New Perspectives on the Late Victorian Economy: Essays in Quantitative Economic History, 1860-1914

Through channels both open and concealed, the Victorian economy continues to influence us powerfully. Much economic thinking today gains support from perceptions of how the nineteenth-century British economy worked and how well it satisfied wants. Contemporary oligopolistic industrial structure is contrasted with Victorian self-regulating competition; the gross inequalities of Victorian laissez-faire are compared with support for the needy provided by the modern welfare state; and some regard Victorian values as vital principles of social organisation which should be regained. By examining the behaviour of the British economy between 1865 and 1914, the present work casts light upon some of these views. It does so in a variety of ways. New methods or evidence are deployed to establish accepted conclusions more firmly; unwarrantedly neglected aspects of the economy are analysed with present day concerns in mind; and traditional conclusions are reassessed. The book focusses upon three central themes: industrial organisation and technology, wages and living standards, and the monetary system. These are at the heart of discussions of productivity growth, the standard of living, well-being and poverty; the criteria by which the Victorian economic system should ultimately be judged.

New perspectives on the late Victorian economy

New perspectives on the late Victorian economy Essays in Quantitative Economic History 1860-1914

Edited by James Foreman-Peck Professor of Economic History, University of Hull

The right of the University of Cambridge to print and sell all manner of books was granted by Henry VIII in 1534. The University has printed and published continuously since 1584.

CAMBRIDGE UNIVERSITY PRESS Cambridge New York Port Chester Melbourne Sydney

Published by the Press Syndicate of the University of Cambridge The Pitt Building, Trumpington Street, Cambridge CB2 1RP 40 West 20th Street, New York, NY 10011, USA 10 Stamford Road, Oakleigh, Melbourne 3166, Australia © Cambridge University Press 1991 First published 1991 Printed in Great Britain by Redwood Press Ltd, Melksham, Wiltshire British Library cataloguing in publication data

New perspectives on the late Victorian economy: essays in quantitative economic history 1860-1914. 1. Great Britain. Economic conditions, history I. Foreman-Peck, James 330.941 Library of Congress cataloguing in publication data

New perspectives on the late Victorian economy: essays in quantitative economic history 1860-1914 / edited by James Foreman-Peck. p. cm. The research in the present volume originated as papers given at two of the annual ESRC Quantitative Economic History conferences (Newcastle and Birkbeck)' Pref. ISBN 0-521-39107-5. 1. Great Britain - Economic conditions - 19th century. 2. Great Britain - Economic conditions - 20th century. 3. Monetary policy - Great Britain - History. I. Foreman-Peck, James. HC255.N39 1990 330.94T081-dc20 89-77375 CIP ISBN 0 521 39107 5

CE

Contents

List of contributors List of

page figures

ix x

List of tables

xi

Preface

xv

1

Quantitative analysis of the Victorian economy

1

James Foreman-Peck

PART I TECHNOLOGY AND INDUSTRIAL ORGANISATION 2

Historical trends in international patterns of technological innovation

35 37

John Cantwell

3

Railways and late Victorian economic growth

73

James Foreman-Peck

4

5

Appendix 3.1 Data for the total factor productivity calculation

90

Appendix 3.2 Derivation of the steady state response of income per head to railway technology

91

Emergence of gas and water monopolies in nineteenth-century Britain: contested markets and public control Bob Millward The expansion of British multinational companies: testing for managerial failure

96

125

Stephen Nicholas

PART II DISTRIBUTION 6 A new look at the cost of living 1870-1914 Charles Feinstein

vii

147 151

viii

Contents

1 Poor Law statistics and the geography of economic distress

180

Humphrey Southall

8 Perfect equilibrium down the pit

218

John G. Treble Appendix 8.1

The bargaining model

PART III THE MONETARY SYSTEM AND MONETARY POLICY 9 Money, interest rates and the Great Depression: Britain from 1870 to 1913

241

249 251

Forrest H. Capie, Terence C. Mills and Geoffrey E. Wood

10 The UK demand for money, commercial bills and quasi-money assets, 1871-1913

285

Paul Turner

11

Appendix 10.1

Econometric techniques

300

Appendix 10.2

Data

301

An analysis of Bank of England discount and advance behaviour 1870-1914 Tessa Ogden

Index

305

344

List of contributors

John Cantwell, Department of Economics, University of Reading. Charles Feinstein, All Souls College, Oxford. James Foreman-Peck, Department of Economic History, University of Hull. Terry Mills, Forrest Capie and Geoffrey Wood, City University Business School. Bob Millward, Department of History, University of Manchester. Stephen Nicholas, Department of Economic History, University of New South Wales. Tessa Ogden, European Commission, Luxemburg. Humphrey Southall, Department of Geography, Queen Mary College, London. John Treble, Department of Economics, University of Hull. Paul Turner, School of Economic Studies, University of Leeds. ix

Figures

1.1

Choice of technique

page 36

1.2

Spatial costs and competition

36

2.1

Galtonian regression with the RTA index

50

3.1

Dynamic social savings

86

II. 1 Cost of living measurement with Laspeyres and Paasche 145 indices II.2 Bilateral monopoly in the labour market 148 Cost of living indices with alternative weights, 1900 = 100 161 Comparison of new cost of living index with Bowley's 172 index, 1900=100 6.3 Comparison of separate components of new index with 173 Bowley's index, 1900= 100 6.1 6.2

7.1

7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 7.11

England and Wales 1857-1899: percentage of total population in receipt of poor relief (a) original monthly series (b) detrended and smoothed Unemployment in three trade unions Power spectra: engineers and carpenters Power spectra: ironfounders and Poor Law County pauperage patterns (Total and ABM), January 1868 Total pauperage, January 1909; poor relief per head, 1831 Percentage increase in annual averages on relief by county (Total and ABM), 1866-1868 Percentage of total population on poor relief, Transect Unions, July 1866 Percentage of total population on poor relief, Transect Unions, July 1863 Rate of change in percentage of total population on poor relief, Transect Unions, 1860-1863 Rate of change in percentage of total population on poor relief, Transect Unions, 1866-1869

184

186 187 188 191 194 196 202 204 206 208

List of

figures

xi

7.12 Rate of change in percentage of able-bodied males on poor 210 relief, Transect Unions, 1866-1869 7.13 Bi-annual time series for four Lancashire unions, 1860- 211 1872 (Ashton-under-Lyne, Preston, Liverpool, Ormskirk) 8.1 The pay-off matrix

230

8.2 The extensive form of the game

231

III. 1 A simple monetary economy 9.1 9.2 9.3 9.4 9.5 9.6

9.7

11.1 11.2 11.3 11.4 11.5 11.6

Narrow money: MO Broad money: M3 Interest rates Output Price level Impulse response functions for (QF, MO, PF, RC) (a) Output responses (b) Money responses (c) Price responses (d) Interest rate responses Impulse response functions for (QF, M3, PF, RC) (a) Output responses (b) Money responses (c) Price responses (d) Interest rate responses Bank of England discount and advance activity 1870-1914: TVB, VAOB and VAOS Bank of England discount and advance activity 1870-1914: TVR Bank of England discount and advance activity 1870-1914: TNT, TNAT and TNR Bank of England discount and advance activity 1870-1914: average daily interest rate Actual and fitted compared - TVB: 1890-1899 Residual and SD compared - TVB: 1890-1899

249 261 261 262 262 263 275

277

316 316 317 317 321 321

Tables

2.1 2.2 2.3

2.4 2.5 2.6 2.7 2.8 3.1 3.2 3.3 3.4 3.5 3.6 3.7 4.1 4.2

A list of 30 leading European companies patenting in page 44 the US before 1914 The total number of US patents granted to residents of the 45 major countries of origin Indices of revealed technological advantage for the major 46 industrialised countries in the periods (i) 1890-1912 and (ii) 1963-1983 The results of the regression of RTA in 1910-1912 on RTA 53 in 1890-1892 The results of the regression of RTA in 1963-1983 on RTA 61 in 1890-1912 The results of the regression of RTA in 1977-1983 on RTA 63 in 1963-1969 The strength of the regression effect over the period 64 1963-1969 to 1977-1983 The results of annual and other subperiod regressions of 66 the RTA index over the 1963-1983 period British railway receipts as a proportion of GDP 1865-1910 Railway returns traffic statistics 1870-1910 British railway output components growth estimates British railway output growth estimates British railway employment growth estimates British railway total factor productivity growth and components 1870-1910 The long-run elasticity of national output per head with respect to the national efficiency index

77 79 80 80 81 81 88

Rates of return on gas, water and other UK securities 103 1870-1913 Numbers of gas and water undertakings in the UK 117 1845-1915

xii

List of tables 5.1 5.2 5.3 5.4 5.5 5.6 6.1 6.2 6.3 6.4

Regional distribution of British FDI in production plants Choice of location for British pre-1914 FDI Interwar investment patterns by pre-1914 MNEs Product group composition of British FDI Bunching of FDI in production branches Logit models for the sales branch-production plant decision

xiii 133 133 133' 135 138 140

Allocation of consumers' expenditure 1900 158 Average weekly expenditure of working-class households 159 1900 Rents of private dwelling-houses, Great Britain, reassess- 166 ment years 1877-1910 Cost of living index 1900= 100 170

7.1

Rank correlations between ASE unemployment and 197 pauperage by union counties 7.2 Variables used in stepwise regression 199 7.3 Stepwise regression results, Suffolk-Lancashire transect 205 for 1863 7.4 Stepwise regression results, Suffolk-Lancashire transect 207 for 1869 8.1

Summary statistics of coal industry arbitration

9.1 9.2 9.3 9.4

Univariate ARIMA models 264 Estimated vector AR( 1) models 266 Proportions of forecast error k years ahead accounted by 270 each innovation Selected structural models 272

10.1 10.2 10.3

288 288 294

Deposits of financial institutions as % of quasi-money Changes in structure of financial sector Exclusion restrictions for M3, NBFI deposits and commercial bills equations 10.4 Demand functions for M3, NBFI deposits and commercial bills 10.5 Demand function for quasi-money and ratio of M3 to NBFI deposits 10.6 Income and interest rate elasticities 1871-1980

239

294 296 297

xiv

List of tables

11.1 Governor regression results (TVA) 11.2 Governor regression results (TVB) 11.3 Governor regression results (VAOS)

329 330 331

Preface

The research in the present volume originated as papers given at two of the annual ESRC Quantitative Economic History conferences (Newcastle and Birkbeck). Since they all concerned the same relatively short period of British history, I thought it would be important and interesting to see what they added up to. Normal academic practice is to scatter such papers among a number of scholarly journals. How they affect our interpretation of the past in total then depends upon individual scholars' sampling of this literature or has to wait upon the work of a synthesising textbook writer. For those not primarily concerned with this particular mode of enquiry, thefindingscould easily be overlooked or underestimated. The following chapters provide a more rigorous analysis of a number of established historical questions, present new evidence, and identify new problems for the maturing industrial economy of the later nineteenth and early twentieth centuries. Readers are likely to have different interests; some will be more concerned with the methods, others will focus on the history. For those interested in the methods, but without a great deal of practice in them, each of the three parts of the book is preceded by a brief statement of the unifying economic concepts. To facilitate an appreciation of the techniques, a glossary is combined with the index. In the introductory chapter, for stylistic reasons, I have used the terms Britain and British to refer to the United Kingdom. I have taken a similar licence with the term 'Late Victorian', here assumed to be a period beginning in 1860 and ending in 1914.

xv

Chapter 1

Quantitative analysis of the Victorian economy James Foreman-Peck

Long after all participants have passed on, the Victorian economy continues to exercise powerful influences, both open and concealed, over economic policy and attitudes, as well as through the legacy of economic structure. Perceptions of how the economy worked and how well, form the background to a number of present-day concerns. Economists allude to the Victorian economy as a period of competition in opposition to today's oligopolistic industry, social commentators contrast the provisions of the modern welfare state with the inequalities generated by Victorian laissez-faire, and Mrs Thatcher appeals to Victorian values to provide the social cement necessary to bind society together. The later Victorians were more prone to question their society than their mid-century forebears. British industry was more often described as uncompetitive (Williams 1896, Stead 1901) and social reformers (Mearns 1883, Booth 1902, Rowntree 1902) provided evidence for criticising the distribution of the gains from their expanding industrial society. Criticism has continued ever since. On the side of production, the Victorians have been blamed for dissipating the head start of the industrial revolution, and neglecting opportunities to lay the foundation for an advanced, technological, competitive British economy in the twentieth century. Inadequate attention to education and industry and too much concern with class, leisure and empire slowed economic progress, while the more vigorous Germany and the United States overtook Britain (Wiener 1981, Barnett 1986). Victorian and Edwardian economic structure was a poor inheritance for the later twentieth century (Crafts 1985). An increasing commitment to the world economy concentrated the British economy on the production of goods from a few sectors, the market and technological prospects for which were not bright. With the stagnation and collapse of the world economy in the years between the world wars, the consequence of this overcommitment was persistent mass employment. In turn unemployment cleared the ground for the acceptance of the ideology of the welfare state and Keynesian economics for a generation or more after the ending of the Second World War. According to Olson (1982), reforms 1

2

New perspectives of the late Victorian economy

after the Second World War must have been a form of 'institutional sclerosis', presumably halted by the advent of Mrs Thatcher's economic policies in 1979. Most of these pessimistic judgements have been strongly contested, among others by Pollard (1989) and McCloskey (1981). Sanderson (1988) and Rubinstein (1988) offer an antidote for the condemnation of British education and class. The present volume casts further light on claims about the performance of the Victorian economy and therefore about its relevance as a model for contemporary society. Some of the essays establish accepted conclusions more firmly by deploying new evidence or methods. Others direct our attention to aspects of the Victorian economy which have been unwarrantedly neglected. A third category reassesses traditional views. The picture offered is a complex and varied one. This was a growing economy, dependent on a rather narrow range of technologies, creating institutions in response to changing aspirations and accumulating experience. At least the monetary sector was still controlled by a doctrine pessimistic about the possibilities of state action and by a small number of influential persons.

1.1 Theory and quantification Not only are the following chapters unified by their concern with the later Victorian economy (which is here taken to include the Edwardian economy as well) but also by a method which the subtitle of the book refers to as 'quantitative'. This term is not especially apt but is adopted here because of the name of the study group to which earlier versions of the chapters were presented. That name in turn reflects the peculiar response of British economic historians to the spread of econometric methods and formal economic modelling into the subject from across the Atlantic during the 1960s {History Today 1985), pp. 36-7). In fact the essays are as united by their use of theory as in their application of numbers. It is theory which gives rise to counterfactuals; what would have happened had things been different? The usefulness of a theory in a historical explanation depends upon the correspondence of its assumptions or premises with the salient facts. Therefore in the present chapter which surveys the main results and approaches of the essays in this book, reference must be made to other studies which do net depend upon any explicit theoretical apparatus. In addition because the chapters only partially cover the later Victorian economy, a brief outline is offered of some other recent quantitative and analytical work on the period.

Quantitative analysis of the Victorian economy Theory is closely linked to quantification. The Victorians poured forth an unprecedented volume of official statistics, and businesses and other private organisations greatly augmented these with their records. Each publisher or keeper of statistics at least originally had in mind a purpose which would be served by the information. Company accounts were originally intended to show shareholders that their capital had not been dissipated or mismanaged. Trade statistics were collected because trade was a source of government revenue. Bank of England lending (bills discounted) was published because the stability of the monetary system which the Bank supervised and, in particular, Bank conformity with the 1844 Banking Act, were matters of public concern. This last is a clear example of theory influencing the generation of data. In the other two cases 'theory' may seem a pretentious term for generalisations such as 'the more imports that enter the country the more revenue can be raised at unchanged tariff rates' and 'credits equal debits'. But even identities, statements which are true by definition, such as credits equal debits, are low level theories which may have profound consequences for the way we view the world (for example Baines 1985, Foreman-Peck 1989a). Money supply statistics and national income accounting are cases in point. The Victorians constructed no official figures for national income, expenditure or output, or for broad money, although a few attempts were made to produce numbers for one or two years by private individuals. Money supply aggregates (Capie and Webber 1985) are in the process of changing our perception of the later Victorian macro-economy. Reconstructions (Deane and Cole 1962, Feinstein 1965, 1972, Greasley 1986) of national income and component aggregates for the nineteenth century, in a manner consistent with modern definitions and theory, have already provided the basis for a continuing reassessment of Victorian economic growth. One of the results of the new data has been the claim that the slowing down of economic growth, in manufacturing at least, was Edwardian, not Victorian (Matthews, Feinstein and Odling-Smee 1982). Cross-national comparisons of national income growth have confirmed the earlier more qualitatively based view of the British economy falling behind those of newer industrialising countries (Maddison 1982).

