LABOUR MARKET EVOLUTION
How have modern labour markets developed? Both labour economists and economic historians agree...
38 downloads
4059 Views
2MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
LABOUR MARKET EVOLUTION
How have modern labour markets developed? Both labour economists and economic historians agree that it is necessary to look at labour markets in their historical context. Labour Market Evolution does just this. The authors examine the operation and development of labour markets in Western Europe and North America since 1500. They address key questions in this complicated process using new quantitative evidence. First, how closely connected were geographically separated labour markets? Second, how flexible were markets in the past—did wages change in response to demand shocks? Did workers move across space and occupations in response to cyclical or seasonal conditions? Third, were relationships between employees and employers short-term or long-term? Why did the relationships change, and what were the implications for the flexibility and integration of markets? In examining these factors, this volume draws on modern labour economic theory and up-to-date quantitative techniques to show how current traditions and systems have evolved. George Grantham is Associate Professor of Economics and Mary MacKinnon is Associate Professor of Economics at McGill University.
LABOUR MARKET EVOLUTION The economic history of market integration, wage flexibility and the employment relation
Edited by George Grantham and Mary MacKinnon
London and New York
First published 1994 by Routledge 11 New Fetter Lane, London EC4P 4EE This edition published in the Taylor & Francis e-Library, 2003. Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 © 1994 George Grantham and Mary MacKinnon All rights reserved. No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Labour market evolution: essays in the economic history of market integration, wage flexibility and the employment relation/edited by George Grantham and Mary MacKinnon p. cm. Includes bibliographical references and index. ISBN 0-415-10865-9 1. Labor market—History. 2. Labor—History. 3. Industrial relations—History. I. Grantham, George, 1941–. II. MacKinnon, Mary HD5706.L22 1994 331.12'09–dc20 93–34913 CIP ISBN 0-203-02177-0 Master e-book ISBN
ISBN 0-203-16195-5 (Adobe eReader Format) ISBN 0-415-10865-9 (Print Edition)
CONTENTS
vii ix xii xiii
List of figures List of tables List of contributors Preface 1
2
3
4
5
6
7
ECONOMIC HISTORY AND THE HISTORY OF LABOUR MARKETS George Grantham BRIDGING THE GAP BETWEEN LABOUR ECONOMICS AND ECONOMIC HISTORY: A LABOUR ECONOMIST’S PERSPECTIVE David A.Green HOW DID PRE-INDUSTRIAL LABOUR MARKETS FUNCTION? Jan de Vries THE DIS-INTEGRATION OF TRADITIONAL LABOUR MARKETS IN FRANCE: FROM AGRICULTURE AND INDUSTRY TO AGRICULTURE OR INDUSTRY Gilles Postel-Vinay REGIONAL LABOUR MARKET INTEGRATION IN ENGLAND AND WALES, 1850–1913 George R.Boyer and Timothy J.Hatton REAL INCOMES IN THE ENGLISH-SPEAKING WORLD, 1879–1913 Robert C.Allen DID LABOUR FLOW UPHILL? INTERNATIONAL MIGRATION AND WAGE RATES IN TWENTIETH CENTURY IRELAND Kevin O’Rourke v
1
27
39
64
84
107
139
CONTENTS
8
9
10
11
12
13
THE DECLINE IN HOURS OF WORK IN US LABOUR MARKETS, 1890–1903 Joshua L.Rosenbloom and William A.Sundstrom
161
JOB TENURE IN THE GILDED AGE John A.James
185
THE GREAT WAR AND THE CANADIAN LABOUR MARKET: RAILWAY WORKERS 1903–39 Mary MacKinnon
205
WAGE WOES IN WEIMAR? A PARTIAL INVESTIGATION David Corbett
225
WAGE BEHAVIOUR IN INTER-WAR BRITAIN: A SCEPTICAL ENQUIRY Mark Thomas
245
RACE, HUMAN CAPITAL, AND LABOUR MARKETS IN AMERICAN HISTORY Warren Whatley and Gavin Wright
270
Bibliography Index
292 317
vi
FIGURES
3.1
Real wage index of southern English building craftsmen 1480-1830 (1451-75=100) 3.2 Daily summer wages in the western Netherlands, 1500–1815 3.3 Skill ratios in the western Netherlands, 1500–1815 3.4 Composite salary index, 1540–1815 3.5 Salary index relative to journeymen’s wage index, 1540–1815 3.6 Cost of living in the western Netherlands, 1480–1819 3.7 Real wage index of master carpenters and masons in the western Netherlands, 1480–1819 (1451–75=100) 3.8 Real wage index of hod carriers in the western Netherlands, 1480–1819 (1451–75=100) 3.9 Real wage indices for building craftsmen in southern England and the western Netherlands (thirteen-year moving averages) 3.10 VOC employees embarking for Asia, decade totals 1630–1795 3.11 Percentage of Dutch-born among total embarking VOC employees 4.1 Index of monthly industrial unemployment in 1860 4.2 Summer shutdowns and economic characteristics of firms 4.3 Monthly industrial employment in 1860 and 1891–3 4.4 Percentage of the male labour force in agriculture, by birth cohort 5.1 Agricultural wages: north, midlands, and east 5.2 Agricultural wages: south, southwest, and Wales 5.3 Carpenters’ wages: midlands and north 5.4 Carpenters’ wages: London, southwest, and Wales 6.1 Interurban-intertemporal consumer price index (Manchester in 1896=1.00) 6.2 Real GNP per capita, CPI deflator (1896 Manchester £) 6.3 Real wage, bricklayers (1896 Manchester pence per day) 6.4 Real wage, labourers (1896 Manchester pence per day) 6.5 Average annual earnings, manufacturing workers (1896 Manchester £)
