 9921, World Bank, Washington, DC. Fenge, R., & Werding, M. (2004). Ageing and the tax implied in public pension schemes: Simulations for selected OECD countries. Fiscal Studies, 25(2), 159–200. Immervoll, H., Levy, H., Lietz, C., Mantovani, D., O’Donoghue, C., Sutherland, H., & Verbist, G. (2006). Household incomes and redistribution in the European Union: Quantifying the equalising properties of taxes and benefits. In: The distributional effects of government spending and taxation (pp. 135–165). New York: Palgrave Macmillan. Immervoll, H., & O’Donoghue, C. (2001). Imputation of gross amounts from net incomes in household surveys. An application using EUROMOD. EUROMOD Working Paper no. EM1/01, Department of Applied Economics, University of Cambridge, Cambridge. Kakwani, N. (1977). Measurement of tax progressivity: An international comparison. Economic Journal, 87, 71–80. Kakwani, N. (1984). On the measurement of tax progressivity and redistributive effect with applications to horizontal and vertical equity. Advances in Econometrics, 3, 149–168. Keenay, G., & Whitehouse, E. (2003a). Financial resources and retirement in nine OECD countries: The role of the tax system. OECD Social, Employment and Migration Working Paper no 8, OECD, Paris. Keenay, G., & Whitehouse, E. (2003b). The role of the personal income tax system in old-age support: A survey of 15 countries. Fiscal Studies, 24(1), 1–21. Kvist, J., & Sinfield, A. (1997). Comparing tax welfare states. Social Policy Review, 9, 249–275. Ministry of Health and Social Affairs. (2003). After tax – the truth of the matter. Social expenditure in Sweden in 1991–2000. Stockholm: Ministry of Health and Social Affairs. O’Donoghue, C., & Sutherland, H. (1999). Accounting for the family in European tax systems. Cambridge Journal of Economics, 23, 565–598. OECD. (1984). Tax expenditures. A review of the issues and country practices, Paris. OECD. (1996a) Tax expenditures. Recent experiences, Paris. OECD. (1996b). Caring for frail elderly people: Policies in evolution, Social Policy Studies no.19, Paris. Sutherland, H. (2001). EUROMOD: An integrated European benefit-tax model, EUROMOD. Working Paper EM9/01, Department of Applied Economics, University of Cambridge, Cambridge. Titmuss, R. (1969). Essays on the welfare state. London: Unwin University Books. Verbist, G. (2002). An inquiry into the redistributive effect of personal income taxes in Belgium. Doctoral dissertation, University of Antwerp.

96

GERLINDE VERBIST

Verbist, G. (2004). Redistributive effect and progressivity of taxes: An international comparison across the EU using EUROMOD. EUROMOD Working Paper EM5/04, Department of Applied Economics, University of Cambridge, Cambridge. Verbist, G. (2005). Replacement Iincomes and taxes: A distributional analysis for the EU-15 countries. EUROMOD Working Paper EM2/05, Department of Applied Economics, University of Cambridge, Cambridge.

APPENDIX A Table A1.

EUROMOD-1998 Datasets.

Country

Base Dataset for EUROMOD

Date of Collection

Austria

Austrian version of European Community household panel (W5) Panel survey on Belgian households (W6) European Community household panel (W2) Income distribution survey Budget de Famille

1999

Annual 1998

1999

Annual 1998

1995

Annual 1994

1998 1994/1995 1998

Annual 1998 Annual 1993/ 1994 Annual 1997

1995

Annual 1995

1994 1996

Month in 1994 Annual 1995

1999 1996

Annual 1998 Annual 1995

1996

Annual 1995

1996

Annual 1995

1997 1995/1996

Annual 1997 Month in 1995/1996

Belgium Denmark Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain Sweden UK

German socio-economic panel (W15) European Community Household panel (W2) Living in Ireland survey (W1) Survey of households income and wealth PSELL-2 (W5) Sociaal-economisch panelonderzoek (W3) European Community household panel (W3) European Community household panel (W3) Income distribution survey Family expenditure survey

Source: Sutherland (2001).

Reference Time Period for Incomes


Table A2.

97

Purchasing Power Parities used in EUROMOD-1998 (31.07.1998).

Country Austria Belgium Denmark Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain Sweden UK

Exchange Rate (a)

PPP (b)

EURO PPP ¼ (a) (b)

13.76030 40.33990 7.51130 5.94573 6.55957 1.95583 340.75000 0.78756 1936.27000 40.33990 2.20371 200.48200 166.38600 8.80730 0.67833

1.0568 0.9852 1.2374 1.1782 1.0434 1.0629 0.8016 0.9903 0.8814 1.0147 0.9505 0.7200 0.8399 1.2249 1.0773

14.54 39.74 9.29 7.01 6.84 2.08 273.15 0.78 1706.63 40.93 2.09 144.35 139.75 10.79 0.73

APPENDIX B Table B1. Proportion of Workers, Old-Age Individuals (65+) and Unemployed in Entire Population (N ¼ All Individuals in Sample). Country (N) Austria (7,386) Belgium (7,057) Denmark (7,044) Finland (26,902) France (29,158) Germany (18,722) Greece (15,062) Ireland (14,585) Italy (23,924) Luxembourg (6,566) Netherlands (11,035) Portugal (14,468) Spain (18,991) Sweden (38,756) UK (16,586) Source: EUROMOD.

Workers

Pensioners

Unemployed

30.3 26.1 30.3 19.6 24.8 29.6 18.3 20.3 21.8 31.2 28.6 27.5 18.1 20.4 30.9

16.4 17.4 13.2 14.3 15.4 17.4 17.6 10.9 16.4 14.9 13.1 14.7 16.5 18.0 15.1

3.9 7.6 9.6 12.5 7.3 5.9 5.5 10.6 6.9 0.8 8.6 3.0 10.1 9.5 3.9

98

GERLINDE VERBIST

Table B2. Gross and Net Pensions Per Old-Age Individual, Personal Taxes (PT%) and Social Insurance Contributions (SIC%) Paid by OldAge Individuals as a % of Their Gross Pension Income, 1st and 5th Quintile, 1998. 1st Quintile


5th Quintile

Gross

Net

PT%

SIC%

Gross

Net

PT%

SIC%

608 403 643 534 497 510 175 354 423 899 755 213 301 573 407

570 400 491 480 492 472 166 354 412 863 653 213 300 479 407

2.6 0.4 23.7 8.4 0.1 0.0 0.8 0.0 2.7 1.3 2.2 0.0 0.4 16.4 0.0

3.7 0.4 0.0 1.7 0.9 7.3 4.0 0.0 0.0 2.6 11.2 0.0 0.0 0.0 0.0

2254 1672 358 1546 1808 1604 827 483 1365 2305 2398 883 1081 1561 1311

1642 1152 228 984 1559 1441 720 405 1060 1763 1806 814 907 1061 1083

23.6 26.3 36.2 32.2 7.0 5.7 9.0 16.1 22.3 21.0 19.4 7.8 16.2 33.9 17.4

3.6 4.8 0.0 4.1 6.7 4.5 4.0 0.0 0.0 2.5 5.2 0.0 0.0 0.0 0.0

Source: EUROMOD.

Table B3. Gross and Net Earnings Per Working Individual, Personal Taxes (PT%) and Social Insurance Contributions (SIC%) Paid by Working Individuals as a % of Their Gross Income From Work, 1st and 5th Quintile, 1998. 1st Quintile

Austria Belgium Denmark Finland France Germany Greece Ireland

5th Quintile

Gross

Net

PT%

SIC%

Gross

Net

PT%

SIC%

1135 901 993 1225 925 975 523 762

913 740 653 861 726 727 447 720

2.2 6.1 22.6 22.3 1.0 5.2 0.3 3.1

17.3 11.7 11.7 7.4 20.5 20.2 14.2 2.3

3001 2629 3025 2308 3012 3049 2093 2371

1942 1648 1691 1427 2216 1871 1587 1679

20.3 25.7 34.0 30.6 7.2 23.7 14.4 25.2

15.0 11.6 10.1 7.6 19.2 15.0 9.8 4.0


99

Table B3.

(Continued).

1st Quintile

Italy Luxembourg Netherlands Portugal Spain Sweden UK

5th Quintile

Gross

Net

PT%

SIC%

Gross

Net

PT%

SIC%

1034 1463 1420 462 825 809 663

851 1272 1071 404 732 543 583

8.9 0.6 4.3 1.6 5.1 26.0 6.1

8.9 12.5 20.3 11.0 6.3 6.9 5.9

2388 4614 3099 2165 2531 2344 2657

1633 3188 2054 1507 1934 1417 1990

22.7 22.1 16.5 19.4 20.8 34.0 18.7

9.0 8.8 17.2 11.0 2.8 5.6 6.4

Source: EUROMOD.

Table B4. Gross and Net Unemployment Benefits Per Unemployed Individual, Personal Taxes (PT%) and Social Insurance Contributions (SIC%) Paid by Unemployed Individuals as a % of their Gross Unemployment Income, 1st Quintile, 1998. 1st Quintile

Austria Belgium Denmark Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain Sweden UK Source: EUROMOD.

Gross

Net

PT%

SIC%

284 432 574 161 183 311 13 400 47 165 62 120 403 27

284 431 402 130 182 311 13 400 43 129 62 117 273 27

0.0 0.4 26.7 18.0 0.1 0.0 0.8 0.0 6.7 3.9 0.0 0.8 25.3 1.1

0.0 0.0 3.2 1.4 0.7 0.0 0.0 0.0 0.0 18.1 0.0 1.8 6.9 0.0

FAMILY TRANSFERS AND CHILD POVERTY IN GREECE, ITALY, SPAIN AND PORTUGAL Manos Matsaganis, Cathal O’Donoghue, Horacio Levy, Manuela Coromaldi, Magda Mercader-Prats, Carlos Farinha Rodrigues, Stefano Toso and Panos Tsakloglou ABSTRACT The paper examines the effect of family transfers on child poverty in Greece, Italy, Spain and Portugal. Family transfers are defined as to include non-contributory child benefits, contributory family allowances and tax credits or allowances. The drive to reduce child poverty is of particular interest in southern Europe, where public support to poor families with children is often meagre or not available at all. The paper uses the European cross-country microsimulation model, EUROMOD, to assess the distributional impact of existing family transfers and to explore the scope for policy reforms, before it concludes with a discussion of key findings and policy implications.


101

102

MANOS MATSAGANIS ET AL.

1. INTRODUCTION Child poverty has risen to prominence as a distinct issue of social policy over the last few years. On the one hand, policies against child poverty enjoy much wider support than general anti-poverty policy. On the other, there is sufficient evidence on the social costs of child poverty and the benefits of early intervention (Esping-Andersen, Gallie, Hemerijck, and Myles, 2002; Kamerman, Neuman, Waldfogel, & Brooks-Gunn, 2003). In the light of that, the Clinton administrations in the United States greatly expanded the scope of earned income tax credit, which has now become the main instrument for the provision of income support to families (Moffitt, 2002). In Britain, the Labour government has committed itself to halving child poverty by the year 2010. A variety of policy instruments are to be employed, including substantial improvements to universal Child Benefit and of the child supplements to means-tested Income Support, as well as the extensive use of in-work benefits (Piachaud & Sutherland, 2001; Brewer, 2003). Furthermore, the European Commission’s contribution to the Lisbon summit in March 2000 included a proposal to halve child poverty by 2010. The proposal was not endorsed by the European Council, but the Social Inclusion Process confirmed the greater visibility of anti-poverty policy at the level of the EU. Indeed, the elaboration of biennial National Action Plans since May 2001 has been in many countries, including those examined here, the occasion for initiatives specifically targeted to children (Ferrera, Matsaganis, & Sacchi, 2002; CEC, 2003; Ruxton & Bennett, 2002). The drive to reduce child poverty is of particular interest in the context of southern Europe. If anything, ‘‘familialism’’ has long been identified as a special ingredient of south European welfare states. At first glance, it might be thought that in such a context families and children are well looked after. Rather paradoxically perhaps, this is not the case. On the one hand, family activism in the domain of social policy has proved far from fully effective in terms of preventing child poverty. The mobilisation of family resources to bail out relatives at risk of poverty requires that such resources are adequate in the first place, even when the existence of families or their willingness to help is not an issue. On the other, the ‘‘subsidiary’’ role of the state in matters of family policy has often implied that formal programmes of public assistance to poor families with children are meagre or not available at all (Ferrera, 1996; Rhodes, 1996; Saraceno, 1997; Matsaganis, Ferrera, Capucha, & Moreno, 2003).

Family Transfers and Child Poverty

103

This paper aims to assess the impact of family transfers on child poverty in Greece, Italy, Spain and Portugal. The analysis relies on EUROMOD, a cross-country comparative benefit-tax model for all 15 members of the EU. Microsimulation models like EUROMOD allow users not only to evaluate the impact of existing tax and benefit measures but also to simulate the effects of alternative policy reforms. Both features are brought to use here. The structure of the paper is as follows. The next section briefly describes the data and methodology used here. Section 3 assesses the distributional impact of existing family transfers. Section 4 estimates target efficiency of family transfers. Section 5 simulates the effects of alternative reforms. The paper concludes with a discussion of key findings and their policy implications.

2. DATA AND METHODOLOGY This paper relies on EUROMOD, a cross-country comparative benefit-tax model. The model simulates a variety of taxes and benefits in each of the 15 countries of the EU. The policy instruments simulated here include income taxes, social insurance contributions, social assistance benefits, unemployed benefits, housing benefits, family benefits and, where possible, social insurance benefits. EUROMOD simulates policy rules on the original micro-data sets drawn from family income surveys. The data used here are derived from the Bank of Italy Household Income Survey (1995) and from the European Community Household Panel for Greece (1995), Portugal (1996) and Spain (1996). Income data have been updated to the year 1998, using appropriate adjustment factors by country and by income source.1 The advantages of a microsimulation model such as EUROMOD are quite obvious. Benefit information is normally collected as part of surveys such as the above. Nevertheless, the benefits of interest (here family transfers) are often difficult to identify because of aggregation. Moreover, income taxes are almost impossible to read off the original data. Finally, the data year is inevitably not the same as the policy year under examination. As a result, it is impossible to estimate the effect of policies introduced or modified after the data were collected – except through microsimulation. In view of all this, the model constitutes a powerful tool for research on reforms to social security and the fiscal system in a comparative perspective. Equally obvious are the disadvantages. EUROMOD is a static model, based upon purely arithmetical calculations. As such, it cannot account for

104


behavioural responses, such as those related to labour supply decisions, when simulating the effects of policy changes. Moreover, due to data limitations, it should be noticed that in-kind benefits and publicly provided services are not examined here. This is an important omission: non-cash benefits have an important effect on family and child welfare and play a major role in the policy debate in many countries. This is an issue we return to later in the paper. A further set of methodological problems is potentially more amenable to treatment. The application of policy rules to a given population implies that these rules are fully adhered to. Of course, this is not true in the real world. On the one hand, not all individuals claim the benefits they are entitled to. It is known that non-take up is caused by fear of stigma, incomplete information about entitlements, administrative errors and other reasons (Atkinson, 1996; Duclos, 1995). It is also known that the extent of non-take up can be considerable with respect to means-tested benefits, though much less so with respect to universal benefits (van Oorschot, 1996; Ripahn, 2001; DWP, 2004). However, non-take up of social benefits in the countries of southern Europe is neglected as a policy issue and relatively overlooked as a research topic – for a review see Hernanz, Malherbet, and Pellizzari (2004). Similarly, there may be ‘‘leakage’’ of means-tested benefits to non-entitled households or individuals. For the purposes of this paper, the impact of family transfers is assessed as if all benefits were perfectly targeted, in the sense of being fully taken up by all legitimate claimants and received by no illegitimate ones. On the other, not all individuals pay the taxes they are liable to. Tax evasion is known to constitute a serious issue, all the more so in the countries of southern Europe. Again, no adjustment is made to the data, as if the incomes reported in the surveys on which the model relies were the same as the incomes declared to the authorities for the purposes of assessing both liability to income tax as well as eligibility to income-related benefits. The implications of the twin assumptions of perfect tax compliance and perfect targeting are discussed in the conclusions. The objective of this paper is to assess the effect of family transfers on child poverty and to examine the role of reforms that are variations on the theme of universal child benefits. In line with the recommendation of UNICEF, children are defined as individuals below 18 years of age. Differences in household size and composition are dealt with by applying the modified OECD equivalence scale, as used by EUROSTAT. Household income is assumed to be equally shared among household members. The poverty line is set at 60 per cent of the median equivalent disposable income in the ‘‘baseline scenario’’ and is held constant in simulated reforms.


105

The term ‘‘family transfers’’ is used broadly to include non-contributory child benefits, contributory family allowances and tax credits or allowances for dependent children. Recently, Levy (2003) and Corak, Lietz, and Sutherland (2005) adopted a still broader definition of child transfers: one that includes all child-related elements, anywhere in the tax and benefit system. The practice of adding special allowances for dependent children to instruments as diverse as unemployment benefits, old-age pensions, housing benefits, etc., is quite diffuse. For instance, as shown in Corak et al. (2005), in Greece pensions contribute up to 25 per cent of all child-related transfers, in Portugal 17 per cent of such transfers is accounted for by child elements in the guaranteed minimum income scheme, while in Italy and in Spain childrelated elements other than child benefits and tax relief seem to be negligible. On the other hand, such elements are significant elsewhere in Europe. For example, 24 per cent of all child-related transfers in Britain is contributed by child elements in Income Support, in Sweden housing benefit and social assistance account for 19 per cent and 4 per cent of such transfers respectively, while in Denmark housing benefit, social assistance and day-care benefit contribute 6–7 per cent each to child-related transfers. In view of these findings, child-related elements other than child benefits and tax relief do seem significant. In contrast, our study looks at income transfers, specifically targeted at children, whether provided via social security or via the tax system. The implications of alternative definitions of family transfers are discussed later in the paper.

3. FAMILY TRANSFERS AND CHILD POVERTY In all four countries of southern Europe income transfers to families include occupational family allowances, non-contributory benefits and tax relief for dependent children. Average values of family transfers for several combinations of family income and family size are shown in Table 1. As Table 1 shows, cash benefits are more substantial than tax relief, except for middle-to high-income families in Spain. The value of cash benefits diminishes with income, though less clearly so in Greece. Tax relief favours higher-income families in all four countries. Lastly, the value of family transfers as a whole generally increases with the number of children. What is the distributional impact of family transfers in southern Europe? Empirical evidence on this question, drawing on estimates produced with the European tax-benefit model EUROMOD, is presented below, followed by a formal analysis of target efficiency.

106

Table 1. Estimated Value of Family Transfers per Child. Family Size

Family Income

Greece

Italy

Spain

Portugal

Cash Benefits Tax Relief Cash Benefits Tax Relief Cash Benefits Tax Relief Cash Benefits Tax Relief 1 Child

2 Children

3 Children

88 71 114 76 72 131 319 313 172 327 457 275

67 68 84 27 71 79 11 83 108 47 88 132

250 267 41 640 390 29 569 664 93 177 895 47

188 202 203 120 165 159 85 147 161 71 129 163

36 16 5 124 9 2 165 24 0 203 95 36

114 138 126 60 158 176 42 185 210 0 113 210

191 188 185 233 202 182 227 225 217 254 308 249

80 87 89 49 86 96 38 76 85 24 24 80

Notes: All values are annual transfer per child in 1998, not adjusted for purchasing power differences between countries. Family income is expressed in terms of average full-time earnings of male employees in each country. Low family incomes: under 75 per cent of average earnings. Middle family incomes: 75–175 per cent of average earnings. High family incomes: over 175 per cent of average earnings. Average full-time earnings of male employees were h10,253 in Greece, h17,300 in Italy, h14,212 in Spain and h9,441 in Portugal. Family income is equivalised. The modified OECD equivalence scale is used, assigning a value of 1.0 to the first adult, 0.3 to children below 14 and a value of 0.5 to other household members. Source: EUROMOD.


4+ Children

Low Middle High Low Middle High Low Middle High Low Middle High


Table 2. Income Decile

1 2 3 4 5 6 7 8 9 10 Total

107

Distribution of Family Transfers by Decile of Equivalent Disposable Income. Greece

Italy

Spain

Portugal

Cash Benefits

Tax Relief

Cash Benefits

Tax Relief

Cash Benefits

Tax Relief

Cash Benefits

Tax Relief

7 9 9 12 8 12 8 11 15 10

2 5 7 11 11 12 12 14 14 12

19 23 25 18 8 4 3 1 0 0

11 12 12 12 10 9 10 8 8 8

66 20 7 4 2 1 1 0 0 0

1 10 13 11 14 11 8 6 13 15

10 11 14 11 9 10 8 8 9 8

1 6 10 12 12 12 12 11 12 12

100

100

100

100

100

100

100

100

Note: Estimates for the year 1998. The unit of analysis is individuals ranked by non-decreasing disposable equivalent household income. Source: EUROMOD.

Table 2 shows the incidence of total expenditure on family cash benefits and child tax relief separately by decile of equivalent disposable income in each of the four countries. Cash benefits to families in Spain and in Italy seem to target the bottom of the income distribution, as the four poorest deciles account for approximately 97 per cent and 85 per cent of all benefit, respectively. Conversely, the proportion of total expenditure received by the four richest deciles is 1 per cent in Spain and 4 per cent in Italy. This effect is less marked in the case in Portugal and reversed in that of Greece: while in the former the ratio of total benefit received by the four poorest and the four richest income deciles is 46–33 per cent, in the latter it is 37–44 per cent (i.e. more benefit to higher-income families). In contrast, child tax relief tends to be more evenly distributed among taxpayers (i.e. except to lower income groups). As a matter of fact, the ratio of the amount received by the upper half of the distribution relative to that received by the bottom half is 1.8 in Greece, 1.4 in Portugal and 1.1 in Spain, while in Italy the distribution of tax relief is skewed in favour of lower incomes. Table 3 shows the incidence of family cash benefits and child tax relief by decile in terms of income share. Focusing on the poorest decile alone, cash

108


Table 3. Income Decile

Income Share of Family Transfers by Decile of Equivalent Disposable Income. Greece

Italy

Spain

Portugal

Cash Benefits

Tax Relief

Cash Benefits

Tax Relief

Cash Benefits

Tax Relief

Cash Benefits

Tax Relief

1 2 3 4 5 6 7 8 9 10

1.1 0.9 0.7 0.8 0.4 0.6 0.4 0.4 0.4 0.2

0.2 0.3 0.3 0.4 0.4 0.4 0.3 0.3 0.2 0.1

6.1 3.2 3.0 1.8 0.6 0.3 0.2 0.1 0.0 0.0

2.2 1.1 1.0 0.8 0.5 0.4 0.4 0.3 0.2 0.1

3.3 0.4 0.1 0.1 0.0 0.0 0.0 0.0 0.0 0.0

0.1 0.6 0.7 0.5 0.6 0.4 0.2 0.2 0.3 0.2

2.6 2.3 2.7 1.9 1.3 1.2 0.8 0.7 0.5 0.2

0.1 0.5 0.7 0.7 0.6 0.5 0.4 0.3 0.3 0.1

Total

0.4

0.3

0.6

0.4

0.1

0.3

0.9

0.3

Note: Estimates for the year 1998. The unit of analysis is individuals ranked by non-decreasing disposable equivalent household income. Source: EUROMOD.

benefits contribute about 1 per cent of total family income in Greece, 3 per cent in Spain and in Portugal, 6 per cent in Italy. Except in Italy (over 2 per cent), the relative value of tax relief to the bottom decile is negligible. On the whole, the distribution of cash benefits to families by income group seems to be more strongly progressive in Italy and, to some extent, in Portugal. In Spain, family cash benefits make an appreciable contribution to the incomes of families in the lowest decile, but taper off rapidly as income rises. The income share of child tax relief is much lower everywhere. Except in Italy, families in the bottom decile take little advantage of tax relief compared to those immediately above them in terms of income. Overall, the weight of tax relief is lower than that of cash benefits everywhere except in Spain. All these findings are in line with the previous discussion of entitlements. Such variations between countries are vividly depicted in Graph 1. The graph contains four lines: the concentration curves of family cash benefits, child tax relief, all family transfers taken together and equivalent disposable income (net of family transfers), where the members of the population are ranked in ascending order according to their equivalent disposable income. The concentration curves show the cumulative distribution of the


109 Greece

Cumulative income share

100 80 60 40 20 0 0

20

40 60 80 Cumulative population share by equivalent income Cash benefits Tax relief All family transfers Disposable income (net of family transfers) Line of perfect equality

100


Italy 100 80 60 40 20 0 0

20

40 60 Cumulative population share by equivalent income

80

Cash benefits Tax relief All family transfers Disposable income (net of family transfers) Line of perfect equality

Graph 1.

Concentration Curves.

100

110



Spain 100 80 60 40 20 0 0

20


80

100



Portugal 100 80 60 40 20 0 0

20


80


Graph 1. (Continued)

100


111

corresponding variables. As the members of the population are ranked according to their equivalent disposable income, the concentration curve of equivalent disposable income, net of family transfers, is the Lorenz curve of this distribution. Furthermore, the (diagonal) line of perfect equality is also depicted in the graph for expositional purposes. Since cash benefits are more important than tax relief in Greece, Italy and Portugal, the location of the concentration curve of all family transfers taken together in these countries is determined to a large extent by the location of the concentration curve of cash benefits. The opposite is observed in Spain, where the income share of cash benefits is relatively low. In Greece, where cash benefits are more evenly spread across the entire distribution, the corresponding concentration curve is close to the diagonal – while in the other three countries it lies above it. In fact, in Italy and Spain the concentration curves approach the top left corner of the graph, implying that these benefits are highly concentrated and clearly progressive. In all countries except Italy the concentration curve of tax relief lies below the diagonal. Actually, since in Greece, Spain and Portugal the poorest households do not benefit from tax relief, the concentration curve of tax relief in these countries crosses the Lorenz curve close to the bottom of the distribution. When all family transfers are taken together, the corresponding concentration curves in all countries lie above the Lorenz curve, implying that such transfers reduce aggregate inequality. However, it should be noted that in Italy and, to a lesser extent, in Portugal and Spain the concentration curve lies above the line of perfect equality. The opposite is true in Greece. Therefore, family transfers are arguably most redistributive in Italy, less progressive in Portugal and Spain, while they appear to be regressive in Greece. Table 4 presents the income share of family transfers and the average transfer per child by household type. In relative terms, the household types whose income increases the most after family transfers are couples with three or more children, followed by single parents with younger children. In absolute terms, the value of family transfers per child rises with the number of children in Greece and, to a lesser extent, in Italy, but the opposite is true in Portugal, while the corresponding profile is rather flat in Spain. Nevertheless, the most striking finding is that the overall value of family transfers in southern Europe is extremely low. For instance, couples with one or two children (i.e. the two family types that account for a majority of all children in southern Europe) seem to receive about h6 per child per month in Spain and in Greece, about h12 in Portugal and no more than h19 in Italy.2

112

Table 4.