1.2 The macroeconomic approach Interest in national income accounting is based in large part upon the belief that what is being exhaustively measured is the actual or potential production of goods and services that people want for consump-

3

4

New perspectives of the late Victorian economy

tion. But, in the long term, economic development could cause a divergence of measured national income from the flow of goods and services actually desired. Ignorance of buyers as to the performance or the effects of increasingly complex products, elimination of previously free facilities, goods or services, such as quietness, rights of way, common lands, and the need to pay for additional, formerly unnecessary facilities, including water and sewage services for towns which could be dispensed with in small villages, each of these might lead national income per head to overstate wellbeing, even when due allowance had been made for the distribution of that income. Broadly speaking, O'Brien and Keyder (1978) take this position in their comparison of British and French economic growth in the nineteenth century. The United Kingdom only appeared much better off than France. The reality was that Britain's road to development required the production of more intermediate goods, things not required for their own sake, which were mistakenly counted as final products. An even more influential reinterpretation based upon accounting identities has been that of Crafts (1985). (Private) national output is composed of the constituent industry outputs. By examining the composition and industry productivity growth of national output and comparing this with the development patterns of other nineteenth-century European economies, Crafts concluded that the industrial revolution took place on a very narrow front; it was confined to a few industries and never pervasively affected the economy. Otherwise agriculture and mining would have been unlikely to experience the productivity slowdown they did from the 1870s. The service sector already employed 30 per cent of the workforce at the mid century, a proportion that had risen to 45 per cent by the outbreak of the First World War. Productivity growth in services, at 0.45 per cent per annum 1873-1913, was lower than the national average and tends to support Crafts' judgement about the narrow spread of the industrial revolution (Millward 1988). The Panglossian view of the Victorian economy, that 'everything was for the best in the best of all possible worlds', emphasises that the British economy was supply constrained and that other countries were catching up by utilising a backlog of technology, that given her human and natural resource endowments, was not efficient for Britain (McCloskey 1981). The assumption of this view is that the market 'worked'. Conceivably as others have asserted (such as Elbaum and Lazonick 1986), there may have been a failure to establish institutions that would invest adequately in the

Quantitative analysis of the Victorian economy human capital that was so much more important in the new industries of electrical engineering and organic chemistry than in textiles, coal and iron and steel. Because they were more competitive, these sectors of late Victorian industry which employed lower than average proportions of skilled labour tended to export more than did the skilled labour-intensive sectors (Crafts and Thomas 1986). If the market did not 'work' (optimally) because of monopoly or monoposony power, conventional uses of total factor productivity indices (TFP) which underpin many of the above claims (and some of those of chapter 3) can be misleading (Nicholas 1982, Thomas 1985, Nicholas 1985). By assuming that the economy was not working optimally, it is possible to calculate how fast the Victorian economy could have grown if more resources had been committed to those sectors capable of most rapid expansion and technological progress, taking the American economy of the period as a standard of comparison. Supposing industries with strong growth potential - chemicals, engineering, instrument production, paper and printing, electricity, telecommunications, construction and construction materials - expanded by more than in fact they did, requires that resources were notionally taken away from less dynamic areas - textiles, agriculture, net property income from abroad, domestic service and manufactured gas production and distribution. Expanding the highproductivity, high-growth prospect sectors by plausible magnitudes yields a gain of 4.55 per cent of average GNP by these sectors at the expense of the others. The shift in sectoral size would have raised aggregate-growth rates by 0.56 per cent per year from 1870 so that by 1913, output would have been 126.8 per cent higher than it actually was, a level not in fact achieved until 1946 (Kennedy 1987, pp. 58-77). More extreme but plausible structural shifts could have raised incomes per head by 150-200 per cent by 1913 according to this method of calculation. The aggregate or macroeconomic approach to the Victorian economy (recently synthesised by Lee 1986) offers explanations and hypotheses at a high level of abstraction. The papers in this volume extend our understanding of the Victorian economy by a more detailed analysis of sectors or themes. This is an essential test of, and sometimes antidote to, macroeconomic generalisations. Although the chapters do not present a comprehensive coverage of the issues of the late Victorian economy (there is no analysis of the demographic slow down for instance and little on international economic relations) they do cover the three crucial themes of industrial organisation and technology, wages and living standards, and the monetary system.

5

6

New perspectives of the late Victorian economy

Industrial organisation and technology underpin the record of economic growth and productivity increase discussed in the macroeconomic approach. Analysis of the experience with particular technologies andfirmscan offer explanations more fundamental than is possible at the macroeconomic level. The nature and distribution of the gains from productive activity ultimately justify or condemn an economic system. Hence the standard and cost of living, the extent of economic distress especially among the weakest members of society, and the method by which earnings are established, are no less fundamental economic issues than the efficiency of production. Neglected during the years of ascendancy of Keynesian economics, monetary history of the Victorian economy is well represented in the present volume. The monetarist counter-revolution was led by Friedman and Schwartz's Monetary History of the United States. It is therefore appropriate that in this reinterpretation of the Victorian economy detailed attention is given to monetary institutions, the functioning of the gold standard economy, and the influences upon monetary policy. Before surveying the following chapters, some mention must be made of recent quantitative findings in two important fields not otherwise represented in this volume.

1.3 Later Victorian demography Late Victorian deceleration of population growth became a phenomenon of particular interest to the world after 1945. Poorer countries then experienced population explosions on an even greater scale than that of the British industrial revolution, so that fears of a Malthusian crisis revived. Britain perhaps offered hope that population growth in these countries also would soon slow down. A range of apparently competing explanations have been offered for the decline in fertility towards the end of the nineteenth century. One of the more imaginative but untested explanations was that the Victorian decline was due to corsets. These were allegedly so tight that female reproductive ability was impaired. Consistent with the pattern of fertility decline between the social classes, tight corsets began as a middle-class fashion before spreading down the social ladder. A number of other influences upon fertility have been tested however. Crafts (1984) showed that work opportunities for women were negatively correlated with birth rates between 1877 and 1938 because they increased the opportunity cost of children. He emphasised the costs of birth-control technology and used

Quantitative analysis of the Victorian economy illegitimacy as a measure. As birth-control costs proxied by illegitimacy fell, so too did marital fertility. Extramarital fertility was related to the organisation of work. According to Humphries (1987) it was the need to control fertility that accounted for Victorian sexual segregation at work. That is how she explains her negative correlation across regions between her index of sexual segregation and illegitimacy rates in 1851. Alternatively segregation might have taken place because of labour market discrimination against women; it was a means of maintaining male earnings. In this explanation lower illegitimacy was merely another consequence of segregation, not primarily a social adjustment to inadequate birth-control technology. Turning to mortality, the second influence upon population size and growth, income distribution played a role in the late Victorian and Edwardian economy but legislation to improve safety at work probably did not. Haines (1985) found that infant mortality in England and Wales in 1911 was higher than in the United States in 1900 because of the greater income inequality in England and Wales. The larger English proportion of the very rich held their position at the expense of the lives of some very poor children. Adult mortality was apparently not reduced by the penalties under the 1878 Factory and Workshop Act imposed upon culpable employers on whose premises fatal accidents took place. The small number of inspectors and the low level of fines for violators rendered the Act rather ineffective (Bartrip and Fenn 1988). By contrast the Workmen's Compensation Act of 1897 raised reported fatal accidents. Possibly the greater chance of compensation made those at risk take less care, but the most likely explanation is an increased incentive to report. The third influence upon population size and growth is migration. Late Victorian emigration assumed remarkably high proportions. By the 1880s falling shipping costs had sharply increased the proportion of single labourers, who were quite likely to return, at the expense of the permanent family migration of farmers and craftsmen of the 1840s. The majority of emigrants were not born in rural areas and emigration rates from rural counties were only marginally higher than those from urban counties, (the greater number of Late Victorians lived in towns, not in the country) (Baines 1985). Agricultural depression (itself a response to changes in the world economy) then can hardly have been a major force pushing workers overseas. Internal migration in Britain, unlike France and Sweden, did not respond to nominal wages but to urbanisation (Soderberg 1985). Regional variations in money wages were in any case lower in nineteenth-century Britain than in France, Prussia or Sweden.

1

8

New perspectives of the late Victorian economy

Unfortunately, how much of this, if any, stemmed from smaller spatial price variations in Britain is unknown because of the absence of data.

1.4 Trade, empire and fluctuations The late Victorian economy was extraordinarily open, not only in terms of the outward flow of labour, but measured by international movements of goods and capital also. The trade/GNP ratio of 1913 was not achieved again until the 1970s. Rapidly falling shipping costs, with the advent of the ocean-going metal hulled steamship from the 1850s, boosted international trade and specialisation (Harley 1980, 1988). A necessary condition for increasing international openness was political acquiescence. Since trade rose so rapidly, the British government could both receive a rising revenue from tariffs and reduce tariff levels and coverage (McCloskey 1981). Elsewhere in late Victorian Europe acquiescence was not so easy to obtain. Cheap agricultural produce, especially wheat, depressed European farm incomes and rents. Those affected reacted with demands for higher tariffs. In Britain the agricultural sector had already been overshadowed in importance, politically and economically, by manufacturing and services. Consequently complaints from landlords went largely unheeded, legitimated by the widely held belief that free trade maximised national income. Yet had Britain maintained a higher tariff in the 1880s, she might in fact have been able to enhance her income at the expense of the rest of the world, by shifting the terms of trade, forcing down import prices by reducing the volume of import bought (McCloskey 1981). In principle it is possible for shifts in the terms of trade with the expansion of the export sector to make a country poorer. Selling more exports drives down their prices. Conceivably more resources might be absorbed by the exports necessary to buy a given quantity of imports after the price fall than before. For this later period of the nineteenth century it was not this possibility of trade immiserisation but of capital exports making the country worse off than it need have been, that has exercised scholars. Pollard (1989) concluded it was a question of'short termism'; that short-term earnings were maximised at the expense of long-run income generation. Kennedy (1987) pointed to the low investment in British industry and the capital market giving the wrong signal to investors, from the viewpoint of long-term development. Contrary to much popular opinion the bulk of investment diverted overseas did not go to the British Empire. This misconception might be

Quantitative analysis of the Victorian economy traced to the conflation of the eighteenth-century mercantilist empire with the liberal empire of the nineteenth century and the coincidence of the vastness of the empire at its height in 1913 with enormous overseas investment and income. Very few administrators ran the empire and the British government tried unsuccessfully to ensure that it was self-financing (Davis and Huttenback 1986). Whereas viewed from Britain, investment and empire are frequently seen as disadvantages in retrospect, from abroad they are often represented as gains made at the expense of the host regions. Examination of the British balance of payments for this period and the balance of payments of India, the jewel in the imperial British crown, shows little evidence of exploitative returns, broadly defined, to British investment. Indeed the returns seem rather low (Foreman-Peck 1989a). International channels can be sources of fluctuations in economic activity, but long cycles are traditionally given a technological origin. Cycles offiftyyears duration, Kondratieffs, have experienced an upswing in popularity in recent years, promising to place economic history in front of a wider audience by providing some form of prediction (Lloyd-Jones 1987). Actually a twenty-five year building/population cycle seems to fit the facts better since 1850. This Kuznets cycle is also more plausible, given the normal lags in human reproduction between generations and the need to house new families (Solomou 1988). The likelihood of the structure of the economy remaining unchanged over periods of more than a century seems rather small. Cycles may be expected to have variable durations depending on the economic structure.

1.5 Production: technology, industrial organisation and competitiveness Contemporaries at home and abroad were enormously impressed by the achievements of mid-Victorian industry, exemplified by the glittering success of the Great Exhibition of 1851. Yet during the last quarter of the nineteenth century the ability of Germany and the United States to surpass Britain in the development of industrially useful technologies was becoming obvious. Whereas Britain had been in the forefront of telecommunications in the 1840s with the electric telegraph, the telephone was invented in the United States and was introduced into the United Kingdom under American patents. Thanks to a number of innovators of which Stephenson was the best known, the railway was spread around the world by British engineers, but it was in Stuttgart that Daimler and Benz

9

10

New perspectives of the late Victorian economy

produced the first practical motor car in 1885 and it was from France that the early British motor industry drew much of its technology. When motor production did get under way in Britain, in at least one major engineering firm it had to compete with far more profitable defence contracts and failed to do so (Irving 1975). The British environment seemed to have become far less congenial to industrial innovation. True, Britain was still at the frontiers of technology in the invention of the electric light bulb and in the compound steam turbine for the generation of electricity. But the most financially successful electrical company, GEC, was founded by a German immigrant, the British radio industry was begun by an Italian, Marconi, and Ferranti, whose Deptford power station of 1888-91 led the world, brought his company to the edge of bankruptcy by 1903 (Wilson 1988). The financial incentives were to encourage mergers without the reorganisation essential to compete with the large German and American advanced technology based enterprises (Hannah 1983, pp. 20-1, Davenport-Hines 1984). In chemicals too, Britain's early lead is generally judged to have been lost in the last quarter of the nineteenth century. The Leblanc soda producers neglected to switch to the ammonia process patented by the Belgian Solvay, perhaps because of the formation of the monopoly United Alkali Company in 1890 (Lindert and Trace 1971). Britain left the synthetic dyestuffs industry entirely to Germany. Unlike the pharmaceutical industry in Germany and the United States, British pharmaceutical firms failed to establish research and development departments and eventually suffered losses of market position accordingly (Liebenau 1984). Only in fairly low technology industry did Britain remain at the frontier of innovation through private enterprise. With Starley's invention of the safety bicycle in 1884 the bicycle industry took off and established a substantial export trade. The surgeon Dunlop's pneumatic tyre of 1888 clearly supported this boom. The Bowden brake was also the invention of an amateur, a journalist. The most successful saddle was designed by a Chelsea shoemaker and an Irish clergyman devised the spring clip for attaching a pump to the bicycle. Even in bicycles Starley's innovating firm, Rover, proved incapable of maintaining their early lead and abandoned bicycle production for the more profitable cars soon after the twentieth century began (Foreman-Peck 1983). Only the interests of national defence conjured up support for Marconi's radio and thus contributed to a temporary British lead in this field (Pocock 1988). The above brief survey of the literature of Britain's alleged industrial decline is impressionistic. Although it suggests late Victorian British economic development may have been retarded by an unwillingness or

Quantitative analysis of the Victorian economy inability of indigenous entrepreneurs to introduce both innovative and profitable advanced products and processes in a number of sectors, the survey does not cover the full range of industry. Cantwell's contribution (chapter 2) in the present volume is to provide an index that compares British innovative potential across industries, between countries and over time. He takes a very large number of US patents granted to residents of the major countries of origin and computes some indices first used by Soete. Each industry in each country has a revealed technological advantage (RTA) index. The numerator of the index is a country's share of all patents in an industry taken out by foreigners in the US. The denominator is all US patents taken out by residents of the (non-US) country as a proportion of all US patents taken out by foreigners (35 per cent for the UK 1890-2 and 24 per cent 1910-12). Britain's index of 1.76 for industrial engines between 1890 and 1912 therefore implies that the United Kingdom was more likely to take out a patent pertaining to the industrial engines sector than it was in agricultural machinery, which had an index of only 0.43. Perhaps not surprisingly in view of differences in industrial structure, the ranking for Canada was the reverse, with an RTA for industrial engines of 0.53 and for agricultural machinery of 2.35. At first sight the retardation of British agricultural productivity growth after 1870 could have been a consequence of this low innovation rate, but equally foreign competition and decline made a poor market for agricultural machinery and therefore for innovation. German agriculture was protected yet the German RTA in agricultural machinery was also low at 0.52. Moreover British innovation in agricultural chemicals was better than average. Agricultural decline therefore seems unlikely to explain Britain's low RTA in agricultural machinery. British firms possessed, and maintained through to the 1963-83 period, a strong technological advantage in cleaning agents, industrial engines and turbines, shipbuilding, textiles, rubber products (tyres) and coal and petroleum products. Despite a relatively strong performance in food based multinationals (Chandler 1980), Britain's RTA for food products was apparently low, as it was in pharmaceuticals, paints and other chemicals. Further disaggregation changes the picture somewhat. British firms did have a higher RTA in certain subsectors within the food industry, especially in drinks. Some pharmaceutical subsectors improved in the later period. German firms held a favourable position in chemicals in general, but especially in dyestuffs and paints, and also in lighting and wiring equipment.

11

12

New perspectives of the late Victorian economy

British exports, unlike those of Germany, were mainly manufactured in a small number of sectors, suggesting a strong comparative advantage in production in those areas. The British pattern of innovation across industries is different however. The variation is rather less for Britain than for Germany; sixteen British industries have indices greater than one compared with nine German. British distribution of patents reveals a much weaker comparative advantage in innovation than in production, a harbinger for British manufacturing in the later twentieth century. Understanding British experience in manufacturing is helped by the idea of a 'technological trajectory', introduced by Dosi (1984). A technological trajectory is a sort of dynamic production function representing the possibilities for improved performance from particular broad technologies or industries. Countries which have specialised in newer technologies, those in the early stages of their trajectories, will by definition have greater productivity growth than those specialising in mature technologies. By the late nineteenth century Britain is conventionally described as falling in the second category. Cantwell's test of the stability of technological advantage over time casts more light on the usefulness of Dosi's idea for understanding Britain's economic position. Between 1890/2 and 1910/12 the pattern of technological advantage remained fairly stable although it became less specialised. Britain was locked in to generally mature technologies perhaps because of an inability to develop the corporate organisation which would foster the new technologies. Among the more famous of these maturer technologies were the steam engine and the railway. The debate about the impact of railways in England, Wales and Scotland has led to either agnostic or confused conclusions in the literature on Victorian economic development (Mathias 1983, p. 245, Church 1975). Perhaps because much of the output of the industry was intermediate, productivity calculations for macroeconomic purposes have paid railways scant attention. But the methodological question of how the impact of a new technology should be calculated is at the heart of any interpretation of productivity calculations. If railways provided major inputs to certain industries, yet these inputs were excluded from a TFP index, then the growth of the industries' efficiency will be overstated by the index and incorrectly attributed. The concept of social saving, by how much national income is increased by an innovation, can be seen as a simple form of productivity measurement. Chapter 3 offers new social savings estimates for British railways in

Quantitative analysis of the Victorian economy

13

the years 1865, 1890 and 1910. The order of magnitude as a proportion of GNP for 1865 (10-11 per cent) is similar to Hawke's (1970) highest estimate for England and Wales. Unlike Hawke's measure, the social saving figure of chapter 3 is not computed as an upper bound, that is ignoring the adjustment of demand in a world without railways. The link between productivity increases and the economy's gain from railways is then utilised to acquire further information about the social savings generated by railways in the Victorian economy for 1890 and 1910. The savings turn out to be large, but Hawke's productivity index overstates technical progress after 1865 and a new index is constructed to 1910. This yields an upward biassed 1890 social savings figure of 21 per cent and 27 per cent for 1910. These measures are overstatements because they implicitly assume the absence of technical progress in competing transport modes. Nonetheless they do give some idea of the importance of railways and further support for Craft's view of the narrow base of the industrial revolution. If British railways had attained US levels of productivity growth between 1890 and 1910, GDP in the second year would have been 3 per cent higher. Chapter 3 shows that these social savings estimates can be embedded in a fairly simple growth model to compute some indirect effects which earlier economic historians maintained were important but which were not previously taken into account by the social savings index. The likely long-run impact of the extra saving and investment generated by railways is calculated, as is the impact on the growth of the labour force, through migration, although the construction labour force is excluded from consideration on the grounds that they were not there in the long run. In so doing the chapter points to the distinction not often noted in the railway literature that economic growth is concerned with income per head and therefore to the extent that the innovation induced a larger number of 'heads', economic growth can be reduced, even though total income is enhanced. The most probable effect of the induced savings and investment was to increase the impact of railways by about half again, over the static social savings. The dynamic model raises the issue of 'How big is large?' where induced labour force growth is concerned. 0.1 per cent per annum is suggested as the most likely dynamic effect, a proportion that sounds small until it is remembered that labour force growth as a whole was only around 1 per cent per annum. Contrary to much opinion, the later Victorian economy saw an advance in state ownership of industry. Railway nationalisation was mooted as early as the 1844 Railway Act and the telegraph system was