vii
41 45 46 48 49 52 53 54 55 59 59 66 70 76 81 95 96 98 98 112 116 119 120 120
FIGURES
7.1 7.2 7.3 8.1 8.2 8.3 8.4 8.5 10.1 10.2 11.1 11.2 11.3 12.1 12.2 12.3 12.4 12.5 12.6 13.1 13.2 13.3 13.4 13.5
13.6 13.7
Irish-British wage ratios and skill differentials, building industry, 1926–79 Irish-British wage ratios and skill differentials, engineering industry, 1926–79 Irish-British wage ratios and skill differentials, printing and publishing industry, 1926–79 Weekly hours by occupational group, United States (1890=100) Weekly hours of skilled and semi-skilled occupations, northeast Weekly hours of unskilled occupations, northeast Hourly wages of skilled and semi-skilled occupations, northeast Hourly wages of unskilled occupations, northeast Employment durations at the Canadian Pacific Railway, 1903–39 Employment and hiring at the Canadian Pacific Railway, 1903–39 Prices, wages, and unemployment in Germany, 1927–31 Net earnings and the civilian labour force in inter-war Germany Actual and simulated German unemployment rates, 1926–36 Nominal wages 1920–38 (1913=100) Unemployment and real wages 1919–25 Real wage series 1920–38 (1913=100) Unit labour costs 1920–38 (1938=100) Reported wage changes and coverage of wage-fixing machinery 1920–38 Nominal wage series 1920–38 (1924=100) Frequency distributions for black and white wages in Virginia, 1908 (manufacturing and building trades) Frequency distributions for black and white wages in Virginia, 1909 (cotton mills, sawmills, tanneries, tobacco) Frequency distributions for black and white wages in Virginia, 1926 (cotton mills, sawmills, tanneries, tobacco) Entry wage profiles, black and white workers at Ford Motor Company, 1918–47 (by age and marital status) Entry wage profiles, black and white workers at Ford Motor Company, 1918–47 (by years of schooling and marital status) Black employment at Rouge and at Rouge foundry 1918–45 Average wages at Rouge, 1920–46
viii
144 145 146 162 169 169 171 172 209 210 231 236 238 246 247 248 250 257 258 276 278 278 285
285 288 288
TABLES
3.1 4.1 4.2 4.3 4.4 4.5 4.6 4.7
4.8 5.1 5.2 5.3 5.4 6.1 6.2 6.3 6A.1 6A.2 6A.3 6A.4 6A.5 7.1 7.2
Unweighted mean daily summer wage paid to labour in the building trades in four countries (in current stuivers) Daily wages in industry and agriculture, 1860–2 Industrial workers’ relative wages in summer and plant shutdowns, 1860 Capital per worker in rural firms in the 1840s Male industrial wages by location of firm (c. 1840) Capital intensity and profitability of firms, 1860 The agricultural labour force, 1862–92 Average wages and coefficients of variation (c.v.) of departmental wages for men in agriculture and industry, 1840–92 Indices of rural and urban male industrial wages, 1840–92 Regressions of coefficients of variation of wages on a time trend Error correction model for agricultural wages, 1856–1903 Error correction model for carpenters’ wages, 1865–1913 Coefficients of variation of wages of unskilled labourers in five countries, 1860–1910 Alternative weights for consumer price indices Expenditure weights for interurban-intertemporal consumer price index Alternative consumer price indices, 1879–1913 Interurban-intertemporal consumer price index (Manchester in 1896=1.00) Real GNP per capita, CPI deflator (1896 Manchester £) Real wage, bricklayers (1896 Manchester pence per day) Real wage, labourers (1896 Manchester pence per day) Average annual earnings, manufacturing workers (1896 Manchester £) Emigration from the Republic of Ireland, 1911–71 Dublin wages relative to London wages ix
51 67 68 74 74 75 77
79 80 93 97 99 100 109 110 114 127 128 129 130 131 139 142
TABLES
7.3 7.4 7.5 7.6 7.7 7.8 8.1 8.2 8.3 8.4 9.1 9.2 9.3 9.4 9.5 9.6 9.7
9.8 9A.1
10.1 10.2 10.3 10.4 10.5 11.1 11.2 11.3 11.4 11.5
Real KHS-adjusted relative wages, 1950–83 Real relative wages, 1926–39 Unemployment rates in Ireland and Britain, 1935–79 Real expected relative wages, 1950–78 Days lost per worker due to strikes, Britain and Ireland, 1938–78 Occupational breakdown of Irish in England and Wales and of labour force in Ireland, 1946, 1951, 1956 Regression coefficients from analysis of variance of weekly hours Regression coefficients from analysis of variance of real hourly wages Elasticities of hours changes with respect to wage changes Decline in cross-city mean hours of work: actual and predicted from wage changes alone Distribution of eventual job tenures by age—Michigan furniture workers, 1889 Percentage of workers in new jobs by age class Percentage of workers by age in jobs of 10+ years— Michigan furniture workers, 1889 Distribution