Income Share of Family Transfers and Average Transfer per Child by Household Type. Income Share of Family Transfers (Per cent)

Couple with 1 child 0–17 Couple with 2 children 0–17 Couple with 3+ children 0–17 Lone parent with all children 0–17 Lone parent with at least 1 child 18+ Couple with at least 1 child 18+ Other household types

Annual Average Transfer per Child (h)

Greece

Italy

Spain

Portugal

Greece

0.9

1.6

0.7

2.0

80

1.8

4.2

1.7

3.9

6.0

8.1

3.2

3.7

5.9

0.5

Italy

Spain

Portugal

212

76

153

72

228

78

133

7.6

131

258

81

115

2.8

6.9

163

314

116

164

1.7

0.1

0.5

127

1,034

37

123

0.5

1.0

0.1

0.7

83

322

20

90

1.3

1.7

0.6

2.3

67

177

48

82

Note: Estimates for the year 1998. Source: EUROMOD.

The impact of family transfers on child poverty by household type is shown in Table 5. The figures can be interpreted as the proportional reduction in the number of children below the poverty line (left-hand panel) and in aggregate child income gap (right-hand panel) due to family transfers. In terms of headcount poverty, family transfers reduce the number of poor children by 19–21 per cent in Portugal and in Italy, and by 7–8 per cent in Spain and in Greece. With respect to household types, family transfers are more effective at taking children out of poverty if these live in large families. On the contrary, the anti-poverty performance of family transfers as regards one-child families is below average in Italy and disappointing in the other three countries: a proportional reduction of 5.9 per cent in Portugal, 2.7 per cent in Spain and no reduction at all in Greece. A similar picture emerges in terms of income gap ratios. On the whole, the aggregate child poverty gap (before family assistance) is reduced by 37 per cent in Portugal, 28 per cent in Italy and 11–12 per cent in Greece and in Spain. This reduction is greatest among families with three children and lowest among those with one child only. As explained earlier, family transfers have two components: cash benefits and tax relief. What is the relative contribution of each to the reduction of


113

Table 5. Impact of Family Transfers on Child Poverty by Household Type. Reduction in Number of Poor Children (Per cent)

Couple with 1 child 0–17 Couple with 2 children 0–17 Couple with 3+ children 0–17 Lone parent with all children 0–17 Lone parent with at least 1 child 18+ Couple with at least 1 child 18+ Other household types All households with children

Reduction in Child Income Gap (Per cent)

Greece

Italy

Spain

Portugal

Greece

Italy

Spain

Portugal

0.0

17.6

2.7

5.9

1.9

14.4

8.5

16.4

3.3

25.6

9.2

17.6

4.8

31.6

12.0

29.6

32.0

23.4

7.8

34.6

34.0

35.7

18.1

58.0

4.1

14.1

8.4

9.8

16.8

29.4

14.4

39.3

0.0

0.0

2.3

4.0

16.7

14.7

10.0

32.6

5.5

10.1

5.7

14.4

10.4

22.3

9.1

21.0

9.4

8.1

9.4

24.1

14.0

16.1

9.7

31.2

8.1

19.0

7.3

20.8

11.4

28.2

12.1

36.7

Note: Estimates for the year 1998. The figures show percentage reduction in the number of children below the poverty line and in aggregate child income gap respectively due to family transfers. Source: EUROMOD.

child poverty? An answer to that question is provided in Table 6. In terms of poverty rates, the impact of tax relief appears to be rather negligible, not exceeding a 1.2 percentage point reduction (in Spain). By contrast, family cash benefits seem to be most effective in Portugal and in Italy (a reduction of over 5 percentage points), much less in Greece (1 percentage point) and almost not effective at all in Spain. Overall, in comparison to what their level would have been in the absence of family transfers, child poverty rates are 6 percentage points lower in Portugal and in Italy, but less than 2 points lower in Greece and in Spain. The estimated values of the FGT index3 shed more light to the distributional impact of family transfers. Tax relief has a negligible impact on the index, causing a proportional reduction ranging from 5 per cent in Portugal and in Italy to 1–2 per cent in Greece and in Spain. In contrast, the corresponding reduction achieved by cash benefits is much stronger: 44 per cent in Portugal, 27 per cent in Italy, 14 per cent in Spain and 11 per cent in Greece. In fact, the case of Spain offers an illustration of the limitations of a

114


Table 6.

Redistributive Impact of Family Transfers. Greece

Italy

Spain

Portugal

Child poverty rate

Before family transfers After tax relief After cash benefits After all family transfers

18.5 18.2 17.4 17.0

32.7 31.9 27.7 26.5

23.4 22.2 23.2 21.6

29.2 28.6 23.7 23.1

FGT index (a ¼ 2)

Before family transfers After tax relief After cash benefits After all family transfers

3.8 3.7 3.3 3.3

6.5 6.2 4.7 4.5

4.3 4.3 3.8 3.7

2.5 2.4 1.4 1.4

Notes: Estimates for the year 1998. The poverty line is held constant at the actual level (i.e. at 60 per cent of national median equivalent disposable income after all family transfers). The poverty rate is the headcount ratio. The FGT index (a ¼ 2) attaches greater weight to larger poverty gaps. Before family transfers excludes the effect of both cash benefits and tax relief. After tax relief includes the effect of tax relief but excludes the effect of cash benefits. After cash benefits includes the effect of cash benefits but excludes the effect of tax relief. After family transfers includes the effect of both cash benefits and tax relief. Source: EUROMOD.

policy approach looking at the headcount rate alone: while tax relief lifts more children over the poverty line than cash benefits, its distributional impact further down the income scale as measured by FGT (a ¼ 2) is much weaker.

4. TARGET EFFICIENCY The preceding discussion of anti-poverty effectiveness raises an obvious question: are family transfers in southern Europe well targeted? The term ‘‘target efficiency’’ is often used loosely, especially in the policy debate, but can be formally measured through a set of indicators. In this paper four indicators are estimated. Vertical expenditure efficiency (VEE) measures the share of total benefit received by individuals below the poverty line. As seen in Fig. 1, VEE ¼ (A+B)/(A+B+C). Poverty reduction efficiency (PRE) is the fraction of total expenditure allowing poor individuals to approach or reach – but not cross – the poverty line. PRE is shown as (A)/(A+B+C). The spillover index (S) is a measure of the excess of expenditure relative to the amount strictly necessary to reach the poverty line, defined as (B)/(A+B). It can be seen that VEE(1S) ¼ PRE.


115

VEE = (A+B)/(A+B+C) PRE = (A)/(A+B+C) S = (B)/(A+B) PGE = (A) /(A+D)

post-transfer disposable income

poverty line

C B D A

pre-transfer disposable income

households ranked by income

Fig. 1. Target Efficiency of Social Transfers. Note: VEE, Vertical Expenditure Efficiency; PRE, Poverty Reduction Efficiency; S, Spillover and PGE, Poverty Gap Efficiency.

The three measures considered so far are useful in measuring vertical efficiency or the proportion of total benefit received by those below the poverty line. However, vertical efficiency cannot evaluate the effectiveness of a programme in fighting poverty. A transfer may be efficient in the sense that it is overwhelmingly targeted on the poor, but may fail to reach all those below the poverty line or its level may be too low to raise the living standards of beneficiaries significantly. This latter aspect is better captured by poverty gap efficiency (PGE), an indicator of horizontal efficiency, measuring the extent to which the transfers succeed in filling the aggregate poverty gap.4 PGE is equal to (A)/(A+D). The poverty gap itself can be either unweighted, when the parameter a of the FGT index is set equal to 1, or weighted to indicate greater concern for the condition of the poorest (higher values of a). All four indicators of target efficiency are presented in diagrammatic form in Fig. 1. The results of our estimation of target efficiency with respect to child poverty, separately for each class of family transfers (cash benefits and tax relief), can be seen in Table 7.

116


Table 7.

Target Efficiency of Family Transfers. VEE

Greece Italy Spain Portugal

Cash benefits Tax relief Cash benefits Tax relief Cash benefits Tax relief Cash benefits Tax relief

23.4 8.3 62.0 26.7 81.5 14.5 43.7 14.0

PRE

21.1 8.2 55.2 26.2 80.5 12.3 38.7 13.6

S

9.9 1.1 10.9 2.1 1.2 14.9 11.5 3.2

PGE a¼1

a¼2

a¼3

4.7 1.0 15.2 4.5 4.9 1.6 20.7 2.6

7.3 1.0 18.1 4.7 8.4 0.7 31.5 2.9

10.2 1.1 18.8 4.5 11.1 0.3 39.6 3.4

Notes: Estimates for the year 1998. All target efficiency indicators are computed with respect to child poverty. The poverty line is held constant at the actual level (i.e. at 60 per cent of national median equivalent disposable income after all family transfers). VEE denotes vertical expenditure efficiency and is defined as (A+B)/(A+B+C) in Fig. 1. PRE denotes poverty reduction efficiency and is defined as (A)/(A+B+C) in Fig. 1. S denotes spillover and is defined as (B)/(A+B) in Fig. 1. PGE denotes poverty gap efficiency and is defined as (A)/(A+D) in Fig. 1. Transfers to those below the poverty line are weighted equally when a ¼ 1, while transfers to the poorest are given more weight when a>1. The unit of analysis is individuals ranked by nondecreasing disposable equivalent household income. Source: EUROMOD.

In terms of vertical efficiency, as measured by PRE, family cash benefits seem to be best targeted in Spain (80 per cent of total expenditure). Targeting is less efficient in Italy (55 per cent of total expenditure), in Portugal (39 per cent) and in Greece (21 per cent). Child tax relief is clearly not targeted. In Italy, 74 per cent of all tax relief for dependent children is aimed above the poverty line. In the other three countries, the equivalent figure is closer to the 90 per cent mark. The other two indicators of vertical efficiency, VEE and S, reiterate the point presenting a similar picture. Our estimates are much less reassuring in terms of horizontal efficiency, as measured by PGE. The reduction of the unweighted aggregate poverty gap (a ¼ 1) caused by family cash benefits ranges from a low but significant of 21 per cent in Portugal and 15 per cent in Italy, to a rather disappointing 5 per cent in Spain and Greece. Obviously, attaching greater weight to the improvement of lowest incomes increases the value of the index: for a ¼ 3, the reduction of weighted aggregate poverty gap is 40 per cent in Portugal, nearly 20 per cent in Italy and just over 10 per cent in Greece and in Spain. In contrast, the anti-poverty impact of tax relief remains negligible in all cases.


117

Summing up the evidence on target efficiency presented here, two findings stand out. In terms of both vertical and horizontal efficiency family cash benefits are better targeted than child tax relief. It is clear that non-refundable tax instruments are unsuitable as a mechanism of income support to the poorest. In comparative terms, family transfers are better targeted in Italy and Portugal than in the other two countries. In Spain, cash benefits appear to exemplify the textbook case of stringent means testing: reserved for the poorest families alone, but not nearly adequate enough to improve significantly their standard of living. In the case of Greece, current policy seems to fail low-income families with children on both counts.

5. REFORMING FAMILY TRANSFERS The previous sections show that a common feature of actual family transfer programmes throughout southern Europe is that many families with children at risk of poverty are left with little or no income support. An obvious response to the problem of coverage gaps is the introduction of universal child benefits. This is a contentious solution, but has the advantage of being easy to explain and simple to implement. Such benefits are assumed to substitute (rather than be added on to) existing family transfer programmes. Universal child benefits differ with respect to parameters such as the value of benefit and eligibility conditions with respect to age. In this section, four versions are explored. In the case of reform I, the (flat) rate of benefit in each country has been chosen so as to match existing family transfers exactly in terms of fiscal costs, i.e. it is ‘‘budget neutral’’. Reforms II–IV mimic the British, Danish and Swedish child benefit schemes respectively. These were chosen to illustrate the effect of different structures of universal child benefit. In the Danish scheme the benefit rate is higher for younger children. Instead, in the British and the Swedish schemes the amount of benefit varies by the number of children in the family, but in opposite ways: the former pays a higher rate to the elder child, the latter rises in value with family size. In order to account for variations in living standards across the four countries, the level of each benefit is fixed as a proportion of average earnings.5 The benefit amount payable under each variation is presented in Table 8. Would universal child benefits of various kinds be more effective than current policies at reducing child poverty? Table 9 shows that, in terms of headcount rates, the impact of those reforms simulated here would be rather mixed. Reform I (budget neutral UCB) would not affect the child poverty rate in Greece, but would increase it by nearly 1 percentage point in Spain

118


Table 8. Child Benefit under Simulated Reforms. Greece Reform I:Budget neutral UCB

Italy

Spain

Portugal

197

582

135

284

Reform II: British child benefit

Eldest child All other children

401 326

676 551

555 452

369 300

Reform III: Danish child benefit

Children aged 0–3 Children aged 4–7 Children aged 8–18

594 540 422

1 003 912 711

824 749 584

547 498 388

Reform IV: Swedish child benefit

First two children

354

597

490

326

Third child Fourth child Fifth+ children

448 637 707

756 1,074 1,194

621 883 981

413 586 651

10,253

17,300

14,212

9,441

Average earnings of male full-time employees

Notes: All values are annual amounts per child in 1998, not adjusted for purchasing power differences between countries. All reforms involve the replacement of existing family transfers for children aged 0–17 by a universal child benefit. In the case of reform I, the (flat) rate of benefit in each country has been chosen so as to match exactly existing family transfers in terms of fiscal costs. In the case of reforms II–IV, the level of benefit in each country has been chosen so as to be exactly equivalent (as a proportion of average earnings of male full-time employees) to the British, Danish and Swedish child benefits. Source: EUROMOD.

and by around 2 percentage points in Italy and in Portugal. Reform III (Danish CB) would reduce the headcount rate by over 3.5 percentage points in Portugal and Spain and by 1.5 point in Greece, but would raise it by 0.5 point in Italy. The effect of reforms II (British CB) and IV (Swedish CB) would be to reduce headcount poverty in Spain and Greece, but raise it in Italy and (slightly) in Portugal. Results for the Foster-Greer-Thorbecke index (a ¼ 2) show that stronger the distributional impact of universal child benefits the greater the weight attached to changes at the bottom of the income distribution. Introducing a budget neutral universal child benefit in place of existing family transfers would have little impact on the FGT index either way. Reform III (Danish CB) would cause a proportional decline of the index in all four countries: by 28 per cent in Spain and Portugal, by 21 per cent in Greece and by 7 per cent in Italy. The other two reforms, II (British CB) and IV (Swedish CB), would


Table 9.

119

Impact of Simulated Reforms on Child Poverty. Greece Italy Spain Portugal

Child poverty rate (per cent) Existing family transfers Reform I: budget neutral UCB Reform II: British CB Reform III: Danish CB Reform IV: Swedish CB FGT index (a ¼ 2)

Existing family transfers Reform I: budget neutral Reform II: British CB Reform III: Danish CB Reform IV: Swedish CB

17.0 17.1 16.0 15.5 15.9

26.5 28.4 28.1 27.1 28.1

21.6 22.5 18.9 17.9 18.9

23.1 25.4 23.5 19.6 23.2

3.3 3.2 2.8 2.6 2.8

4.5 4.7 4.6 4.2 4.5

3.7 3.9 3.0 2.6 2.9

1.4 1.3 1.3 1.0 1.2

Notes: Estimates for the year 1998. As reforms are simulated the poverty line is held constant at 60% of national median equivalent disposable income after all existing family transfers. The poverty rate is the headcount ratio. The FGT index for a ¼ 2 attaches greater weight to larger poverty gaps. For more detail on the reforms simulated see Table 8. Source: EUROMOD.

Table 10.

Fiscal Effects of Simulated Reforms.

Existing family transfers Reform I: budget neutral UCB Reform II: British CB Reform III: Danish CB Reform IV: Swedish CB

Greece

Italy

Spain

Portugal

0.9 0.9 1.6 2.1 1.7

1.3 1.3 1.4 1.8 1.4

0.5 0.5 1.8 2.4 1.8

1.5 1.5 1.8 2.3 1.8

Notes: Estimates for the year 1998. Fiscal costs are expressed as proportion of aggregate nonequivalised disposable income. For more detail on the reforms simulated see Table 8. Source: EUROMOD.

reduce the value of the index in Spain (by 18–20 per cent), Greece (14–15 per cent) and Portugal (8–10 per cent), but would cause a small increase in Italy (1–2 per cent). These results make it clear that, provided it is pitched at a level high enough, a universal child benefit could have a considerable redistributive impact in southern Europe. That raises the question of cost. Clearly, the fiscal effect of introducing a universal child benefit would be a function of the level and scope of the benefit itself. However, it would also depend on the demographic profile of each country and the generosity of the family transfer programmes it would replace. The fiscal implications of existing programmes and simulated reforms are all presented in Table 10 below.

120


As the current cost of family transfers is low (0.5–1.5 per cent of aggregate disposable income), the reforms simulated here appear relatively costly. Reform III (Danish CB) would be the costliest of all, raising expenditure between 1.8 per cent and 2.3 per cent in the four countries. Reforms II (British CB) and IV (Swedish CB) would have a softer fiscal impact, bringing expenditure on income transfers to families with children to between 1.4 per cent and 1.8 per cent of disposable income.

6. CONCLUSION As the preceding analysis showed, existing programmes seem to perform at best modestly in terms of poverty reduction. Many poor families with children are ineligible for benefit (as in the case of Italy), or receive low sums (as in Spain and, to some extent, in Portugal) or both (as in Greece). Needless to say, this effect is even more pronounced with respect to tax benefits, as nonrefundable schemes exclude poor families by design.6 In the paper, family transfers are defined as child benefits plus family allowances plus tax relief for dependent children. Would our results have been different had we adopted a broader definition of family transfers, accounting for all child-related elements anywhere in the tax-benefit system? For all the differences between the two approaches, the results of Corak et al. (2005) confirm our key finding, that compared to the rest of the EU-15 family transfers in southern Europe are poorly funded and their distributive impact is weak – as well as several other findings, as for example that highincome families in Spain and in Greece receive more public support per child than low-income ones. Moreover, while the analysis presented here assumes 100% take up of all benefits, the take up of benefits in real life may be incomplete, often significantly so. Non-take up is known to be high when benefits are of low value, means-tested, or poorly administered. However, family transfers in southern Europe are likely to have at least one of these characteristics. In a few words, the redistributive performance of income transfers to families with children, shown to be weak under the assumption of full take up, is probably weaker still. In view of the above, the scope for improving the redistributive performance of income transfers to families with children through redesigning the structure of benefits is ample. Would universal benefits improve on the antipoverty performance of existing schemes?


121

Our results show that replacing current policies by universal child benefits would not reduce the number of children in poverty by much – and could even increase it. This can happen if current policies provide relatively generous benefits to a substantial proportion of families on low incomes (e.g. in Italy, where family transfers are both income tested and categorically targeted). Nonetheless, where existing policies leave coverage gaps, universal child benefits will improve the position of families at the bottom of the income scale but ineligible for current assistance. Bringing in the FGT index reveals that the performance of reforms improves when a poverty measure is adopted that registers improvements at the bottom of the income distribution. Among the child benefits simulated that are actually in operation elsewhere in Europe, the Danish scheme clearly emerges ahead of the others in terms of generosity: it would be the costliest, but also the one with the highest impact on child poverty in all countries of southern Europe. On the other hand, the British and Swedish schemes, although different in terms of structure, would have quite similar effects on child poverty and fiscal costs in the four countries studied. In general, a trade off between fiscal cost and poverty reduction operates, since more generous benefits have stronger distributional impact at a higher fiscal cost. In any case, current expenditure on family transfers in southern Europe is too low to expect significant improvements in terms of poverty reduction through a simple reallocation of public spending within this policy area alone. On the whole, a judicious combination of a universal, even if low, income base with targeted but non-categorical policies could be an effective way to improve coverage and reduce poverty at a reasonable cost to the tax payer. As explained earlier, the analysis presented here is restricted to income transfers alone. Ignoring benefits in kind and their distributional effect is common practice in current research on distributional issues. Although this must be partly driven by convenience, the methodological complexities of accounting for the distributional impact of benefits in kind should not be underestimated. Whatever the reason, the omission is regrettable. According to Abramovici (2002), transfers in kind are quantitatively less important than cash transfers in the field of family policy in the countries under examination, with the exception of Spain. Moreover, Abramovici (2002) demonstrates that non-cash transfers play a far more important role within the framework of family policies in Denmark, Sweden and, to a lesser extent, in the UK, than in southern Europe. No detailed studies of the distributional impact of such non-cash transfers can be found in the literature of any of the four countries under consideration. However, studies examining the overall

122


distributional impact of non-cash transfers in a comparative perspective suggest that their impact on reducing inequality and alleviating poverty should not be underestimated (Smeeding et al., 1993, van de Walle & Nead, 1995). Therefore, it is not unlikely that the inclusion of non-cash benefits in our analysis would have produced stronger distributional effects in the ‘‘baseline scenario’’. On the other hand, it seems equally likely that simulating the non-cash family benefits encountered in the social protection systems of Denmark, Sweden and, probably, the UK would have resulted in even more expensive but more redistributive outcomes in the four southern European countries. All in all, there can be little doubt that a concerted policy effort aimed to combat child poverty in Europe must assign a high priority to the universal provision of family services alongside cash benefits. To mention the most obvious example, high-quality publicly provided child care to allow mothers to work may well be a more promising route out of child poverty than relying on cash benefits alone, however generous. The final conclusion concerns the methodology used. Important policy questions, such as that posed here (‘‘what is the effect on child poverty of income transfers to families?’’), are too complex – and to some extent counterfactual – to answer without recourse to a benefit-tax model like EUROMOD. While microsimulation models are clearly not immune from limitations of their own, some of which were discussed here, the ability to simulate the full impact of policy reforms is their unique advantage.

NOTES 1. Space constraints do not allow for a discussion of policy developments since 1998 and their implications for our work, nor for a comparison of our results with official statistics. For a detailed analysis of both issues see Matsaganis et al. (2004), where it is shown that the findings of the paper broadly hold for later periods and are largely consistent with official statistics. 2. Low sample size seems to affect the estimates for some household types, as in the case of lone parent with at least one older child (i.e. aged over 18) in Italy. 3. The index proposed by Foster, Greer and Thorbecke (1984) simultaneously takes into account the poverty rate, the income gap and the extent of inequality among the poor for values of the poverty aversion parameter (a) greater than one. 4. In general, vertical efficiency measures poverty reduction due to a benefit as a proportion of all spending on that benefit. Instead, horizontal efficiency measures poverty reduction as a proportion of pre-benefit poverty. For an early analysis see Beckerman (1979). 5. For example, the eldest child rate under reform II (British CB) is 3.9 per cent of average male full-time earnings in all four countries, as in Britain in 1998. More


123

information on the child benefit programmes in Britain, Sweden and Denmark can be found in the EUROMOD country reports (http://www.iser.essex.ac.uk/msu/ emod/countries/). 6. In the meantime, refundable tax credit schemes have been actually introduced in two of the four countries: Greece (in 2002, for low-income families with children aged 6–16 at school) and Spain (in 2003, for working mothers with children aged below 3). Although assessing their impact is a subject of future research, these schemes seem unlikely to alter the regressive nature of tax relief for dependent children in the two countries.

ACKNOWLEDGMENTS The paper was written as part of the Micro Analysis of the European Social Agenda (MICRESA) project, financed by the European Commission under the Improving Human Potential programme (SERD-2001-00099), co-ordinated by Holly Sutherland, now at the University of Essex. The authors wish to acknowledge the comments and suggestions offered by participants to the various seminars, meetings and conferences where earlier versions were presented. Special thanks are due to Michael Fo¨rster, Olivier Bargain and an anonymous referee for their thorough review of the manuscript. Obviously, the authors alone are responsible for errors concerning use of the model and interpretation of results. As EUROMOD is continually being improved and updated, the results presented here should be viewed as work in progress. The version used here was 17A.

REFERENCES Abramovici, G. (2002). Social protection expenditure on cash benefits and on benefits in kind. Statistics in focus Theme 3 – 16/2002 population and living conditions. Luxembourg: Eurostat. Atkinson, A. (1996). On targeting and family benefits. In: A. Atkinson (Ed.), Incomes and the welfare state. Cambridge: Cambridge University Press. Beckerman, W. (1979). Impact of income-maintenance payments on poverty in Britain, 1975. Economic Journal, 89(354), 261–279. Brewer, M. (2003). The new tax credits. IFS briefing note 35. London: Institute of Fiscal Studies. CEC. (2003). Draft joint inclusion report. Commission Staff Working Paper. Com (2003). 773 final. Brussels: Commission of the European Communities. Corak, M., Lietz, C., & Sutherland, H. (2005). The impact of tax and transfer systems on children in the European Union. Innocenti Working Paper 2005-04. Florence: Unicef Innocenti Research Centre. Duclos, J.-Y. (1995). Modelling the take-up of state support. Journal of Public Economics, 58, 391–415.

124


DWP. (2004). Income-related benefits: Estimates of take-up in 2001/2002. London: Department for Work and Pensions. Esping-Andersen G., Gallie D., Hemerijck A., & Myles J. (2002). Why we need a new welfare state. Oxford: Oxford University Press. Ferrera, M. (1996). The ‘‘southern model’’ of welfare in social Europe. Journal of European Social Policy, 6(1), 17–37. Ferrera, M., Matsaganis, M., & Sacchi, S. (2002). Open co-ordination against poverty: The new EU ‘‘social inclusion process’’. Journal of European Social Policy, 12(3), 227–239. Foster, J. E., Greer, J., & Thorbecke, E. (1984). A class of decomposable poverty measures. Econometrica, 52, 761–766. Hernanz, V., Malherbet, F., & Pellizzari, M. (2004). Take-up of welfare benefits in OECD countries: A review of the evidence. OECD Social, Employment and Migration Working Paper 17. OECD, Paris. Kamerman S.B., Neuman M., Waldfogel J., &Brooks-Gunn J. (2003). Social policies, family types and child outcomes in selected OECD countries. OECD Social, Employment and Migration Working Paper 6. OECD, Paris. Levy, H. (2003). Child-targeted tax-benefit reform in Spain in a European context: A microsimulation analysis using EUROMOD. Working Paper EM2/03. Microsimulation Unit, University of Cambridge, Cambridge. Matsaganis, M., Ferrera, M., Capucha, L., & Moreno, L. (2003). Mending nets in the South: Anti-poverty policies in Greece, Italy, Portugal and Spain. Social Policy and Administration, 37(6), 639–655. Matsaganis, M., O’Donoghue, C., Levy, H., Coromaldi, M., Mercader-Prats, M., Farinha Rodrigues, C., Toso, S., & Tsakloglou, P. (2004). Child poverty and family transfers in southern Europe. EUROMOD Working Paper EM2/04. Microsimulation Unit, University of Cambridge, Cambridge. Moffitt, R. (2002). Means-tested transfer programs in the US. Chicago: University of Chicago Press. Piachaud, D., & Sutherland, H. (2001). Child poverty in Britain and the new Labour government. Journal of Social Policy, 30(1), 95–118. Rhodes, M. (1996). Southern European welfare states: Identity, problems and prospects for reform. South European Society and Politics, 1(3), 1–22. Ripahn, R. T. (2001). Rational poverty or poor rationality? The take-up of social assistance benefits. Review of Income and Wealth, 47(3), 379–398. Ruxton, S., & Bennett, F. (2002). Including children? Developing a coherent approach to child poverty and social exclusion across Europe. Brussels: The European Children’s Network. Saraceno, C. (1997). Family change, family policies and the restructuring of welfare. In: Family, market and community:Equity and efficiency in social policy. OECD Social Policy Series 21. OECD, Paris. Smeeding, T. M., Saunders, P., Coder, J., Jenkins, S. P., Fritzell, J., Hagenaars, A. J. M., Hauser, R., & Wolfson, M. (1993). Poverty, inequality and living standard impacts across seven nations: The effects of non-cash subsidies for health, education and housing. Review of Income and Wealth, 39, 229–256. van de Walle, D., & Nead, K. (Eds) (1995). Public spending and the poor: Theory and evidence. Baltimore: Johns Hopkins University Press. van Oorschot, W. (1996). New perspectives on the non-take-up of social security benefits. TISSER Studies. Tilburg: Tilburg University Press.