14

New perspectives of the late Victorian economy

nationalised under acts of 1868 and 1869 (Foreman-Peck 1989b). Moreover the supply of local services was gradually municipalised. In view of the Thatcher government policy of reversing the tendency towards state ownership, the reasons for the Victorian local government acquisitions of gas and water are of particular interest (Foreman-Peck 1989c). Two general possibilities present themselves: that the market did not work or that the state was captured by interest groups, such as rate payers, who saw an opportunity to enhance their own wellbeing at the expense of others (such as share holders) by state ownership and regulation. Millward (chapter 4) opts for the first alternative. Distribution networks for both gas and water created a high proportion offixedcosts in total costs so that initially unit costs fell, as both customers and the volume of gas or water consumed increased. As the geographical boundaries of the supplier expanded, unit costs eventually began to rise because of the distance the gas or water needed to be piped from the storage point to the household. That created the opportunity for competing suppliers to enter the geographical market and charge lower prices. Competitors' unit costs were lower because they supplied only a portion of the market. The resulting duplication of facilities raised costs and prices above the level they need have been. Private statutory monopoly proved an unsatisfactory solution because armslength regulation did not give the necessary control over quality, safety and price. Regulation was therefore 'internalised' by municipalisation, and public welfare was increased. Millward's account draws upon recent work on 'contestable markets' (Baumol 1982). This focuses upon the threat of entry as a means of holding down a monopolist's price. A contestable market is one in which entry and exit are so easy that the incumbent firm, though a monopoly because of the nature of industry costs, will price as cheaply as a competitive firm of the same size. The contestability literature is dependent upon the incumbent maintaining the same price after a competitor enters as before. That might seem implausible for a profit maximiser who would be more likely to cut prices as low as marginal costs to see off any new competition even if that entailed short run losses. However when the price of the established supplier was regulated, as was likely because of monopoly power, freedom to change prices even downwards could have been curtailed. Even with temporary price competition the resulting industry equilibrium was likely to have higher costs than necessary; there is some evidence of this in the Victorian railways (Foreman-Peck 1987) and in American late nineteenth-century oil industry experience (Telser 1987). Declining competitiveness has been measured primarily in loss of

Quantitative analysis of the Victorian economy export and home market shares and by alleged failures to introduce best practice technology. Nicholas's (chapter 5) study of over 400 multinational enterprises (MNEs), of which over 200 were founded between 1870 and 1914, introduces another dimension into the debate. The ability to establish a successful subsidiary in another country indicates a degree of competitiveness because at least, in developed countries, local firms have the advantage of knowing the environment better and one country's MNE subsidiaries have to compete against those of other countries. Nicholas examines the claim that foreign competition shifted British foreign direct investment (FDI) into the allegedly 'soft' markets of the British Empire. Canada unilaterally introduced Empire tariff preference in 1897. This favouring of British over, say, American imports must have reduced the incentive for the British to undertake foreign direct investment in Canada, since the desire to jump tariff barriers has been a major motivation for establishing multinational subsidiaries. Restrictions on the investment of other countries'firmsin Empire countries will have created a less competitive environment for British MNEs but such restrictions were mainly a phenomenon of the interwar years (Svedberg 1981). Exclusion of British investors by other world powers was a fear that encouraged colonial acquisitions in West Africa towards the end of the nineteenth century. To the extent that there was such exclusion then British MNEs would have been crowded into British territories, rather than choosing them for fear of more intense competition elsewhere. Nicholas's sample shows that 28 per cent of MNE's production plants went to the developed Empire and 6 per cent to the undeveloped empire before 1914. In the more protectionist interwar years the Empire total increased from 34 per cent to 52 per cent. Before 1914, the single most popular first choice for production branches of British companies was Germany and the second was the United States. Of course the economic importance of these markets was greater than any single Empire country, but these new data at least contradict the simple generalisation about inversion of British companies before 1914, that most direct investment went to the Empire. Chemicals was the most important product category upon which British foreign direct investment concentrated, followed by food and textiles. Nicholas's archival evidence shows that FDI 'bunched' into a very few years, which he hypothesised was due to perceived managerial economies in setting up overseas subsidiaries. Nicholas' model seeks to explain the decision in his sample to change from a sales branch overseas to a production branch in terms of production, location and transactions costs variables. Transactions costs are the costs of monitoring and polic-

15

16

New perspectives of the late Victorian economy

ing a particular organisational form of economic relation. They clearly differed between an agency without sole loyalty to the company and a wholly owned production subsidiary. These costs therefore influenced the optimum organisational form. Production costs are those associated with particular technologies. They determine whether a British enterprise will have had an advantage in a particular sector. Location costs include tariff barriers and local market size which influence the desirability of choosing a particular national location for a plant. The qualitative (binary) dependent variable requires the use of logit regression since the dependent variable is bounded between zero and unity, contrary to the assumption of ordinary least squares regression. In addition to the production cost variables which confirmed the competitive advantage in chemicals and food, 'psychic proximity', proxied by whether or not the English language was native, was a significant positive location cost variable, and the number of branches a parent firm operated abroad and product diversification were significant positive transactions costs variables. Each of these influences increased the chances that a production branch rather than a sales branch would be chosen. If Britain had less diversified or fewer multiplant companies than her industrial competitors, as seems likely, then these results explain or predict a lower degree of multinationality of her firms.

1.6 Distribution: wages, the cost of living and the social security system Most people in industrial societies primarily work to live rather than live to work. How industrial and technological changes influenced the possibilities for consumption is therefore of fundamental importance to wellbeing. Recent studies of living standards have concentrated on outward manifestations rather than on problems of measuring dimensions. Thus the impact of greater longevity on generalised living standards has been computed using techniques for estimating the willingness to pay for reductions in risk of death (Williamson 1984). Others have measured increases in height as an indicator of living standards (Floud and Wachter 1982). Estimates of the distribution of money income show Victorian Britain was a land of great inequality of income, wealth and opportunities in comparison with present day Britain (Soltow 1968, 1971, Lindert and Williamson 1982, Williamson 1982, 1985), as well as in comparison with other nineteenth-century European countries and the United States

Quantitative analysis of the Victorian economy

17

(Kuznets 1966, pp. 208-11). Skill differentials were wider in the nineteenth than in the twentieth century, although they were being reduced between 1880 and 1911 (Jackson 1987). Studies of inequality are concerned with the interpersonal distribution of income, which in turn is related to the functional distribution. If the functional distribution is tilted towards labour and away from capital and land, the interpersonal distribution will tend to become more equal because capital and land ownership are very unequally distributed. Since the bulk of the population gained their income from labour, there were two principal elements to living standards, money wages and the cost of living. Like all index numbers, the cost of living at best measures only the position of the average family, not those with slightly different tastes or habits, or in different circumstances. Charles Booth's survey of London life and labour, beginning in 1886, postulated a London poverty line based on the bare necessities and therefore effectively proposed using fixed weights, independent of actual expenditure patterns in a cost of living index. Subsequent researchers have preferred a more empirical approach based upon actual consumption patterns. The historical problem has been to obtain a large enough sample of appropriate price series with suitable weights (reflecting expenditure patterns) to make an accurate judgement about changes in the average workers' real purchasing power. That in turn raises questions of appropriately allowing for varying expenditure patterns (which strictly requires changing weights), of allowing for quality change and quality differences in the products consumed, of including new products that enter consumption for the first time within the period of study and of the exclusion of obsolete goods and services. Even the basic items of consumption create difficulties. Bread is a fairly simple case. The average retail price of a 4 lb loaf in London fell from 7-ld in 1880-2 to 5-2d in 1895-7 rising to 5-7d in 1910-12. By contrast, because of changes in fashion and durability, clothing raises far greater problems. Clothing, food, heating and lighting and housing were the bare necessities of life. The Ministry of Labour estimated that in 1914 the average working-class family in the United Kingdom distributed their expenditure across food, fuel and light, rent and clothing in the proportions 15:2:4:3. Although tobacco and alcohol were not necessities they formed a considerable portion of, generally underreported, expenditure which slowly changed with technology, income and attitudes. Beer was an important source of nutrition, sold in a number of forms. In 1890 one brewery charged 4d a quart for 'running ale', 6d for bitter ale, 3^d for

18

New perspectives of the late Victorian economy

porter and 5d for light bottled beer (New Survey 1930, p. 95). National British beer consumption in 1882 according to official figures amounted to 27 million barrels of 36 gallons, equivalent to about 220 pints per year for every man, woman and child in the United Kingdom. Or if only one quarter of the population drank beer, almost two and a half pints a day. At the 1899 peak, annual recorded consumption reached 260 pints per head. Tobacco prices after 1900 began to rise with the imposition of taxes but tobacco consumption increased even more rapidly. Tobacco consumption underlines at least one interesting major problem about living standards. Is the consumer the best judge of what constitutes a higher living standard? If so then working-class living standards, measured by this index, must have been rising in Edwardian Britain, unless there was a great deal of substitution away from other products. Elementary school education was not free although it was compulsory after 1880 and therefore in principle entered the working-class budget. Rising subsidies over the nineteenth century increased working-class access to state schools charging fees at levels below comparable private schools (Mitch 1986). From 1891 universal access to elementary schools was made easier by the abolition of fees in the state sector. Feinstein's (chapter 6) reassessment of changes in the cost of living between 1870 and 1913 originated in a concern that his earlier estimates for Edwardian national income and expenditure were too low. The output data show no Edwardian slowdown relative to the last twenty years of the nineteenth century but the income data reveal an extraordinary deceleration and the retardation indicated by the expenditure series is also sharp. The largest single component of the income series is wages and these data imply that money wages rose more slowly in the Edwardian period of rising prices than they had in the previous phase of declining prices, despite an increase in trade union membership. Falling labour productivity in manufacturing, mining and construction was a major contributor. Contrary to Arthur Lewis (1978), movements in the terms of trade do not seem to have been important. In the present work Feinstein revises Bowley's cost-of-living index, which had a limited coverage and related to a low family income. Feinstein uses 1900 household expenditure weights from Prest's (1954) estimate of consumers' expenditure in that year. Feinstein employs a wider range of prices than Bowley, although still mainly from London. The new series does not fall as sharply as Bowley's between 1880 and the mid 1890s (about 6-7 per cent) but there is little difference between the indices in the upswing. When combined with his new wage data (Feinstein 1990a,

Quantitative analysis of the Victorian economy 1990b), Feinstein can show that Edwardian employment real income increased at about 0.6 per cent per annum instead of declining at 0.5 per cent per annum. An important dimension of living standards is not just real wages but the likelihood of unemployment and the conditions and amount of unemployment benefit. A variety of working-class savings institutions, friendly and cooperative societies, savings banks, and industrial life insurance companies, provided afinancialbuffer to absorb the uncertainties of economic life (Johnson 1985). When savings were exhausted, credit was supplied by pawnbrokers, money-lenders and local shopkeepers. Workers needed resources for illness and old age, as well as for unemployment. State old age pensions, introduced in 1909, raised life-time living standards, increasing private savings for retirement. Pensions did not encourage the taking of greater leisure by earlier retirement (Johnson 1984). The effects of state unemployment benefits cannot be judged until the interwar years, since they were only available from 1913. For those permanently unable to work, the chronically sick, the insane, or children under working age, the state social support system was a crucial element in their wellbeing. The Victorian Poor Law nominally at least provided for all these categories, although the able bodied unemployed were supposedly discouraged from seeking Poor Law relief by the infamous principle of iess eligibility' embodied in the 1834 Poor Law Act. By the middle of the nineteenth century conditions were not invariably as bad as Dickensian caricatures suggest. The mentally disturbed agricultural labourer poet John Clare spent a considerable portion of his later life in a Poor Law institution in Northampton where he was encouraged to write, paint and wander round the village. The social stigma of the Poor Law, and the vigilance of the Guardians, was nonetheless sufficient to deter all but the desperate from applying for relief. Southall (chapter 7) turns to Poor Law statistics for evidence of the spatial distribution of economic distress. He is particularly concerned to establish whether Poor Law data confirm his earlier results from trade union unemployment series (Southall 1986, 1988) that a North-South divide already existed in Victorian Britain and did not first emerge in the interwar years. When detrended and smoothed, the Poor Law series resembled the trade union data, with a similar power spectrum despite their different client groups. (A power spectrum indicates for each periodic component of a stationary time series the share of the overall variance in the time series that can be attributed to the periodic components or frequencies.) With only 5 per cent of the population in receipt

19

20

New perspectives of the late Victorian economy

of relief even in depressions and most of these not able-bodied males, it might be questioned whether one ought to expect similar behaviour of trade union and Poor Law series. However in 1868/9 there were riots in Blackburn over the Guardians' policy and therefore a considerable number of able bodied men must have thought themselves eligible for relief. The proportion was however very small. In July 1868 able-bodied males in receipt of relief amounted to only 0.65 per cent of all employable males, although contrary to the mythology of the Poor Law, 85 per cent of these were on 'out-relief, they were not obliged to enter the workhouse. By 1909 the Poor Law had virtually ceased supporting able-bodied males, the task had been assumed by the trade unions, if at all. In January the Amalgamated Society of Engineers alone paid relief to two thirds of the numbers of able-bodied supported by the Poor Law. National time series do show industrial workers and paupers with not dissimilar seasonal and cyclical patterns. Building workers and paupers were greatly affected by the seasons though the metal trades were not. To isolate the regional pattern mainly for the 1860s, Southall employs stepwise regression to identify the sources of pauperdom and concludes from the explanatory power of his trend variable that the policy of individual Boards of Guardians was the major factor. The Lancashire Cotton Famine caused by the Northern blockade of Southern ports during the American Civil War was a break in Poor Law policy. In 1863 10.3 per cent in Lancashire were on relief. In general though, Poor Law relief for able-bodied males was centred in the low wage, wheat growing, rural south; Wiltshire ranked behind Lancashire in 1863 with 7.9 per cent on relief. This observation is consistent with a continuing high degree of local rate-payer control of Poor Law policy. Farmers maintained a pool of labour on relief to ease agricultural labour supply at periods of peak demand especially harvest time. In industrial areas the need was quite different and so were the resources. Merely to preserve order during troughs of severe depressions the Guardians would be obliged to pay relief to the able-bodied unemployed, since the majority were not covered by trade union benefits and in any case union resources would be quickly exhausted. Bargaining institutions can exercise considerable influence over the share of wages in national income as well as over the efficiency with which the economy worked. Nominal wage rigidity fell (Hatton 1988), encouraged by the growth of trade union membership and strength in the late Victorian and Edwardian years; unions were concerned to maintain or enhance real wages, which in periods of rising prices, needed frequent

Quantitative analysis of the Victorian economy

21

increases in nominal wages. New statutes gave unions greater ability to bargain their members' wages upwards. Disraeli's government in 1871 and 1875 legislated that unions could legally engage in any actions 4in contemplation or furtherance of a trade dispute' which would not be illegal for individuals. Unions might still be liable for civil conspiracy actions though. During the 1880s an increasing public concern with social problems such as the sweated trades and overcrowding in housing, and the formation of socialist organisations, culminated at the end of the decade in what came to be called the New Unionism, the organisation of unskilled workers. The London dockers' strike, the formation of the Miners Federation of Great Britain by a number of local miners' organisations and the establishment of gas workers' and general labourers' unions all took place in 1889. In response, a Royal Commission sat between 1891 and 1894. The Report of 1894 concluded, largely on the basis of the leading employers' testimony, that industrial relations were at their best when strongly organised trade unions voluntarily negotiated with stable employers' associations. In this way unofficial strikes were avoided. Official strikes and lockouts were inevitably industry wide in such a bargaining regime. The Engineering Employers' Federation lockout of the Amalgamated Society of Engineers in 1897 arose over the introduction of new technology by employers and its resistance by the Union (Lewchuk 1987, pp. 70-7). Unions seemed to employers to be holding back productivity growth (Phelps Brown 1983, pp. 44-6). Trade unions were not incorporated and therefore could not be sued for the damage such restrictive practices or strikes caused until the courts introduced a change of procedure, with the Taflf Vale judgement of 1900, and the House of Lords confirmed that decision in 1901. Immediately after Taff Vale, strikes were less frequent, but the economy was not booming either. Regression analysis of strike frequency in these years shows that strikes increased with business activity and profits, as well as with unionisation (Bean and Peel 1976). A one per cent increase in the rate of change of union membership was associated with a one half per cent increase in the number of strikes, holding other influences constant. The 1906 Act introduced complete immunity for trade unions from actions at tort, against the initial wishes of the Liberal government. The king can do no wrong; neither can a trade union' it was said. Labour militancy reached a pre-war peak in 1912, vividly described by Dangerfield in the Strange Death of Liberal England. Coal was a crucial industry not only as the principal fuel source for

22

New perspectives of the late Victorian economy

industry but because, as Britain's productivity growth in manufacturing fell behind that of the newly industrialising states, her comparative advantage shifted towards exploitation of her natural resources where her inability to match the industrial efficiency of Germany and the USA mattered much less (somewhat analogously to the shift towards North Sea oil at the end of the 1970s). Pencavel (1976) calculated that the spread of formal trade unionism on the coal fields from about two thirds of the labour force unionised in 1900 to four fifths in 1913 reduced efficiency through the adoption and maintenance of restrictive practices so that coal output was lower by 2-3 per cent in 1913 than it would have been at 1900 levels of unionisation. Extrapolating his results to compare a completely non-unionised industry with a 100 per cent union membership industry, he concluded output was lowered by 22 per cent by unionisation. Treble (chapter 8) examines the wage bargaining arrangements in this vital industry. There was rent dependent upon the fluctuating price of coal to be divided between the coal mine owners and the workforce. The value of the marginal product of the minprs also varied with the price of coal. Both were potential sources of industrial strife. The arbitration scheme analysed by Treble emerged from the 1893 strike as a means of resolving conflict in a manner that would be mutually advantageous. Such was the economic importance of the coal industry that within two weeks of the breakdown of talks in 1893, the Prime Minister wrote to the parties suggesting that the Foreign Secretary chair another negotiating session. Treble adopts the assumption of neoclassical institutional economics that institutions are optimal responses to the economic environment and seeks to understand what the properties of the arbitration system were. The stability of the bargaining machinery was a measure of success, he maintains. The two parties in the bargaining process had access to different information. This gave rise to a problem of designing bargaining institutions so as to ensure each had incentives to tell the truth even when the other could not know whether the truth was being revealed. The conciliation board functioned as a means of committing the parties and this was part of the solution to the bargaining problem. Whereas the boards were solutions to cooperative games, the second stage was a non-cooperative game, the rules of which were part of the solution to the cooperative problem. The two tier bargaining structure of the rules for the non-cooperative game consisted of, first, the parameters chosen every three years and second, the bargaining over actual agreements with those parameters, when the price of coal changed. Some informal evidence suggesting that perfect equilibrium may have obtained is that there was

Quantitative analysis of the Victorian economy

23

usually only one offer and one counter offer in each negotiation. Thanks to the institution of the conciliation board, each player knew the chain of possible actions and reactions that would follow from playing a particular strategy. Therefore the players could and did reach a solution without going through all those moves. In contrast with much neoclassical institutional economics, Treble's analysis has the merit of providing a testable prediction - the number of bargaining rounds - and a test of efficiency or optimality - the number of strikes under the regime. Measured by this last criterion conciliation boards performed well. Frequency of resort to the arbitrator is also a prediction that drops out of Treble's model, when coupled with further information about the possibilities for delay in the arbitration process. The terms of an implicit sliding scale linking coal prices and wage rates were the subject of negotiations. When coal prices rose, employers were able to delay passing on the increase in higher wages in Northumberland and South Wales and therefore there were frequent appeals to the arbitrator. By contrast the Federated Area, Scotland and Durham had few opportunities for delay and low rates of appeal to the arbitrator.