ASSESSING THE IMPACT OF TAX/TRANSFER POLICY CHANGES ON POVERTY: METHODOLOGICAL ISSUES AND SOME EUROPEAN EVIDENCE Tim Callan, Kieran Coleman and John Walsh ABSTRACT A method of systematically assessing the ‘‘first-round’’ impact of tax and transfer policy changes on the income distribution and the incidence of relative income poverty is proposed. It involves the construction of a ‘‘distributionally neutral’’ policy, which can be approximated by a policy that indexes tax allowances, credits and bands and welfare payment rates in line with a broad measure of income growth. The impact of actual policy changes in five EU countries over the 1998–2001 period is then measured against this benchmark, using the EUROMOD tax-benefit model.

1. INTRODUCTION1 A view that poverty must be conceptualised and measured relative to the particular society is common in Europe, and has been formally endorsed at Micro-Simulation in Action: Policy Analysis in Europe using EUROMOD Research in Labor Economics, Volume 25, 125–139 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(06)25005-2

125

126

TIM CALLAN ET AL.

EU level on a number of occasions.2 Recent developments have built on the decision of the European Council meeting at Laeken in 2001, which endorsed a portfolio of 18 statistical indicators to monitor progress in the fight against poverty and social exclusion. Prominent among these ‘‘Laeken indicators’’3 are a number of measures of relative income poverty (see Atkinson et al., 2002) and some measures of income distribution.4 The incidence of relative income poverty can be monitored using cross-sectional income surveys.5 But in addition to monitoring outcomes in this way, we need to analyse and understand the impact of past policy decisions and gauge the likely impact of alternative policy choices faced today. If analysis is to assist policy makers, then it must be possible to assess how past policy choices have affected relative income poverty, and how current policy choices may influence poverty in the future. Analysis using tax-benefit models is essential for both the backward-looking assessment of policy impact and for exploration of the likely impact of future policy in order to inform policy choices. The central question examined in this paper is how to assess the impact of tax and transfer policy changes across the income distribution. For example, how should we measure the impact of tax and transfer policy changes implemented in an annual budget? One natural approach is to ask what would have happened if budget day changes in income tax, social security contributions and social benefits had simply not taken place. This counterfactual is of particular relevance for certain purposes – for example, keeping track of changes in government expenditures and tax revenues. It will be argued, however, that this is a flawed benchmark against which to assess the distributional impact of policies actually implemented. Instead, a ‘‘distributionally neutral’’ benchmark is proposed, which can be approximated by indexation of tax and welfare parameters in line with growth in wages. This provides a more accurate picture of the impact of policy changes on the income distribution and on measures of relative income poverty. The EUROMOD model is used to apply this method in assessing the ‘‘cash’’ or ‘‘first-round’’ impact of policy changes over the 1998 to 2001 period in five European countries. The choice of a framework for the assessment of the distributional or poverty impact of budgetary changes in tax and welfare policy is examined in Section 2. Issues arising in the application of the analysis to five European countries using EUROMOD are described in Section 3, which goes on to set out the results of the analysis. The main findings are drawn together in Section 4.

Assessing the Impact of Tax/Transfer Policy Changes on Poverty

127

2. MEASURING THE DISTRIBUTIVE IMPACT OF POLICY CHANGES In this section, we consider three alternative benchmarks against which the distributive impact of changes in income tax and social security or welfare policies may be assessed. Most studies in this area use one of these benchmarks: (a) No nominal change: No change in the nominal value of tax and welfare policy parameters – policy is ‘‘frozen’’ in nominal terms. (b) Price indexation: Tax and welfare parameters are ‘‘price indexed’’ in line with consumer price inflation. (c) Wage indexation: Tax and welfare parameters are ‘‘wage indexed’’ in line with growth in nominal wages. As will be shown, this leads to a ‘‘distributionally neutral’’ benchmark, in contrast to both (a) and (b). In this section, we first outline briefly some of the key features of each benchmark. We then construct a simple economic backdrop, using SWITCH, the Irish tax-benefit model,6 which serves to highlight the key differences between the three benchmarks. Based on this analysis, we argue that the wage indexation benchmark is to be preferred in identifying the distributional impacts of past or proposed future policy changes. (a) No nominal change in tax and welfare parameters Under this benchmark, tax and welfare rates would remain fixed in nominal terms. (This is in line with the conventions governing the ‘‘opening budget’’ in some countries.) What would happen if such a budget were implemented? On the income tax side, there would be no change in rates, allowances or bands. Under a progressive income tax system, this would involve a rise in the tax take as a proportion of income or ‘‘fiscal drag’’. The amount of fiscal drag would depend inter alia on the rate of growth of nominal incomes. On the welfare side, constant nominal payments in the face of price inflation would lead to a fall in real incomes for those depending on welfare payments. (b) Price indexation of tax and welfare parameters Under this benchmark, welfare payment rates, tax bands and other ‘‘money’’ parameters in the tax and benefit system would be indexed in line with price inflation. Thus, the real value of welfare payments would be held constant. If real wages were growing, then welfare payments would be lower in relation to average wages. On the tax side, growth in

128

TIM CALLAN ET AL.

real wages when tax bands were indexed in line with prices would mean that some ‘‘fiscal drag’’ would still occur, raising the average tax rate. (See Immervoll et al., 2005 for an analysis of the impact of growth in earned incomes on average disposable income and relative income, when tax and benefit parameters do not keep pace with real earnings growth). (c) Wage indexation of tax and welfare parameters When examining the impact of budgetary policy on income distribution and relative income poverty, it is useful to have a benchmark which can be regarded as ‘‘distributionally neutral’’. Under such a benchmark, major population groups would share equally in the benefits of economic growth. Growth in disposable income would be the same for all major population groups, and shares of income for different groups in the population would remain the same after the budget as in the year before. While some would argue that the government should undertake more redistribution, and others that it should do less, the ‘‘distributionally neutral’’ benchmark at least provides a yardstick against which changes can reasonably be measured. A number of choices arise in implementing such a benchmark. The approach implemented here involves indexing tax and social welfare to the growth in gross wage income, the predominant element in national income.7 In effect, then, the benchmark represents a budget which is neutral in terms of the share of wages going in tax, and in terms of the relationship between wages and the incomes of social welfare recipients. For wage earners, this is achieved by increasing tax-free allowances and tax bands in line with the growth in gross wages. For those depending on social welfare payments for their income, an increase in welfare rates equal to the rate of increase in pretax wages would, in general, ensure that they shared equally in the growth in income.8 It is worth noting that this ‘‘wage indexation benchmark’’ can also be viewed as a ‘‘neutral’’ option in macroeconomic perspective: indexing policy to wage growth would keep government revenue and expenditure roughly constant as a proportion of national income.

2.1. Which Benchmark? A natural approach to the assessment of policy impact is to ask what would have happened to the values of income tax parameters and welfare payment rates in the absence of an explicit policy change. The default option for what happens to welfare payment rates and tax bands (which we will term the


129

‘‘opening budget’’) varies across countries. For some countries (e.g., Ireland) the answer is very simple: tax and welfare parameters would remain unchanged in nominal terms if no changes were announced in the annual budget statement by the Minister for Finance. Other countries have some elements of price and/or wage indexation built into parts of the tax and welfare code (see Immervoll, 2000). In the UK, for example, key parameters in the income tax code are indexed to changes in the retail price index (consumer prices). Pensions in some countries are indexed to changes in wages, and in others to prices. For some countries indexation (whether to price or to wage growth) may be part of the legislative framework; but for others it may instead be something which has become an accepted practice, although not formalised into legislation. While national practices as regards what happens in the absence of a budget differ considerably, the assessment of distributional impacts by reference to the ‘‘opening budget’’, indexed at most by price inflation, appears rather common.9 This means that the assessment of distributional impacts across countries could depend heavily on the nature of the ‘‘opening budget’’ in the country in question. We argue that the natural standard against which to measure the distributional impact of policy changes is the wage-indexed benchmark, which approximates distributional neutrality. If accepted, this would also provide a common standard which could be applied across countries as illustrated in Section 3. In order to examine the question of which benchmark is the most appropriate for measuring distributional impact, we start by exploring the implications of using each of these benchmark. In the real world, identifying budgetary impacts can be hampered by concurrent changes in economic and social structures and by difficulties in identifying behavioural responses to tax and welfare policy changes. Here we construct a much simpler economic backdrop in which the direct impact of budgetary changes on the income distribution and on relative income poverty can be measured. Differences between the alternative benchmarks in assessing budgetary impacts emerge much more clearly against this backdrop. The simplified economic backdrop is one in which, the economy can be regarded as being in a ‘‘steady state’’, with prices and wages growing at fixed rates, and economic and social structures perfectly stable e.g., employment and unemployment rates, the age distribution of the population and so on. All earnings (by employees and by the self-employed) are assumed to grow at the same rate – there are no shifts in the earnings distribution towards greater or lesser inequality, or shifts in relativities. Occupational pensions are also set to rise by the same proportion. Essentially, the only changes are

130

TIM CALLAN ET AL.

Table 1. Prices rise by Real wages rise by Nominal wages rise by

An Illustrative Scenario. 5% 5% 10.25%

to incomes and prices, and not to employment, unemployment or household composition. (Table 1). Our analysis is conducted using SWITCH, the Irish national model drawing on Callan et al. (1998). The baseline year used is 1999, though similar results would be expected for any other year. The projection to the year t+1 involves growth of 10.25% in wages, self-employment income, occupational pensions and all other ‘‘market’’ incomes, as distinct from state transfers.10 The increase, if any, in state transfer incomes such as oldage pensions, unemployment benefits and child benefit depends on the benchmark policy: it may be unchanged in nominal terms, indexed to prices, or indexed in line with wage growth. We first examine how real disposable income growth varies across the income distribution under an ‘‘unchanged policy’’ benchmark (simply freezing policy in nominal terms) and under the wage indexation alternative. Family units are ranked from poorest to richest, based on income per adult equivalent (where the first adult counts as 1, other adults as 0.66 and children as 0.33). Families are then divided into 10 equal sized groups or ‘‘deciles’’, and the growth in income for each decile is shown. Fig. 1 shows that under the wage indexation benchmark real disposable income growth is the same for all 10 income groups. Furthermore, that rate of growth is equal to the growth in real earnings before tax. If the ‘‘unchanged policy’’ budget were actually implemented, this would be far from neutral in its effects across the income distribution. With tax and welfare parameters frozen in nominal terms, there would be real income losses for those dependent on welfare. Higher up the income distribution, the average tax rate would rise because of the progressivity of the income tax system i.e., there would be ‘‘fiscal drag’’ due to both inflation and real growth in incomes. As a result, growth in real disposable incomes would be somewhat below real earnings growth in the upper income groups. Once again, the fact that it is itself distributionally skewed – albeit less so than a budget ‘‘frozen’’ in nominal terms – makes it inappropriate as a benchmark for distributional analysis. The price indexation benchmark would also involve non-neutrality. Real incomes rise by less than 1% for the bottom three deciles, but grow by 3%

Assessing the Impact of Tax/Transfer Policy Changes on Poverty 6%

Poorest 10%

2

3

4

5

6

7

131

8

Richest 9 10%

% growth in disposable of income

4%

2%

0%

-2%

No nominal change intax/welfare Price-indexed budget

-4%

Wage-indexed budget

-6% Deciles of income per adult equivalent

Fig. 1.

Real Income Growth under Alternative Budgetary Benchmarks, Illustrative Scenario.

or more for the top half of the income distribution. This non-neutrality means that the price-indexed policy is also unsuitable as a benchmark for distributional impact. Similar considerations apply to the measurement of budgetary impact on poverty. We illustrate this in terms of one of the most commonly used indicators of poverty, the proportion of households below half of mean income.11 We use the same illustrative scenario, involving steady growth. Table 2 shows that just under 20% of households fell below half of mean disposable income per adult equivalent in the base year (year zero). Under the benchmark involving no change in nominal values of tax and welfare parameters this proportion would rise by 2.3 percentage points. There would be a smaller rise (0.7 percentage points) under a price-indexed budget. Under a wage-indexed budget the relative income poverty rate would remain constant. If actual policy followed the route of price indexation, this would be seen as a budget with a favourable impact on the poverty rate compared to the ‘‘no change in nominal policy’’ benchmark (a fall of 1.6 percentage points), but a rise in poverty compared to the wage-indexed benchmark. If actual policy simply froze tax and welfare parameters in nominal terms, this would be seen as having no effect on poverty relative to the ‘‘no nominal change’’

132

TIM CALLAN ET AL.

Table 2.

Relative Income Poverty Rate at Half Average Income Under Alternative Policy Scenarios.

No nominal change in tax and welfare parameters Price indexation Wage indexation

Year Zero

Year One

Change in Percentage Points

19.8%

22.1%

+2.3

19.8% 19.8%

20.5% 19.8%

+0.7 +0.0

Source: Calculations using SWITCH, the ESRI tax-benefit model for Ireland.

benchmark as the actual policy and the benchmark policy would then be identical. Critically, however, this does not seem to be a sensible characterisation of the impact of the budget. Relative income poverty would increase between the base year and year one, This would not be due to any economic shock or downturn increasing unemployment, but would simply reflect the decision made in the budget to freeze welfare in nominal terms, thereby leading to losses in real and relative income for those at the bottom of the income distribution. It seems more accurate to characterise such a rise in the relative income poverty rate as due to the tax welfare policy package chosen. This is what happens when a ‘‘distributionally neutral’’ budgetary policy is used as the starting point. The application of price indexation to welfare payments could be interpreted as delivering a constant real standard of living to welfare recipients. Wage indexation of tax and welfare policies can be seen as providing a distinctive, distributionally neutral benchmark, which under steady-state conditions would imply no change in the proportion of persons falling below relative income poverty lines. In what follows, we compare actual policy changes with the benchmark constructed by indexing tax and welfare policies to growth in wages in five EU countries.

3. IMPACT OF POLICY CHANGES ON INCOME DISTRIBUTION AND POVERTY The five countries selected (mainly on the basis that, at the time this analysis was undertaken, suitably validated policies were available for use with the EUROMOD tax-benefit model for both 1998 and 2001) were Austria,


133

Greece, Ireland, Portugal and the United Kingdom. In uprating tax and welfare parameters, we used the same estimates of nominal wage growth, which were used in uprating the basic data between the relevant years: to do otherwise would introduce an inconsistency into the analysis. These estimates of nominal wage growth (see Table 3) varied from 7% growth in Austria to 26% in Ireland. This growth rate was used to index the 1998 tax and welfare parameters in order to provide a distributionally neutral benchmark against which the actual 2001 system could be compared. The fact that this indexed policy is being used as a benchmark does not imply a judgement about its merits or demerits as a policy: it is a benchmark simply because it would see incomes at different income levels grow at the same rate, keeping the distribution of income constant. Policy impacts relative to this baseline (and calculated on a ‘‘static’’ basis, before any behavioural change – for simulations including estimated behavioural changes see Callan et al. (2003)) are reported in Table 4. The results suggest that tax-benefit policy changes over the 1998–2001 period Estimated Nominal Wage Growth and Consumer Price Inflation, 1998–2001.

Table 3.

Nominal Wage Growth (%)

Consumer Price Inflation (%)

7 17 26 18 11

5 9 12 10 3

Austria Greece Ireland Portugal United Kingdom

Table 4. Quintile

1 2 3 4 5 All

Policy Impact on Income by Quintile, Five EU Countries, 1998–2001.

Austria (%)

1.4 1.6 0.7 0.4 0.1 0.5

Greece (%)

9.7 4.5 2.0 1.1 0.3 1.9

Ireland (%)

0.3 2.8 8.4 9.0 8.6 4.6

Portugal (%)

UK (%)

3.8 3.0 3.0 2.7 2.0 2.6

7.8 4.9 2.7 1.7 0.6 2.3

Note: Figures show percentage change in income under actual policy relative to a wage-indexed baseline policy.

134

TIM CALLAN ET AL.

were progressive in four of the five countries. Ireland is the exception, with policy changes boosting the incomes of the top 60% of families by close to 9%, as against modest gains or a small loss for lower income groups. The greatest gains for the bottom quintile were in the UK and Greece, with declining gains for other quintiles. For the UK, the main policy factors driving these changes were substantial increases in income support rates for pensioners, coupled with sharp increases in income support rates for children, particularly younger children.12 In Greece, the boost to the income of low-income groups is also associated mainly with marked increases in the rates of payment for certain retirement benefits.13 The pattern in Portugal was rather more even across the income distribution, while in Austria the changes were rather small for all income groups. Table 5 presents initial estimates of impact of policy changes on headcounts of relative income poverty, for a range of poverty lines based on both median and mean incomes. These results show that the greatest reductions in head-counts of poverty have been achieved in Greece and the UK – the countries where the distributional pattern of policy impact was most progressive. For the UK, a reduction of between 2 and 3 percentage points is estimated for this 1998–2001 period. In Ireland, on the other hand, the head counts for relative income poverty as a proportion of median income rose by between 3 and 5 percentage points. This reflects a distributional impact, which saw losses for the poorest quintile, and strong gains for the top three quintiles. The background was one in which real wage growth was strong, and income tax cuts added to disposable income for those in employment. Welfare payment rates, on the other, failed to keep pace with the growth in disposable income so that the relative position of welfare recipients disimproved. Table 5.

Policy Impact on Relative Income Poverty, Five EU Countries, 1998–2001.

Poverty Line Cut-Off

Percent of median 50 60 70 Percent of mean 40 50 60

Change in Head Count Measure of Poverty (% Points) Austria

Greece

Ireland

Portugal

UK

0.5 0.2 0.7

1.2 1.1 1.4

3.8 3.3 4.8

0.4 0.1 0.2

2.5 2.8 2.8

0.1 0.2 0.3

1.1 1.0 1.6

0.3 2.7 3.0

0.6 0.0 0.2

2.4 3.2 2.5


135

4. CONCLUSIONS The main purpose of this paper is to argue for a ‘‘distributionally neutral’’ policy benchmark against which to measure the distributional impact of actual policy changes. This runs counter to the common practice of using a fixed nominal policy (or a price-indexed policy) as a benchmark against which to assess distributional impact. A number of practical issues arise in constructing a ‘‘distributionally neutral’’ benchmark. Broadly speaking, a tax and benefit policy indexed for growth in wages is sufficient to ensure similar net income growth across the income distribution. However, different measures of wage or income growth may have a claim on our attention, and forecasts are almost certain to overestimate or underestimate actual wage growth. Different choices are possible in applying the method – one might use an initial forecast of wage growth, and revise it as new information comes to light. Is wage indexation an appropriate benchmark in a context where tax or welfare policy either makes no pre-commitment to a particular level of indexation, or makes a commitment to price indexation? We argue that it is, because the policy counterfactual created by wage indexation has a unique claim on our attention: it is the only one that comes close to providing a distributionally neutral benchmark. Attempting to measure the distributional impact of policy relative to a skewed benchmark is inappropriate. An initial application of the method suggests that over the 1998–2001 period, four of the five countries analysed (Austria, Greece, Portugal and the UK) had tax/transfer policy changes which were particularly favourable to lower income groups. The exception was Ireland, where welfare payment rates over the years 1998 to 2001 failed to keep pace with growth in disposable income. The greatest gains for low-income groups were in Greece and the UK. Correspondingly, these were the countries for which head counts of relative income poverty were most reduced – with a reduction of between 2 and 3 percentage points in the UK. The ‘‘open method of coordination’’ adopted for social policy at Lisbon lays particular stress on comparisons of national policies with ‘‘best practice’’ in the EU in addressing social issues such as unemployment and social exclusion. To date, this has tended to centre on specific schemes or initiatives dealing as examples of ‘‘best practice’’ in dealing with certain client groups. While valuable in itself, this needs to be complemented by ‘‘systemwide’’ comparisons across countries, which take account of the big picture. A selected number of schemes or policy initiatives may work well in terms of reducing relative poverty, but may be ‘‘swimming against the stream’’ in

136

TIM CALLAN ET AL.

terms of the impact of broader policy developments. Analyses of the type conducted here can help to ensure that individual policy initiatives and broader policy developments receive due weight in assessing ‘‘best practice’’ in countering poverty.

NOTES 1. This paper was written as part of the Micro Analysis of the European Social Agenda (MICRESA) project, financed by the Improving Human Potential programme of the European Commission (SERD-2001-00099). We are indebted to all other past and current members of the EUROMOD consortium. In particular, we would like to thank Christine Lietz, for her work in making the model more accessible, and Horacio Levy for valuable technical advice. However, the views expressed in this paper, as well as any errors, are the responsibility of the authors. In particular, this applies to the interpretation of EUROMOD results and any errors in its use. EUROMOD is continually being improved and updated and the results presented here represent the best available at the time of writing. EUROMOD relies on microdata from 12 different sources for 15 countries. This paper uses the European Community Household Panel (ECHP) User Data Base made available by Eurostat; the Austrian version of the ECHP made available by the Interdisciplinary Centre for Comparative Research in the Social Sciences; the Living in Ireland Survey made available by the Economic and Social Research Institute and the Family Expenditure Survey (FES) made available by the UK Office for National Statistics (ONS) through the UK Data Archive. Material from the FES is Crown Copyright and is used by permission. Neither the ONS nor the Data Archive bear any responsibility for the analysis or interpretation of the data reported here. An equivalent disclaimer applies for all other data sources and their respective providers cited in this acknowledgement. Thanks are also due to all those involved in the development of the EUROMOD model. We are grateful to the editors and to an anonymous referee for helpful comments. Comments received from Tony Atkinson, Manos Matsaganis, Holly Sutherland, and from participants in the EUROMOD meeting at Lisbon in April 2004 are gratefully acknowledged. The usual disclaimer applies. 2. See, for example, the definition of poverty adopted by the European Commission (1985), which identified poverty with the situation of people whose ‘‘resources (material, social and cultural) are so limited as to exclude them from the minimum acceptable way of life in the countries in which they live’’. 3. The Council met at Laeken in December 2001. 4. The relative income poverty measures are relabelled as identifying those ‘‘at risk’’ of poverty; it is accepted in this context that not all of those below the income cut-off are in poverty, but the ‘‘at risk of poverty’’ measure is accepted as a useful one in monitoring social exclusion. 5. The new EU-SILC (Statistics on Income and Living Conditions) will provide a base for monitoring of the indicators in EU countries. In practice, the use of such surveys means that one obtains a picture of relative poverty with a significant time lag. For example, the Joint Report on Social Inclusion (European Commission, 2004) relies on data gathered some three years earlier.


137

6. SWITCH is a static tax-benefit model, like the Tax Policy Center model in the US and the IFS model in the UK. Such models do not take behavioural responses into account, but are a commonly used starting point for the assessment of distributive effects of policy changes. In any event, the key focus here is on differences in the benchmark against which policy is assessed: the same issues arise for both static and behavioural response models in this context. Further details on the SWITCH model are provided in the appendix. 7. Incomes from self-employment are more variable from year to year than wages, so indexing taxes and social welfare to wage growth provides a more stable benchmark. 8. If tax cuts over and above indexation were implemented, then welfare payments would have to rise faster to keep pace with growth in net wage incomes. 9. See, for example, Giles and Johnson (1995), Myck (2000), Brewer et al. (2005) in the UK and Mitrusi and Poterba (2000) in the US. 10. This illustrative scenario is not meant to be a representation of the year 2000, but is a counterfactual construct which helps to clarify differences in the nature of the alternative benchmarks. 11. Similar results obtained for poverty lines framed in terms of median income, which are used in the second part of this paper. 12. There were also substantial increases, well ahead of the rate of wage growth, in the additions to payment rates for carers and in respect of disabled children. See Sutherland (2001) and Sutherland and Guttierez (2004) for details. 13. Specifically, the rates increased sharply for the farmers’ basic pension, social pension and pensioner social solidarity benefit. See Matsaganis and Tsakloglou (2004a, b) for details.

REFERENCES Atkinson, A. B., Cantillon, B., Marlier, E., & Nolan, B. (2002). Social indicators: The EU and social cohesion. Oxford: Oxford University Press. Brewer, M., Goodman, A., Shaw, J., & Shephard, A. (2005). Poverty and inequality in Britain: 2005. London: Institute for Fiscal Studies Commentary Commentary 99. Callan, T., Nolan, B., & Walsh, J. (1998). Income tax and social welfare policies. In: T. Callan (Ed.), Budget Perspectives 1999. Dublin: The Economic and Social Research Institute. Callan, T., Layte, R., van Soest, A., & Walsh, J. (2003). Taxes, benefits and labour market responses: New evidence for Ireland. Dublin: The Economic and Social Research Institute Policy Research Series No. 48. European Commission (1985). On specific community action to combat poverty (Council Decision of 19 December 1984). Official Journal of the EEC 2/24 (85/8/EEC). European Commission. (2004). Joint report on social inclusion. Office for Official Publications of the European Communities, Luxembourg. Giles, C., & Johnson, P. (1995). Tax reform in the UK and changes in the progressivity of the tax system, 1985–1995. Fiscal Studies, 15(3), 64–86. Immervoll, H. (2000). The impact of inflation on income tax and social insurance contributions in Europe. EUROMOD Working Paper EM2/00. Immervoll, H., Levy, H., Lietz, C., Mantovani, D., & Sutherland, H. (2005). The sensitivity of poverty rates in the European Union to macro-level changes. Cambridge Journal of Economics (advance access published June 2005, doi: 10.1093/cje/bei054).

138

TIM CALLAN ET AL.

Matsaganis, M., & Tsakloglou, P. (2004a). EUROMOD Country Report: Greece 1998. http:// www.iser.essex.ac.uk/msu/emod/countries/greece/cr98gr0205.pdf. Matsaganis, M., & Tsakloglou, P. (2004b). EUROMOD Country Report: Greece 2001. http:// www.iser.essex.ac.uk/msu/emod/countries/greece/cr01gr0205.pdf. Mitrusi, A., & Poterba, J. (2000). The distribution of payroll and income tax burdens, 1979–99. National Tax Journal, LIII(3), Part 2. Myck, M. (2000). Fiscal reforms since May 1997. IFS Briefing Note no. 14, Institute for Fiscal Studies, London, www.ifs.org.uk/taxben/fiscalreform.pdf. Sutherland, H. (2001). EUROMOD Country Report: United Kingdom 1998. http://www.iser. essex.ac.uk/msu/emod/countries/uk/cruk0701.pdf. Sutherland, H., & Guttierez, R. (2004). EUROMOD Country Report: United Kingdom 2001. http://www.iser.essex.ac.uk/msu/emod/countries/uk/cr01uk0704.pdf.

APPENDIX. SWITCH – THE ESRI TAX-BENEFIT MODEL Tax-benefit models are needed for a comprehensive assessment of the effects of tax and welfare policy changes, taking into account the wide variation in individual and family circumstances relevant to welfare entitlements and tax liabilities. SWITCH, the ESRI tax-benefit model, is a well-established tool for analysing the ‘‘first-round’’ effects of tax and welfare policy changes. It has been based on large-scale nationally representative survey of households undertaken by the ESRI (most recently the Living in Ireland Survey). The model database is adjusted each year to ensure that it reflects recent changes in incomes, employment, unemployment and population and draws on projections of such changes for some years ahead, in order to provide a framework for medium-term analysis of budgetary issues. The model uses detailed information on individual and family circumstances (including information on wages and hours of work for those in paid employment, and on labour force status and receipt of social welfare benefits for those not in paid employment) to assess the social welfare entitlements and tax liabilities of each family in the database. The model can therefore simulate for each family the disposable income they would receive under actual policy, or under alternative policies of interest. Using these detailed calculations it is possible to summarise the impact of policy changes in many different ways. Here we focus in particular on how the average gain or loss varies depending on the income of the family. Family units are ranked by income, adjusting for differences in family size and composition using a simple equivalence scale: 1 for the first adult in the family, 0.66 for a second adult and 0.33 for children. Thus, a married couple


139

with a disposable income of £200 per week would have an ‘‘equivalised’’ income of just over £120 (i.e., £200 divided by 1.66). A married couple with one child would have an equivalised income of just over £100 (i.e., £200 divided by 1.99 ( ¼ 1+0.66+0.33)). Families are then divided into 10 equalsized groups or ‘‘deciles’’, from poorest to richest. One underlying technical assumption is that labour market behaviour and wage rates are the same under each policy; but the model can shed light on how such behaviour may change by identifying the impact of policy changes on financial incentives to work. For structural estimates of labour supply and estimates of the impact of tax policy changes on labour supply behaviour see Callan et al. (2003).

FISCAL DRAG – AN AUTOMATIC STABILISER?$ Herwig Immervoll ABSTRACT By reducing the real value of nominally fixed tax band limits, deductions and tax credits, inflation can lead to higher real tax burdens (‘‘fiscal drag’’). The traditional view is that this reduces aggregate demand and thus acts as an automatic stabiliser. Yet, this familiar reasoning ignores the supply side and, in particular, possible effects of higher tax burdens on labour costs. Recent work on imperfect labour markets has shown that such effects can indeed arise as employees are able to bargain for higher wages that partly compensate for tax increases. In this case, the resulting upwards pressure on real labour costs can be inflationary. To illustrate this mechanism, this article analyses labour tax burdens in four European countries and how they are altered if tax systems are not adjusted for inflation. This is then combined with available results on the effects of tax

$

This chapter was written when the author was based at the University of Cambridge. He is now Economist at the OECD Directorate of Employment, Labour and Social Affairs ([email protected]). The chapter is part of the EUROMOD project, financed by Targeted Socio-Economic Research programme of the European Commission (CT97-3060). The views expressed in this chapter as well as any errors are the author’s responsibility. In particular, the chapter does not represent the views of the EUROMOD consortium or the institutions to which the author is affiliated.


141

142

HERWIG IMMERVOLL

changes on wages in imperfect labour markets. The results suggest that, in an unadjusted tax system, inflation can produce a moderate upward pressure on wages. It is argued, however, that more detailed empirical work on the role of taxes in the wage-setting process is needed as existing work ignores the substantial heterogeneity of workers and the tax rates they face.

1. INTRODUCTION Inflation can alter the structure of tax systems and, thus, the incidence and overall levels of tax revenues. There are several channels through which such distortions can operate. The most evident and thoroughly researched one is the influence of inflation on the tax schedule commonly known as ‘‘bracket creep’’ (Immervoll, 2005; OECD, 1986; Aaron, 1976). In the absence of offsetting adjustments, a progressive tax schedule will produce higher real tax burdens as nominal incomes rise (‘‘fiscal drag’’). The dependence of the tax system on an a priori unknown variable can be problematic. For example, inflation-induced changes by-pass the democratic process that is otherwise mandatory for enacting tax changes. On a more pragmatic level, they can lead to unintended effects that may run counter the originally intended objectives of fiscal policy. As a result, there have been numerous proposals to protect the tax system from inflation-induced distortions and many countries have, in fact, introduced automatic inflation adjustments into their tax systems (Messere, 1998; OECD, 1976; Tanzi, 1980). It has therefore been shown that any alleged practical or administrative difficulties of neutralising the effects of inflation can be overcome. However, there is one powerful economic argument that is commonly put forward in favour of inflation-induced tax increases. According to this view, increasing real tax burdens reduce consumption and thereby aggregate demand, acting as an automatic stabiliser, which helps to cool down the economy during inflationary spells.1 This reasoning, however, only looks at the demand side, ignoring any effects that higher tax burdens may have on the cost of production. If employees have some market power and bear less then the full burden of higher taxes then real labour costs will go up as well, generating a cost-push upwards pressure on prices and opening up the possibility of a wage-price spiral (Dernburg, 1974; Jackson, Wilkinson, & Turner, 1972; Malcomson & Sartor, 1987). During most of the 1990s, inflation has not been a major problem in OECD countries. For instance, Heinemann (2001) argues that declining

Fiscal Drag – An Automatic Stabiliser?

143

inflation rates have limited the potential for employing fiscal drag as a means of increasing real tax revenues. Yet, it has been widely accepted that recent developments in Europe and, especially, the European Monetary Union (EMU), have contributed to considerable variability of inflation rates between countries. As long as there remain important structural differences between different EMU countries, one would expect a more unified monetary policy to lead to differing price developments across countries. Indeed, the process of ‘‘convergence’’ can itself contribute to such differences, at least temporarily (European Commission, 1999). At the same time, individual EMU countries have now arguably less control over monetary policy parameters. Smaller countries or those whose economic environment differs markedly from the EMU ‘‘average’’ may be particularly affected (e.g. Ireland). In this scenario, the stabilising properties of the fiscal system are of particular relevance (see Brunila, Buti, & in ’t Veld, 2003). The present chapter aims to illustrate how inflation-induced tax distortions can affect wage levels. It builds on the recent literature explaining the consequences of changes in the level and progressivity of (effective) tax rates for the wage-setting process. In particular, I draw on results showing that changes in marginal and average tax rates influence wages in opposite directions. Using a European tax-benefit microsimulation model, I then decompose the effects of inflation-induced bracket creep into changes in marginal and average tax rates. The chapter then discusses how the resulting tax changes can be used to quantify likely cost-push effects of bracket creep. The approach improves on previous studies in that it is able to capture the full range of effects caused by tax changes in different parts of the population. Section 2 reviews the channels through which taxes can influence the price level and thus provides the theoretical framework for the analysis that follows. Section 3 describes the microsimulation approach used in the empirical part of the chapter. Both the model and the data are discussed here. The results are presented in Section 4 while Section 5 discusses how they can be used to approximate the impact on wage levels. Some caveats are highlighted along with possibilities to address them in future research.

2. THEORY The effects of labour tax burdens, whether discretionary or caused by inflation, on price levels depend on a number of factors. The most obvious is the income elasticity of demand.2 Yet, even where demand responds to

144

HERWIG IMMERVOLL

changes in net incomes, tax increases may have unintended macroeconomic effects in cases where excess domestic demand is not the primary cause of inflation. If inflation is not due to domestic overheating but is ‘‘imported’’, curbing domestic demand is not the optimum counter measure. The same holds if price increases result from supply disturbances. Rather than inducing price adjustments, tax increases in such a situation are likely to raise the level of unemployed resources (Dernburg, 1974). Another reason why pushing up the tax burdens faced by private households may not have the desired stabilising effects is that price levels are determined by aggregate demand. As is well known, tax-financed public expenditure actually increases aggregate demand so long as the marginal propensity to consume is smaller than one. As a result, higher taxes do not translate into lower aggregate demand if lower private consumption is outweighed by increased public consumption made possible by the additional tax revenues. To prevent expansionary multiplier effects, discretionary tax increases aimed at reducing inflation have, in fact, occasionally been accompanied by measures aiming to ‘‘freeze’’ additional revenues.3 For non-discretionary tax increases caused by inflation, however, such attempts have not become known (it is tempting to interpret the lack of such measures as suggesting that stability considerations are not the primary reason for allowing the fiscal drag to operate). On the supply side, the responsiveness of wages to tax burden changes is critical. In the absence of fiscal illusion, wage earners will want to compensate higher taxes by demanding wage increases from their employers. If they are successful, tax increases will lead to higher production costs and, in this case, inflation-induced tax increases in the past ‘‘may well have contributed to higher inflation rates’’ (Hersoug, 1984, p. 50). Consider an individual subject to average tax rate ta and marginal tax rate tm. With inflation p, nominal gross wages w have to increase by an amount Dw. It is straightforward to show that Dw 1 ta ¼ w 1 tm

(1)

For instance, consider an individual with a pre-inflation wage w ¼ 100; ta ¼ 30%, tm ¼ 50% and p ¼ 10%. This individual’s nominal wage would need to increase by Dw ¼ 14 in order to keep real after-tax earnings at the original level of 70.4 This shows that the required wage increase does not primarily depend on the magnitudes of ta and tm but on the difference between them. Note that (1ta)/(1tm) is the inverse of the local progressivity measure proposed by


145

Musgrave and Musgrave (1976). The more progressive the tax schedule, the larger the additional tax burden caused by nominal wage increases (or, equivalently, by the inflation-induced erosion of the real value of tax band limits) and, thus, the larger the wage increase necessary to compensate for higher tax burdens. To the extent that wages increase by more than inflation – even if they fall short of fully maintaining pre-inflation real after-tax income as in Eq. (1) – real labour costs will go up and exert upward pressure on prices. In the first instance, the macro-economic consequence of this cost-push effect is an upwards shift in the aggregate supply schedule.5 Combined with lower aggregate demand due to higher taxes (i.e. assuming that the reduction in household demand is not undone by increased public spending), an upwards-shifting supply curve unambiguously results in lower output. The effect on the price level is less clear. It depends on the slopes of the relevant sections of the supply and demand curves (see Blinder, 1973). If aggregate demand remains unchanged (for instance, because of increased public spending), we will see both lower output and higher prices.6 But to what extent do wages actually respond to changing tax burdens? Obviously, employees must be aware of them before attempting to obtain compensatory wage increases. Without fiscal illusion, the resulting wage developments depend crucially on the structure of the labour market and, more specifically, on the wage-setting mechanism. In a perfectly competitive labour market, involuntary unemployment is ruled out and the only way labour taxes can alter the equilibrium wage level is through changes in labour supply. This is illustrated in Fig. 1. Given hourly wages w, B0 is the individual’s initial budget constraint. With a progressive tax system, the gain in after-tax income from working an additional hour declines with longer working hours (the budget constraint becomes flatter). Initially, the individual chooses the optimum number of working hours h0. At the associated level of gross income a+b the average tax rate is a/(a+b). The marginal tax rate, illustrated by the slope of the tangency line, is tm. The effects of increased tax progressivity (higher marginal tax rates but unchanged average tax rates) can be shown by changing the budget constraint to Bm, which reduces working hours to hm (pure substitution effect). Conversely, lower tax progressivity (a higher average tax rates while keeping the marginal rate unchanged) would shift the budget line downwards (Ba).7 For the individual with initial working hours h0, this downward shift represents a pure income effect. Assuming that leisure is a normal good, this will increase working hours (to ha in the figure). The effects on wage levels are of course fully

146

HERWIG IMMERVOLL

1 a

w B0

net income

Bm

Um

Ba

U0 Ua 1

b

hm

(1− m)·w

leisure

h0 ha

Fig. 1.

Changing Marginal and Average Tax Rates in a Competitive Labour Market.

determined by these labour supply changes as, in a competitive labour market, wages adjust to market-clearing levels. With unchanged demand, higher marginal (average) tax rates will, ceteris paribus, therefore lead to higher (lower) wages. For most segments of existing European labour markets, however, the standard competitive model is probably unrealistic. Indeed, findings by several authors strongly suggest that there are fundamental differences in wage-setting mechanisms between Europe and the US (see for instance Blanchard & Katz, 1999; Cahuc & Zylberberg, 2004). Where there is involuntary unemployment, individuals are, by definition, unable to alter their working hours freely. Hence, markets do not clear. In this situation, modern theories of imperfect labour markets offer alternative, and more differentiated, analyses of links between tax burdens and wage levels.


147

Starting in the mid-1980s, there have been several influential theoretical chapters on how taxes impact on the wage-setting process in imperfect labour markets (Hansen, Pedersen, & Slok, 2000; Koskela, 2001; Pissarides, 1998; Sorensen, 1997; an overview is provided by Røed & Strøm, 2002; Section 4). Results from relatively recent empirical studies for a considerable number of European countries tend to be consistent with these theories – indicating that the imperfect labour market model is indeed appropriate in the European context (Aronsson, Wikstro¨m, & Bra¨nnlund, 1997; Graafland & Huizinga, 1996; Holmlund & Kolm, 1995; Lockwood & Manning, 1993; Malcomson & Sartor, 1987; Wulfsberg, 1996). In terms of wage formation, the focus in these models is on the respective incentives of employees and employers. That is, workers have some market power and are not seen as mere price-takers as in the competitive model. While higher marginal tax rates lead to lower employment (and, therefore, higher equilibrium wages) in the textbook competitive scenario, a trade-off between unemployment and wage levels can explain why a more progressive tax system may turn out to be good for employment if it moderates wage demands. The focus on tax progressivity in this literature implies that it is important to analyse marginal and average tax rates separately.8 This is relevant in the context of the present chapter as it suggests that, by analysing how inflation alters average and marginal tax burdens, it is possible to say something about the impact on wage levels. A commonly used theoretical framework for analysing imperfect labour markets is the union bargaining model. This assumes a trade-off between unions’ desire for lower unemployment among the workers they represent and higher (real after-tax) wages (Hersoug, 1984; Holmlund & Kolm, 1995; Lockwood & Manning, 1993; Malcomson & Sartor, 1987). Intuitively, by trading wages for unemployment, unions can influence wages and, thus, the slope of the budget constraint in Fig. 1. Assuming constant real net unemployment benefits and unchanged progressivity at the relevant point of the tax schedule, increased average tax rates reduce the relative value of labour income vis-a`-vis unemployment benefits (i.e. the replacement ratio increases).9 Unemployment thus appears less painful to union members and the union’s preferences tend to shift away from avoiding unemployment towards securing higher wage increases. Increasing marginal tax rates, on the other hand, are predicted to lead to moderated wage demands, since changes in before-tax wages are now worth less in terms of union members’ after-tax income.10 Note that both these predictions are contrary to those deriving from the competitive model.

148

HERWIG IMMERVOLL

The final outcome of the bargaining process is driven by the respective target wages of employers and unions as well as their relative bargaining powers. Since taxes on employment income do not change the target wage of the employer ‘‘it is not unreasonable to assume that, as long as the union has some bargaining power, the effect of such changes on the actual wage will be in the same direction as the effect on the union’s target wage.’’11 While the remainder of this chapter will focus on the union bargaining model, the theoretical results on the role of tax progressivity in the wagesetting process in an imperfect labour market are in line with those obtained from an alternative theoretical framework; the efficiency-wage model (Sorensen, 1997). It is clear from the above discussion that it is essential to distinguish carefully between marginal and average tax rates when assessing possible effects of tax changes on wage levels in an imperfect labour market setting. This, however, is not straightforward for several reasons. First, tax rules are often complicated. Apparently minor details of the tax code often can cause very substantial ups and downs of people’s budget constraints. Ignoring these details can seriously distort the incentives faced by employees and their representatives in the wage-setting process. Second, it is desirable to characterise these incentives in terms of effective tax rates by accounting for both income taxes and social insurance contributions.12 In addition, take-home pay and effective tax rates can be influenced by transfer payments as well. The rules of these benefit payments are frequently even more complicated than tax rules. Another practical difficulty arises from the fact that different types of taxpayers face very different effective tax burdens and that, moreover, the effects of inflation on tax burdens will vary between individuals as well. Previous studies have simply computed the relevant tax rates for ‘‘typical’’ household types with average earnings (e.g. married couple with a single breadwinner and two children). Such an approach ignores the obvious influence of the household structure on applicable effective tax rates (number of income earners, number and ages of any dependants, etc.). Equally important, it ignores the fact that different earnings levels will be affected differently by any real-world change of the tax system. Different employees therefore favour different target wages. Hence, a critical question is how unions aggregate these different preferences of the employees they represent. The resulting average preferences may be very different from one that would be optimal for someone with average earnings.13 As Lockwood and Manning (1993) note, ‘‘the average marginal and the average average tax rates will in general not be equal to the marginal and average tax rates evaluated at the


149

average earnings’’ (p. 11; italics are in the original). The choice of the most appropriate aggregation rule in practice is subject to some controversy. A method that is able to characterise the relevant parameters for all observations in a full sample of employees is attractive as it maximises analytical flexibility in this regard.

3. METHODOLOGY: MEASURING TAX BURDENS BEFORE AND AFTER INFLATION I use a tax-benefit microsimulation model to establish how inflation alters marginal and average tax rates for each employee in representative samples of the household sector. The model contains detailed tax and benefit rules, which fully account for interactions between different instruments, such as the tax-deductibility of social insurance contribution payments or income testing of transfer payments. Using representative micro-data it also captures the heterogeneity of household types and earnings levels. The approach therefore addresses some of the potential empirical difficulties described above. The model is used here to calculate distributions of average effective tax rates (AETRs) and marginal effective tax rates (METRs). In order to determine the effects of inflation, tax rate calculations are performed for a ‘‘before inflation’’ baseline corresponding to the policy rules in place in the reference year (here 1998); and a simulated ‘‘after inflation’’ scenario with inflated incomes (and inflated values of all other monetary amounts reported in the data, such as housing or commuting costs) but tax policy parameters, such as tax band limits, held constant in nominal terms. This simple simulation approach permits holding ‘‘everything else’’ constant so that changes in calculated tax rates are only due to the inflation effect of interest. Indeed, simulating a hypothetical scenario is necessary since the aim is to assess the effects of allowing the inflation-induced fiscal drag to operate (i.e. assuming that tax systems are not adjusted for inflation) while three of the countries considered here do in fact automatically adjust most tax policy parameters for inflation (Finland, the Netherlands and UK).14 Results are reported for an arbitrarily chosen inflation rate of 7% – sizeable enough for identifying meaningful bracket creep effects (results for other inflation rates, leading to similar conclusions, can be obtained from the author). Using results from the literature on the elasticity of wages with respect to average and marginal tax rates as a starting point, I then discuss

150

HERWIG IMMERVOLL

how the measured tax-rate changes may be used to obtain an indication of the potential wage pressures caused by inflation-induced tax-burden changes. It should be noted that, rather than providing a realistic account of how inflation affects prices and quantities, a uniform increase of all incomes is used here as an analytical tool in order to isolate the effects of ‘‘pure’’ inflation, i.e. a general price increase with constant relative prices. Keeping all other individual characteristics unchanged amounts to the assumption that all prices and wages move in parallel and that taxpayers do not rearrange their portfolios in response to inflation. Moreover, the scenario rules out changes to income-reporting behaviour. These may result from actual changes of real gross income, or because higher nominal incomes, and the higher marginal tax rates associated with them, lead taxpayers to report less; Saez (2003), provides estimates for these latter effects in the United States. While there are good theoretical reasons to take these aspects into account, there is little empirical basis for doing so. Abstracting from transmission issues, it is possible to focus attention on the mechanical links between ‘‘pure’’ inflation and effective tax rates.

3.1. Tax-Benefit Model and Data The specific model used in this chapter is EUROMOD, an integrated European tax-benefit model currently covering the 15 EU countries prior to the 2004 enlargement. The model is designed to examine, within a consistent comparative framework, the impact of national policies on national populations or the differential impact of co-ordinated European policy on individual member states. Within the context of the present chapter, the most relevant feature of the model is that it can produce conceptually consistent and comparable output across countries. EUROMOD is built around separate but partly harmonised household datasets describing the population of each country. Using detailed algorithms representing existing tax and benefit legislation, the model is able to compute a range of tax and benefit amounts for each observation unit. These algorithms capture the full range of institutional features including, for instance, accurate income definitions (such as taxable income or the ‘‘means’’ relevant for computing means-tested benefits), definitions of family and assessment units (such as who counts as a child for the purpose of particular tax or benefit rules), thresholds, floors, ceilings and relevant tax rates as well as specific eligibility rules, claw-back rates or income disregards used in computing benefit


151

entitlements. This level of detail makes it possible to derive a finely grained picture of tax burdens and how these vary with earnings and individual or family characteristics. Currently, the main policy instruments that are relevant for this chapter and that can be simulated with EUROMOD are income taxes, mandatory social contributions as well as universal and means-tested social benefits. Income components that are not simulated and are required as an input into the calculation of taxes and benefits are taken directly from the data. These include earnings and capital income. Further information on EUROMOD and the simulated tax-benefit instruments, including simulation details as well as validation of model results against other sources, is provided by Sutherland (2001) as well as on the Internet at www.iser.essex.ac.uk/msu/emod.php. Simulations are run for four countries for which EUROMOD modules were available at the start of this project: Finland, Ireland, the Netherlands and the UK. The micro-data underlying the simulations are derived from the 1996 Finnish Income Distribution Survey (IDS), the Living in Ireland Survey, 1994 (Wave 1), the 1996 wave of the Dutch Socio-Economic Panel (SEP) and the UK 1995/1996 Family Expenditure Survey. In each case, samples are weighted to adjust for non-response bias and to bring the results up to population levels. The simulations are based on the systems of tax and benefit rules current in June 1998 and all monetary variables in the microdata are brought forward to this year using an appropriate uprating index for each type of income (explained in detail in Sutherland, 2001).

3.2. Definition of Tax Rates A small ‘‘margin’’ of +3% is used for computing marginal tax rates (corresponding approximately to one additional working hour per week for a full-time employee). In other words, the METR represents the fraction of a small earnings increase of 3% that is effectively ‘‘taxed away’’ as a result of higher taxes and any reduced benefit entitlements. AETRs, ta, and METRs, tm, are thus computed as follows: ta ¼ ðT BÞ=E

(2)

tm ¼ DðT BÞ=DE

(3)

where T is the sum of income taxes and own social contributions, B the sum of all benefit payments and E denotes gross employment income.15

152

HERWIG IMMERVOLL

AETRs and METRs are computed for all employees in the data, where ‘‘employee’’ is defined as a person of any age working more than 10 hours per week and whose employment income represents the main source of earnings (i.e. it exceeds the value of any transfers or incomes from selfemployment). Tax rates are measured at the individual level. This means that one needs to decide sharing rules for instruments that operate at a level other than the individual. For example, family benefits are paid to the family as a whole but as a negative tax they potentially influence effective tax rates of each person in the family. In the simulations I make the assumption that all social benefits that do not clearly accrue to one specific person are shared equally among all adults in the relevant family unit. Any taxes that are paid jointly are assumed to be split in proportion to the tax base of each individual of the tax unit (this is important in Ireland, where there is joint income taxation for married couples).

4. SIMULATION RESULTS The resulting distributions of METRs and AETRs for employees are presented in Figs. 2 and 3. The distribution of marginal rates is more 70% 60% FI

Share

50%

IR UK

40%

NL 30% 20% 10%

Fig. 2.

5% 60 -6

0% 55 -6

55 % 50 -

50 % 45 -

40 -4 5%

40 % 35 -

5% 30 -3

30 % 25 -

25 % 20 -

15

-2 0%

0%

Distribution of Marginal Effective Tax Rates (Employees). Source: EUROMOD.


153

35% 30% FI

Share

25%

IR UK

20%

NL 15% 10% 5%

Fig. 3.

5%

-5 0% 45

0%

-4 40

-4 35

-3 5%

0%

30

-3

5%

25

-2

0%

20

-2

-1 10

15

5%

% 10 5-

5% 0-

%

0% =

-0 -5

-1

0-

-5 %

0%

Distribution of Average Effective Tax Rates (Employees). Source: EUROMOD.

Table 1.

Mean Marginal and Average Effective Tax Rates Before and After Inflation in an Unadjusted Tax System. 1998

Finland Ireland The Netherlands United Kingdom

Following Inflation of 7%

Mean METR (%)

Mean AETR (%)

Relative Change in Mean METRs (%)

Relative Change in Mean AETRs (%)

50.0 39.1 41.2 33.0

30.2 21.1 23.8 20.1

2.1 3.4 0.2 0.3

5.3 6.0 4.0 4.5

Note: Reported changes assume that the fiscal drag is allowed to operate fully, i.e. in the absence of any offsetting adjustments to tax systems. Changes are relative to 1998 values, not in percentage points. Source: EUROMOD.

concentrated in the UK than in the other countries. Ireland shows the widest dispersion of rates. Table 1 shows that at 33%, METRs are lowest in the UK on average (first column) followed by Ireland and the Netherlands. They are highest in Finland. Turning to average rates (Fig. 3 and second

154

HERWIG IMMERVOLL

column of Table 1), one finds again a relatively pronounced contrast between Finland and other countries. Most Finnish employees are faced with effective tax rates of between 30–35%. More than a quarter face AETRs above 35%, whereas in the other countries this is only true for a much smaller minority (0.5% in the UK, 3% in Ireland and 9% in the Netherlands). Although it would be interesting to explore these results further, the main interest here is how inflation can cause these rates to change in an unadjusted tax system (a detailed discussion of distributions and driving factors of effective tax rates in 15 EU countries can be found in Immervoll, 2004). In a progressive tax system, one would expect inflation to cause increases in average tax rates for most taxpayers: their tax burdens will increase because of eroded deductions, tax credits and tax band limits. On the other hand, marginal tax rates will only increase for those ‘‘slipping’’ across the upper limit of their current tax band. In a large number of cases taxpayers will remain in the same tax bracket but will be confronted with higher average rates. In these cases, the theoretical and empirical results discussed above would point towards increasing real labour costs. Whether and to what extent inflation raises marginal tax rates for a significant number of tax payers depends on the inflation rate and the proximity of taxpayers’ nominal incomes to the next higher tax band limit. An increase in marginal rates will obviously be more likely if tax bands are narrow or inflation is high.16 The tax bands of the countries considered here are rather wide. In most cases, one would therefore expect the number of taxpayers with unchanged marginal tax rates to dominate in cases of low-to-moderate inflation. This is confirmed by the results reported in the last two columns of Table 1. The fiscal drag caused by uniformly rising incomes has a much bigger effect on average tax burdens than on marginal rates. Indeed, with AETRs, and therefore total tax burdens on earnings, going up by between 4% and 6%, fiscal drag has a substantial impact on both household incomes and real tax revenues. Fig. 4 shows that AETRs go up for almost everybody. In most cases, any negative influence on AETRs, which may result from regressive social contribution systems, less than compensates for the rise in income tax burdens. At the same time, the majority of employees see unchanged METRs in Finland (68%), the Netherlands (60%) and the UK (85%). This is shown in Fig. 5. The fraction of employees with constant METRs is smaller in Ireland (42%). Interestingly, we also see reduced METRs for a small number of employees. This can result from upper contribution limits built into social insurance contribution schedules. In


155

90% 80%

FI

70%

IR

Share

60%

NL

50%

UK

40% 30% 20% 10%

>

+2

+2

0%

0%

5% +1

+1

5

0

to

to

+1

+1

0%

% +5

0

to

to

=

+5

0%

0%

-5

to 0

to

-5

%

0% -1

-1

-1

-2

5

0

to

to

+4 0

%

5%

% +3

0

to

+3

+3 0

+2

5

to

+2 to 0

+2

to

+2

5%

0%

% +1 5

to 0 +1

+5

to

+1

+1 5

0%

% 0

to

+5

0% =

40%) for a considerable number of taxpayers. In the UK, where tax bands are wide, the fraction of employees with unchanged METRs is much higher than in the other countries. If, as is assumed in this chapter, benefit payments are not adjusted for inflation either, AETRs also increase as the real value of benefit payments is eroded.18 Family benefits have the largest impact in this respect: around 52% of Irish and 44% of Finnish employees live in families entitled to these benefits. In the Netherlands (40%) and the UK (34%) the fractions are lower. METRs are only affected by benefit payments if they are incomerelated. In this case, inflation can push individuals out of the income range where benefits are withdrawn. As withdrawal rates are often very high, METRs can decline substantially for the individuals concerned (METRs can drop all the way to zero if the incomes of those originally entitled to means-tested benefits are below the tax threshold). However, this mechanism will only be relevant for a minority of employees as most of them have incomes that make them ineligible for means-tested benefits even in the ‘‘before inflation’’ scenario.