1.7 Money: the gold standard, monetary institutions and monetary policy Around 1873 prices and interest rates began a downward drift which was not halted until the mid 1890s. Then began a rise lasting until the outbreak of the First World War. The first period is often thought of as a depression because of the effects on arable farming, well represented in the House of Lords, but nowadays depression is generally thought to have been confined to monetary variables, the price level and interest rates, not real variables such as output and employment (Saul 1969). There is some evidence of a climacteric in economic growth but on statistical grounds it does not seem robust (Crafts et al. 1989). Although the bulk of money in circulation consisted of bank deposits the period is known as the age of the gold standard, for the Bank of England was legally obliged to exchange its notes for gold. Gold sovereigns, with a weight which defined the pound sterling, were much more widely used than notes. Bank deposits themselves could be exchanged for notes or coin and so the entire monetary edifice rested upon a fairly small stock of gold and widespread confidence that the public would not all at once demand gold instead of other forms of money. If confidence in the convertibility of deposits or notes into gold began to slip, the rising

24

New perspectives of the late Victorian economy

demand for gold was known as an 'internal drain'. The corresponding 'external drain' arose when sterling assets were exchanged for foreign assets and the sterling foreign exchange rate slid below the lower 'gold point'. That was the exchange rate at which it was profitable to sell sterling for gold and incur the insurance and transport costs necessary to ship the gold abroad for sale. In the light of the international turmoil between the world wars, the nineteenth-century gold standard regime has long been regarded as a success story, although the evidence points to the 1950-72 regime as more effective (Foreman-Peck and Michie 1986). The fundamental objective of gold standard monetary policy was to maintain convertibility and the principal policy instrument was the Bank Rate, the interest rate at which the Bank of England might be prepared to lend to selected banking institutions which were experiencing temporary difficulties in meeting the drain. Other institutions would also raise interest rates and the greater benefits to be gained from lending, together with the confidence imparted by the Bank's lending, was generally sufficient to restore the status quo. If other assets, such as building society deposits or trade bills, were close substitutes for money, then interest rates on these assets would also rise during a crisis. Contractions of the money supply could be offset by expansion of such 'near money'. The Bank of England's success in preventing external drains sufficient to deplete Bank gold stocks maintained Britain on a fixed exchange rate relative to other gold standard countries. Supporting the external value of the currency was made easier by the United Kingdom's net asset position and the deposits of foreign banks in London, (Collins 1988). On the other hand, measurement of the effective money supply, which should include these foreign deposits, is made more difficult, since their total volume is unknown. World gold stock growth slowed at the same time as there was a scramble for monetary gold, when Germany abandoned silver for gold in 1873 and the US prepared for the return to the gold standard in 1879. Early monetary explanations for the depression of prices often foundered on the parallel movement in prices and interest rates dubbed the Gibson paradox. It was thought that if prices were falling because of monetary scarcity then interest rates should have risen, not fallen. In order to examine this 'paradox' and to cast further light on the working of the late Victorian economy, Capie, Mills and Wood (CMW, chapter 9) construct a four equation model explaining the changes in output, the interest rate, the price level and the money stock in terms of lagged values of these variables. They consider two hypotheses. The first is that the late Victor-

Quantitative analysis of the Victorian economy

25

ian economy was a small country (with a fixed exchange rate) in the sense that prices and interest rates were determined by world conditions, and the British money stock was determined by these and the growth of domestic income. The second hypothesis is that the British economy was large enough to influence world conditions and therefore the growth of the domestic money supply influenced domestic output, prices and interest rates. CMW use an autoregressive system to examine these hypotheses. A descriptive account of such systems can be obtained from an analysis of the system's response to typical random shocks (Sims 1980). The residual in an estimated money autoregression equation has been labelled a 'money innovation' because it is new in the sense that it has not been predicted from past values of the variables. The 'innovation' in, say, money is assumed to disturb all other variables of the system instantly according to the strength of the contemporaneous correlation of other residuals with the money residual. The endogenous variable residual (innovation) is only allowed to affect the endogenous variable in the initial period. Both of the hypotheses give broadly similar results with this method; interest rates do not appear to affect other variables, prices affect interest rates substantially to begin with but as the period wears on, output and money come to be stronger influences upon interest rates. Money and output appear to influence prices, but prices have a weaker effect upon money and output. (For criticisms of Sims's econometric approach see Malinvaud 1988 and Pagan 1987.) On the basis of these innovation accounting decompositions and the coefficient estimates, CMW specify and reestimate two more restricted structural models, one for Mo, the money definition that includes only notes and coin, the other for M3, which in addition includes domestic bank deposits. Contrary to much traditional economic historiography of the period (for example Mathias 1983, p. 367), money did strongly influence prices. Interest rates did influence output. Interest rates responded to M o monetary shocks indirectly through the effects on output and prices. Monetary expansion did not lower interest rates initially unless the effect was over within one year. Irving Fisher's explanation for the Gibson paradox is borne out; that expectations of price declines made lenders willing to accept lower interest rates, since the purchasing power of their money, when it was returned, was enhanced. Turner (chapter 10) addresses the question, to what extent were near money assets substitutes for money in the late nineteenth century? Some of the interest in this question arises because the answer determines the

26

New perspectives of the late Victorian economy

possibilities open for monetary policy to control the economy. Late Victorian Britain was unusual among the more advanced economies of the time in not exhibiting a decline in the ratio of money to income. In the US, Canada, Norway and Sweden the fall in the money-income relationship was due to a shift towards industry, which was an intensive user of money, away from agriculture, which was not, as income grew (Bordo and Jonung 1981). Subsequently in Britain and these other countries the money-income ratio began to rise as money substitutes developed. In fact money substitutes in 1880 accounted for almost one fifth of Turner's definition of broad money (Bank deposits + notes and coin + Trustee Savings Bank (TSB) deposits + Post Office Savings Bank deposits + Building Society (BS) deposits). By 1913 these money substitutes had increased their percentage of the total by 4 per cent. National monetary institutions followed different courses often by historical accident and exercised differing influences upon national economic development. Because owners of the stock exchange building were not brokers in London but were in New York, the nineteenthcentury London stock exchange became a much larger and more influential institution than its New York counterpart (Michie 1987). New York aimed to restrict entry so as to maintain profit margins, London wanted more brokers to bid up rent. British bank deposits and Post Office deposits were growing more rapidly than the British average while TSB and BS deposits grew less rapidly between 1880 and 1913. Each component of near money was supplied by non-profit organisations, a condition which some might suggest would have restricted their growth relative to comparable profit organisations. In other words with a different institutional structure the rise in the money to income ratio might have begun earlier. Alternatively Britain's institutional structure may have been unusually favourable to branch banking, lacking the prohibitions on interstate banking of the United States. Turner finds non-bank financial intermediary (NBFI) deposits were close substitutes for money. Expansion of these NBFI deposits because of institutional innovation therefore carried very similar implications for gold flows as did money. A declining form of near-money, commercial bills, was supplied by private institutions. Turner finds these were not close substitutes for money defined as bank deposits plus notes and coin. He also tests Nishimura's (1971) hypothesis that the decline was a consequence of the spread of the telegraph which enhanced the effectiveness of bank deposits. Hefindsno significant relationship between the volume of telegrams sent and the volume of commercial bills, holding constant

Quantitative analysis of the Victorian economy the volume of world trade, the rate of interest and GNP. Rather, bills appear to be an 'inferior' asset; fewer were demanded at higher levels of incomes. Turner's use of cointegration techniques are explained in his appendix. The key idea is that if there is a stable long-run relationship between two or more variables then the residuals in their (cointegrating) regression must show no trend over time. When such a relationship exists it will be possible to describe the short-run links between the variables by an 'error correction mechanism' (ECM), an equation which explains changes in the dependent variable by means of changes in the independent variables, their lagged levels and lagged values of the level of the dependent variable. The lagged levels variables in the ECM serve to identify the long-run equilibrium relationship. An alternative estimation procedure therefore is to substitute for these variables the lagged residuals from the cointegrating regression. Although the rules of the central bank policy eventually hardened into dogma in the first half of the nineteenth century, the market never knew whether the Bank would follow Bagehot's later maxim and lend copiously in times of panic or internal drain'. Formally, the Bank of England was still a privately owned profit seeking institution. During the 1890s the Bank was competing aggressively for business against the banks that it was supposed to regulate. Yet the Bank of England's monopoly privileges were exchanged for some obligations which extended beyond managing the government's debt. Lovell (1957) has claimed in one of the earliest essays in econometric history that, on the basis of the positive correlation of lending and bankruptcies, the Bank during the eighteenth century began to act as a lender of the last resort in a way which would not be expected of a simple profit maximiser, even though there was no announced policy. As part of an investigation which employs both written primary sources from the Bank archives and a massive weekly data set, Ogden (chapter 11) finds a similar, but not statistically significant, relationship in the years between 1870 and 1914. In this instance she is obliged to use annual data to establish a positive association with bank failures rather than bankruptcies. She reports similarfindingsfor correlations with Bank rate. Her tests raise a question about the meaning of Lovell's correlation and an issue in econometric methodology about the proper treatment of autocorrelation. Autocorrelation in a time series regression equation is the correlation of successive residuals which violates the assumptions of the ordinary least squares statistical model, rendering suspect tests of sig-

27

28

New perspectives of the late Victorian economy

nificance on the estimated coefficients of the equation. Ogden removes the autocorrelation by a method due to Cochrane and Orcutt, and finds that the coefficients are no longer significantly different from zero. Another approach (Hendry and Mizon 1978) is to regard the presence of autocorrelation as a symptom that the equation is not correctly specified. Instead of employing the Cochrane-Orcutt transformation, the proper response, Hendry and Mizon maintain, is to respecify the equation by experimenting with the lag structure and/or introducing new variables. A statistically insignificant relationship in the first approach conceivably may become a significant one with the second. A much larger experiment Ogden conducts is the attempt to identify 'periods of tension', heavy discounting, which are not referred to in the laconic Board minutes. She fits polynomial time trends to data on discounts and identifies twenty-six points more than three standard deviations from the mean as periods of tension. Only five of these can be linked with previously recognised crises in 1873, 1878/9, 1890, 1907 and 1914. A support operation for the Yorkshire Penny Bank constitutes a sixth identifiable period of heavy discounting. Thanks to the rational expectations revolution in economics, the unwillingness to state the circumstances under which the Bank would intervene, or to announce when support had been given, that persisted to the days of Bagehot, may be given a rational explanation (Begg 1982). Commercial banks that could count on the Bank bailing them out when in difficulties would lose incentive to avoid becoming illiquid. Either by its publicly revealed behaviour, its announcements or lack of announcement, the Bank may have wished to convey uncertainty about whether or not it would intervene in a crisis, so as to encourage prudent banking. The City of Glasgow Bank was allowed to fail in 1878 because it was insolvent (asset values were less than liabilities) whereas the Clydesdale Bank in the same year, Barings in 1890 and the Yorkshire Penny Bank in 1911, were rescued because they were merely illiquid (they could not convert their assets into cash quickly enough).

1.8

Conclusion

US patent data point to a British economy locked into a mainly mature technological trajectory which served the country less well as the twentieth century progressed. The economy was heavily reliant upon railways, yet this was a sector with rather low productivity growth, certainly in comparison with American railways. In many respects this

Quantitative analysis of the Victorian economy

29

dependence mirrored that of exports, where cotton was king. Such concentration might indicate that gains from specialisation according to comparative advantage were being reaped, but it also left the economy vulnerable to changing technological and political circumstances, as events of the 1920s were to show. Until the shock of the First World War the economy continued to grow at rates which looked reasonable by historical standards. British firms demonstrated their competitiveness by establishing subsidiaries abroad in large numbers, not merely in protected Empire markets. Nonetheless, dissatisfaction with the workings of Victorian private enterprise in some sectors was sufficient to encourage creeping state ownership, in gas and water for instance. New cost of living data, when linked with a revised wage series, allow the Edwardian period to be reinterpreted as years of growth of real income, rather than decline. Conditions diverged markedly between the industrial North and the rural South (and London was quite different from both). A significant proportion of the agricultural workforce were on Poor Relief in the 1860s because of the seasonal nature of their work. But only during cyclical crises were even small numbers of industrial workers able to exercise any claim on the social security system. Wages were now more rarely set by reference to custom; trade unions increasingly participated in the process. In the coal idnustry, the potentially indeterminate bargain between union and employers' association was rather efficiently resolved by the mechanism of arbitration boards, a neglected area of Victorian institutional success. By contrast with the later twentieth century, the late Victorian economy was untroubled by inflation. Only at the end of the century did prices begin a sedate rise. The domestic money supply did influence the price level though, not merely the balance of payments. The link of the currency with gold, or the accompanying ideology of small, balanced government budgets, underpinned the stability of monetary growth and therefore also the stability of the price level. Most financial transactions were conducted by means of assets other than gold. Continuing financial innovation (rather than the spread of the telegraph) ensured that traditional assets, especially the inland bill, were replaced by newer instruments, such as bank deposits. Control of the monetary sector was, despite this innovation, still vested in a fairly small group of people. That meant that an influential governor at the Bank of England could, by moral suasion, implement monetary policy at a lower cost than when Bank decisions were taken by less powerful men.

30

New perspectives

of the late Victorian

economy

References Baines, D. (1985) Migration in a Mature Economy: Emigration and Internal Migration in England and Wales 1861-1900, Cambridge University Press. Barnett, C. (1986) The Audit of War: the Illusion and Reality of Britain as a Great Nation, London: Macmillan. Bartrip, P. W. J. and Fenn, P. T. (1988) 'Factory fatalities and regulation in Britain 1878-1913', Explorations in Economic History 25: 60-74. Baumol, W. J. (1982) 'Contestability: an uprising in the theory of industrial structure', American Economic Review, 72: 1-15. Bean, R. and Peel, D. A. (1976) 'Business Activity, labour organisation and industrial disputes in the United Kingdom, 1892-1938', Business History, 18: 205-11. Begg, D. K. H. (1982 ) The Rational Expectations Revolution in Macroeconomics: Theories and Evidence, Oxford: Philip Allan. Booth, C. (1902) (ed.) Life and Labour of People in London, London: Macmillan. Bordo, M. D. and Jonung, L. (1981) T h e long run behaviour of the income velocity of circulation of money in five advanced countries, 1870-1975' Economic Inquiry, 19: 96-116. Capie, F. and Webber, A. (1985) A Monetary History of the United Kingdom 1870-1982, London: Allen and Unwin. Chandler, A. D. (1980) T h e growth of the transnational firm in the United States and the United Kingdom: a comparative analysis', Economic History Review, 33: 396-410. Church, R. A. (1975) The Mid-Victorian Boom 1850-73, London: Macmillan. Collins, M. (1988) Money and Banking in the United Kingdom: A History, London: Croom Helm. Crafts, N. F. R. (1984) A time series study of fertility in England and Wales 1877-1938', Journal of European Economic History, 13: 571-90. (1985) British Economic Growth During the Industrial Revolution, Oxford: Clarendon Press. Crafts, N. F. R. and Thomas, M. (1986) 'Comparative advantage in UK manufacturing trade, 1910-1935', Economic Journal, 96: 629^45. Crafts, N. F. R., Leybourne, S. J. and Mills, T. C. (1989) T h e climacteric in late Victorian Britain and France: a reappraisal of the evidence', Journal of Applied Econometrics, 4: 103-17. Dangerfield, G. (1966) The Strange Death of Liberal England, London: Macgibbon and Kee. Daunton, M. J. (1985) Royal Mail. The British Post Office Since 1840, London: Athlone Press. Davenport-Hines, R. P. T. (1984) Dudley Docker: The Life and Times of a Trade Warrior, Cambridge University Press. Davis, L. E. and Huttenback, R. (1986) Mammon and the Pursuit of Empire: the Political Economy of British Imperialism 1860-1912, Cambridge University Press.