157

In 1998, around 7% of UK employees working more than 10 hours received means-tested benefits (income support, housing benefit and council tax benefit). A further 3% benefited from the income (and working-hours) dependent in-work benefit (family credit). As a result, we see relatively large numbers of employees with decreasing marginal rates following the stimulated increase in nominal incomes. In the Netherlands, about 1% and 4% of employees are, respectively, in receipt of social assistance and rent subsidy. The fraction of Irish and Finnish employees receiving income-related benefits (family income supplement and social assistance, respectively) is below 1%.

5. INTERPRETATION AND DISCUSSION Empirical results on the responsiveness of wage levels to changes in average and marginal tax rates are available for three of the countries covered in this chapter. These are presented in Table 2. Studies for all three countries confirm the theoretical predictions of a positive wage effect of average tax rates (a negative elasticity with respect to 1–ta) and a wage-moderating effect of marginal tax rates (a positive elasticity with respect to 1–tm). How might these results be combined with the detailed tax-rate information, presented in the previous section? Unfortunately, there is no straightforward way to do this. Empirically, the main problem is that the elasticities shown in Table 2 are not available at the same disaggregated level as the Table 2.

Sensitivity of Real Wages With Respect to Marginal and Average Tax Rates. Elasticity of Real Pre-Tax Wages with Respect to:a

Finland The Netherlands United Kingdom a

Average Retention Ratio (1–ta)

Marginal Retention Ratio (1–tm)

1 0.61 (0.64, 1.40)

+0.3 +0.18 (+0.65, +1.40)

Predicted Cost-Push Effect of Fiscal Drag: Change in Real Pre-Tax Wageb +1.6% +0.7% (+0.6%; +1.4%)

Sources for reported elasticities are as follows: Finland: Tyrvaïnen (1995) as reported in Sorensen (1997); Netherlands: Graafland and Huizinga (1996); UK: Lockwood and Manning (1993). No known published studies for Ireland. Where intervals are shown, they represent the maxima and minima of different estimation approaches or measurement periods. b Computed based on changes in mean AETRs and METRs as reported in Table 1.

158

HERWIG IMMERVOLL

above tax rates. The studies cited in the notes to Table 2 use a representative agent framework. In this setup, the variation in tax rates, required for an empirical identification of links between tax rates and wage outcomes, is measured for one representative worker only. The typical approach taken to estimate the wage equations starts with the assumption of a non-segmented labour market where there is only one relevant distribution of tax rates (or the distributions are the same in all unions). It is therefore not necessary to identify who belongs to which union. Second, the union is assumed to base its target wage on the tax rates faced by an employee with average earnings living in a household that is considered to be ‘‘typical’’.19 Hence, there is no need to specify an explicit aggregation rule for the target wages of different union members.20 Put differently, the currently available elasticities of wages with respect to marginal and average tax rates do not support the level of disaggregation at which changes in AETRs and METRs have been computed here. Since the bracket creep phenomenon increases tax burdens for practically everybody, the degree of tax-rate heterogeneity is clearly less of an issue than in the case of redistributive tax reforms with large differences between gainers and losers. Nevertheless, the calculations show that the spread of tax rates (and changes thereof) is considerable. This highlights the need to go beyond aggregate studies relying on single representative individuals. By providing detailed estimates of tax rates on an individual level, microsimulation models can be used to enrich the specification of microeconometric studies of wage-setting behaviour.21 This seems necessary to gain a better understanding of wagesetting processes in segmented labour markets. Even looking at a single trade union, ‘‘members have different labour and non-labour incomes, different family characteristics, etc. and face different marginal and average tax rates [y] The empirical studies of tax effects on wage formation normally use aggregate time-series and do not allow for the microeconomic diversity that microeconometricians would like to emphasize’’ (Christiansen, 1997, p. 260). Using currently available results on the responsiveness of wages, one can merely speculate about the extent of any inflation-induced cost-push effects. One possible scenario is reported in the third column of Table 2. In line with the empirical wage elasticities, they are based on the assumption of a nonsegmented labour market. In addition, I assume that unions based their target wages on mean AETRs and METRs so that the change in mean tax rates would be the parameter that drives the union’s target wage. The resulting real wage increases between 0.6% and 1.6% suggest that inflationinduced distortions of tax systems do indeed cause cost-push effects through real wage increases. Relative to an inflation rate of 7%, real wage increases


159

in excess of 1% can be expected to cause some concern, particularly if they are the result of fiscal mechanisms that are meant to act as an automatic stabiliser. For reasons stated above, however, these numbers are subject to a number of qualifications. More research is needed to understand how disaggregated tax-rate information as presented here can be used to deepen our understanding of wage-setting processes of different groups of employees and in situations where unions represent a heterogeneous membership. Indeed, one would expect the wage-moderating effect of tax progression to be less important for groups of employees where the degree of unionisation is weak. Lockwood, Sløk, and Tranaes (2000) present an approach that recognises that wage determinants are likely to differ between different groups of workers. Using data for Denmark, they find that effects of tax progressivity on wages vary by income group. However, the tax-rate calculations used in that chapter are approximations do not account for relevant determinants of tax burdens such as household structure. The use of microsimulation models would provide more accurate tax-rate measures and a better-founded basis for studying the impact of tax burdens on wage determination. By providing detailed tax rate-data for the full range of different employment situations, these models give researchers the variation needed to identify wage equations for different parts of the labour market.

NOTES 1. See, for instance, the model presented in Zilberfarb (1981). Nowotny (1980) reviews the macroeconomic issues. 2. Since the ‘‘bracket creep’’ caused by inflation is less visible than discretionary tax policy measures, there is a distinct possibility that individuals do not appreciate the full extent of tax hikes. With ‘‘fiscal illusion’’, demand will be less responsive to tax increases than the elasticity of demand would suggest. 3. For instance, Piltz (1977, p. 175) reports that the German ‘‘Stability Law’’ dating from 1973 has introduced a temporary increase in the corporate income tax for the years 1973 and 1974. The law specified that resulting additional tax receipts could not be spent but had to be kept on reserve with the Bundesbank. 4. (114–14 0.5–100 0.3)/1.1 ¼ 70. 5. Whether and to what extent higher wages are, in fact, passed on to consumers in the form of price increases depends on the relative degrees of market power of producers and consumers. See Tarling and Wilkinson (1985) for a discussion of the mechanisms of mark-up pricing. 6. Smyth (1983) formalises these relationships in a simple model. 7. Note that it is impossible to change the average tax rate while keeping the marginal rate constant for all income levels. In Fig. 1, marginal tax rates are left

160

HERWIG IMMERVOLL

unchanged for all upper income levels, while for lower incomes, they are increased (right-hand part of Ba is flatter than B0). 8. A third relevant parameter is the opportunity cost of participation in the labour market, usually represented by out-of-work benefits, and how they are affected by the tax change. 9. For countries where unemployment benefits are non-taxable, a sufficient condition for the replacement ratio to increase is that benefit levels change by the same proportion as pre-tax wages (which is the case if the former is a fixed percentage of the latter). If benefits are taxable then the replacement rate will still go up if the tax system is progressive (and the inflation-induced tax increase is therefore larger for working individuals than for the unemployed whose taxable income is lower). Note also that the actual impact of the replacement rate on the wage outcome is likely to depend on the unemployment rate. See e.g. Graafland and Huizinga (1996). 10. Most earlier union-bargaining models of the ‘‘right-to-manage’’ type take working hours per union member as exogenously given. A few more recent studies have introduced working hours as an endogenous parameter of the bargaining. See Aronsson and Sjo¨gren (2004) and Fuest and Huber (2000). 11. Malcomson and Sartor (1987, p. 1583). As far as the bracket creep phenomenon is concerned, this statement could be made more forcefully if social contribution schedules are not adjusted for inflation. Since contribution schedules are usually regressive, inflation potentially reduces the contribution burden faced by employers. If labour costs are indeed reduced then this further strengthens the argument that the union’s target wage and actual wage outcomes can be expected to move in parallel. 12. This chapter assumes that unions and employers bargain over take-home pay. All mandatory social contributions are therefore considered a tax while employer contributions are disregarded. See Immervoll (2004) for effective tax rate measures including employer contributions. That chapter also provides a discussion of the appropriate treatment of mandatory contributions in tax-rate measures. 13. It is also possible, and quite common in some countries, that a particular union would negotiate different wage settlements for different wage levels. Nonetheless, the question remains how preferences would be aggregated over employees in each group. 14. The precise adjustment mechanisms are described in Immervoll (2005) which includes an analysis of the remaining inflation-induced tax distortions in countries operating automatic, but incomplete, inflation adjustment regimes. 15. The focus here is on the taxes paid on employment incomes E. However, where other income components (such as capital income) are taxed jointly with earnings, taxes are first determined for taxable income as a whole and then attributed to earnings E in proportion to their overall share in taxable income. 16. For tax schedules such as in Germany, where the marginal tax rate rises continuously rather than in steps, most taxpayers will face higher marginal tax rates as a result of rising nominal incomes. 17. This is because the regressive social insurance contributions are considerably more important than income taxes in terms of total revenue collected in the Netherlands. They are also much more regressive than in other countries. Employee and employer health insurance contributions are no longer payable at all once


161

employment income exceeds a certain upper limit as higher-earning employees are expected to find private health insurance. See Immervoll (2005) for a description of relevant Dutch and British income tax and contribution rules. 18. In order to capture the pure ‘‘bracket creep’’ effect due to unadjusted income tax and social insurance contribution rules, simulations were also run for a scenario where all benefit rules are indexed for inflation. These results are not reported here but are available on request. Essentially, adjusting benefit rules reduces changes of both AETRs and METRs but differences turn out to be small. 19. For instance Lockwood and Manning (1993). 20. See Booth (1985), Oswald (1982, 1985) for discussions of different aggregation rules. 21. Numerous studies examining the role of tax-benefit systems in labour supply decisions have successfully done this. See for example Blundell, Duncan, McCrae, and Meghir (2000), Callan (2000), Callan and Van Soest (1993) and Duncan and Weeks (1997).

ACKNOWLEDGMENTS I am grateful for access to micro-data from the Finnish Income Distribution Survey (IDS) 1996 made available from Statistics Finland; the Living in Ireland Survey, 1994 (Wave 1) made available by the Economic and Social Research Institute (ESRI), Dublin; the Socio-Economic Panel Survey (SEP) made available by Statistics Netherlands through the mediation of the Netherlands Organisation for Scientific Research – Scientific Statistical Agency; and the UK Family Expenditure Survey (FES), which have been made available by the Office for National Statistics (ONS) through the Data Archive. Material from the FES is Crown Copyright and is used by permission. Neither the ONS nor the Data Archive bear any responsibility for the analysis or interpretation of the data reported here. An equivalent disclaimer applies to the other data sources and their respective providers. I am grateful for helpful comments on an earlier version of this chapter from Tim Callan, Cathal O’Donoghue, Frank Wilkinson and Klaas de Vos.

REFERENCES Aaron, H. J. (1976). Inflation and the income tax. Washington, DC: Brookings Institution. Aronsson, T., & Sjo¨gren, T. (2004). Is the optimal labor income tax progressive in a unionized economy?. Scandinavian Journal of Economics, 106, 661–675. Aronsson, T., Wikstro¨m, M., & Bra¨nnlund, R. (1997). Wage determination under non-linear taxes: Estimation and an application to panel data. Oxford Economic chapters, 49, 404–418.

162

HERWIG IMMERVOLL

Blanchard, O., & Katz, L. F. (1999). Wage dynamics: Reconciling theory and evidence. American Economic Review. chapters and Proceedings, 89, 69–74. Blinder, A. S. (1973). O1_MRKO1_MRKCan income tax increases be inflationary? An expository note. National Tax Journal, 26, 295–301. Blundell, R., Duncan, A., McCrae, J., & Meghir, C. (2000). The labour market impact of the working families tax credit. Fiscal Studies, 21, 75–100. Booth, A. L. (1985). The free rider problem and a social custom model of trade union membership. Quarterly Journal of Economics, 100, 253–261. Brunila, A., Buti, M., & in ’t Veld, J. (2003). Fiscal policy in Europe: How effective are automatic stabilisers?. Empirica, 30, 1–24. Cahuc, P., & Zylberberg, A. (2004). Labor economics. Cambridge, MA: MIT Press. Callan, T. (Ed.) (2000). Taxes, transfers and labour market responses: What can microsimulation tell us? Dublin: Economic and Social Research Institute. Callan, T., & Van Soest, A. (1993). Female labour supply in farm households. The Economic and Social Review, 24, 313–334. Christiansen, V. (1997). Discussion in public finance solutions to the European unemployment problem?. Economic Policy, 25, 223–264. Dernburg, T. F. (1974). The macroeconomic implications of wage retaliation against higher taxation. IMF Staff chapters (Nov.). Duncan, A., & Weeks, M. (1997). Behavioural tax microsimulation with finite hours choices. European Economic Review, 41, 619–626. European Commission. (1999). Study 4: Market integration and differences in price levels between EU member states. In: E. Commission (Ed.), EU Economy: 1999 review. Brussels: European Commission. Fuest, C., & Huber, B. (2000). Is tax progression really good for employment? A model with endogenous hours of work. Labour Economics, 7, 79–93. Graafland, J. J., & Huizinga, F. H. (1996). Taxes and benefits in a non-linear wage equation. Research Memorandum, 126. Hansen, C. T., Pedersen, L. H., & Slok, T. (2000). Ambiguous effects of tax progressivity – theory and Danish evidence. Labour Economics, 7, 335–347. Heinemann, F. (2001). After the death of inflation: Will fiscal drag survive?. Fiscal Studies, 22, 527–546. Hersoug, T. (1984). Union wage responses to tax changes. Oxford Economic chapters, 36, 37–51. Holmlund, B., & Kolm, A. S. (1995). Progressive taxation, wage setting and unemployment – theory and Swedish evidence. Tax Reform Evaluation Report, 15. Immervoll, H. (2004). Average and marginal effective tax rates facing workers in the EU: A micro-level analysis of levels, distributions and driving factors. OECD Social, Employment and Migration Working chapter No. 19, available through www.oecd.org/els/ workingchapters. Immervoll, H. (2005). Falling up the stairs: The effects of bracket creep on household incomes. Review of Income and Wealth, 151(1), 37–62. Jackson, D., Wilkinson, F., & Turner, H. A. (1972). Do trade unions cause inflation? Cambridge: Cambridge University Press. Koskela, E. (2001). Labour taxation and employment in trade union models: A partial survey. Bank of Finland Discussion chapter, 19. Lockwood, B., & Manning, A. (1993). Wage setting and the tax system. Theory and evidence for the United Kingdom. Journal of Public Economics, 52, 1–29.


163

Lockwood, B., Sløk, T., & Tranæs, T. (2000). Progressive taxation and wage setting: Some evidence for Denmark. Scandinavian Journal of Economics, 102, 707–723. Malcomson, J. M., & Sartor, N. (1987). Tax push inflation in a unionized labour market. European Economic Review, 31, 1581–1596. Messere, K. (Ed.) (1998). The tax system in industrialized countries. Oxford: Oxford University Press. Musgrave, R., & Musgrave, P. B. (1976). Public finance in theory and practice (2nd Ed.). New York: McGraw-Hill. Nowotny, E. (1980). Inflation and taxation: Reviewing the macroeconomic issues. Journal of Economic Literature, 18, 1025–1049. OECD. (1976). The adjustment of personal income tax systems for inflation. Paris. OECD. (1986). Personal income tax systems under changing economic conditions. Paris. Oswald, A. J. (1982). The microeconomic theory of the trade union. Economic Journal, 92, 269–283. Oswald A. J. (1985). The economic theory of trade unions: An introductory survey. Scandinavian Journal of Economics, 87, 160–193. Piltz, D.J. (1977). Deutscher landesbericht In: International Fiscal Association (Ed.), Studies on international fiscal law: Inflation and taxation (pp. 161–179) Vienna. Pissarides, C. A. (1998). The impact of employment tax cuts on unemployment and wages; the role of unemployment benefits and tax structure. European Economic Review, 42, 155–183. Røed, K., & Strøm, S. (2002). Progressive taxes and the labour market: Is the trade-off between equality and efficiency inevitable?. Journal of Economic Surveys, 16, 77–110. Saez, E. (2003). The effect of marginal tax rates on income: A panel study of ‘bracket creep’. Journal of Public Economics, 87, 1231–1258. Smyth, D. J. (1983). Taxes and inflation. In: N. Schmukler & E. Marcus (Eds), Inflation through the ages: Economic, social, psychological and historical aspects (pp. 326–339). New York: Brooklyn College Press. Sorensen, P. B. (1997). Public finance solutions to the European unemployment problem?. Economic Policy, 25. Sutherland, H. (2001). Final Report. EUROMOD: An integrated European benefit-tax model. EUROMOD Working chapter EM9/01, available through www.iser.essex.ac.uk/msu/ emod.php. Tanzi, V. (1980). Inflation and the personal income tax: An international perspective. Cambridge: Cambridge University Press. Tarling, R., & Wilkinson, F. (1985). Mark-up pricing, inflation and distributional shares: A note. Cambridge Journal of Economics, 9, 179–185. Tyrvaïnen, T. (1995). Real wage resistance and unemployment: Multivariate analysis of cointegrating relations in 10 OECD countries. OECD Job Study Working chapter, 10. Wulfsberg, F. (1996). Do progressive taxes reduce wage pressure? Panel data Evidence from Norway. Memorandum 38/96, Department of Economics, University of Oslo. Zilberfarb, B. Z. (1981). Partial tax indexation and economic stability. In: J. M. Flanders, & A. Razin (Eds), Development in an inflationary world. Proceedings of the 1979 Pinhas Sapir Conference, New York (pp. 163–171).

BEANS FOR BREAKFAST? HOW EXPORTABLE IS THE BRITISH WORKFARE MODEL? Olivier Bargain and Kristian Orsini ABSTRACT Social assistance and inactivity traps have long been considered as one of the main causes of the poor employment performance of EU countries. The success of New Labour in the UK has triggered a growing interests in instruments capable of combining the promotion of responsibility and selfsufficiency with solidarity with less skilled workers. Making-work-pay (MWP) policies, consisting of transfers to households with low earning capacity, have quickly emerged as the most politically acceptable instruments in tax-benefit reforms of many Anglo-Saxon countries. This chapter explores the impact of introducing the British Working Families’ Tax Credit (WFTC) in three EU countries with rather different labor market and welfare institutions: Finland, France and Germany. Simulating the reform reveals that, while first-round effects on income distribution is considerable, the interaction of the new instrument with the structural characteristics of the economy and the population may lead to counterproductive second round effects (i.e. changes in economic behavior). The implementation of the reform, in this case, could only be justified if the social inclusion (i.e. transition into activity) of some specific household


165

166

OLIVIER BARGAIN AND KRISTIAN ORSINI

types (singles and single mothers) is valued more than a rise in the employment per se.

1. INTRODUCTION From a labor supply perspective, high-tax wedges coupled with generous minimum income and unemployment benefits are perceived as one of the main causes of persistently lower employment rates of less skilled workers, since they make work financially unattractive. The picture is particularly dramatic when employment rates are broken down by gender. According to recent Labor Force Survey statistics, female employment rate in EU15 is just above 65% in 2004 (i.e. 20 percentage points lower than the male employment rate), while the employment rate for low skilled women is as low as 46%, with almost half of active women being marginally or part-time employed. Any attempt to increase employment rates in Europe should therefore primarily target this group. Employment policies in Europe have traditionally relied on labor demand levies, such as subsidies and targeted social security contribution (SSC) reductions (Martin & Grubb, 2001). Recently, however, the unquestionable success of New Labour in the UK has triggered an increased interest in supply side measures. In recent years, while unemployment rates have increased in many EU countries and employment rates have been stagnant, the UK has seen its unemployment rate falling to just under 5%. Even more remarkably, the employment rate of females has increased by almost 3 percentage points. Making-Work-Pay (MWP) policies – i.e. policies that tackle low employment from a labor supply perspective through employment conditional transfers – are argued to play a significant role in the success of British labor market dynamics. Following the pioneering measures in the US and the UK – the Earned Income Tax Credit (EITC) and the Working Families’ Tax Credit (WFTC) respectively – several European countries are now looking with renewed interest at such policy instruments. Indeed as argued by Banks, Disney, Duncan and van Reenen (2003), the globalization of knowledge has lead to increased opportunities to learn from the experiences of other countries. According to the authors, this has been particularly evident in the case of labor market policies, retirement and saving policies and indeed in ‘MWP policies’. The analysis of successful experiences in other countries have recently gained the interest of policymakers is witnessed by several studies carried out by their advisors. For example, in a government report on employment and

How Exportable is the British Workfare Model?

167

economy perspectives for the next decades, Ilmakunnas, Kiander, Parkkinen and Romppanen (2000) do no exclude that Finland may recur to earned income tax credits similar to those employed in other countries. While underlining the success of such policies the board of advisors are cautious about the effects in a different economic and institutional settings. In the report of the advisory board of the German Ministry of Economics and Technology on employment of the low skilled (Bundesministerium fu¨r Wirtschafts und Technologie, 2002), the authors analyze in detail the British WFTC and the American EITC. They even simulate the introduction of the EITC in the German context, but only on some typical households. Finally they argue that the extent to which a German reform of low skilled employment should resemble other international experiences is mainly a political issue. In the French context, the British WFTC experience has been cited twice in the reports of the economic advisory board of the prime minister (see Bourguignon & Bureau, 1999; Ferry, 2000). Most striking are perhaps the words of Maarek – member of the economic advisory board of the French prime minister – in the final remarks of the report of Bourguignon and Bureau (1999): It is now time to seriously look at some form of negative income tax. This is no longer a simple intellectual speculation. The reforms could be inspired by the American EITC, or the British WFTC [translated from French]

Indeed, policymakers are mostly attracted by the capacity of MWP to combine solidarity and the promotion of self-sufficiency. However, the extent to which these instruments may be successfully implemented in the context of western and northern European countries has not yet been thoroughly analyzed. This chapter aims at assessing the validity and applicability of MWP policies to the European context, by looking at three EU countries – France, Germany and Finland – that despite being characterized by different institutions and so called ‘framework conditions’, all characterized by inactivity traps. France and Germany have indeed struggled to increase employment rates in the past years, while unemployment rates have kept increasing. Germany, moreover, continues to show lower employment rates for mothers, and a higher share of part-time work. On the other hand, in the years following the early 1990s’ recession, Finland has been characterized by declining unemployment rates and increasing employment rates. Yet unemployment among the low skilled population remains extremely high, as compared to other Scandinavian countries. It is not surprising that these countries, as well as several other EU countries have looked with interest at the British policies. The relative consensus on the need for implementing MWP policies is however muddied by concerns

168


about efficient policy design, with respect to the structural characteristics of the population and of the economy and the general objectives pursued. The institutional context in which the policy implementation takes place (i.e. existing tax-benefit system, presence of a minimum wage, etc.) as well as the shape of wage and income distributions and the size of labor supply sensitivities are known to have a deep impact on the effectiveness of such instruments.1 Moreover it is dubious that redistributive and efficiency objectives may be pursued with equal intensity by a single-policy measure.2 Being means-tested on household income, policies in the fashion of the WFTC and the EITC are known to be well targeted at households in need but may indeed discourage the work of women in two-earner couples. All these aspects will be analyzed in a comprehensive way by simulating the implementation of a pseudo-WFTC in the tax and benefit legislation of the three aforementioned countries. The structure of the chapter is as follows: Section 2 describes the British experience and explain how the WFTC was modified to be adapted to the countries at stake. Section 3 describes the pre reform tax-benefit systems and analyzes how the budget constraints were modified. Section 4 provides a static analysis of the reform in terms of budgetary costs and effects on poverty, while Section 5 analyzes different indicators to derive plausible expectations of labor supply effects. Section 6 concludes.

2. SIMULATING THE BRITISH EXPERIENCE In-work benefits such as the EITC and the Family Credit and it successor – the WFTC – have been in place for over 20 years in the US and in the UK, respectively. Canada, Ireland and New Zealand have also had a relatively long experience with such schemes. In recent years, MWP are rapidly gaining policymakers’ interest in continental Europe as well – where there is a widespread consensus on the need for such policies (Banks et al., 2003). The official objective set forth by policymakers is double: (i) to expand employment by increasing work incentives, (ii) to increase income of disadvantaged groups (Pearson, 2002). The second objective is clearly redistributive and in-work benefits seem an interesting way to redistribute to a subset of the poor population known as the ‘working poor’. Such instruments can be seen as more politically acceptable than a rise in social assistance – given the feared increase in work disincentives – and more efficient than an increase in the minimum wage, which might push up wage rates above the market equilibrium and hence lower the employment rate.


169

The British WFTC has attracted considerable attention from researchers and policymakers. From a political point of view, the tax credit represents one of the most celebrated achievements of the Blair administration in the current and passed legislation, and a concrete application of the New Labour ideology. Its (potential) effects have been evaluated in several studies from a variety of perspectives and using different methodologies (see Brewer & Shephard, 2004 for a review). The WFTC represents therefore a ‘natural candidate’ for transposing in-work benefits in Continental and Scandinavian welfare systems. At the same time, as it will become clearer in the discussion, some minor adaptations will be needed in order to ensure coherence between the simulated benefit and the tax-benefit system in each country.