Quantitative analysis of the Victorian economy

31

Deane, P. and Cole, W. A. (1962) British Economic Growth 1688-1959: Trends and Structure, Cambridge University Press. Dosi, G. (1984) Technical Change and Industrial Transformation, London: Macmillan. Elbaum, E. and Lazonick, W. (1986) The Decline of the British Economy, Oxford University Press. Feinstein, C. H. (1965) Domestic Capital Formation in the United Kingdom 1920-38, Cambridge University Press. (1972) National Income, Expenditure and Output in the United Kingdom, 18551965, Cambridge University Press. (1990a) 'What really happened to real wages? Trends in wages, prices and productivity in the United Kingdom, 1880-1913', Economic History Review, 43. (1990b) 'New estimates of average earnings in the United Kingdom, 1880-1913' Economic History Review, 43. Floud, R. and Wachter, K. W. (1982) 'Poverty and physical structure: evidence on the standard of living of London boys 1770-1870', Social Science History, 6. Fogel, R. W. (1964) Railroads and American Economic Growth, Baltimore: Johns Hopkins University Press. Foreman-Peck, J. S. (1983) 'Diversification and the growth of the firm: the Rover Company to 1914', Business History, 25: 179-92. (1987) 'Natural monopoly and railway policy in the nineteenth century' Oxford Economic Papers, 29: 699-718. (1989a) 'Foreign investment and imperial exploitation: balance of payments reconstruction for Britain and India 1815-1914', Economic History Review, 42: 354-74. (1989b) 'Competition, cooperation and nationalisation in the nineteenth century telegraph system', Business History, 312: 81-101. (1989c) 'Privatization of industry in historical perspective', Journal of Law and Society, 16: 129-48. Foreman-Peck, J. S. and Michie, R. (1986) 'The performance of the nineteenthcentury gold standard' in W. Fischer, R. Marvin Mclnnis and J. Schneider (eds.) The Emergence of a World Economy 1500-1914, Pt II. London: Franz Steiner. Friedman, M. (1962) Capitalism and Freedom, University of Chicago Press. Friedman, M. and Schwartz, A. (1963) A Monetary History of the United States, Princeton University Press. Greasley, D. (1986) 'British economic growth: the paradox of the 1880s and the timing of the climacteric', Explorations in Economic History, 23: 416-44. Haines, M. (1985) 'Inequality and childhood mortality: a comparison of England and Wales 1911 and the United States 1900', Journal of Economic History, 45: 885-912. Hannah, L. (1983) The Rise of the Corporate Economy, London: Macmillan.

32

New perspectives of the late Victorian economy

Harley, C. Knick (1980) Transportation, the world wheat trade, and the Kuznets Cycle, 1850-1913', Explorations in Economic History, 17: 218-250. (1988) 'Ocean freight rates and productivity 1740-1913: the primacy of mechanical invention confirmed', Journal of Economic History, 48: 813-50. Hatton, T. J. (1988)'Institutional change and wage rigidity in the UK 1880-1985', Oxford Review of Economic Policy, 4, 1: 74-86. Hawke, G. R. (1970) Railways and Economic Growth in England and Wales 1840-1870, Oxford University Press. Hendry, D. and Mizon, G. (1978) 'Serial correlation as a convenient simplification, not a nuisance', Economic Journal, 88: 549-63. History Today (1985) 'What is economic history?', 35: 35-43. Humphries, J. (1987) \ .. Most free from objection... the sexual division of labour and women's work in nineteenth-century England', Journal of Economic History, 47: 929-49. Irving, R. J. (1975) 'New industries for old: some investment decisions of Armstrong-Whitworth 1900-1914', Business History, 17: 150-75. Jackson, R. R. (1987) 'The structure of pay in nineteenth-century Britain', Economic History Review, 40: 561-70. Johnson, P. (1984) 'Self-help versus state help: old age pensions and personal savings in Great Britain 1906-1937', Explorations in Economic History, 21: 329-50. (1985) Savings and Spending: the Working-Class Economy in Britain 1870-1939, Oxford: Clarendon Press. Kennedy, W. P. (1987) Industrial Structure, Capital Markets, and the Origins of British Economic Decline, Cambridge University Press. Kuznets, S. S. (1966) Modern Economic Growth: Rate, Structure and Spread, New Haven: Yale University Press. Lee, C. H. (1986) The British Economy since 1700: A Macroeconomic Perspective, Cambridge University Press. Lewchuk, W. (1987) American Technology and the British Vehicle Industry, Cambridge University Press. Lewis, W. A. (1978) Growth and Fluctuations 1870-1913, London: Allen and Unwin. Liebenau, J. (1984) 'Industrial R&D in pharmaceutical firms in the early twentieth century', Business History, 26: 329-46. Lindert, P. H. and Trace, K. (1971) 'Yardsticks for Victorian entrepreneurs' in D. N. McCloskey (ed.) Essays on a Mature Economy: Britain after 1840, London: Methuen. Lindert, P. H. and Williamson, J. G. (1982) 'Revising England's social tables, Explorations in Economic History, 19: 358-408. Lloyd-Jones, R. (1987) 'Innovation industrial structure and the long-wave: the British economy c. 1873-1914', Journal of European Economic History, 126: 315-37. Lovell, M. C. (1957) 'The role of the Bank of England as lender of the last resort in

Quantitative analysis of the Victorian economy

33

the crises of the eighteenth century', Explorations in Entrepreneurial History, 10:8-21. McCloskey, D. N. (1981) 'Did Victorian Britain fail?' in Trade and Enterprise in Victorian Britain: Essays in Historical Economics, London: Allen and Unwin. Maddison, A. (1982) Phases of Capitalist Development, Oxford University Press. Malinvaud, E. (1988) 'Observation in macroeconomic model building', Presidential Address European Economic Association, Bologna. Mathias, P. (1983) The First Industrial Nation, second edition, London: Methuen. Matthews, R. C. O., Feinstein, C. H. and Odling-Smee, J. C. (1982) British Economic Growth 1865-1973, Oxford: Clarendon Press. Mearns, A. (1883) The Bitter Cry of Outcast London, London. Michie, R. C. (1987) The London and New York Stock Exchanges, 1850-1914, London: Allen and Unwin. Millward, R. (1988) 'The UK services sector, productivity change and the recession in long-term perspective', Services Industry Journal, 8: 263-76. Mitch, D. F. (1986) 'The impact of subsidies to elementary schooling on enrolment rates in nineteenth-century England', Economic History Review, 39:371-91. New Survey of London Life and Labour (1930) Forty Years of Change, vol. 1, London: P. S. King. Nicholas, S. J. (1982) 'Total factor productivity growth and the revision of post-1870 British economic history', Economic History Review, 35: 83-98. (1985) 'British economic performance and total factor productivity growth 1870-1914', Economic History Review, 38: 576-82. Nishimura, S. (1971) The Decline of Inland Bills of Exchange in the London Money Market 1855-1913, Cambridge University Press. O'Brien, P. K. and Keyder, C. (1978) Economic Growth in Britain and France 1780-1914: Two Paths to the Twentieth Century, London: Allen and Unwin. Olson, M. (1982) The Rise and Decline of Nations, New Haven: Yale University Press. Pagan, A. (1987) 'Three econometric methodologies: a critical appraisal' Journal of Econometric Surveys, 1: 3-24. Pencavel, J. H. (1976) 'The distributional and efficiency effects of trade unions in Britain', British Journal of Industrial Relations, 15: 137-56. Phelps Brown, H. (1983) The Origins of Trade Union Power, Oxford: Clarendon Press. Pocock, R. F. (1988) The Early British Radio Industry, Manchester University Press. Pollard, S. (1985) 'Capital exports 1870-1914: harmful or beneficial', Economic History Review, 38: 489-514. (1989) Britain's Prime and Britain's Decline: the British Economy 1870-1914, London: Edward Arnold. Prest, A. R. (1954) Consumer's Expenditure in the United Kingdom, 1900-1919, Cambridge University Press.

34

New perspectives of the late Victorian economy

Rowntree, B. Seebohm (1902) Poverty: a Study of Town Life, third edition, London: Macmillan. Rubinstein, W. D. (1988) 'Social class, social attitudes and British business life', Oxford Review of Economic Policy, 4, 1: 51-8. Sanderson, M. (1988) Technical education and economic decline: 1890-1980s, Oxford Review of Economic Policy, 4, 1: 38-50. Saul, S. B. (1969) The Myth of theGreat Depression 1873-1896, London: Macmillan. Saxonhouse, G. R. and Wright, G. (1984) 'New evidence on the stubborn English mules and the cotton industry 1878-1920', Economic History Review, 37: 507-19. Sims, C. A. (1972) 'Money, income and causality', American Economic Review, 62, 4: 540-52. (1980) 'Macroeconomics and reality', Econometrica, 48: 1-48. Soderberg, J. (1985) 'Regional economic disparity and dynamics 1840-1914: a comparison between France, Great Britain, Prussia and Sweden', Journal of European Economic History, 14: 273-98. Solomou, S. (1988) Phases of Economic Growth 1850-1973: Kondratieff Waves and Kuznets Swings, Cambridge University Press. Soltow, L. (1968) 'Long run changes in British income inequality', Economic History Review, 21: 17-29. (1971) 'An index of the poor and rich of Scotland 1861-1961', Scottish Journal of Political Economy, 18: 17-29. Southall, H. (1986) 'Regional unemployment patterns among skilled engineers in Britain', Journal of Historical Geography, 12: 268-86. (1988) 'The origins of the depressed areas: unemployment, growth and regional economic structure in Britain before 1914', Economic History Review, 41: 236-58. Stead, W. T. (1901) The Americanisation of the World, London: Review of Reviews. Svedberg, P. (1981) 'Colonial enforcement of foreign direct investment', Manchester School, 39: 21-38. Telser, L. G. (1987) A Theory of Efficient Cooperation and Competition, Cambridge University Press. Thomas, M. (1985) 'Accounting for growth 1870-1914: Stephen Nicholas and total factor productivity measurement', Economic History Review, 38: 560-75. Wiener, M. J. (1981) English Culture and the Decline of the Industrial Spirit 1850-1980, Cambridge University Press. Williams, E. E. (1896) Made in Germany, London: Heinemann. Williamson, J. G. (1982) 'The structure of pay in Britain 1710-1911' in P. Uselding (ed.) Research in Economic History, 7: 1-54. (1984) 'British Mortality and the Value of Life 1781-1931', Population Studies, 38: 157-72. (1985) Did British Capitalism Breed Inequality? London: Allen and Unwin. Wilson, J. F. (1988) Ferranti and the British Electrical Industry, 1864-1930, Manchester University Press.

PARTI

Technology and Industrial Organisation

The idea that 'where you start from determines where you end up' has had considerable currency in theories of technological change. An economy with resources concentrated in one particular industry will gain experience which will allow further improvements in the technology of that industry. In terms of figure I.I ray OA might be the technology for chemicals and ray OB the technology for industrial engines. As drawn chemicals are more capital intensive than industrial engines. For a given state of technical knowledge, points on the rays closer to the origin O, using fewer inputs of capital and labour, produce lower outputs. Technical progress shifts towards O at least one point on the constant output curve joining CD. A technological trajectory (chapter 2) implies that movements down one ray will be easier or faster than movements down another. An interpretation of the problem of measuring the impact of railways is a shift outwards along one ray or even the elimination of one ray (chapter 3). Railways might be ray OA and competing transport modes, OB. At point E railways produce more transport with less need for both labour (F) and capital (G) than the other modes. Abolishing railways would move the economy to D with greater demands for labour (H) and capital (J) to supply the same volume of transport. Capital

0

H

Figure 1.1 Choice of technique 35

Labour

36

New perspectives on the late Victorian economy Costs C

0

A

D

Distance

Figure 1.2 Spatial costs and competition

Competition between firms in space, such as are the concerns of chapters 4 and 5, can be represented in figure 1.2. A is a gas company, the production cost of whose gas is given by the AB. Distribution costs rise with distance from the point of production (BC). How far it is worth extending the distribution network depends upon whether another firm, D, is already supplying gas, or may enter the market at that point. If D does enter, then at all points to the right of E, firm D's costs will be less than firm A's. Since firm A does not know where D will enter the market, A is likely to avoid substantial investment in a large network, because the larger the network, the higher the chance that some portion of it may be rendered redundant by later entry. Where setting up a subsidiary overseas is concerned, company headquarters will prefer lower cost location A to D, other things being equal. With the same distribution costs, A will be able to compete in a larger market, at a given world price. But even though an existing plant D is less efficient than A, it will still undercut A's products in its own market unless A establishes a subsidiary next to D. A measure of the vitality of an economy's industry is the willingness and ability to set up profitable plants like A in foreign locations close to D.

Chapter 2

Historical trends in international patterns of technological innovation John Cantwell

2.1 Introduction It is frequently argued by economic historians that Britain's technological decline relative to countries which industrialised later can be dated to the Victorian period (Habakkuk 1962, Landes 1969).1 One factor which has been suggested to explain this relative decline is the disadvantages which were due to British industry's early start, making a switch to newer methods of production more costly (a theme discussed by Jervis 1947, Frankel 1955, Saville 1961, Kindleberger 1961, Habakkuk 1962, Ames and Rosenberg 1963, Aldcroft and Richardson 1969, amongst others). Generally underlying the early start thesis is the belief that production methods in different parts of an industry or an economy are interrelated, such that it is costly to change one component of a production process or the methods prevailing in an individual sector without complementary changes elsewhere. This idea can be allied to an argument which has become popular more recently in the literature on the economics of technological change. The proposition has been advanced that technological innovation proceeds as a cumulative process, and that as a consequence it tends to iock in' to a particular course, once that course has become established. This proposition has been seen as relevant in a variety of contexts. One suggestion is that where there is competition between rival technologies for adoption it is possible that adopting firms become 'locked in' to a potentially less efficient technology following the choice of the earliest adopters (Arthur 1989). The decisions of the earliest adopters affect the decisions of the firms that come later. One reason is the role of technological interrelatedness, whereby a supporting infrastructure of complementary technologies become better developed, lowering the costs and increasing the benefits of adopting a technology which is already more widely diffused. Other reasons for cumulative gains from adoption are the improvements made through learning by using; the network externalities that are created by a growth in the network of users such that each benefits from a 37

38

New perspectives on the late Victorian economy

greater availability and variety; and the presence of increasing returns in information about the technology as it diffuses. These factors may combine to set in motion a process of cumulative causation in the adoption of one type of technology ahead of another. Despite the technical advantages possessed by alternative systems, the petrol driven motor vehicle won out over the steam driven, and the VHS video recorder is gradually displacing the Sony Betamax (Arthur 1989). David (1985) refers to the locking in of technological development to a particular standard model or design, using the example of the QWERTY keyboard used in typewriters and word processors. These are all illustrations of cumulative or path-dependent processes in technological change, even if their origins lie in small historical events in which chance plays a role. They are path-dependent in the sense that the probabilities (of adoption of a certain kind) are influenced by past decisions which constrain the limits of existing choices, as portrayed in statistical theory by a Markov process or a Polya process (Arthur, Ermoliev and Kaniovski 1987). The choices available may vary with location, according to differences in the past experience of thefirmsbased in each country or area. From this it can be seen that the firms of a given country may find themselves 'locked in' to a path of technological change which ultimately offers them less opportunities than are available to the firms of other countries who have begun along alternative paths, but after a time it is very costly and difficult for them to switch. The proposition that technological change is cumulative is therefore relevant to explaining the historical evidence of differential rates of innovation across countries or national groups of firms. It suggests that innovation is liable to 'lock in' to a particular industrial pattern or configuration in any location, and that this pattern is likely to change only gradually over time since a shift towards sectors in which technological opportunities are rising most rapidly may not be easy to achieve. It is in this context that the proposition of cumulativeness is examined in this paper. The proposition of cumulativeness is examined using historical evidence, by testing the extent of continuity in the industrial pattern of innovation. The issue is whether the sectoral composition of technological activity 'locks in' to a particular course for each national group of firms, and if so over what periods of time this tends to persist before shifts in the pattern of innovation become apparent. This is related to the theory of technological accumulation, which has been most clearly articulated by Pavitt (1987), although similar theories which build upon the idea of

Historicaltrendsininternationalpatternsoftechnologicalinnovation

39

cumulativeness in technological change can be found in other recent literature (most notably Rosenberg 1982, Nelson and Winter 1982 and Dosi 1984). Pavitt argued that technology isfirm-specific,cumulative and differentiated, and consequently that the industrial composition of innovative activity in a given location or amongst a given national group of firms reflects past technological accumulation. This suggests that international patterns of technological advantage, having been established, will remain relatively stable over time at least in the short or medium term. The sectors in which each group offirmsis technologically strongest changes only gradually. In the theory of technological accumulation, the view that innovation is cumulative is just one of three connected but separate propositions. The others are that innovation proceeds incrementally and that it is differentiated between firms and locations as described below. It is the iock in' effect associated with cumulativeness, however, which suggests the hypothesis that the sectoral composition of innovation amongst thefirmsof an industrialised country is by and large stable over periods often or twenty years. This hypothesis is statistically tested here, against the alternative hypothesis that technological change follows a random course, in which the relative strength of innovative activity is likely to regularly switch between industries. The alternative hypothesis would suggest that there are no substantial problems of interrelatedness in switching between different types of technology. In association with these statistical tests, changes over time in the degree of technological specialisation of each national group offirmsare examined using a related statistical procedure. The degree of technological specialisation is a measure of whether the innovation of the firms in question is highly concentrated in a few industries, or more broadly spread across a wider range. It should be noted that the proposition under investigation here refers to the cumulativeness of technological development in terms of the industrial composition of innovation, rather than in terms of its overall rate or rapidity for each national group of firms. The question of why German firms on average enjoyed a faster rate of innovation than British firms at the turn of the twentieth century is only addressed indirectly. The principal issue is the nature of the comparative advantage held by each national group of firms in technology creation, and the stability of that pattern of comparative advantage over time. This may then, as has been argued above, have some bearing on which national group innovates most rapidly if US and German firms historically and Japanese firms today have a comparative advantage in sectors in which technological