2.1. The Working Families’ Tax Credit in the UK In October 1999, the WFTC replaced the Family Credit (FC), by maintaining the main framework of the older benefit and substantially increasing the income range of eligibility. The basic WFTC transfer is conditioned on at least one of the adults in the household being in paid work (employment or self-employment) for at least 16 hours per week. Once household income reaches the threshold level, the maximum amount is tapered away, at a rate of 55% as household’s income increases. Income, however, is assessed after income tax and contributions have been paid, so that, in the phase out range, the withdrawal rate with respect to gross income is 37%. The maximum amount of benefit increases with the number of children (the basic transfer increases by 49% for every child). Moreover a premium corresponding to 20% of the basic transfer is awarded if one household member works at least 30 weekly hours. The new WFTC also introduced a deductability of 70% of approved child care, up to a maximum of 150 pounds/week. While MWP policies in the fashion of the WFTC unequivocally encourage the participation of singles, it is often the case that they discourage second-earners in couples, bringing about a gender bias against the participation of women. Moreover, the generosity of the reform implies a sharp taper rate in the phase-out range of the measure. Average tax rates are hence lowered at the cost of an increase in effective marginal tax rates (EMTRs hereafter) generating disincentives at the intensive margin (see Eissa & Hoynes, 2004; Blundell, Duncan, McCrae, & Meghir, 2000). Ex-ante predictions based on microsimulation software and structural models of labor supply were partially confirmed by ex post analysis using the ‘difference in differences’ methodology (Brewer & Shephard, 2004) as well as

170


by official statistics. Using the Family Resource Survey and the tax-benefit model TAXBEN3, Blundell et al. (2000) have evaluated ex-ante the labor supply responses to the WFTC. It is found that nearly 80% of lone parents in part-time employment (between 16 and 30 hours per week) benefit from the reform. As for couples, the credit is more generous for one-earner households, a third of which would be entitled to the tax credit. The impact on hours is ambiguous as the number of households with an EMTR above 70% decrease by around 450,000 while households with an EMTR above 50% increase by about the same amount. This is due to a lower taper rate (55% instead of 70% with the FC) entailing a smaller positive impact on EMTRs but for a larger number of people. The net change in participation rate would consist of an increase by 2.2 percentage points for single mothers (34,000 individuals) and a decrease by 0.57 percentage points (20,000 individuals) for married women with employed partners. Combining all the behavioral effects, the WFTC leads to a small increase in overall participation, by just above 27,000 individuals. Labor supply responses to the WFTC should act to reduce the cost of the program by around 14%. Brewer et al. (2003) used a more general model (with selective program participation) and find similar results. However, comparing the results with ex-post data, the authors find that the overall participation rates of both females and males in couples decreased. Indeed, the policy mix between April 1999 and April 2000 also increased social aid for households with children under the age of 11. The generous policies aiming at a drastic decrease of child poverty have therefore partially reduced the positive effect of the WFTC. Yet the employment rates of lone mothers increased by almost 3.4 percentage points. Few works assess the impact of the measure from a distributional perspective. Pichaud and Sutherland (2002) decompose the effect of different policy measures in a static framework and conclude that while Labour policies have been rather effective in reducing poverty risk among children (with almost one million children being lifted out poverty from tax-benefit reform), other groups have been significantly neglected, so that overall poverty rates were not characterized by a significant reduction.

2.2. Simulating a Working Tax Credit in France, Germany and Finland Policies aiming at increasing the financial incentives of employment have been implemented in most EU countries in the previous decade. Germany and Finland have aimed at reducing the tax burden on labor income through allowances and tax rebates, whereas France has gone one step


171

further, by introducing a refundable tax credit (see Laine, 2002; Haan & Steiner, 2005a; Bargain, 2004 for details on Finnish, German and French reform respectively). The budget allocated to these recent reforms, however, is well below the level of transfers implied by the British schemes. Even the most generous reform, i.e. the German tax reform, will imply a maximum net gain of slightly over 1000 EUR, well beneath the British WFTC. An additional motivation for the present study is to question what could have happened, had these countries dedicated the same budgetary expenses as the UK to MWP transfers. The rules and the amounts of the new tax credit to be simulated in France, Germany and Finland match closely those of the British 2001 WFTC. However we apply the tax credit to the tax and benefit system of 1998.3 Eligibility requires that at least one of the adults is in paid work for at least 16 hours per week. The formula to compute total household entitlement to the tax credit is as follows: TC ¼ B maxð0; ðz yÞtÞ

(1)

The maximum theoretical amount of the benefit B is composed of a basic transfer of EUR 3880 plus the additional 49% of the basic transfer per dependent child and the 20% premium for working more than 30 hours. The threshold or disregard, y, amounts to EUR 6671 while the taper rate t equals 55%. Note that while this corresponded roughly to 37% on gross income in the UK, the corresponding percentage will vary across the countries we examine, due to differences in the level of SSC and income tax. Fig. 1 shows the amount of the credit in EUR/year when total weekly working hours increase from 0 to 80 (at a net hourly wage rate of 6 EUR/ hour). The 16 and 30 hours thresholds are particularly evident, as they mark the eligibility and premium conditions respectively. The basic idea behind the British WFTC is to design a benefit that increases the financial incentives for poor households of being in paid work. The transfer, which is conditioned on at least one adult being in paid work, is substantially higher than the minimum income. The structure of the benefit, including unit of assessment, definition of qualifying child and income concept, is the same as for the minimum income scheme. The potential gains of WFTC eligibility are hence homogeneously extended to all households (with adults in working age) in social assistance. In assessing income all main sources of income are included net of tax and SSC. These include earnings, self-employment income, unemployment benefits, pensions, irregular incomes, capital income and maintenance income.

172


1000 WTC (singles/couples)

900

WTC (singles/couples with two children) 800

WTC (EUR/year)

700 600 500 400 300 200 100 0 0

10

Fig. 1.

20

30 40 50 Hours Worked (hours/week)

60

70

80

Amount of Benefit by Weekly Hours Worked.

The main excluded incomes are the Child Benefit, the Maternity Benefit, the Statutory Maternity Pay, and some minor UK-specific disregards on maintenance payments and war pensions. At the same time, the amount of the WFTC enters the assessment of the Income Support so that when receiving the WFTC, a households looses eligibility to the Income Support. Moreover, households eligible to income assistance are automatically eligible for the full housing benefit, whereas the latter is tapered away for higher levels of household income (including the WFTC transfer). The net effect of the WFTC tends therefore to be much smaller, as eligibility to the in work benefit reduces the transfers from other schemes. In implementing the reforms in the countries we consider, we have tried to preserve the same logical structure, while keeping the tax credit as close as possible to the WFTC. Eligibility conditions and income assessment rules were therefore adapted in an effort to balance cross-country comparability and coherence with each country’s social assistance legislation. In France, the assessment is the same as for the social assistance schemes (Revenu Minimum d’Insertion and Allocation pour Parents Isole´s), but similar to the WFTC the generous child benefit (Allocation Familiale) where


173

excluded from the income assessment. In Finland and Germany, on the other hand, income assessment for existing social assistance benefits (Toimeentulotuki in Finland, Sozialhilfe in Germany) are already net of the universal child benefit. Other minor income disregards existing in each country’s social assistance legislation were not modeled. Similarly to the UK, we retain the definition of a qualifying child used for social assistance in each country. While this choice might in theory generate slightly more advantageous schemes in some countries (e.g. in Germany, where there is no age limit for qualifying child), it is unlikely that the latter will play a major role, given that there are very few households with grown up children. A third issue concerns the interaction of the credit with other meanstested instruments. As in the UK, the WFTC enters the assessment of the national social assistance schemes.4 Differently from the UK, however, we decided not to model any ‘direct’ interaction with the housing benefit, in order to avoid multiple interferences.5 Due to the latter, the net transfer will be comparatively higher than in the UK, since eligibility to the WFTC does not cause the housing benefit to be withdrawn. This choice partially reflects the different role of the housing allowances in continental Europe and in Scandinavia. Although means-tested, these benefits are designed to enhance affordability of housing costs. In the UK, housing benefits mainly function as an additional safety net, which explains why they are withdrawn as households become eligible for WFTC.6 Other than the aforementioned country-specific adaptations, the main difference with the WFTC is that childless households are also eligible to the credit. The credit thus anticipates some of the features of the 2003 reform of the British system. The new structure involves two separate credits: a refundable Child Tax Credit (CTC) to support children in low-income families, regardless of the work status of the parents, and a Working Tax Credit (WTC) now extended to childless singles and couples.7

3. A FIRST LOOK AT THE RESULTS In the current section we will concentrate on each’s country tax and benefit system and analyze how the budget constraint is modified when introducing the tax credit. This analysis, although based on hypothetical households, allows for a first intuition of the effects of the interaction of the different instruments and its potential impact on income distribution and work

174


incentives. In addition, differences in eligibility rates and entitled amounts across countries may be explained by structural differences in terms of employment rates and household composition.

3.1. Pre- and Post-Reform Budget Constraints The budget curves in Fig. 2 shows pre-and post-reform budget constraints for a single earner and for a secondary earner in a couple with two children. Gross earnings are on the horizontal axis and disposable income on the vertical axis. Each unit on the X axis represents an increase in gross income corresponding to a unitary increase in weekly working hours, times a fixed gross hourly wage (in this case 6 EUR).8 In the case of couples, the increase in earnings of the secondary earner adds up to the earning of the other spouse, whose earnings correspond to working full time at the same wage rate. The analyzed countries differ profoundly with respect to shape of personal income tax, weight of SSC and relative importance of family-related and other benefits.9 However, the crucial aspect here is the relatively flat region which characterizes single earners in all three countries and clearly illustrates the inactivity trap. Despite some differences in eligibility rules, disregarded incomes and benefit level, minimum income schemes usually share a common structure: they are means-tested on net income. This implies that every additional euro of net income is automatically offset by the loss of a euro of social assistance. In term of effective marginal tax rates, households on welfare are then characterized by an implicit 100% taxation of their earnings. In year 1998, German minimum income was more generous than in Finland and in France. The minimum income for a single was 4,650 EUR per year in Germany, and about 4,000 in Finland and France, whereas the amounts for a lone parent with two children corresponded respectively to 9,627, 6,283 and 5,432 EUR.10 France, however, relies on a combination of minimum income and generous housing benefits. The latter is phased out at a very slow rate but does not enter the income assessment of social assistance, which explains why the first segment is slightly negatively sloped. More generous minimum income in Germany makes the flat segment longer than in France and Finland. However, small amounts of labor income (70 EUR per month) are entirely disregarded, which explains the small portion of the curve displaying a 45 degree slope; additional earnings are partially disregarded (30%) up to a maximum of 140 EUR of disregard, which corresponds to the second positively slopped portion of the curve. Beyond such threshold, however, the withdrawal rate is of 100%.

Disposable income (EUR/year)


25000

France: singles without children


40000

France: Couples, two children

35000 20000

30000 25000

15000

20000 10000

15000 10000

5000

5000 0 Gross labour income (EUR/year)


175

25000

Germany: singles without children

0

40000

Gross labour income (EUR/year) Germany: Couples, two children

35000 20000

30000 25000

15000

20000 10000

15000 10000

5000

5000 0

25000

Gross labour income (EUR/year) Finland: singles without children

0

40000

Gross labour income (EUR/year) Finland: Couples, two children

35000 20000

30000 25000

15000

20000 10000

15000 10000

5000

5000 0

Gross labour income (EUR/year) Post reform disposable income

Fig. 2.

0

Gross labour income (EUR/year)

Pre reform disposable income

Pre- and Post-Reform Budget Curves.

Gross income

176


Working full time at an hourly wage rate of 6 EUR is enough to earn an income above the social assistance threshold (for a couple with two children) in France and in Finland, but not in Germany. This explains the small flat segment even in the graph on the right-hand side. As the secondary earner increases the labor supply, household disposable income increases in all countries, but the rate of the increase is different due to differences in SSC and personal income tax (PIT). In the case of Finland, taxation of low incomes is quite low, which explains a steeper budget curve (i.e. closer to the 45% line). In France, PIT is typically low, but SSC are high (and contrary to income tax, not progressive). Moreover, the aforementioned phasing out of the housing benefit accounts for the somewhat flatter profile of the budget constraint. In Germany, on the other hand, both SSC and PIT tend to be high even at low incomes. In the case of singles, the aforementioned 16 and 30 hours thresholds are particularly evident in the shape of the budget constraint. The difference between pre- and post-reform disposable income appears to be much smaller than the amount of the tax credit, which is due to the fact that the tax credit enters income assessment for social assistance, and thus crowds out such benefit. This effect occurs in all countries and is maximum in the case of Germany, due to a more generous safety net. The maximum net gains are therefore in a range between 10,000 and 15,000 EUR. It appears, however, that in all countries, the reform creates a significant financial difference between non-participation and full-time activity, unambiguously enhancing the probability of participation for single individuals or lone parents. Also the phasing out of the benefit is much slower in the case of Germany: this is due to the fact that heavy SSC reduce the net income. When looking at the secondary earner case, the kinks in the budget curve are no longer visible: indeed it is the first earner that opens up eligibility to the tax credit. This implies that when the secondary earner starts working, he or she are already in the phase out range of the benefit, which implies that increased working effort and higher gross earnings are transformed into negligible increases in household disposable income. The relative sharpness of the phase out is moreover affected by the slope of the pre-reform budget constraint.

3.2. Household Structure and Employment Participation rates give a first indication on the proportion of households potentially eligible to the in-work benefit. More precisely we may look at the proportion of households with at least one member working a minimum of


177

Table 1. Distribution of Households According to Income Ranges in WTC Rules.

Not eligible Potentially eligiblea Maximum benefit In benefit phase-out range Out of benefit range

France (%)

Germany (%)

Finland (%)

29.8 70.2 1.1 13.1 56.0

43.2 56.8 0.3 11.7 44.8

33.5 66.5 0.4 14.3 51.8

Source: Authors’ computations using EUROMOD. a Households with one adult employee working at least 16 hours.

16 hours per week. Table 1 shows the proportion of these ‘potentially eligible’ households, in percentage of the total number of households with head or spouse in working age. Other than the employment rate, also the distribution of the potentially eligible households in the benefit range is a matter of primary concern. The number of eligible units, and their distribution in the benefit range determines not only the cost of the measure, but also the size of potential behavioral responses. The share of non-eligible units is highest in Germany, whereas in France and in Finland only one household out of three is not eligible The distribution of eligible households in the benefit range, on the other hand, appears to be very similar across the three countries, given that the share of households out of the benefit range is substantially higher in France and in Finland (56.8% and 44.8% respectively), than in Germany (44.8%).11 With respect to the population in the benefit range, Finland shows a slightly higher concentration of households in the phase out range (14.3% as compared to 13.1% and 13.7% of France and Germany respectively), whereas France has the highest share of households in the maximum benefit range (1.1%, i.e. from 4 to 3 times higher than Germany and Finland). The previous picture, nevertheless, hides substantial differences in the structure of the eligible population. The latter not only will affect the redistributive power of the reform, but will also have a substantial impact on the size and direction of the potential behavioral adjustment. Crucial elements of the analysis are (i) the share of singles and couples, (ii) the relative share of single (male) breadwinner and double earner households and (iii) the share of households with and without children. This information is summarized in Table 2; household types are classified in accordance to the structure of the household and female employment status (since there is very limited variation in the employment status of males). The households considered are headed by

178


Table 2.

Distribution of Households Types According to Female Employment Status. France

Single males Single females Single mothers

Total couples

Finland

Not Employed (%)

Employed (%)

Not Employed (%)

Employed (%)

Not Employed (%)

Employed (%)

6.6 2.1 0.3

6.6 5.0 2.6

10.3 2.9 1.4

10.3 7.0 2.2

11.4 3.2 0.7

11.4 6.7 2.7

Total single adult households Couples, no children Couples with children

Germany

16.6

23.8

24.7

4.6

8.0

3.9

9.1

3.6

11.9

6.3

17.7

5.4

7.9

3.8

15.0

36.5

26.3

34.3

Source: Authors’ computations using EUROMOD.

potentially active singles or couples, only. Elderly households as well as more complex household structures are not included in the sample as this is a rather heterogeneous group, characterized by law eligibility rates. A first-striking aspect is the difference in the share of singles: Finland has almost twice the share of single adult household than France (24.7% and 16.6% respectively), whereas Germany is closer to Finland. Between half and one-third of single female households are inactive, but the share of inactive single mothers is extremely low in France and in Finland, but substantially higher in Germany. France has therefore the lowest share of singles, but shows also the highest activity rate among this group. For sake of comparison, in year 1998, the share of single-mother households was slightly under 5% in the UK, with almost half of the single-mothers being inactive. Potentially eligible households are also encountered among couples and especially couples with children, where the benefit range is widest. France has the highest share of couples with children, but again the employment rate of secondary earners (i.e. mostly females) is rather high, which implies that a higher shares of households might be pulled out of the benefit range. The share of active women in couples with and without children is highest in Finland (15.0% and 11.9%, respectively), and lowest in Germany (7.9% and 9.1%, respectively). France has slightly lower employment rates than Finland, but the share of working mothers is as high as 17.7%.


179

4. BUDGETARY COST AND EFFECTS ON POVERTY In this section we will review the budgetary cost of the reforms and compare them across countries. We will also analyze the first-round distributional effects of the reform. In other words, we will look at the reduction in poverty rates brought about by the change in the budget constraint, without taking into consideration possible behavioral reactions. 4.1. Aggregate Statistics The aggregate impact of the reform is summarized in Table 3. The apparent cost is aggregated expenditure on the instrument, whereas the net cost is the net increase in public expenditure. The latter is indeed smaller than the former, due to the fact that spending on the tax credit crowds out spending on other transfers, mainly social assistance transfers, but also (in the case of Germany) housing benefits. Apparent cost is slightly smaller in Germany (0.36% of GDP) than in France or Finland (around 0.40% of GDP). In absolute terms, the cost of 5.8 billion EUR in France and 7.6 billion in Germany can be compared to the £5 billion spent in 2001 in the UK on the WFTC (7.3 billion EUR). However, net costs are much smaller in the case of Germany (0.29% of GDP). The reason for this, is the high generosity of social assistance. Indeed, as shown in Fig. 2, a single-breadwinner household with two dependent children (with the breadwinner working full time at a low-wage rate), and a single adult with two dependent children working more than 16 hours per week are still entitled to social assistance benefits. The introduction of the tax credit, then simply shifts part of the spending on social assistance to spending on the working tax credit. It should be noted here that our simulations assume a 100% benefit take up rate of the tax credit and of social assistance. Indeed, several households in Germany have a positive labor supply, despite being in the social assistance benefit range (with its 100% EMTR). Such behavior is consistent with the supposed high social stigma of claiming benefits.12 As expected from the discussion on framework conditions, the number of actual recipients is of comparable size. Eligibility is slightly higher in Finland (10.6% of the households) and in France (10.1%) than in Germany (9%). The difference between France and Finland is indeed quite small, which is consistent with differences in potential eligibility and income distribution. The distribution of the budgetary cost across the different household typologies is indeed driven by the differences in the structure of the

France Apparent cost Net cost Recipients Net average amount Net maximum amount

Net cost Recipients Net average amount Net maximum amount Finland Apparent cost Net cost Recipients

Net average amount Net maximum amount

Aggregate Impact of Introducing the Tax Credit. All

Singles

Lone Parents

Couples

Couples with Children

(Billion euros/year) (% of 1998 GDP) (Billion euros/year) (% of 1998 GDP) (No. of hh) (% of population) (Euros/week) (Euros/week)

5.859 0.404% 5.741 0.395% 2,316,233 10.1% 49 205

22.5%

20.3%

2.3%

54.9%

22.5%

19.9%

2.3%

55.4%

34.3%

16.3%

4.5%

44.9%

32 90

60 194

25 62

59 205

(Billion euros/year) (% of 1998 GDP) (Billion euros/year) (% of 1998 GDP) (No. of hh) (% of population) (Euros/week) (Euros/week)

7.662 0.356% 6.221 0.289% 3,443,208 9.0% 43 149

20.9%

23.7%

2.2%

53.2%

23.9%

21.4%

2.0%

52.8%

31.6%

15.7%

4.2%

48.5%

28 90

65 120

23 57

47 149

(Billion euros/year) (% of 1998 GDP) (Billion euros/year) (% of 1998 GDP) (No. of hh) (% of population)

0.506 0.394% 0.492 0.383% 248,663 10.6%

41.9%

18.0%

1.3%

38.9%

42.1%

18.4%

1.1%

38.4%

53.7%

14.4%

1.7%

30.2%

(Euros/week) (Euros/week)

39 248

31 90

49 159

30 53

50 248



Germany Apparent cost

180

Table 3.


181

population: Germany and France share a similar structure of spending, with over 50% accruing to couples with children and over 40% shared more or less equally between singles and lone parents. Compared to Germany, France is characterized by a significantly lower share of singles, but also by a higher share of eligible households, while the high employment rate of females in couples implies lower expenditure on this household type. The same arguments apply to explain the different structure of spending in Finland: the share of singles is slightly higher than in Germany, but the employment rate is also rather high. On the other hand the share of females (and especially non-mothers) in employment is higher than in France which implies that even fewer couples are entitled to the benefit.

4.2. Effects on poverty The distributive impact of the reform has been analyzed by comparing poverty indicators before and after the reform.13 Table 4 shows that the reform achieves significant poverty reduction in France as the poverty rate declines from 7.03% to 6.38%. In Germany, the reduction is not as large: poverty rate is reduced from 11.18% to 10.69%. Such result is consistent with less net per capita spending. In Finland, the reduction is almost null when looking at the 50% poverty line, which suggests that the working poor phenomenon is less developed in the Nordic country. There is however a significant poverty reduction if the 60% poverty line is considered, which means that the reforms redistribute relatively more to the ‘richest’ among the poor households in all countries. The latter shows that the tax credit is indeed not the most appropriate instrument, if the aim of the policy is to redistribute income, as the WTC only targets the ‘working poor’ which are significantly less poor than inactive households. The efficiency cost of the reform, however, i.e. the cost per year per household taken out of poverty, is extremely high. With a more conservative definition of poverty (i.e. fixing poverty line at 50%) the efficiency cost ranges from over 38,000 EUR in France to about 113,000 EUR in Germany and over 840,000 EUR in Finland. With a more liberal definition of poverty line (i.e. 60% of median equivalent income), the net cost per household out of poverty are substantially lower: 14,600 EUR, 33,800 EUR and 54.600 EUR for France Germany and Finland respectively. Arguably, efficient activation measure could reach a significantly higher poverty reduction target at a much lower efficiency cost.

182


Table 4.

Distributive Impact of Introducing the Tax Credit. France

Germany

Finland

PreReform

PostReform

PreReform

PostReform

PreReform

PostReform

1,222

1,225

1,246

1,249

1,090

1,124

Poverty rate – 60% of median Number of hhs out of povertya

14.00%

12.29%

11.18%

10.69%

11.97%

11.59%

Poverty rate – 50% of median Number of hhs out of povertya

7.03%

Median equivalent income (EUR/month)

393,110 6.38% 150,121

183,836 5.65%

5.51%

8,966 3.75%

54,654

3.72% 583

Source: Authors’ computations using EUROMOD. a Poverty rates in post-reform scenarios are computed on the basis of the pre-reform poverty line.

Yet these are only first-round effect. In the next paragraph we will briefly discuss potential labor supply reactions, and argue that efficiency cost might be even higher.

5. POTENTIAL LABOR SUPPLY REACTIONS 5.1. Effective Marginal Tax Rates A common way to characterize potential work disincentives at the intensive margin consists in computing the distribution of EMTRs in the population. EMTRs indeed measure the size of the distortions generated by the taxbenefit system, that is, the fraction which is levied from marginal additional income. This includes increased taxes to be paid but also the partial loss of means-tested benefits. Changes in EMTRs then correspond to changes in implicit wages.14 Fig. 3 shows the distribution of mean EMTR by decile of equivalent disposable income for the whole population, EMTRs being averaged over active and potentially active households of each decile only (i.e. household head must be between 25 and 60 year old, neither disabled, nor in full-time education or retired).15


183

FRANCE 0.7 0.6

EMTRs

0.5 0.4 0.3 0.2 0.1 0.0 1

2

3

4 5 6 7 8 Equivalent income decile

9

10

9

10

GERMANY

0.7 0.6

EMTRs

0.5 0.4 0.3 0.2 0.1 0.0 1

2

3

4 5 6 7 8 Equivalent income decile FINLAND

0.7 0.6

EMTRs

0.5 0.4 0.3 0.2 0.1 0.0 1

Fig. 3.

2

3

4 5 6 7 8 9 10 Equivalent income decile Post reform Pre reform

Pre- and Post-Reform Distribution of EMTRs by Equivalent Income Decile.

184


The U-shaped distribution we find in France and Finland is typical in several EU countries. In the upper part of the distribution, these rates are explained by the progressiveness of the income tax schedule, while they are due to the means-testing of social assistance at the bottom. In France, the overall level of taxation is lower so that EMTRs are lower than in Finland, except in the first decile (which is hardly concerned by income taxes). The Finnish curve is fairly smooth while the anomalies in the French curve are mainly due to thresholds in means-tested transfers to families and to income tax rebates. In Germany, the aforementioned disregard of labor income for social assistance assessment is very small but sufficient to explains lower EMTR for the first decile which contains most of the inactive households,16 while ceiling on SSC explain the decreasing profile of EMTRs in the top of the distribution.17 In Germany and Finland, high EMTRs in lower deciles 2–4 are due to the means-testing of generous transfers observed on budget curves above. This not the case in France where housing benefits present lower withdrawal rates and assure a relatively larger part of the transfers compared to the other countries. This feature, together with a lower level of income taxation, mainly explains the difference with Finland and Germany. It is important to recall that the EMTRs computed in this paper account for SSC in addition to direct taxes and transfers.18 This explains the lower general level of EMTR curves compared to Bourguignon (1997) and turns out to be important in the German case.19 A priori, with the tax credit, average tax rates are decreased at the expense of an increase in EMTRs over the whole range of the benefit. EMTRs should decrease only for those households close to the eligibility threshold, and increase for all households in the phase-out range as the amount of the transfer decreases with increments in gross income.20 Numerical simulations confirm this intuition: EMTRs increase substantially for deciles 3–5, given the large phasing-out of the transfer. Despite the fact that the amount of the benefit is the same in the three countries, the rise is effectively more important in France where taxation is relatively lower. Finland starts out from a situation of high marginal income tax rates so that a smaller part of the gross increase will be taxed away in the phase-out region. The same applies to Germany, although the increase in EMTRs starts only at the third decile (the second in Finland). This is probably related to the greater importance of (flat rate) SSC rather than of progressive income taxation in Germany, so that low earnings are relatively more taxed than in Finland (hence EMTR rise relatively less at the bottom).


185

Two reasons explain why EMTRs decrease in the first two deciles in Germany. First, the 16-hours threshold for eligibility may induce negative EMTRs in the cases where the increment makes the household eligible for the tax credit; this is more often the case in Germany due to a larger proportion of households below the 16-hours threshold. Second, the tax credit crowds out social assistance for low-income household in activity; this way, the implicitly taxation of additional gross earnings switches from 100% (withdrawal rates of minimum income schemes) to dz=dyt: In all countries, the reform mostly targets the first-half of the distribution so that EMTRs of deciles 6–10 hardly change. Table 5 shows the distribution of households, according to the range of EMTRs they face. The share of households facing high EMTRs (over 60%) is considerable in the case of Germany and Finland, although the highest concentration of households is found in the bracket 40–50%. In France, on the other hand, the greatest majority of the household face EMTRs in the 30–40% bracket. Introducing the reform, however, considerably changes the distribution: the share of household facing high marginal tax rates increases by a few percentage points in Finland and in Germany (+7.2 and +8.1 respectively) and by over 12 in France. Interestingly, the most significant increase is in the share of household facing extremely high EMTRs. The difference is due to the fact that in the WTC phase-out range, the withdrawal of the working tax credit cumulates Distribution of Households by EMTR Bracket.