40

New perspectives on the late Victorian economy

activity is growing fastest, and each group is 'locked in' to its own particular pattern of comparative advantage. This supplementary issue is also examined in what follows. Which industries offer the greatest technological opportunities is conditioned by the prevailing techno-economic paradigm that characterises innovation in each historical period (Freeman and Perez 1988). The second proposition of the theory of technological accumulation is that innovation develops incrementally, so that firms tend gradually to move between related sectors. Although the underlying technology and skills continue to build upon the past, the industrial applications may gradually change, a particularly extreme case of which may come about with the formation of new industries. This proposition calls for further extensive historical research, and so it is not explored in any depth here, but it is simply noted that it is in line with thefindingsof Rosenberg (1976 and 1982). However, the proposition implies that the industrial composition of innovation amongst each national group of firms may shift over longer historical periods. Statistical tests are conducted to show for which groups of firms this happens and the extent to which it happens in each case. In addition, descriptive evidence is presented on how the sectoral pattern of innovative activity of particular national groups of firms has actually evolved, paying particular attention to historical evidence on the technological development of British firms. The final proposition of the theory of technological accumulation is that innovation is differentiated between firms and locations. That is, the path of technological development followed by a particular firm or in a particular location is distinctive and characterised by elements that are specific to that firm or location. This proposition itself is not examined here, as to do so thoroughly would require intra-industry and microeconomic evidence which lies beyond the scope of the chapter. It is simply noted that even where British firms performed relatively well in innovation, the characteristics of their technological development may well have been different from those of their American or German competitors. Returning, then, to the central proposition of cumulativeness, it is implied that day-to-day adaptation of technology, through an interaction between its creation within a firm and its use in production, has a more pervasive influence than the major technological breakthroughs which give rise to entirely new production processes. Even radically new technologies, once they move beyond the purely scientific and experimental stage, often rely upon or are integrated with earlier technologies in the

Historicaltrends in internationalpatterns oftechnologicalinnovation

41

course of their development (Usher 1954). For this reason, innovation tends to gather a certain logic of its own through the continual refinement and extension of established technologies. As specific technological experience is accumulated, the further development of production within the firm throws up new requirements, which its research and engineering departments must try and meet. Improvements tend to set the stage for their own future problems, which compel further modification and revision through the adaption of production by innovative firms. Until there is a new stream of innovations based on a different set of fundamental discoveries, firms at the existing frontier of progress tend to establish dynamic advantages over others in the same industries. This helps to explain why, for example, German firms in the chemical industry have maintained a strong tradition for a period of at least a hundred years. It may also explain why, despite their failure to move as fast as others into the newer science-based sectors, Britishfirmsat the turn of the century remained dominant in technological development in textile machinery, railway engines and shipbuilding (Walker 1980). They were locked in to areas of innovation which had once ensured the success of British industry, but at a time when technological opportunities had begun to rise more rapidly elsewhere. In any given industry, firms based in certain leading countries where a tradition has been established tend to push forward with a sequence of innovations conditioned by the prevailing technological opportunities. These firms are geared up to problem-solving activity (increasingly through corporate R&D in the twentieth century) and production engineering in areas of technology in which they have accumulated a wealth of practical experience. There is a strong interaction between those responsible for production, who are utilising the latest processes available as they become feasible, and engineering and research support staff. Such firms are capable of generating strong technological advantages, in part through the assimilation of the relevant features of complementary foreign technologies. However, the patterns of technological advantage across national groups offirmscan also be expected to change gradually or incrementally over time. In order to examine statistically the combined significance of cumulativeness and incremental change in innovation, an index of technological advantage was calculated for selected years spanning a period stretching back into the last century. The index was drawn up for each of the major industrialised countries. For a cross section of industries, the index measures the strength of innovative activity of the firms of each

42

New perspectives on the late Victorian economy

country, and it is constructed for the period 1963-83, and for selected years before 1914. It is an index of 'revealed technological advantage', and is calculated in much the same way as the index of 'revealed comparative advantage', familiar from the literature on international trade (see Balassa 1965). In this case the index measures comparative advantage in innovative activity rather than comparative advantage in trade. For the industrial sector of any country, its revealed technological advantage (RTA) is given by the country's share of US patents taken out by foreigners in that sector, divided by its total share of US patents due to non-US residents. Hence, when the RTA index assumes a value greater than one the country concerned is relatively advantaged in that sector, while a number less than one indicates that its firms are relatively disadvantaged. Such an index was first used by Soete (1980). The revealed technological advantage in industry i for the firms of country j is thus defined as RTA;, = (PijttjPijyGiPi/ZZjPtj) where Ptj is the number of US patents in industry / granted to residents of country/

2.2 A description of the data The suitability of patent data as a measure of technological advantage is now quite well documented (for a review of the literature see Pavitt 1985). While it is true that some innovations are never patented, and that some patents either have little qualitative impact or are never used, this leads principally to systematic industry-specific and countryspecific differences, as it seems that firms from the same sector in any country have a similar propensity to patent. Scherer (1983), using US patent data, found that most of the variation between firms in the propensity to patent was to be explained by the extent of their research effort (as measured by R&D expenditure). Once interindustry differences are allowed for, patenting as a measure of innovative output is strongly correlated across firms with a widely used measure of innovative input. Allowing that for thefirmsof a given country certain interfirm intrasectoral differences in the propensity to patent do exist, it seems reasonable to conclude that their variance is systematically lower than intersectoral differences. It can then be hypothesised that on relatively large numbers, the propensity to patent of a given national group of firms cannot be

Historicaltrends in internationalpatterns oftechnologicalinnovation

43

expected to have any systematic bias as compared with a notional industry average. A similar formulation could be applied equally well in the case of intranational group differences in the propensity to patent in a given industry. Here, however, this additional assumption is not required as the RTA index is examined separately for the firms of each country, as will become clear. The RTA index is normalised for both interindustry and intercountry differences in the propensity to patent. The use of foreign patent data is further supported by the findings of Soete and Wyatt (1983), that there is a strong intercountry correlation between foreign patenting and R&D, and a strong interindustry correlation between foreign and domestic patenting. Moreover, the USA as a host to foreign patenting represents an important market for firms from the countries under comparison, so that they regularly take out patents there. The USA grants a higher number of patents to non-residents than any other country. A highly significant correlation between national shares of R&D and patenting in the USA was observed by Pavitt (1982). Data on foreign patenting in the US, organised by industry and country of origin, has been compiled by the Office of Technology Assessment and Forecast (OTAF) for years from 1963 onwards. Before 1963, the US Index of Patents provides a list of all patents granted in the USA by alphabetical order of the patentees, showing their state or country of origin. A brief description of each patent follows, with a patent number, such that if this is inadequate the complete description can be consulted in a separate catalogue. Using this source, US patents granted to foreigners can be allocated to the relevant countries and industries, just as OTAF has done for recent years. For this paper, patent counts were made for the years 1890-2 and 1910-12. Table 2.1 indicates the importance of patenting in the USA to a number of technologically advanced European firms that had interests in the USA before 1914. It has to be admitted that the use of foreign patenting activity in the USA before 1914 in the assessment of technological advantage is likely to be much less reliable than it is when working only with recent data. For a start, the assumption that for the firms of a given country the propensity to patent innovations can be treated over large numbers of observations as varying only with industry-specific factors is much more questionable. Moreover, it cannot be assumed that before 1914 there was the same interindustry correlation between foreign and domestic patenting that exists today. The importance of international activity, and of the USA as a market and a source of competition, may well have varied between firms

44

New perspectives on the late Victorian economy

Table 2.1

A list of 30 leading European companies patenting in the US before 1914

Chemicals Badische Anilin & Soda Fabrik (Germany)** Bauer & Co. (Germany) Bayer (Germany)** Burroughs & Wellcome (UK) Cassella & Co. (Germany)* Deutsche Gold & Silber-ScheideAnsalt (Germany)* Geigy & Co. (Switzerland)* Heyden (Germany)* Kalle & Co. (Germany)* Lever Brothers (UK) Merck (Germany)* Nobel (France) Society of the Chemical Industry (Switzerland)* Solvay (Belgium)* United Alkali Co. (UK)*

Electrical equipment Bosch (Germany) Howard & Bullough (UK)* Marconi (UK)* Orenstein (Germany) Siemens & Halske (Germany)* Smidth & Co. (Denmark) Motor Vehicles Daimler Motor Co. (UK)* Fiat (Italy) Rolls (UK) Rubber tyres Dunlop (Ireland/UK)* Michelin (France)* Textiles Courtaulds & Co. (UK) Linen Thread Co. (UK)

Metals Deutsche Metallpatronen Fabrik (Germany) Metallurgische Gesellschaft (Germany) Notes:

Denotes granted over 10 patents. ** Denotes granted over 100 patents. Source: US Index of Patents, various issues.

in the same industries. Indeed, individual inventors, with no formal affiliation to any firm, played a more prominent role at that time. For these reasons, historical comparisons must be treated with caution. However, it would be unfortunate if a valuable source of evidence on historical patterns of technological advantage were to be entirely ignored for the purposes of statistical analysis. The RTA index can be conveniently compared over long periods of time, providing it is recognised that it is likely to change due to an improved representation of innovative activity, as well as due to actual changes in the sectoral pattern of technological advantage. Dutton (1984) provides evidence that patenting in the nineteenth century was linked to economic activity, since even though individual inventors played a greater role they frequently had links with industry. Table 2.2 sets out the numbers of US patents granted to residents of the

Historical trends in internationalpatterns oftechnological innovation Table 2.2

45

The total number of US patents granted to residents of the major countries of origin

Country of origin

1890-2

%

1910-12

%

1963-83

%

USA

66,766 6,084 2,145 1,378

91.6

95,022 12,285 2,970 3,961 1,673 1,031

88.6 11.4

929,133 416,113 55,028 101,864 22,160 38,956 4,560 4,072 23,733 14,621 5,125

69.1 30.9

Non-US total

UK Germany Canada France Austria (-Hungary) Australia Switzerland Sweden Belgium and Luxembourg Ireland Italy Denmark Netherlands Spain Japan

975 548 198 147 139 101 54 44 31 22 19 17 6

8.4 2.9 1.9 1.3 0.8 0.3 0.2 0.2 0.1 0.1 0.1 0.0 0.0 0.0 0.0 0.0

439 284 310 318 149 37 175 94 56 35 34

2.8 3.7 1.6 1.0 0.4 0.3 0.3 0.3 0.1 0.0 0.2 0.1 0.1 0.0 0.0

309 13,299 2,760 12,317 1,316 94,046

4.1 7.6 1.6 2.9 0.3 0.3 1.8 1.1 0.4 0.0 1.0 0.2 0.9 0.1 7.0

Source: Cantwell (1989b).

major countries of origin in 1890-2, 1910-12 and 1963-83. The ten foreign countries (counting Belgium and Luxembourg together) whose firms were granted the most patents before 1914 are today still all amongst the top thirteen. They are the UK, Germany, Canada, France, Austria, Switzerland, Australia, Sweden, Italy and Belgium and Luxembourg. They have been joined by Japan and the Netherlands, as shown, and by the USSR which is excluded from the analysis as it raises a quite different set of institutional considerations. When considering the most recent period it is also feasible to include the US (for which RTA is calculated as US firms' share of all patents granted, rather than all patents granted to foreigners) and three other EC countries (Denmark, Ireland and Spain). Portugal and Greece are excluded as they have not attained very high levels of patenting in the USA even in recent years. It should be made clear that for the recent period (1963-83), the cross-industry variation in the RTA index for any country predominantly reflects the structure of innovative activity amongst national groups of firms (especially the largest firms including multinationals), rather than that country's domestic industrial structure. This is because in the data on patent counts used for 1963-83, patents were in general registered by and thus attributed to parent companies, irrespective of the location of

Table 2.3

Indices of revealed technological advantage for the major industrialised countries in the periods (i) 1890-1912 and (ii) 1963-1983 UK

Sector

(i)

Food products 0.63 Chemicals, n.e.s. 0.70 Synthetic resins 1.03 Agricultural chemicals 1.21 Cleaning agents 1.47 Paints, etc. 0.17 Pharmaceuticals 0.72 Ferrous metals 1.11 Non-ferrous metals 1.05 Fabricated metal products 0.87 Mechanical engineering, n.e.s. 1.04 Agricultural machinery 0.43 Construction equipment 0.90 Industrial engines 1.76 Electrical equipment, n.e.s. 1.25 Transmission equipment 1.02 Lighting and Wiring 0.90 Radio and TV receivers 1.21 Motor vehicles 1.09 Shipbuilding 1.53 Transportation equip, n.e.s. 1.05 Textiles and clothing 1.34 Rubber products 1.75 Non-metallic mineral products 0.97 Coal and petroleum products Professional instruments Other manufacturing

Germany (ii)

Canada (ii)

France (ii)

Austria

Australia (ii)

Switzerland (ii)

(>)

00

(i)

0.75 0.99 0.77 1.18 1.22 0.65 1.20 0.93 0.93

1.37 0.83 0.00 1.17 1.32 0.09 1.55 1.14 0.76

0.74 0.50 0.48 0.62 0.40 0.92 0.70 2.58 2.00

0.85 0.43 0.00 0.00 0.69 0.00 0.90 1.99 0.40

1.80 0.56 0.55 0.98 0.64 0.99 0.60 1.54 1.52

1.90 1.01 7.87 0.00 1.29 1.78 1.41 0.59 0.74

1.36 1.85 0.78 2.05 1.31 0.93 1.55 0.61 0.73

0.78

1.10

0.97

1.26

1.12

1.66

0.58

0.85

0.96 2.19

1.05 0.34

0.93 1.01

1.09 0.81

1.25 1.20

0.98 3.14

0.99 3.62

1.12 1.18

1.01 1.37

1.18 0.53

1.70 0.70

0.20 0.99

1.06 1.00

1.64 0.67

2.41 0.76

1.47 0.82

1.51 0.93

0.34 0.99

0.85 0.59

0.86

0.95

0.74

1.29

1.11

1.49

0.63

0.64

0.53

0.97

0.68

0.97 1.38 1.72 0.70 0.74

0.90 0.94 0.62 1.20 0.64

0.69 0.40 0.26 1.56 0.72

1.18 1.78 0.60 1.04 2.19

1.29 1.40 0.00 1.35 0.99

1.17 1.13 0.63 1.15 1.29

0.00 1.81 0.00 0.56 0.00

0.47 0.35 0.71 0.80 0.12

1.28 0.53 0.00 1.15 1.11

0.51 0.97 0.28 0.93 1.90

1.60 0.50 0.00 0.89 0.00

0.91 0.64 0.34 0.65 0.47

1.44 1.17 1.03

0.62 0.85 0.62

1.07 1.18 1.06

1.53 0.90 0.60

1.00 0.48 0.93

1.16 0.94 1.21

1.35 0.89 0.94

1.12 0.86 0.44

0.93 0.44 0.92

1.24 1.83 1.19

0.80 0.80 0.97

0.2! 1.26 0.32

0.76 1.64 0.64

1.27

0.99

0.85

1.19

0.84

0.71

1.11

1.89

1.15

1.10

1.13

0.83

0.52

(ii)

(i)

0.96 0.90 0.73 1.11 1.54 0.93 1.02 0.81 1.01

1.26 1.68 0.47 0.28 0.78 2.82 1.41 0.90 1.03

0.61 .18 .29 .06 .02 .24 ().88 13.82 (3.70

0.34 0.27 1.68 1.47 0.99 0.16 0.63 1.03 0.87

1.39 0.62 0.32 0.49 0.30 0.64 0.57 1.01 1.50

1.34 1.22 0.57 0.50 0.19 0.28 0.91 0.78 1.23

1.10

1.01

193

1.45

1.53

0.98 0.95

0.89 0.52

1.12 0.86

0.91 2.35

1.15 1.27

0.89 0.46

0.97 1.06

1.04

0.70

1.10 1.17 0.80 1.02 1.30

(i)

(i)

(i)

1.47

1.48

1.15

0.66

0.17

2.77

0.89

1.32

1.39

0.58

0.00

0.75

0.00

0.25

0.96 0.98

0.84 0.87

0.96 0.90

0.94 0.80

1.02 1.68

0.71 2.26

0.94 0.82

0.84 0.96

0.97 0.82

0.87 1.91

1.14 0.98

1.09 1.74

2.43 0.70

0.94 0.84

(ii)

0)

(ii)

~Belg. & Lux. (i) (ii)

US

(i)

(ii)

Japan (ii)

1Neth. (ii)

Denm. (ii)

Spain (ii)

Ireland (ii)

0.92 1.06 2.02 0.00 2.65 0.14 0.58 0.71 1.02 1.21 1.19 3.48 1.41 2.26 1.05 1.23 1.02 0.00 0.26 0.54 0.84 0.10 0.33 0.57 0.00 1.39 0.62

0.96 0.31 0.08 0.50 0.76 0.30 0.62 .61 .09 .67 .29 .52 .74 .13 ().84 1.22 1.03 0.32 0.99 1.82 1.25 0.40 0.71 1.14 0.38 0.82 1.54

0.44 0.95 0.00 7.64 1.43 0.00 0.63 .31 .92 ().52 .15 1.31 1.53 1.22 1.44 3.56 1.84 3.39 1.40 3.48 1.97 0.45 0.00 0.46 1.13 0.75 0.38

0.71 1.29 1.98 1.21 0.72 0.94 1.50 0.75 0.62 0.85 1.20 0.73 1.06 0.60 0.76 0.94 1.29 0.64 0.80 0.77 0.87 0.95 1.17 0.74 0.63 0.62 1.10

1.80 1.57 0.00 0.00 1.45 0.92 0.00 0.89 0.28 1.12 0.87 0.33 0.78 0.99 1.36 0.90 0.37 0.00 1.43 0.59 0.77 0.34 1.08 2.02 4.60 0.26 0.82

1.08 0.91 0.92 0.84 1.01 1.01 0.79 0.83 0.88 1.10 0.96 1.10 1.04 0.87 1.01 1.03 1.09 0.92 0.94 1.03 0.98 0.92 1.01 1.00 1.26 0.99 1.13

1.06 0.87 1.29 0.75 0.79 1.28 1.02 1.13 1.18 0.74 0.81 0.35 0.46 1.14 1.31 1.05 0.72 2.18 1.06 0.66 0.78 0.99 1.15 1.03 0.65 1.41 0.81

2.32 1.01 1.10 0.76 1.38 0.91 0.69 0.57 0.49 0.94 0.83 1.79 1.22 0.46 1.76 1.18 1.95 2.23 0.31 1.66 0.22 0.64 0.92 1.02 2.72 0.73 0.63

2.86 0.68 0.07 1.10 0.58 0.31 1.39 0.24 0.32 1.36 1.17 1.31 1.21 1.07 0.81 0.92 0.68 0.72 0.35 0.86 0.21 0.91 0.96 1.64 0.55 1.08 1.53

1.72 0.75 0.12 1.52 0.30 0.32 .37 .20 ().55 .37 .29 1.01 0.87 1.08 0.51 0.64 1.75 0.36 1.02 2.60 1.32 0.85 0.67 0.84 0.28 0.55 2.41

3.53 0.00 1.71 0.90 0.66 1.43 0.00 0.00 0.00 1.23 1.19 3.62 1.70 0.84 0.69 0.63 1.45 0.00 1.41 0.88 0.69 0.00 1.34 1.24 0.00 0.91 2.29

Sweden Sector Food products Chemicals, n.e.s. Synthetic resins Agricultural chemicals Cleaning agents Paints, etc. Pharmaceuticals Ferrous metals Non-ferrous metals Fabricated metal products Mechanical engineering, n.e.s. Agricultural machinery Construction equipment Industrial engines Electrical equipment, n.e.s. Transmission equipment Lighting and Wiring Radio and TV receivers Motor vehicles Shipbuilding Transportation equip, n.e.s. Textiles and clothing Rubber products Non-metallic mineral products Coal and petroleum products Professional instruments Other manufacturing

Note: (n.e.s. = not elsewhere specified.)