Table 5. EMTR

o0 in [0.0; 0.1] in [0.1; 0.2] in [0.2; 0.3] in [0.3; 0.4] in [0.4; 0.5] in [0.5; 0.6] in [0.6; 0.7] in [0.7; 0.8] > 0.8

France

Germany

Finland

PreReform (%)

PostReform (%)

PreReform (%)

PostReform (%)

PreReform (%)

PostReform (%)

0.1 0.5 3.2 21.3 53.0 12.9 3.6 1.0 0.5 3.8

0.2 0.5 3.1 20.5 44.1 10.5 3.1 3.1 3.6 11.3

0.6 4.3 5.5 2.5 10.9 17.2 50.0 3.1 1.6 4.3

0.8 4.3 5.4 2.2 10.7 15.6 44.0 3.2 6.8 7.1

0.0 0.5 0.2 2.6 4.9 32.9 45.6 7.6 1.5 4.2

0.0 0.5 0.2 2.6 4.6 28.5 43.1 7.1 6.3 7.1


186


with the withdrawal of the housing benefit in France, but not in the other countries. These results are not comparable with the effects found in the UK, since one of the objectives of the WFTC was to partly reduce the disincentive effect embedded in the high taper rate of the Family Credit. Following the decrease in the taper rate the number of households with an EMTR above 70% decrease by around 450,000 while households with an EMTR above 50% increase by about the same amount, since the lower taper rate extended the phasing out of the benefit to higher ranges in the income distribution.

5.2. Financial Gains to Work Financial gains to work when moving from inactivity to full time or part time employment provide further insights on the interaction of the reform with each country’s tax and benefit system. Differently from the analysis of the EMTRs, the analysis of the financial gains to work helps understanding the reshaping of financial incentives to work at the extensive margin. To compute financial gains to work we simulate the disposable income when women in working age work 0, 20 or 40 hours. The financial gain to take up work is then just the ratio of disposable income when working part time or full time to income out of work. Gross income when working part-time or full time have been imputed assuming constancy of the current wage rate for women in employment. In the case of inactive women, the wage rate was imputed using the usual Heckman technique. Table 6 shows the regression coefficients of female hourly wage (in logs) using a two-step Heckman selection model. The coefficients all have the expected signs, both in the wage equation and in the selection equation.21 Tables 7 and 8 show the result for single and married women respectively, under the prereform and post-reform scenario. For women in couples, the gains of working full time are on a similar average level for Germany and France and slightly higher for Finland (principally due to the individualized tax system), but the share of low gains is higher in France and in Germany, than in Finland. On the other hand, the share of single women with low financial gains to work is incredibly high in Germany (owing to generous assistance toward single mothers), and to a less extent in Finland, whereas the concentration is much lower in France. The tax credit transfers lead to a significant increase in financial gains to work for single women. This is particularly striking as regards the gain of

France

Germany

Finland

Coefficient

Standard Error

Coefficient

Standard Error

Coefficient

Standard Error

Hourly wage (ln) Secondary education Tertiary education More than tertiary education Age Age2 Age3 Number of children Constant

0.2068 0.4538 0.6627 0.0616 0.0007 1.0E-06 0.0211 2.5458

0.0173 0.0248 0.0226 0.0315 0.0008 5.9E-06 0.0080 0.4246

0.0049 0.1504 0.4772 0.1927 0.0045 3.5E-05 0.0255 0.3828

0.0323 0.0389 0.0477 0.0761 0.0018 1.4E-05 0.0201 1.0190

0.0658 0.2208 0.4940 0.1965 0.0043 3.2E-05 0.0182 1.0617

0.0659 0.2261 0.5013 0.1974 0.0043 3.1E-05 0.0198 1.0136

Employment status Married Number of children Child o2 Child between 3 and 5 Age Age2 Secondary education Tertiary education More than tertiary education Other income (ln) Constant

0.4147 0.2858 0.3262 0.1868 0.2050 0.0029 0.4785 1.2889 1.1882 0.1807 0.2457

0.0637 0.0212 0.0643 0.0545 0.0209 0.0002 0.0495 0.0787 0.0665 0.0228 0.4935

0.0403 0.7533 0.7553 0.3062 0.1429 0.0020 0.5254 0.6971 0.6537 0.1547 0.4769

0.1835 0.0840 0.0800 0.0308 0.0259 0.0003 0.0576 0.0750 0.1052 0.0137 0.5310

0.0274 0.9180 0.0465 0.0609 0.2795 0.0034 0.0365 0.5150 0.7075 0.2413 3.4901

0.0612 0.0669 0.0582 0.0181 0.0147 0.0002 0.0438 0.0510 0.0574 0.0121 0.3425

0.0371

0.5543

0.0579

0.1073

Lambda Number of obs. Censored obs. Uncensored obs. Wald (w2, 11) Prob>w2

0.1166 5210 1625 3585 2602.37 0

3732 1013 2719 459.15 0

0.0201 7925 4798 3127 1396.75 0

187

Bold figures indicate significance at the 1% level and italic figures refer to the 5% level. Source: Authors’ computations using EUROMOD.


Table 6. Heckman Two-Step Wage Regression.

188

Table 7.


Impact of the Reform on Financial Gains to Work (Singles).

Percentage increase in disposable income when working full time

o20% in [20%; 40%[ in [40%; 60%[ in [60%; 80%[ in [80%; 100%[ in [100%; 120%[ in [120%; 140%[ in [140%; 160%[ >160% Average gain to work full time Average gain to work part time

France

Germany

Finland

PreReform (%)

PostReform (%)

PreReform (%)

PostReform (%)

PreReform (%)

PostReform (%)

1.1 6.5 11.1 15.8 14.0 10.5 7.7 6.6 26.7

0.5 1.7 3.3 9.2 23.6 19.6 8.3 6.6 27.3

12.8 10.4 9.1 11.0 8.6 6.6 7.3 7.3 26.9

4.9 8.6 12.1 11.5 10.4 9.5 6.4 7.5 29.1

4.4 8.9 9.4 9.4 9.6 7.9 7.7 8.4 34.3

1.7 4.6 11.8 8.4 10.3 10.6 7.2 8.9 36.5

132.4 47.0

143.9 78.7

143.0 67.2

153.2 90.6

134.9 56.2

142.2 81.0

Gains are computed as percentage increase in disposable income when the female head is working part time or full time (20 and 40 hours week respectively) with respect to not working at all. Hourly wage is kept constant for population in employment or imputed using regression techniques (see text for details). Source: Authors’ computations using EUROMOD.

working part time, which rises from 47% to 79% (respectively 67% to 91% and 56% to 81%) in France (respectively Germany and Finland) after implementation of the reform. The average gain of working full time increases by 7 percentage points in Finland and by more than 10 points in Germany and France. Similarly, the proportion of very low gains (less than 40%) decreases substantially when the tax credit is introduced, and this for all three countries. For married women the picture is completely different. First, it is noticeable and expected that the gain is much smaller than for single individuals; this is especially the case in France and Germany where the earnings of the second-earner are taxed away at the marginal tax rate of the first earner, as a consequence of the joint income taxation system. The same holds for the tax credit for which income is jointly assessed at the household level. As a result, additional earnings by wives may lead to a loss in the tax credit entitlement for their working partners. The gain of working full time thus shifts from 58% to 54% (respectively 58% to 53% and 65% to 61%) in France (respectively Germany and Finland). Besides average figures, the distribution in Table 8 reveals that the proportion of small gains associated to working


Table 8.

Impact of the Reform on Financial Gains to Work (Couples).

Percentage increase in disposable income when working full time

o10% in [10%; in [20%; in [30%; in [40%; in [50%; in [60%; in [70%; >80%

189

20%[ 30%[ 40%[ 50%[ 60%[ 70%[ 80%[

Average gain to work full time Average gain to work part time

France

Germany

Finland

PreReform (%)

PostReform (%)

PreReform (%)

PostReform (%)

PreReform (%)

PostReform (%)

0.0 2.2 7.3 15.1 18.3 17.8 13.8 9.1 16.4

0.7 5.9 13.4 17.9 18.0 14.8 12.1 6.4 10.8

0.5 5.4 11.4 14.6 17.0 14.6 12.4 7.6 16.5

2.6 7.9 12.7 14.6 17.7 12.9 10.8 7.0 13.9

2.5 1.7 3.5 7.8 13.9 15.4 16.3 13.6 25.4

0.9 3.6 6.3 10.0 15.1 16.6 14.0 12.4 21.1

57.6

50.4

57.7

53.4

65.3

61.3

29.7

24.5

31.1

28.6

37.0

34.6

Gains are computed as percentage increase in disposable income when the female spouse is working part time or full time (20 and 40 hours week respectively) with respect to not working at all. Hourly wage is kept constant for population in employment or imputed using regression techniques (see text for details). Source: Authors’ computations using EUROMOD.

full time (less than 30%) will increase drastically in France (from 9.5% to 20% of the selected women) and more moderately in Germany and Finland (from 17% to 23% and 7.7% to 10.8% respectively). 5.3. Labor Supply Elasticities The increase in the EMTRs might indeed have a negative impact on labor supply at the intensive margin, however, many workers are indeed constrained by demand side rigidities and may not adjust their labor supply in a fully flexible way. The incentives to take up work increase significantly for singles and single mothers, but are equally substantially reduced for women in couples. How household members would adjust to such a modification in the budget constraint is however difficult to anticipate. A reliable analysis of potential behavioral reactions relies on the estimation of unobserved structural parameters which relate to preferences and in

190

Table 9.

Labor Supply Elasticity: A Survey. Females in Couple

France Bargain (2004) Chone´ et al. (2003) Bargain and Orsini (2004)

Haan and Steiner (2005b) Bargain and Orsini (2004) Finland Ilmakunnas (1992) Kuismanen (1997) Bargain and Orsini (2004)

Sample Selection

Wage Elasticity

Sample Selection

Married and cohabitating females (either inactive or employed) Couples with one child under the age of 7 Married and cohabitating females (either inactive or in employment) with partners in employment

0.30

–

Married and cohabitating females in working age Married and cohabitating females in working age Married and cohabitating females in working age Married and cohabitating females (either inactive or in employment) with partners in employment Married and cohabitating females in working age Married and cohabitating females in working age Married and cohabitating females (either inactive or in employment) with partners in employment

0.30 0.60

– Inactive or employed single women

0.27

–

0.32

Single females in working age Single females in working age Inactive or employed single women

0.39 0.40

0.29

–

0.10

–

0.15

Inactive or employed single women

Wage Elasticity

0.12

0.13 0.13 0.16

0.28


Germany Bonin Kempe and Schneider (2002) Haan (2006)

Single Females


191

fixed costs of labor supply. These aspects are usually summarized in indicators of labor supply responsiveness (i.e. labor supply elasticities). A fully structural labor supply model would allow to identify preference parameters and fixed cost of labor supply, which would then be used to anticipate optimal behavior in the post-reform scenario. Bargain and Orsini (2004) follow Blundell et al. (2000) and evaluate the second-round effect of difference in work policies using a structural model of labor supply. Such procedure, however, relies heavily on the specified functional form and on the assumed distribution of disturbances. The ex-ante analysis by Blundell et al. (2000), for example, correctly evaluated the direction of potential responses to the WFTC, but significantly underestimated the effect of the reform, when compared with the ex-post analysis of Brewer and Shephard (2004), who rely on a ‘difference in difference’ estimation methodology. Table 9 summarizes some recent estimates of female labor supply elasticities. The main conclusion is that overall labor supply elasticities are indeed quite small in continental Europe and in Finland; the range of variation of the estimate, however, is rather high. Some broad conclusions may nevertheless be drawn: consistently with economic theory, wage elasticities of single women are significantly lower than elasticities of married or cohabiting women. The significant exception of Finland, reported by Bargain and Orsini (2004) should deserve further analysis. In Finland, moreover, elasticities seem to be significantly lower than in Germany and in France, whereas the differences in the size of labor supply of France and Germany are not consistent across studies. It is interesting to remark that behavioral reactions are likely to be stronger among the population negatively affected by the reform, and weakest among singles and single mothers.

6. CONCLUDING REMARKS In this study, we have compared ‘framework conditions’ relevant for introducing a pseudo WFTC in three EU countries – Finland, France and Germany – afflicted by potential inactivity traps. Such ‘framework conditions’ not only help understanding the differences in the first-round effects of introducing a refundable tax credit, but may also be helpful in speculating over potential second-round effects. The analysis suggests that while firstround redistributive effect might indeed be significant, the impact on overall employment rate could be reasonably expected to be very small, or even negative due to different composition effect.22 In this case, the relatively

192


high efficiency cost of lifting a household over the poverty line might become even larger, whereas the objective of increasing employment rates would clearly be out of reach. On the other hand, if the objective is not an increase in the employment rate per se, but an improvement for most disadvantaged groups, like indeed lone mothers, the tax credit could be a desirable policy for Germany, but indeed not for Finland and even less so for France, which shows already high employment rates for single mothers. Increasing the employment rate of singles and lone mothers – even at the expense of a decrease in the employment rate of women in couple, might in turn bring to positive externalities as described by Phelps (2000).23 Should the rising trend in lone motherhood continue in the near future and approach the British figures, moreover, the evaluation on the need of such policy could change dramatically – especially in Germany where inactivity traps for lone mothers are significantly higher. An alternative approach, could then be to design a policy measure which only benefit lone mothers as is in the case in the Canadian experimental Self Sufficiency Program (SSP).24 Another alternative to avoid the potential negative impacts on women in couples would be to completely individualize the credit. The debate on the unit of assessment for such in work benefit is extremely up to date, and a second generation of MWP policies seem to depart more and more from the original Anglo-Saxon scheme. This is partially the case in France, where inwork benefits are simultaneously conditioned on households means being under a rather generous income threshold and individual wage rate being smaller than 1.4 times the minimum wage.25 In the Netherlands, a recent tax reform has introduced a tax credit on earned income that not only is purely individualized, but is also not tapered away as income increases. The potential high cost of such reform, however, has pushed policymakers to prefer a nonrefundable tax credit, which reduces considerably the scope for positive labor supply reactions.26 In the Belgian tax reform, finally, the tax credit is also totally individualized, but differently from the Dutch case it is phased in and phased out. The small amount of the credit is however not likely to induce to a massive participation effect, but it is the idea of the government to substantially increase the level of the tax credit in the future.27 Bargain and Orsini (2004) use a structural labor supply model to assess the second-round effects the tax credit described in this paper and of an individualized measure having the same budgetary cost in the same three countries. They conclude that, although only the individualized measure may be realistically expected to achieve an increase in employment, the efficiency cost with respect to both the employment and poverty reduction objectives remain extremely high. In fact, while the lack of targeting of an


193

individualized credit avoids the disincentive effect on second earners, the dispersion of the benefit over low skilled workers living in middle-class and well-off households implies significantly lower amounts. It therefore seems that, despite the great attention triggered by the reform of the Family Credit in the UK, the current interest in ‘MWP’ policies is destined – in the medium run – to fade away. At least in Continental Europe and in Scandinavia the attention should be shifted toward demand-side policies (given the level of structural involuntary unemployment) and active labor market policies that aim at permanently increasing the productivity of less skilled workers. A short-sighted alternative strategy, triggered by the current budgetary difficulties in France and in Germany, is to substantially reduce entitlement to social assistance, by requiring some form of participation in the labor market. The debate over the Revenue Minimum d’Activite´ in France or the Harz IV reform in Germany point indeed in this direction. Rather than ‘MWP’, however, these reforms will most probably further inflate the number of ‘working poor’.

NOTES 1. The importance of these initial conditions has been stressed in previous studies, but the analysis has not been exploited in large-scale analysis using microsimulation models and even less so within multi-country comparative framework. See Pearson and Scarpetta (2000) and Bertola (2000). 2. This difficulty is illustrated by the recent changes in the UK tax system where the WFTC has indeed been split into two distinct policies serving different objectives: a working tax credit to create incentives to work and subsidize working poor and a child tax credit to redistribute toward families and children (Brewer, 2003). 3. To this extent British amounts have just been deflated to 1998 and multiplied by the exchange rate (0:679 £/EUR). Since the construction of an integrated microsimulation model such as EUROMOD is extremely labor intensive, policy rules are not immediately updated for all countries and are hence available only for some reference year (e.g. 1998, 2001 and in the near future 2003). Moreover even if more updated policy rules were available, in order not to draw biased conclusions, one should use tax and benefit rules that are not too distant in time from the year in which the micro-data was collected. Indeed some important reforms were introduced in France between 1998 and 2001, and Germany too has implemented a significant tax reform and a change in the unemployment insurance, that are likely to modify the labor supply of households. The use of an older system implies a more consistent estimation of the effects of the reform. The main drawback is of course that the analysis is not fully up to date with respect to changes in the tax and benefit system and the current economic situation. 4. Like in the UK, the credit does not enter the means test for child benefits in France, while no such means-tested child allowance exist in the other countries.

194


5. In Finland, the amount of housing benefit enters the assessment of social assistance eligibility (and hence of the tax credit), while in Germany the interaction is reversed since social assistance determines the type and level of housing benefits. In France the interaction between social assistance and housing benefit is even more complex: if a household is theoretically eligible to a positive level of social assistance, labor income or replacement incomes (unemployment benefits) are not accounted for in the income assessment for the computation of housing benefits, but the level of the housing benefit then determines the actual amount of the social assistance. 6. Stephens, Burns and MacKay (2002) critically assess the different instruments of the British housing policy with respect to the experience of other EU countries, including Germany, France and Finland. The authors stress the different ideology underlying housing subsidy in the UK with respect to most other countries. 7. It should be noted that our reform does not correspond to the reform implemented in the UK in 2003. In the new British system, the child premia for couples is universalized in a new instrument (the CTC) available to both active and inactive households. Simulating the new system would have implied designing an additional harmonization of family benefits and social assistance scheme in all three countries. Indeed, the basic idea of the WFTC is to recreate significant financial difference between social assistance and paid work. Since the former is scaled on family size in France, Germany and Finland, the tax credit must also include a child dimension in order to tailor the financial gains from taking up work to the income level guaranteed by minimum schemes. Moreover the new British system simultaneously increase income out of work (through the refundable child tax credit) and in employment (through increased working tax credit). In this sense, distributional (towards households with children) and efficiency concerns are clearly addressed by two separate instruments, whereas it is the scope of this paper to assess weather a single measure may attain both objectives simultaneously. 8. An additional assumption was made concerning the cost of housing. While this information is reported in the survey data, in draw the budget set of typical households, we assumed housing cost in the order of 500 EUR/month. 9. See in-depth descriptions of the tax-benefit systems in Bargain and Terraz (2001) for France, Grabka (2001) for Germany and Vittama¨ki (2001) for Finland. 10. The level of ‘safety net’ is very similar in all three countries, due to the fact that housing benefits play a larger role in France (3,000 EUR for an inactive person, i.e. 44% of her disposable income) and Finland (2,000 EUR, i.e. a third of her disposable income) than in Germany (1,200 EUR, 20% of her disposable income). 11. The distribution of the population in the benefit range determines the first round distributional impact of the reform. The fact that the share of the population in the benefit range is of comparable size and distribution implies comparable aggregate costs. However, the share of non-eligible units and the share of units in the phaseout range or out of the benefit range provide intuitions on the size of potential behavioral responses. Households in the phase out range and to a lesser extent households whose income is higher than the benefit range may in fact reduce their labor supply, whereas non-eligible households might increase it to enter the benefit range. The scope for positive reactions is clearly smallest in France and in Finland, and largest in Germany.


195

12. Other dynamic considerations may explain such behavior as well, namely the fact that working brings entitlement to pension rights or career expectations dependent on past employment records. 13. Mantovani and Sutherland (2001) compare poverty rates obtained using EUROMOD and national official figures. Figures are quite in-line for France (the 1997 Fiscal Data give 6.9% with a poverty line at 50%) and Finland (the 1999 Income Distribution Survey gives 4%). Differences are somewhat larger in Germany (the 1999 GSOEP gives 8.2%), which is due to assuming 100% take-up of welfare assistance (Sozialhilfe). Notwithstanding, we are primarily concerned here with relative moves in and out of poverty under both reforms than with the absolute level of head count ratios. 14. In principle, EMTRs could be computed analytically as one minus the first derivative of the budget constraint. However, the complexity of the tax-benefit system forces us to rely on a numerical approximation. We therefore rely on EUROMOD to simulate the increase in disposable income (dC) following an increase in gross employment income (dy) of household heads (defined as the main earner in the household). The formula is simply: EMTR ¼ 1

dC dy

We choose a uniform gross income increment dy ¼ 1,500 EUR per year for all three countries. Note also that the increment is larger than what one may think of as ‘marginal’. Yet, this choice corresponds to an additional productive effort that can be seen as more realistic than an additional euro of income. Indeed, it corresponds to around 5 additional hours per week for a worker paid at the minimum wage. In particular, the choice to link the ‘marginal’ increase in income to an increase in working hours allows to capture the negative effect of the tax-benefit system on an increase in working effort, rather than an increase in productivity (the wage rate is in fact not a variable under direct control of a worker – at least not in the short run). 15. It is important to note that the distribution of EMTRs is quite heterogenous within each decile, due to the complexity of tax-benefit systems and the fact that deciles are computed on income per unit of consumption. To illustrate this point, notice for instance that the EMTR of the first decile does not reach 100%, which would be the case if it was composed only by inactive households. 16. This point depends crucially on the EMTR definition. In our computations, the 1,500 EUR yearly increment implies that inactive households of the first decile are assumed to switch from 0 to 125 EUR per month of earnings, only 55 EUR of which will be considered in the income assessment due to the 70 EUR disregard. In that case, the benefit withdrawal rate is 30%, so that EMTR will be only 13.2%. Additional disregards for workers with children may bring the withdrawal rate down. This effects naturally explain the relatively lower mean EMTR in the first decile in Germany. A ‘truly marginal’ increment would imply full disregard hence 0 EMTRs for inactive households. 17. The funding of the German social security system is characterized by contribution assessment limits, i.e. maximum amounts from which the contributions to

196


the state insurances fund are calculated, even if an employee earns more. In 1998, the contribution assessment limit was around 4,000 EUR per month (the assessment limit varies in accordance to whether the region is in East or West Germany and to the risk ensured). 18. The choice to treat these contributions as taxes is highly debatable but can be justified to some extent. We do not go into details here and refer to Bourguignon (1997) for a more in-depth discussion. 19. In addition, some instruments have changed since the reference year in Bourguignon (1997), which explain that the mean EMTR has decreased for the first decile. In particular, the additional disregard for social assistance has been brought up from 15% to 30% of income higher than 70 EUR. 20. First, the EMTR increases by the level of the taper rate t ¼ 55%, corrected by the fact that this withdrawal rate applies to the income concept z (labor income net of tax and social security contributions plus some benefits) rather than to gross earnings y. 21. As stressed by Ermisch and Wright (1994), negative selection bias is consistent with both positive and negative inverse mill ratio terms. In all countries average predicted wage using Heckman selection correction was significantly higher, then when using the coefficients of a standard regression model. 22. Naturally, the poverty criterion is only one among several distributional aspects, and the evaluation of the policy changes significantly according to the definition of the poverty line. In this respect, it is evident that the working tax credit achieves an important transfer to the first-half of the income distribution but concentrates on the ‘working poor’, thus neglecting the households in the very bottom of the income distribution. 23. Phelps argues that there are potentially important social and economic externalities associated with entering the labor market (he insists on the fact to hold a full-time job): ‘‘Bringing marginalized groups, including those who work in the underground economy, into mainstream economic activities may generate beneficial outcomes for society as a whole, for example through the amelioration of problems like crime, social destitution, drug etc.’’ 24. See Foley et al. (2002) for a description of the program and an evaluation of its effects. 25. See Bargain and Terraz (2003) for a description of the reform and an evaluation of its potential labor supply effects. 26. For a detailed description of the reform and an analysis on the potential labor supply effects, see Van Soest and Das (2000). 27. Orsini (2005) analyzes the Belgian tax reform and models its impact on female labor supply. The author finds that the effects are moderate, but they are mostly coming from other aspects of the reform, and not from the credit itself.

ACKNOWLEDGMENTS We are grateful to Tony Atkinson, Denis Beninger, Franc- ois Bourguignon, Tim Callan, Andre´ Decoster, Marc Gurgand, Seija and Pekka Ilmakunnas,


197

Herwig Immervoll, Thomas Picketty, Cathal O’Donoghues, Holly Sutherland, Heikki Vittama¨ki and Sile O’Dorchai for useful comments and advices. All errors or omissions remain ours. This chapter was written as part of the Micro Analysis of the European Social Agenda (MICRESA) project, financed by the Improving Human Potential programme of the European Commission (SERD-2001–00099). For the countries considered in this paper, EUROMOD relies on the following micro-data: the Income Distribution Survey made available by Statistics Finland; the Household Budget Survey made available by INSEE and the German Socio-Economic Panel Study made available by DIW.

REFERENCES Banks, J, Disney, R, Duncan, A, & van Reenen, J. (2003). The internationalisation of public welfare policy. The Economic Journal, 115, 62–81. Bargain, O. (2004). Aides au retour a` l’emploi et a` l’activite´ des femmes en couple. Revue de l’OFCE, 88, 59–87. Bargain, O., & Orsini, K. (2004). In-work policies in Europe: Killing two birds with one stone? IZA Discussion Paper Series 1445, forthcoming in Labour Economics. Bargain, O., & Terraz, I. (2001). EUROMOD country report for France. Euromod Working Paper. Bargain, O., & Terraz, I. (2003). Evaluation et mise en perspective des effets incitatifs et redistributifs de la Prime pour l’Emploi. Economie et Pre´vision, 160/161, 124–149. Bertola, G. (2000). Policy choices and interactions with existing instruments. Economic Studies, 31, 185–196 OECD. Blundell, R., Duncan, A., McCrae, J., & Meghir, C. (2000). The labour market impact of the working families tax credit. Fiscal Studies, 21(1), 75–104. Bonin, H., Kempe, W., & Schneider, H. (2002). Household labor supply effects of low-wage subsidies in Germany. IZA Discussion Paper Series 637. Bourguignon, F. (1997). Fiscalite´ et redistribution. Report, Conseil d’Analyse Economique. Bourguignon, F., & Bureau, D. (1999). L’architecture des pre´leèvements en France: e´tat des lieux et voies de re´forme. Report, Conseil d’Analyse Economique. Brewer, M. (2003). The new tax credits. Briefing note 35, IFS. Brewer, M., Duncan, A., Shephard, A., & Sua´rez, M. J. (2003). Did working family tax credit work? Analysing the impact of in-work support on labour supply and programme participation. Working Paper 2, Inland Revenue. Brewer, M., & Shephard, A. (2004). Has labour made work pay? Discussion Paper, Joseph Rowntree Foundation. Bundesministerium fu¨r Wirtschafts und Technologie (2002). Reform des Sozialstaats fu¨r mehr Bescheftigung im Berich gering qualifizierter Arbeit. Report, Bundesministerium fu¨r Wirtschafts und Technologie. Chone´, P., le Blanc, D., & Robert Bobeé, I. (2003). Female labor supply and children. CESifo Working Paper 1059. Eissa, N., & Hoynes, H. (2004). Taxes and the labor market participation of married couples: The earned income tax credit. Journal of Public Economics, 88, 1931–1958.