Italy

0.87 1.11 1.72 1.11 2.50 1.02 1.24 2.16 1.51 0.84 0.92 1.49 0.72 0.18 0.56 0.78 0.83 0.58 0.42 0.21 0.59 1.07 1.27 1.99 1.05 1.53 0.84

48

New perspectives on the late Victorian economy

innovations on which they were based, and because large firms account for a substantial proportion of total patenting activity in the USA. For this reason the-analysis refers to national groups of firms (owned by residents of the country in question), rather than simply national patterns of innovation. In order to calculate the RTA index for years before 1914, each patent granted to a foreign resident in 1890-2 or 1910-12 has been classified to an industry using the description provided by the US Index of Patents. Time and budget constraints prevented the extension of this patent count (which already ran to over 18,000 patents) for the entire period 18901914, or for an additional period in the interwar years, which would have been the ideal. Historical comparisons are reported below for the top ten foreign countries, each of which accounted for over 200 US patents in the years 1890-2 and 1910-12 combined. The relevant values of the RTA index, calculated across twenty-seven industries, are shown in table 2.3. For other countries, the distribution of the RTA index for these early years would be heavily influenced by the small numbers of patents in any category. This is illustrated by the rise in the number of zeros in the index as the number of patents falls. Taking the index for 1890-1912, Italy with 206 patents had three industries with an RTA of zero, Belgium and Luxembourg (203 patents) had four zeros, Denmark (116 patents) eleven zeros, and Japan (40 patents) no less than eighteen zeros. Countries outside the top ten must therefore be excluded from the historical aspect of the study. It may well be that the historical analysis should be restricted to the top three countries (Germany, the UK and Canada), each of which had over 2,000 patents in the early period, and which were the only countries with no zeros in the RTA index for these years. The RTA index can be calculated for any number of consecutive years. Here, it has been calculated for 1890-2, 1910-12, 1890-1912 (from a combination of 1890-2 and 1910-12), 1963-83, 1963-9 and 1977-83. The stability of the RTA index, and changes in the extent of technological specialisation, are examined over different time periods. Section 2.3 describes the statistical methodology used to achieve this end. Sections 2.4 and 2.5 consider the historical evidence, sections 2.6 and 2.7 discuss the results for later periods, each referring to trends in technological advantage over different time span, while section 2.8 discusses some conclusions that might be drawn.

Historicaltrends in international patterns oftechnologicalinnovation

49

2.3 The statistical methodology The proposition that innovation is cumulative would suggest that for the firms of any given country, the sectoral distribution of the RTA index is likely to remain fairly stable over time. This means that if the RTA index is calculated for a national group of firms at two different points in time perhaps twenty years or so apart, then these two sectoral distributions of technological advantage should be positively correlated with one another. However, since the nature of innovative activity will change gradually over time, the degree of correlation is likely to fall, the further apart are the two groups of years under consideration. Over longer periods of time firms on a cumulative path of development may still move across technically related sectors. The relevant statistical methodology is the Galtonian regression model, a statistical technique devised for the analysis of bivariate distributions. This approach was originally applied to economic problems in the context of work on the size distribution of firms by Hart and Prais (1956), and other useful applications have since been developed by Hart (1976) and Creedy (1985) in investigating changes in income distribution in the UK, and by Sutcliffe and Sinclair (1980) in the case of the seasonality of tourist arrivals in Spain. To adopt a similar procedure, the correlation between the sectoral distribution of the RTA index at time / and at the earlier time / — 1 is estimated through a simple cross-section regression of the form RTA,, = a + (3 RTA/,-! + e/7

(2.1)

This is estimated for a particular country, and the subscript refers to industry / at time /. The standard assumption of this analysis is that the regression is linear and that the error term e,, is stochastic and independent of RTA/7_ i. This is valid if the cross-industry index at each point in time approximately conforms to a normal distribution. The regression line will pass through the point of means and in figure 2.1, for convenience of exposition, this is illustrated for the case where the mean of each distribution happens to be the same, the expected values of the distributions being given by E(RTA,) = E(RTA,_i). The analysis does not depend upon the values of the means being identical, but it can be seen that a sizeable difference between them may be indicative of a substantial measure of skewness in one of the indices, which consequently departs significantly from a normal distribution. In figure 2.1, the regression line (2) is drawn in such a way that the estimated coefficient p takes the value of one. This implies not only that

50

New perspectives on the late Victorian economy RTA,

(3)

E(RTA,)

E(RTA,_!) Figure 2.1 Galtonian regression with the RTA index

the ranking of industries remains unchanged (advantaged industries remain advantaged, while disadvantaged industries remain disadvantaged), but also that they retain the same proportional position (advantaged industries do not become any more advantaged, and disadvantaged industries do not become any more disadvantaged). Where P > 1, as in line (3), then there is a proportional shift in which already advantaged industries tend to become still more advantaged, while disadvantaged industries are increasingly disadvantaged. In the case of regression line (1) where p < 1 disadvantaged industries improve their position, and advantaged industries slip back. This is what has elsewhere been termed 'regression towards the mean' (Galton 1889, cited in Hart 1976). Where this is a true representation of 0 < p < 1 then industries remain in the same ranking, but they come closer to one another. The magnitude of (1 - p ) therefore measures the size of what is here called the 'regression effect', and this is the interpretation placed on the estimated coefficient p. In the case of p < 0 then the very ranking of industries would be reversed, contrary to the prediction of cumulativeness in technological change. The expectation that P > 0, such that the RTA index is positively correlated across two points in time, can be readily tested for each country. The relevant test of p being significantly different from zero is the t-test. In a regression equation with only one independent variable the t-test is equivalent to the F-test, which refers to the significance of the correlation of the regression as a whole. The test for whether p is significantly greater than zero is a test of the proposition of cumulativeness against the alternative that sectoral com-

Historical trends in in ternationalpatterns oftechnological inno vation position of innovation is random. However, the second proposition to be set alongside that of cumulativeness in the industrial pattern of innovation is that of incremental change. If firms generally innovate in order to gradually adapt their existing technological strengths, they may still begin to shift the industrial nature of their activity. As the pattern of demand changes, and technology evolves, the sectoral distribution of innovation in a country may change, even though still drawing on a similar set of underlying technological skills. This effect is likely to be more pronounced the further apart the RTA distributions are in time. The condition under which cumulativeness in the industrial distribution of innovation outweighs incremental change is that p > 1. Strictly speaking, if there were a path-dependent cumulative process with no change in the technological relatedness between sectors and therefore no shift in the underlying industrial structure of innovation (no incremental change), it would evolve towards a position where the proportion of innovations accounted for by each industry was stable and fixed (Arthur, Ermoliev and Kaniovski 1987). This would correspond to p = 1, and to a regression effect (1 - (3) exactly equal to zero. The test of whether cumulativeness outweighs incremental change in the period in question is hence the t-test of p not being significantly less than one (equivalent to a regression effect which is negative or not significantly different from zero). Where p is significantly greater than zero but significantly less than one then elements of cumulativeness and incremental change are combined. If cumulativeness dominates initially over relatively shorter periods (P > 1), tests should reveal the length of time that it takes for incremental change to begin to play a significant role (0 < p < 1). What is then also required is that the regression analysis is supported by a more detailed inspection of the actual shifts in the RTA index, to investigate the actual evolution of sectoral strengths and weaknesses, and to decide whether the change really is of an incremental kind. The other feature conveniently arising from the regression analysis of RTA distribution is a simple test of changes in the degree of technological specialisation. The degree of technological specialisation in a country can be measured by the variance of its RTA index, which shows the extent of the dispersion of the distribution around the mean. Pavitt (1987) used the standard deviation of the index, which is the square root of the variance, as a measure of such specialisation. The original work of Soete (1980) also analysed the variance of RTA indices. The procedure for estimating changes in the variance of a distribution over time follows from Hart

51

52

New perspectives on the late Victorian economy

(1976). Taking equation (2.1) above, if the variance of the RTA index at time t is denoted by o>2 then a,2 = p 2 * , - , 2 * ^ 2

(2.2)

Now the square of the correlation coefficient (R2) is given by R2 = 1 -(a e 2 /a, 2 ) = (a, 2 -a e 2 )(l/a, 2 )

(2.3)

Combining equations (2.2) and (2.3) it follows that (x, 2 -a e 2 = p 2 a / _, 2 = RW

(2.4)

Equation (2.4) may be rewritten to show the relationship between the variance of the two distributions as follows: a,2/ R2, and it falls where p 2 < R2. A high variance indicates a high or narrow degree of specialisation, while a low variance indicates that the country has a broad range of technological advantage or a low degree of specialisation. Using the estimated regression values, the extent of specialisation rises where |p| > \R\, and it falls where |p| < \R\. The estimated Pearson correlation coefficient, R, is a measure of the mobility of industries up and down the RTA distribution. A high volume of R indicates that the relative position of industries is little changed, while a low value indicates that some industries are moving closer together and others further apart, quite possibly to the extent that the ranking of industries changes. The magnitude of (1 -R) thus measures what is here described as the 'mobility effect'. It may well be that, even where the regression effect suggests a fall in the degree of specialisation due to a proportional move in industries towards the average (p < 1), this is outweighed by the mobility effect, due to changes in the proportional position between industries (p > R).

2.4 The stability of technological advantage before 1914 The results of the regression of the RTA index in 1910-12 on the index in 1890-2 are reported in table 2.4. Each distribution represented a cross-section of thirty-one industries for the ten major countries patenting in the US before 1914. Because of the problems created by a small number of observations,

Historical trends in internationalpatterns of technological innovation Table 2.4

53

The results of the regression of RTA in 1910-1912 on RTA in 1890-1892

Country UK Germany Canada France Austria-Hungary Australia Switzerland Sweden Italy Belgium and Luxembourg

&

P

0.870 0.493 0.553 0.926 0.865 1.067 1.066 0.726 0.993 0.614

0.262 0.533 0.457 -0.073 -0.004 0.009 0.252 0.047 0.060 0.230

6.06** 4.38** 3.80** 9.23** 5.36** 1.07 2.91** 4.75** 5.86** 3.97**

h

R

Pi

2.13* 5.24** 3.41** -1.75 -0.04 0.01 0.77 0.80 2.92** 2.81**

0.367 0.697 0.535 -0.308 -0.008 0.009 0.141 0.145 0.476 0.462

0.935 0.969 0.962 0.790 0.933 0.858 0.869 0.929

Notes: * Denotes coefficient significantly different from zero at the 5% level. ** Denotes coefficient significantly different from zero at the 1 % level. Number of observations = 31.

alluded to in section 2.2 above, the critical assumptions required by the regression analysis are questionable for all but the top three countries. The assumption that for the firms of a given country the propensity to patent varies more systematically between sectors than within them relies as a minimum on a large number of observations (patent counts). In addition, the assumption that the regression is linear may be invalid if the crosssectoral RTA distributions are not approximately normal. Outside the distributions of the UK, Germany and Canada, the Pearson tests for normality (Pearson and Hartley 1976) reveal that the RTA index for the other thirteen countries is significantly skewed in both 1890-2 and 1910-12, except for Austria-Hungary, Sweden and Belgium and Luxembourg in 1910-12. By contrast, for the firms of the UK, Germany and Canada the only significant measure of skewness occurs in the case of Canada in 1910-12, and here it is still a great deal lower than for most other countries. A total number of patents greater than the minimum required to construct an index that roughly conforms to a normal distribution obtained historically only in the cases of the UK, Germany and Canada. Accordingly, the results shown in table 2.4 are only really trustworthy for the first three countries. In the case of France, which had the fourth largest number of foreign patents in the US before 1914, the negative estimate of the p coefficient is indicative of the fact that the sectoral distribution for France in 1890-2 had a much higher mean value than in 1910-12. The mean value, and the skewness of the distribution, is

54

New perspectives on the late Victorian economy

increased by the RTA of the cleaning agents sector in 1890-2, which reached 11.54. This result arises due both to the relatively small number of French patents in the US, and the relatively small number of all foreign patents in the cleaning agents industry in 1890-2. If the basic assumptions of the regression are not met, then the /-statistics are not valid tests of the significance of the values of the estimated coefficients. However, for the three countries for which the results can be trusted with a reasonable degree of confidence, the findings appear to be broadly in line with expectations. The p coefficient assumes a positive value for all of these three, and indeed for eight out of ten countries, demonstrating a positive correlation between the pattern of technological advantage in the early 1890s and just before the First World War. The t-test suggests that this correlation is significant infiveof these eight cases. With reference to the three countries for which the data are most reliable, the correlation is significant for all three. Where the value of p was positive, the implicit value of p for a one year regression was calculated as Pi in the final column. This is calculated on the assumption that the structure of the relationship was unchanged throughout the period 1890-2 to 1910-12, and is the value of p that year on year would generate the observed p for a twenty year period. Although this procedure must not be taken too far, as the structure of the relationship was not unchanged (that is, the mobility effect or 1 - R was greater than zero), it helps to illustrate that the value of the regression effect (measured by 1 - P) is very much weaker than would be suggested by the estimated value of p if considering periods of less than twenty years. However, while for the three national groups for which there are a sufficient number of observations - British, German and Canadian firms randomness in innovation must be rejected in favour of cumulativeness, it is clear that the sectoral distribution of their technological development was also subject to a degree of change in the twenty years around the turn of the century. The regression effect was positive and significant (P was significantly less than one on a t-test) in all three cases. Although there is some evidence of incremental change during these years, it is difficult to draw conclusions at this level of analysis, given that many firms were only just beginning to establish themselves as US patenters at the time. This must itself account for a certain amount of change in the distribution of patenting activity. For eight out often countries the value of R exceeded p, signifying a fall in the degree of technological specialisation. That is, patenting activity in the USA tended to become broader in its sectoral scope. In part, this

Historicaltrends in internationalpatterns oftechnological innovation

55

result may reflect an improvement in the representativeness of the data, and the number of US patents granted to foreigners rose substantially in the twenty year period in question (see table 2.2). Once again, this also reflects the fact that many firms began patenting in the USA for the first time around the turn of the century. However, as the newer technologies pioneered at this time began to take root, there may well have been a broadening in the sectoral distribution of innovation in the leading industrialised countries. The possibility that the sectoral distribution represented by the RTA index confirms more closely to a lognormal than a normal distribution was also investigated. The major reason for supposing that this might be the case is that while there is a lower bound to the distribution (disadvantaged industries are constrained to take RTA values between zero and one), there is no upper bound. Although there is some evidence which at first glance might support such a view, it in fact only strengthens the case against it. Distributions that are more closely lognormal than normal as judged by the Pearson tests of skewness and kurtosis are found only where the sample of patent counts is really insufficiently large, creating an artifically high degree of dispersion in the index. This is what tends to lead to a skewed distribution, rather than any inherent property of the RTA index itself. Considering the five countries that were excluded from the historical analysis due to lack of data (Ireland, Denmark, the Netherlands, Spain and Japan), all have significantly skewed RTA distributions for both 1890-2 and 1910-12. However, in every case these are distributions in which the longer tail lies towards the lower values of RTA (to the left), which is a consequence of the large number of zeros. In other words, the theoretical justification for expecting a lognormal distribution is that there is no upper bound so it will be skewed to the right, but tests demonstrate the existence of lognormality only in instances where due to the paucity of data the distribution is skewed to the left. Therefore, although assuming a logarithmic rather than a linear functional form in such cases sometimes improved the value of the /-statistic on p, and the estimated correlation coefficient, R, these results are not discussed here. It is worth noting those sectors in which the leading three countries (the UK, Germany and Canada) were strongly advantaged in both 1890-2 and 1910-12 (see also table 2.3). British firms appear to have maintained a strong technological advantage at this time in cleaning agents (mainly in soaps and detergents; the early French strength here mentioned above was in perfumes and cosmetics), industrial engines and turbines, ship-

56

New perspectives on the late Victorian economy

building, textiles, rubber products (tyres) and coal and petroleum products. German firms enjoyed a favoured position in chemicals in general, but especially in dyestuffs and paints, and in lighting and wiring equipment. Canada's strength lay in agricultural chemicals, railways and other transportation equipment, paper products, and other manufacturing (including furniture and wood products). In the case of these three countries at least, as far as can be told, the sectoral pattern of technological advantage remained fairly stable (elements of cumulativeness were present) in the twenty-five years before 1914, although the degree of specialisation fell. Finally, returning to the theme of the opening remarks of the paper, some assessment of the relationship between the industrial composition of innovation and overall technological performance can be made. This concerns the extent of representation of each national group offirmsin the sectors of fastest (or slowest) growing technological activity. For any country j denote the proportion of patents held in industry / by pij9 the share of total world patenting by w7, and the mean value of the RTA index by Mj. That is (supposing there are altogether n industries) PiJ

= PijKjPij

Wj =

%Pi//X^jPu

RTA,7 = The ratio of the mean value of RTA in an earlier period t-\ to that in the next period t is then given by Mjt- xIMJt = (wJt/wJt- iVGiPijittpijt-1)

(2.6)

Now since the regression equation (2.1) must pass through the point of means it is also known that Mj, = d + (W /7 _,

(2.7)

This may be rewritten MJt-XIMj, = (MJt - d)/flW/7)

(2.8)

From equations (2.6) and (2.8) it follows that the lower is (Mjt — a)/$Mjt the slower is the rise (or the greater is the decrease) in the world share of patents (ny) compared with the average proportion held in industries at the chosen level of disaggregation (ZjPy/n). It can be shown that this happens either because the firms in question are particularly advantaged in industries with the slowest growing technological activity,

Historical trends in internationalpatterns of technological innovation

57

or because of a shift in the structure of industry proportions and thus in the cross-industry pattern of RTA (what has been termed above the mobility effect).2 Of British, German and Canadian firms over the period 1890-2 to 1910-12, British firms recorded the lowest value of(MJt — a)/$Mjt, while German firms had the highest such ratio. While this may be partly attributed to a greater mobility effect for UK firms (a lower R), it also suggests that they had their comparative advantage in industries in which fewer new technological advantages were opening up. In other words, to the extent that they were 'locked in' to a particular path of innovation in the twenty to twenty-five years before 1914 this was to their overall detriment, but to the benefit of their German rivals.

2.5 Were British firms 'locked in' to innovation in sectors with fewer technological opportunities? It is worth pausing at this juncture to consider the pattern of technological activity amongst British firms prior to 1914 in greater detail. The RTA index as shown in table 2.3 roughly matches what is known from other historical evidence. British firms in this period continued to develop in a cumulative fashion many of their earlier traditional strengths. They retained their leadership in the fields of shipbuilding, heavy mechanical engineering, armaments and industrial engines (Mathias 1969). They also performed well in branches of the food, drink or distilling and brewing sectors (through Cadbury, Fry, Guinness and Bass), in consumer chemical and pharmaceutical products (through Beecham and Lever), in synthetic fibres (through Courtaulds) and in coal and petroleum products. They built upon an early lead in agricultural chemicals and in radio and telegraphic equipment (through Marconi). However, even in most of these cases it is clear that innovation in British companies was of a rather specialised or differentiated kind, which while it had suited them well in the past became a source of weakness in a newly emerging age of mass production. They concentrated on particular types of electrical equipment (such as the development of the radio and international telegraphic communication) rather than on the broader electrical engineering sector which became responsible for spreading economy-wide electrification. In electric traction, the main underground railway lines constructed in London between 1900-14 were built with American plant and expertise (Mathias 1969). British firms held their place in transport equipment and special kinds of electrical equipment,

58

New perspectives on the late Victorian economy

but did not establish themselves in the operating systems on which they came to depend. Indeed, there is clear argument for saying that the comparative technological advantage of British firms in the motor vehicle sector was far from being a sign of strength. This is a good example of a wide cluster of British family firms which were small, differentiated and emphasised quality and skilfulness losing out to the larger US corporations which went down the route of organising mass production. There were over 200 tiny British firms in the motor vehicles sector, led by competent engineers and producing highly sophisticated and very expensive products (Mathias 1969). By comparison, Ford's greater reliance on production engineering and Taylorist work practices led to a reduction of technical standards on occasions. Organisational innovation may then have brought technological innovation in its wake, but aimed at solving a different set of problems and conditioned by a different set of criteria than the technologies developed by British firms (Freeman, Clark and Soete 1982). US and German firms went down a path of innovation more suited to the needs of mass production and a mechanised age which characterised a new technological system, with a corresponding shift in the types of technological development in which opportunities were greatest. This is another aspect of firms in a particular location 'locking in' to a certain form of innovation. The textiles industry offers a different example of how even a group of formerly strong firms may lose their position when their competitors in other countries make cumulative gains at their expense. Although British firms such as Platts, the supplier of automatic looms, continued to lead the way in textile machinery, British textile firms failed to adopt the new equipment (Mathias 1969). By 1919 half the looms in the US cotton industry were automatic, while British firms had only just begun to introduce them, due in part to a lesser pressure to introduce labour-saving devices (Habakkuk 1962, Aldcroft and Richardson 1969). British firms also had alternatives, not only by being better able to turn to wage cost reductions, but by switching to Empire markets and entering into collusive arrangements with other firms. In these terms the initial introduction of the latest technology may well have been uneconomic and so company decisions can be justified as rational. However, in the medium and longer term, due to the existence of cumulative gains in innovation through learning by doing and learning by using (Nelson and Winter 1982, Pavitt 1987) their failure to adopt meant that they were left behind. As well as falling behind within certain industries, as has been sug-

Historicaltrends in internationalpatterns oftechnologicalinnovation

59

gested above British firms became 'locked in' to innovation in sectors with fewer opportunities. Textiles and shipbuilding remained important in British industrial structure in the early twentieth century, and mechanical engineering was better developed than the electrical industry (Hannah 1976). However, Mowery (1984) argues that the problem was not so much a failure to shift the industrial pattern of output, as a failure to take up innovative activity in the most dynamic sectors (which particularly affects one component of output, namely export performance). British companies were frequently left behind in science-based areas, due to the difficulties of institutional adjustment as well as cumulative technological progress amongst the leaders elsewhere. In industrial chemicals the deficiencies of the educational and training systems and their weak links with industry meant a shortage of highly qualified scientists and technicians (Haber 1958, Liebenau 1984). In heavy engineering the consulting engineer played a particularly important role in British firms which reduced their incentive to invest in corporate R&D and lessened the potential for the development of innovative linkages between firms (Mowery 1984). As a symbol of their lagging behind in science-based fields, it was only in the interwar period that corporate R&D increased substantially amongst larger British firms (Hannah 1976). Where British firms did consolidate or establish themselves in these areas, their scientific traditions tended to lie outside the mainstream. For example, in Pharmaceuticals the British strength emanated from medical and biological research, as illustrated by the Evans Medical Company, the Lister Institute and the Wellcome Company, rather than as in Germany spinning off from the chemicals revolution which began with artificial dyestuffs (Liebenau 1984). This is an interesting case, as it perhaps helps to explain the recent British revival in Pharmaceuticals innovation at a time when opportunities are increasingly coming from biotechnology rather than chemicals (again, building upon past traditions). Likewise, British firms enjoyed an early lead and consolidated their position in agricultural chemicals, the pioneer being J.B. Lawes in superphosphates and other fertilisers (Freeman 1982). In the electrical industry an example of specialised success was provided by Marconi, set up in London by an Italian entrepreneur with British and Irish connections. The technological development of this British company was initially related to communications in shipping, the Royal Navy being the first major purchaser of equipment (Baker 1970). Marconi led an active research team, which successfully competed with German and US groups despite their lead in other types of electrical

60

New perspectives on the late Victorian economy

equipment, due to its emphasis upon international transmissions. It was obliged to concentrate on international communications with the Post Office having a monopoly over land-based telegraphy in Britian, while it was excluded from other national markets. With the German firm Telefunken, Marconi controlled most of the ship- and shore-installations across the world, including those in the USA. In 1912 they reached an agreement on patent rights and cross-licensing. A more detailed analysis of innovation amongst British firms at the turn of the century therefore suggests that cumulativeness in technological development is important in explaining the pattern of their success and failure. It also emphasises the role played by institutional characteristics and constraints, alongside the cumulative innovative and learning activity which builds upon existing technological traditions and experience. The problem was only partly that British firms were 'locked in' to innovation in sectors with fewer opportunities. They were also 'locked in' to dated institutions and organisation practices which adversely affected their technological performance in all sectors, but especially in those with the greatest opportunities, the science-based sectors. This is in addition to the other general considerations affecting Britishfirmsin this period, such as their access to Empire markets and their greater scope for holding down wage costs, factors which have been discussed elsewhere.

2.6 The stability of technological advantage over the last hundred years Table 2.5 shows the results of the regression of the RTA index in 1963-83 on the index in 1890-1912 (the years 1890-2 and 1910-12 combined). These relied on the cross-section of twenty-seven industries given in table 2.3. Of the original thirty-one industries in the historical series, to ensure comparability with the recent data, drink and tobacco were subsumed under food products, leather products were subsumed under textiles, leather and clothing, and paper products included under other manufacturing. In constructing the RTA index for the whole span 1890-1912, the German distribution joined those that were significantly skewed away from normality. This problem arises because of the industrial reaggregation, which causes a number of industries whose RTA values were similar to be grouped together or lost under other headings, such that some information which was important historically is lost in moving to the more recent classification.

Historical trends in internationalpatterns oftechnological innovation 61 Table 2.5 The results of the regression ofRTA in 1963-1983 on RTA in 1890-1912 Country

UK Germany Canada France Austria Australia Switzerland Sweden Italy Belgium and Luxembourg

a.

P

ta

'P

R

Pi

0.803 0.957 1.013 0.912 0.731 0.537 0.901 0.804 0.967 1.066

0.256 -0.147 0.120 0.127 0.241 0.632 0.026 0.163 -0.016 0.125

6.97* 11.22* ' 3.88* 10.41" 3.19" 3.27" 8.02" 5.50" 11.03" 6.66*'

2.48** -0.19 0.50 1.38 1.13 4.44** 0.43 1.43 -0.40 0.10

0.445 -0.037 0.099 0.266 0.219 0.664 0.084 0.276 -0.077 0.063

0.981 0.971 0.972 0.980 0.994 0.951 0.975 0.972

Notes: * Denotes coefficient significantly different from zero at the 5% level. Denotes coefficient significantly different from zero at the 1% level. Number of observations = 27.

A positive correlation between the two distributions was obtained for eight out of ten countries. However, on the whole the extent of this correlation was poor, and it was significant in only two cases. It is the firms of the UK and Australia that appear to have shifted least from the sectoral patterns of technological advantage that prevailed historically. In these cases as well, though, the regression effect is still significant (P is significantly less than unity). It is not terribly surprising that the industrial composition of innovation shifts substantially over long historical periods. The value of Pi generally remains high, but the distributions are over seventy years apart, and it is unrealistic to suppose that the structure of the model has remained unchanged throughout this time. The other noticeable feature of table 2.5 is that the estimated value of a is high for every country, and everywhere significantly different from zero. This might suggest a low value of p and a strong regression effect. However, it seems that what it actually represents is the weakness of correlation between the two RTA distributions, which is associated with a strong mobility effect. Leaving aside the two countries for which correlation was significant and the degree of specialisation fell, in four of the remaining eight countries the mobility effect outweighed the regression effect leading to an apparent rise in the degree of specialisation. For the firms of Germany, Canada, Austria and Belgium a low magnitude of p was offset by a still lower value of R.

62

New perspectives on the late Victorian economy

Of course it might be that these results would be altered if patent counts were made over the full period 1890-1914, rather than for only six years. However, it seems unlikely that this would affect the conclusion that the sectoral distribution of technological advantage has tended to shift quite substantially since the turn of the century. It is possible to say a little on what might be the long-term effects of incremental changes in the composition of innovation. British firms seem to have moved from synthetic resins towards paints and Pharmaceuticals, from ferrous metals towards metal products, and from general mechanical engineering and industrial engines towards agricultural and construction equipment. German firms appear to have improved their position in mechanical engineering, particularly in industrial engines, and in motor vehicles and other transportation equipment. This might be related to traditional strengths in metal products and certain categories of electrical goods. The improvement of the German group in synthetic resins, agricultural chemicals and cleaning agents may be similarly linked to their long-standing overall strength in general industrial chemicals. Shifts in the RTA index for other countries can be determined from an inspection of table 2.3, though they are not as reliably indicated due to the influence of small numbers of patents in the earlier period.

2.7 The stability of technological advantage since 1963 The estimates obtained from the regression of the RTA distribution in 1977-83 on the distribution in 1963-9 are set out in table 2.6. They were derived from the same cross-section of twenty-seven industries listed in table 2.3. The only serious data problem in the recent period arises in the case of Ireland, whose firms had a low level of patenting activity even in the recent period, and which has a sizeable number of zeros in the RTA index for the latest years. As a result, the RTA distribution for Ireland particularly in 1963-9 was substantially skewed. Pavitt (1987) reported on a similar set of correlations for the regression of the RTA index for ten countries in 1975-80 on the equivalent RTA in 1963-8. He also used a twenty-seven industry disaggregation, though he adopted a rather different sectoral classification. The other important difference here is that the original patent data have now being reworked by OTAF, based on a 'fractional' allocation of those patents that had previously been allocated to more than one industrial group. In his study, Pavitt found a positive and significant correlation between

Historical trends in international patterns of technological innovation

63

Table 2.6 The results of the regression of RTA in 1977-1983 on RTA in 1963-1969 Country US Germany Japan UK France Switzerland Canada Sweden Italy Netherlands Belgium and Luxembourg Austria Australia Denmark Spain Ireland

a

P

>«

-0.513 0.341 0.586 1.124 0.758 0.168 -0.143 0.130 0.681 0.327 0.806

1.513 0.642 0.339 -0.216 0.306 0.905 1.141 0.979 0.231 0.712 0.401

-2.97* 2.66* 4.22* 4.15* 3.71" 1.38 -0.77 0.79 6.10*' 1.82 2.82*'

0.399 0.149 0.063 0.384 0.636

0.572 0.794 0.889 0.855 0.543

2.58* 0.77 0.29 1.43 2.01

R

Pi

8.72** 4.85** 2.88** -0.08 1.54 7.49** 7.64** 6.11** 2.36** 5.40** 1.55

0.868 0.696 0.500 -0.017 0.295 0.832 0.837 0.774 0.428 0.734 0.297

1.030 0.969 0.926 0.919 0.993 1.009 0.998 0.901 0.976 0.937

4.55** 5.56** 4.71** 3.30** 2.85**

0.673 0.745 0.686 0.551 0.495

0.961 0.987 0.992 0.989 0.957

Notes: * Denotes coefficient significantly different from zero at the 5% level. Denotes coefficient significantly different from zero at the 1 % level. Number of observations = 27.

the two RTA distributions for nine out of the ten countries. He also found that the sectoral pattern of technological advantage for each country was distinctive, in the sense that there was little association between the distributions of any two countries. A similar state of affairs applies to the sixteen countries listed in table 2.6. A positive correlation holds for fifteen of the sixteen countries, and for thirteen of these it is significant at the 5 per cent level (it is significant for the other two at the 15 per cent level). Only the UK demonstrates little relationship between the distributions characterising the two periods in question, and in this instance the hypothesis of randomness (0 = 0) as against cumulativeness O > 0) cannot be rejected. Unlike in the historical application of the analysis, the estimate of the a coefficient is negative in the case of two countries (the US and Canada), though in the case of Canada it is not significantly different from zero. The corollary, given that the means of the two distributions are not too far apart, is that |3 exceeds unity, implying a negative regression effect, which is particularly strong for the US. Indeed, for eleven of the sixteen countries, despite generally high values of R, the mobility effect (which is measured by 1 - R) exceeds the regression effect (which is measured by 1 - P). This means that there has been a tendency for the degree of

64

New perspectives on the late Victorian economy

Table 2.7

The strength of the regression effect over the period 1963-1969 to 1977-1983

Country

t$

US Germany Japan

2.96** ~2-70** - 5.62**

UK

~4JKl

France —3.49** Switzerland -0.79 Canada 1.07 Sweden -0.13 Italy -7.86** Netherlands -2.18* Belgium and Luxembourg -2.32* Austria - 3.40 Australia —1.44 Denmark -0.59 Spain -0.56 Ireland -2.40* Notes: * Denotes coefficient significantly different from one at the 5% level. Denotes coefficient significantly different from one at the 1 % level. Number of observations = 27 technological specialisation to rise over the last twenty or twenty-five years. In other words, not only has the pattern of technological advantage remained fairly consistent, but advantaged industries have often acquired a stronger position than was the case for high-ranking sectors in the 1960s. The strength of the regression effect over the period is interpreted as indicating the extent of gradual change as against cumulativeness, and it can be separately measured by testing whether p is significantly different from one (which amounts to a test on whether the regression effect, 1 — P, is significantly different from zero). The results of this test are reported in table 2.7. This shows that the regression effect was insignificant for half of the sixteen countries. This includes the case of the US, for which p was significantly different from one, but it was substantially above one, and hence the regression effect was significantly negative. US firms have become increasingly specialised in areas in which they were already advantaged. In the historical analysis the regression effect was everywhere significant, so this helps to demonstrate the way in which it gradually increases over longer periods. However, between the early 1960s and the early 1980s it fails to reach significant proportions for eight

Historical trends in internationalpatterns oftechnological innovation

65

countries. Thus for half the national groups cumulativeness outweighs incremental change over this period. As has been mentioned above, the calculation of pi from the estimated value of p over a longer period is unreliable as it assumes a zero mobility effect, which is known to be untrue. For the recent period the greater level of patenting activity makes it feasible for the firms of some countries to estimate the actual magnitude of successive values of Pi from year-on-year regressions. This was done for the ten countries whose residents were granted over 10,000 patents in the US between 1963 and 1983, each averaging over 500 patents per year. The first column of table 2.8 reports the average value of the estimated slope coefficients from a series of annual regressions for each of these ten countries. Alongside the mean of the annual Pi (denoted by jipr), the standard deviation around the mean (<xp,) is shown in the second column. The regression estimates of Pi are uniformly smaller than the calculated pi of table 2.6, since the annual mobility effect does not always work in the same direction. This implies that for each national group of firms there is to some extent a random short-run fluctuation of industries up and down the ranking of innovative activity, in addition to any gradual but sustained shift in ranking that may be slowly taking place. Of the groups of firms of all countries, Japan is the only one whose pi over the period 1963-83 appears to have been on a significant upward trend over time. Regressing the time series of Pi on a simple time trend, the coefficient on the trend was positive and significantly different from zero at the 5 per cent level for Japan, but insignificant for all other countries. Noting that Japan is one of the minority of countries whose firms have experienced a recent fall in their degree of technological specialisation (p < R in table 2.6), what this suggests is that the broadening of innovative activity by Japanese firms was especially strong early in the period. Further support for this view emerges when dividing the 1963-83 period into three subperiods (1963-9, 1970-6, and 1977-83), and running regressions for consecutive subperiods. The estimate p 7 coefficients from the (seven year period) regression of the RTA index in 1970-6 on that in 1963-9, and of RTA in 1977-83 on that in 1970-6 are also given in table 2.8. It can be seen that they are not tremendously different from the individual year-on-year Pi, and that for most countries p does not change very much over the period as a whole. The only real exception as before is Japan, for which the value of P7 nearly doubles, confirming that p has been on something of an upward trend. It should also be noted that the

66

New perspectives on the late Victorian economy

Table 2.8 Country US

Germany Japan UK

France Switzerland Canada Sweden Italy Netherlands

The results of annual and other subperiod regressions of the RTA index over the 1963-1983period M