198


Ermisch, J. F., & Wright, R. E. (1994). Interpretation of negative sample selection effect in wage offer equations. Applied Economic Letters, 1, 187–189. Ferry, J. P. (2000). Plein Emploi. Report, Conseil d’Analyse Economique. Foley, K., Ford, R., Gyarmati, D., Michalopoulos, C., Miller, C., Morris, P., Redcross, C., Robins, P. K., & Tattrie, D. (2002). Making work pay. Final report on the self-sufficiency project for long-term welfare recipients. Discussion Paper, SRDC, Ottawa. Grabka, M. (2001). EUROMOD country report for Germany. Euromod Working Paper. Haan, P. (2006). Much ado about nothing: Conditional logit vs. random coefficient models for estimating labour supply elasticities. Applied Economics Letters, 13, 251–256. Haan, P., & Steiner, V. (2005a). Distributional effects of the German tax reform 2000: A behavioral microsimulation analysis. Journal of Applied Social Science Studies, 125, 39–49. Haan, P. and Steiner, W. (2005b). Labor market effects of the German tax reform 2000. DIW Discussion Paper 472. Ilmakunnas, S., Kiander, J., Parkkinen, P., & Romppanen, A. (2000). Globalisation: The end of work. The economy and employment in Finland to 2030. Report, Government Institute for Economic Research. Ilmakunnas, S. (1992), Income taxation, hours restrictions and labour supply. Labour Institute for Economic Research, 113. Kuismanen, M. (1997). Labour supply, unemployment and income taxation: An empirical application. VATT Discussion Paper 145. Laine, V. (2002). Changing poverty post-1997. Working Paper 63, Centre for Analysis of Social Exclusion. Mantovani, D., & Sutherland, H. (2001). Social indicators and other income statistics using the Euromod baseline: A comparison with Eurostat and national statistics. Euromod Working Paper, em1/03. Martin, J., & Grubb, D. (2001). What works and for whom: A review of OECD countries’ experience of active labour market policies. Working Paper, OECD. Orsini, K. (2005). The 2001 Belgian tax reform: Equity and efficiency. Ces Discussion Paper no. 05–04. Pearson, M. (2002). International experience of ‘making work pay’ policies. Working Paper, OECD. Pearson, M., & Scarpetta, S. (2000). An overview: What do we know about policies to make work pay? OECD Economic Studies, no. 31, 2000/ii. Phelps, E. (2000). The importance of inclusion and the power of job subsidies to increase it. OECD Economic Studies, 31, 85–114. Pichaud, D., & Sutherland, H. (2002). Evaluating Tax and Benefit Reforms in 1996–2001. Discussion paper 280, VATT. Stephens, M., Burns, N., & MacKay, L. (2002). Social market or safety net? British social rented housing in a European context. Bristol: The Policy Press. Van Soest, A., & Das, M. (2000). Family labor supply and proposed tax reforms in the Netherlands. De Economist, 149, 191–218. Vittama¨ki, H. (2001). EUROMOD country report for Finland. Euromod Working Paper.

IMPLICIT SOCIAL PREFERENCES AND FISCAL REFORMS: A MICROSIMULATION ANALYSIS FOR SPAIN Amedeo Spadaro ABSTRACT In this paper, I support the usefulness of using microsimulation models for the normative analysis of real redistribution system. Drawing from three recent works (Bourguignon & Spadaro, 2000, 2005; Oliver & Spadaro, 2004), I propose an application consisting in analyzing how social preferences on inequality have changed since the introduction of the 1999 reforms to the Spanish personal income tax (PIT). The starting point is the observed distribution of a population’s gross and disposable incomes and the observed marginal tax rates as computed in standard microsimulation models. I show that, using a set of simplifying assumptions, it is possible to identify the social welfare function that would make the observed marginal tax rate schedule optimal. I apply this methodology to the 1998 and 1999 Spanish PIT, using the Eurostat (ECHP) dataset on the income and socio-demographic characteristics of Spanish households.


199

200

AMEDEO SPADARO

1. INTRODUCTION Microsimulation models are powerful tools to evaluate the redistributive and incentive impacts of public policies. With the development of the computational capacities of PCs and the increasing availability of large data sets, it has become common practice in many countries. Bourguignon and Spadaro (2006) provide an overview of what can be achieved in several branches of the literature. In Spain, microsimulation studies are becoming the references for academic and political debates about the future of the redistribution system (see Castan˜er, Onrubia, & Paredes, 2000; Levy & Mercader-Prats, 2002; Sanchis & Sanchis, 2001, among others). One important aspect is the usefulness of microsimulation analyses for normative evaluations of public policies. In particular, a new branch of the empirical public economics literature has emerged with the operationalization of optimal tax theory (Mirrlees, 1971) based on microdata and taxbenefit microsimulation. In particular, if the optimal redistribution problem is simple (e.g. a linear tax system or a piecewise linear tax system with a small number of tax brackets) and can be solved numerically, it is possible to derive optimal tax rates from the specification of a social welfare function and for different levels of social inequality aversion. Comparing the distribution of optimal tax rates to actual ones may reveal the range of social preferences (see Bourguignon & Spadaro, 2000; Spadaro, 2005). Alternatively, it is possible to assume that the actual marginal tax rates characterize the true preferences of the social planner. Provided that the optimal taxation problem can be inverted, it is then possible to recover the implicit social welfare function embedded in the actual tax-benefit system. This is the methodology suggested by Bourguignon and Spadaro (2005). In this paper, I apply this methodology in order to retrieve the social preferences in Spain for the years 1998 and 1999. I question whether the 1999 changes in the Spanish redistribution system reveal a change in social inequality aversion. For this purpose, I use the European Community Household Panel (ECHP) dataset for Spain together with the microsimulation model EUROMOD. The structure of the paper is as follows. Section 2 deals with the theoretical model and its empirical implementation. Section 3 describes the dataset and important features of the Spanish tax system for the years 1998 and 1999. Section 4 presents the empirical results and Section 5 concludes.

Implicit Social Preferences and Fiscal Reforms

201

2. THE MODEL AND ITS IMPLEMENTATION This section largely draws on the methodology developed in Bourguignon and Spadaro (2005). In the original optimal taxation framework proposed by Mirrlees (1971), agents choose the consumption y-labor L combination that maximizes their welfare, given the budget constraint which is affected by the tax-benefit function T. This can be expressed as follows: Max Uðy; LÞ

(1)

s:t:y ¼ wL TðwLÞ

(2)

y;L

where w is the productivity of the agent and U the well-behaved utility function. If f(w) is the density distribution of individual productivities defined over the interval [w0, A], the government’s optimal taxation problem is written as: Z A Max G V ½w; TðÞ f ðwÞdw (3) TðÞ

w0

s:t: V ½w; TðÞ ¼ Uðy ; L Þ

(4)

ðy ; L Þ ¼ arg max½Uðy; LÞ; y ¼ wL TðÞ

(5)

Z

A

TðwL Þf ðwÞdw B

(6)

w0

where B is the government’s budget and G{ } the social welfare function that transforms individual indirect utility V, defined in (4), into social welfare. In words: the planner maximizes social welfare under a national budget constraint (6) and under incentive compatibility constraints (4) and (5). Note that expression (5) is an alternative formulation of the agent maximization problem (1) and (2). Further assumptions are required for the model to be tractable (see Tuomala, 1990; Bourguignon & Spadaro, 2005). Additional components are required at this stage. First, it is a standard practice to focus on the special case where the utility function is quasi-linear in income and iso-elastic with respect to labor, that is: 1 1 1þ1 Uðy; LÞ ¼ y 1 þ L

(7)

202

AMEDEO SPADARO

where e is the elasticity of labor supply. This leads to a simple labor supply function: L ¼ w ½1 T 0 ðwL Þ

(8)

Second, it is usually assumed that agents’ productivities are unknown (first best taxation would be achievable otherwise). It is still possible to obtain the agents’ productivity as a function of observables and for a given elasticity by inverting expression (8):

wL w¼ ð1 t Þ

1 þ1

(9)

This requires the local linearization of the budget constraints in order to use the actual effective marginal tax rate t* for each agent.1 We also assume that agents are rational so that observed earnings coincide with optimal choices (wL*). A necessary condition for the inversion of (8) is that disposable income is a strictly increasing function of productivity so that actual labor choices ‘reveal’ the ability level of the agent. This ‘agent monotonicity’ condition is also required for the optimal taxation model. Note that the household dimension is ignored in the model for the sake of simplicity. Further details on the calibration of productivities, on the treatment of family dimension and on the treatment of non-participating households are discussed in Spadaro (2005) and Bourguignon and Spadaro (2000). With the previous framework, the optimal tax schedule is characterized by the following equation: tðwÞ 1 1 F ðwÞ SðwÞ ¼ 1þ 1 (10) 1 tðwÞ wf ðwÞ Sðw0 Þ as proved by Atkinson (1995) and Diamond (1988). In this expression, F(w) is the cumulative distribution function, t(w) the marginal tax rate for an agent with productivity w and S(w) the average marginal social utility of all agents with a productivity higher than w: Z A 1 G 0 ½V ðw; TðÞÞf ðwÞdw (11) SðwÞ ¼ 1 F ðwÞ w We normalize the welfare function G in such a way that the mean marginal social welfare is equal to 1, i.e. S(w0) ¼ 1. The two expressions (10) and (11) characterize the duality between the effective marginal rate of taxation and the social welfare function. The combination of these equations and some algebra lead to an expression of the marginal social welfare reached by


203

a household with productivity w: 1 þ tðwÞ wt0 ðwÞ wf 0 ðwÞ þ þ1 G ½V ðw; TðÞÞ ¼ 1 þ 1 tðwÞ tðwÞð1 tðwÞÞ f ðwÞ 0

(12)

This formula gives us the marginal social welfare weight of an agent characterized by productivity w under a certain redistribution system t(w), a given distribution of productivities f(w) and some hypotheses regarding e. A recent survey by Blundell and MaCurdy (1999) gives a range of estimates of the elasticity of labor supply concentrated in the interval [0,1]. In the following, we shall use a middle value of 0.5 which seems plausible for Spain. Sensitivity analyses around this value show that the results do not change much qualitatively. Note finally that the implementation of formula (12) is far from easy since it requires not only the knowledge of the marginal tax rates and the density of the productivity distribution but also their elasticity with respect to the level of productivity. Moreover, some technical conditions must be respected in order to guarantee the possibility and the uniqueness of the inversion procedure described above. For lack of space, we shall not get into further details here and refer to Bourguignon and Spadaro (2005) for an extensive discussion of the implementation of the model.

3. DATA AND REFORM OF THE SPANISH TAX SYSTEM 3.1. Data and Selection I use the 1996 Spanish database from the ECHP published by EUROSTAT. It contains information at both individual and household levels on sociodemographic characteristics, incomes and labor status for 6,119 households comprehending 18,991 individuals. More information is provided in Levy and Mercader-Prats (2001). The original dataset was updated, using a correction factor including inflation and the growth rate for the period up to 1998 and 1999. Changes in the socio-demographic structure were not taken into account. To retain the logic of the optimal taxation model, all households for which unearned income (including pension and unemployment benefits) represented more than 10% of their total income were eliminated from the sample. In this sense, the analysis is restricted to the main labor force.

204

AMEDEO SPADARO

3.2. The 1998 and 1999 Spanish Tax Systems The microsimulation model EUROMOD is used to simulate several taxbenefit scenarios and in particular the income tax legislation in force in 1998 and 1999.2 The Spanish personal income tax (PIT) system has been introduced in 1979. In 1989, the rules have been changed to include the possibility of separate tax declarations for spouses. The tax system is based on annual income: while income tax is paid during the year – withheld at the income source – on all types of income, an adjustment occurs at the end of the tax year once the exact tax liability is calculated by the treasury department. The tax schedule in force until 1998 is provided in Table 1. The Spanish PIT system has undergone a dramatic change with the reform of 1999. The main features concern the reduction in the number of tax brackets from 9 to 6 (see Table 1) and the cut in the maximum and minimum marginal tax rates, from 56% to 48% and from 20% to 18%, respectively. Another important change concerns the treatment of family dimension: direct tax credit (25,000 ptas for the first child, 35,000 ptas for the second and 50,000 ptas for every additional child) has partly been replaced by tax allowances which decrease the level of taxable income (200,000 ptas for each of the first and second child, 300,000 ptas for each additional child; 100,000 ptas for each dependent relative over 65 years of age). In addition, the reform has followed the German philosophy of a subsistence level of income: only the income that exceeds the basic needs is taxable. The flat 5% deduction of gross income from dependent employment has been replaced by new complex deductions that depend on the level of gross wages. No relevant changes have been made in self-employment income and property income taxes. Fig. 1 shows the distribution of effective marginal tax rates across income levels for the years 1998 and 1999. The main change occurs for higher incomes where the 1999 curve is systematically lower than for 1998, illustrating the important reduction in the progressivity of the system due to the reduction in the number of brackets and the considerable decrease in the highest marginal tax rate. The moderate decrease in the lowest marginal rate displays little effects.3 The reform has been realized after the Popular party came to power in 1996, following 14 years of Socialist government. The new tax law was introduced in 1999 and the political and academic debate on its effects on equity and efficiency is still open. Interestingly, the following results characterize the possible change in social preferences that has occurred following

Tax Rates Schedule (in Euros).

1998 Single Person’s Income Tax Return Bracket 0–2,806.73 2,806.73–6,977.75 6,977.75–13,793.23 13,793.23–21,005.37 21,005.37–30,621.57 30,621.57–40,838.77 40,838.77–51,837.29 51,837.29–63,106.27 >63,106.27

1999 Family Income Tax Return

Single Person’s and Family Income Tax Return

Tax Rate

Bracket

Tax Rate

Bracket

Tax Rate

0 0.2 0.23 0.28 0.32 0.39 0.45 0.52 0.56

0–5,415.12 5,415.12–13,492.72 13,492.72–19,028.04 19,028.04–26,390.44 26,390.44–35,255.37 35,255.37–47,485.97 47,485.97–59,716.56 59,716.56–72,938.83 >72,938.83

0 0.2 0.246 0.29 0.33 0.39 0.45 0.53 0.56

0–3,606.07 3,606.07–12,621.25 12,621.25–24,641.50 24,641.50–39,666.08 39,666.08–66,111.33 >66,111.33

0.18 0.24 0.283 0.372 0.45 0.48


Table 1.

205

206

AMEDEO SPADARO 0.6

Marginal Tax

0.5 0.4 0.3 0.2

KernelTM98

KernelTM99

0.1 0 0

20000

40000

60000

80000

100000

120000

Gross income (in euros) Fig. 1.

Marginal Taxes for Workers, 1998 and 1999 (Kernel Estimation).

the change of government and the implementation of a less redistributive tax system. Recently, the Socialist party has proposed a possible amendment, consisting of replacing the current PIT system with a partial basic income – flat tax scheme (Atkinson, 1995). This reform is analyzed by Oliver and Spadaro (2004) using the present methodology.

4. RESULTS The calibration of household productivities on gross income, as summarized by expression (9) and described extensively in Spadaro (2005), leads to the distribution in Fig. 2, with mean productivity normalized to one. Recall that this distribution is obtained for the quasi-linear form and under the assumption of a moderate elasticity of labor supply (e ¼ 0.5). The marginal social welfare consistent with the previous framework is shown in Fig. 3. The first result is that the marginal social welfare is positive and declines with the level of productivity almost everywhere. This is reassuring since it suggests that the redistribution systems analyzed exhibit some minimum optimality features in the sense that they maximize an increasing and concave social welfare function.4 The main result is that the social welfare weight of the poorest group of the population decreases with the reform, and is more than compensated by an increase in the weight of the richest group. Implicitly, and under the


207

3.5000E-05 3.0000E-05

frequence

2.5000E-05 2.0000E-05 1.5000E-05 1.0000E-05 5.0000E-06 0.0000E+00 0

0.5

1

1.5

2

2.5

3

3.5

4

productivity

Fig. 2.

Kernel Density Estimation of Productivities (Mean of w ¼ 1).

Marginal Social Welfare

1.8

1.3

0.8

0.3

-0.2 0

0.2

0.4

0.6

0.8

1

Productivity Quantiles

Fig. 3. Marginal Social Welfare Functions Whole Sample. Note: The Thick Black Line Corresponds to the 1998 Scenario and the Thick Gray Line Corresponds to the 1999 Scenario.

acceptation of the preceding framework, the 1999 reform can be interpreted as a move toward more utilitarian preferences. The analysis performed on the whole sample does not consider that in reality redistribution systems are concerned not only with income differences but also with other dimensions, e.g. family size. The theoretical model at use cannot explicitly incorporate this dimension. A possible way to take into account the size and composition of the households is to apply the previous methodology to separate household groups with homogeneous

208

AMEDEO SPADARO

demographic compositions. This approach has strong limitations since it assumes implicitly that the redistribution taking place across these groups is exogenous (i.e. independent of productivity and income). Figures 4–7 show the results of the inversion of the marginal tax rates into marginal social welfare for singles, couples without children, with one child


1.4 1.2 1 0.8 0.6 0.4 0

0.2

0.4

0.6

0.8

1


Fig. 4. Marginal Social Welfare Functions Singles. Note: The Thick Black Line Corresponds to the 1998 Scenario and the Thick Gray Line Corresponds to the 1999 Scenario.


2

1.5 1

0.5 0 0

0.2

0.4

0.6

0.8

1


Fig. 5. Marginal Social Welfare Functions Couples. Note: The Thick Black Line Corresponds to the 1998 Scenario and the Thick Gray Line Corresponds to the 1999 Scenario.


209


2

1.5

1

0.5

0 0

0.2

0.4

0.6

0.8

1


Fig. 6. Marginal Social Welfare Functions Couples+1 Child. Note: The Thick Black Line Corresponds to the 1998 Scenario and the Thick Gray Line Corresponds to the 1999 Scenario.


2

1.5

1

0.5

0 0

0.2

0.4

0.6

0.8

1


Fig. 7. Marginal Social Welfare Functions Couples+2 Child. Note: The Thick Black Line Corresponds to the 1998 Scenario and the Thick Gray Line Corresponds to the 1999 Scenario.

and with two children, respectively. The properties of the marginal social welfare curves are comparable to what is obtained for the population as a whole. Importantly, it seems that the reform has a less negative impact on the progressivity among families with children, which implies that the other changes that occurred (notably the replacement of tax credits by tax

210

AMEDEO SPADARO

allowances to account for family size) have played a significant role. More remains to be done to understand the interdependence of the treatment of family size and implicit preferences concerning vertical equity. The specification of a multidimensional optimal tax model explicitly considering family size as a redistribution variable would be necessary to go further in this domain.

5. CONCLUSIONS The main objective of this paper is to illustrate the use of arithmetical microsimulation models enriched by economic theory to perform normative analyses of public policies. Using the approach developed by Bourguignon and Spadaro (2005) and based on Mirrlees’ optimal tax theory, I have derived the social preferences underlying the 1998 and 1999 Spanish taxbenefit systems under simple assumptions on preferences and provided that the actual distribution of marginal tax rates reflects the optimal schedule. The distribution of effective marginal tax rates is computed using EUROMOD. I have observed that, in general, the social welfare function is increasing and concave. Then results do not contradict the fact that taxbenefit legislation could correspond to what is recommended by optimal taxation theory. Of course, this is just one of many possible interpretations, but this allows rationalizing observed tax rates and reinterpreting changes in the tax schedule as a change in social preferences. As a matter of fact, the recent reform of the tax system in Spain, coinciding with a political change after the last election, can be seen as a move toward more utilitarian preferences. Our results clearly show that social weights put on the more productive section of the population have increased, which is coherent with the declared objectives of the reform to reduce disincentive effects of the tax system and improve efficiency. This approach is a first attempt to build a bridge between the use of tax-benefits models for policy evaluation and the more formal normative analysis of public policy.

NOTES 1. The effective marginal tax rate is the local derivative of the budget constraint. This variable is not present in the survey but can be computed easily with numerical simulations: we assume that earnings x increase by a uniform amount – fixed here at 10% of the total population average gross labor income – and use the microsimulation model to compute the new disposable income y, so that tðwÞ ¼ 1 Dy=Dx:


211

2. Social security contributions have not been affected much by the reform. Minor changes essentially account for inflation. Unlike in many other European countries, contributions are uneven between employers (35% of total gross income) and employees (6.4%). The EUROMOD module for Spain is described by Levy and Mercader-Prats (2001). 3. See other analyses of the reform: Castan˜er et al. (2001) use the panel data of the Spanish Institute of Fiscal Studies and characterize the reduced redistribution implied by the reform. Levy and Mercader-Prats (2002) focus on the analysis of the withholding mechanism and the effects of the new income tax system on the efficiency showing that the 1999 reforms fail in reducing the compliance costs of taxpayers. Using another database, Sanchis and Sanchis (2001) simulate the new PIT system, taking into account the effects of a VAT increase on household consumption introduced to compensate for the fall in income tax revenue that the reforms involved. 4. Note, however, that the conclusion does not apply to the extreme of the distribution where Kernel estimates have limited validity.

ACKNOWLEDGMENT This paper draws on several works with Franc- ois Bourguignon and Xisco Oliver that I acknowledge. I also acknowledge financial support of the Spanish Government – MCYT (SEC2002-02606) and Fundacioń FBBVA. Usual disclaimers apply.

REFERENCES Atkinson, A. B. (1995). Public economics in action: A basic income-flat tax proposal. Oxford: Clarendon Press. Blundell, R., & MaCurdy, T. (1999). Labour supply: A review of alternative approaches. In: O. Ashenfelter & D. Card (Eds), Handbook of labour economics (Vol. 3a, pp. 1559–1695). Amsterdam: North-Holland. Bourguignon, F., & Spadaro, A. (2000). Redistribution and labour supply incentives: An application of the optimal tax theory. Paper presented at the 56th congress of the IIPF, 28–31 August, Seville. Bourguignon, F., & Spadaro, A. (2004). Microsimulation as a tool for evaluating redistribution policies. Journal of Economic Inequality, 4(1), 77–106. Bourguignon, F., & Spadaro, A. (2005). Tax-benefit revealed social preferences. PSE Working Paper No 22-2005. Castan˜er, J. M., Onrubia, J., & Paredes, R. (2000). Efectos de la reforma del I.R.P.F. sobre la renta disponible, su distribucioń y sobre el bienestar social. Economistas, 84, 183–199. Diamond, P. (1998). Optimal income taxation: An example with a U-shaped pattern of optimal marginal tax rate. American Economic Review, 88(1), 83–95. Levy, H., & Mercader-Prats, M. (2001). EUROMOD country report: Spain. EUROMOD Working Paper.

212

AMEDEO SPADARO

Levy, H., & Mercader-Prats, M. (2002). Simplifying the personal income tax system: Lessons from the 1998 Spanish reform. Fiscal Studies, 23(3), 419–443. Mirrlees, J.A. (1971). An exploration of the theory of optimal income taxation. Review of Economic Studies, 39, 175–208. Oliver, X., & Spadaro, A. (2004). Are Spanish governments really averse to inequality? A normative analysis using the 1999 Spanish tax reform. Investigaciones Econo´micas, XXVIII(3), 551–566. Sanchı´ s, J. A., & Sanchı´ s, A. S. (2001). Ana´lisis de simulacioń de los efectos distributivos de la reforma del IRPF de 1999, en Desigualdad, redistribucioń y bienestar: Una aproximacioń a partir de la microsimulacioń de reformas fiscales. Coord. J.M. Labeaga y Magda Mercader. Instituto de Estudios Fiscales. Spadaro, A. (2005). Microsimulation and normative policy evaluation: An application to some EU tax-benefits system. Journal of Public Economic Theory, 7(4).

Micro-Simulation in Action, Volume 25: Policy Analysis in Europe using EUROMOD

Housing Policy In Europe

Making Policy in Europe

B-25 Mitchell in Action

Advances in Parasitology Volume 25

Advances in Genetics Volume 25

Authority in Contention, Volume 25

Advances in Agronomy, Volume 25

Advances in Immunology Volume 25

Advances in Parasitology Volume 25

M24 Chaffee in action - Armor No. 25

Using Corpora in Discourse Analysis

Perception and Action in Medieval Europe

WebWork in Action (In Action)

Regional Policy and Planning in Europe

Objectivism, Subjectivism, and Relativism in Ethics: Volume 25, Part 1 (Social Philosophy and Policy) (v. 25)

Early Medieval Jewish Policy in Western Europe

Transcultural Europe: Cultural Policy in a Changing Europe

Advances in Chemical Engineering, Volume 25

Advances in Heterocyclic Chemistry, Volume 25

Advances in Insect Physiology, Volume 25

Advances in Insect Physiology, Volume 25

Advances in Marine Biology: Volume 25

Reviews in Computational Chemistry, Volume 25

Transcultural Europe: Cultural Policy in a Changing Europe

Advances in Heat Transfer, Volume 25

Communications in Algebra, volume 25, number 12

Advances In Quantum Chemistry, Volume 25

Advances in Experimental Social Psychology, Volume 25

Advances in Botanical Research, Volume 25

Annual Reports in Medicinal Chemistry, Volume 25

Micro-Simulation in Action, Volume 25: Policy Analysis in Europe using EUROMOD

Housing Policy In Europe

Making Policy in Europe

B-25 Mitchell in Action

Advances in Parasitology Volume 25

Advances in Genetics Volume 25

Authority in Contention, Volume 25

Advances in Agronomy, Volume 25

Advances in Immunology Volume 25

Advances in Parasitology Volume 25

M24 Chaffee in action - Armor No. 25

Using Corpora in Discourse Analysis

Perception and Action in Medieval Europe

WebWork in Action (In Action)

Regional Policy and Planning in Europe

Objectivism, Subjectivism, and Relativism in Ethics: Volume 25, Part 1 (Social Philosophy and Policy) (v. 25)

Early Medieval Jewish Policy in Western Europe

Transcultural Europe: Cultural Policy in a Changing Europe

Advances in Chemical Engineering, Volume 25

Advances in Heterocyclic Chemistry, Volume 25

Advances in Insect Physiology, Volume 25

Advances in Insect Physiology, Volume 25

Advances in Marine Biology: Volume 25

Reviews in Computational Chemistry, Volume 25

Transcultural Europe: Cultural Policy in a Changing Europe

Advances in Heat Transfer, Volume 25

Communications in Algebra, volume 25, number 12

Advances In Quantum Chemistry, Volume 25

Advances in Experimental Social Psychology, Volume 25

Advances in Botanical Research, Volume 25

Annual Reports in Medicinal Chemistry, Volume 25

Recommend Documents