of eventual job tenures by age—Kansas, 1899–1900 Percentage of workers by age in jobs of 10+ years—Kansas, 1899–1900 Mean years in occupation and years with present employer by age class—Michigan furniture workers, 1889 Percentage of workers with years with present employer greater than years in occupation by age class and skill level— Michigan furniture workers, 1889 Average number of days lost due to no work (selective lay-offs)—Michigan furniture workers, 1889 Percentage of workers by age in jobs lasting or that will last at least 5, 10, or 20 years—Michigan furniture workers, 1889, unadjusted for industry growth Reason for separation, CPR sample, 1903–39 Skill margins at the CPR Characteristics of workers hired, 1903–28 Occupational composition of the CPR workforce Occupational retention and upgrading at the CPR Indices of industrial unit labour costs Industrial profit and investment rates Industrial labour demand: neoclassical specification Industrial labour demand: ‘Keynesian’ specification Total labour demand: neoclassical and ‘Keynesian’ x
146 147 148 148 151 153 167 170 178 180 191 192 193 194 195 197
198 199
202 212 216 218 219 220 228 229 232 233
TABLES
specifications 11.6 Industry-specific labour demand equations, 1925–38 11.7 Labour supply equations, 1925–38 11.8 Counterfactual total unemployment rates: neoclassical specification 11.9 Counterfactual changes in total employment: neoclassical specification 11.10 Counterfactual changes in total labour force: neoclassical specification 11.11 Counterfactual total unemployment rates: ‘Keynesian’ specification 12.1 Changes in wage rates, 1920–38 12.2 Coverage of wage-fixing machinery, 1920–38 12.3 The union mark-up model 12.4 Earnings dispersion by industry, 1924–38 12.5 Changes in wage rates by industry 13.1 Average hourly wage rates for entry-level workers in twenty industries, July 1937
xi
234 235 237 239 240 240 241 255 256 260 263 265 280
CONTRIBUTORS
Robert C.Allen, Department of Economics, University of British Columbia. George R.Boyer, Department of Labor Economics, Cornell University. David Corbett, Department of Economics, Williams College. Jan de Vries, Departments of History and Economics, University of California at Berkeley. George Grantham, Department of Economics, McGill University. David A.Green, Department of Economics, University of British Columbia. Timothy J.Hatton, Department of Economics, University of Essex. John A.James, Department of Economics, University of Virginia. Mary MacKinnon, Department of Economics, McGill University. Kevin O’Rourke, Department of Economics, University College, Dublin. Gilles Postel-Vinay, Institut national de la recherche agronomique, Paris. Joshua L.Rosenbloom, Department of Economics, University of Kansas. William A.Sundstrom, Department of Economics, Santa Clara University. Mark Thomas, Department of History, University of Virginia. Warren Whatley, Department of Economics, University of Michigan. Gavin Wright, Department of Economics, Stanford University.
xii
PREFACE
The McGill Conference on the Evolution of Labour Market Structures, of which this book is a product, grew out of discussions at the Second World Congress of Cliometrics in Santander in 1989, where the large number of papers dealing with nineteenth and early twentieth century labour markets revealed an emerging critical mass of work that may in time alter the way economists and historians view the development of modern labour markets. The intention of the organizers of the McGill conference was to provide a venue for discussing this work, in the hope that it might be possible to draw from it new historical generalizations about the economic history of modern labour markets. The conference was held in February 1991, in Montreal. Authors were asked to focus on the quantifiable characteristics of labour market structures and to analyse their behaviour over time. The papers in this book thus privilege the analysis of regional labour market integration, job attachments, wage flexibility, hours of work and turnover, neglecting such features as the development of unions and the impact of early twentieth century social legislation on the performance of labour markets, topics that by agreement have been reserved for a later conference. Conferences, like armies, depend on finance. We thank the Economics Department of McGill University and especially its Chairman, John McCallum, for making it possible to bring twenty-five scholars from North America and Europe for a three-day conference. We are also indebted to the Faculty of Arts and its Dean, Michael Maxwell, for providing additional funds for the conference and to defray the cost of preparing the present manuscript. The work of arranging hotels, airline tickets, meals, and printing the programme was expertly and cheerfully executed by Cathy ChandlerOliveira and Anne Renaud. Louis Bouchard contributed his time and energy to ensure that papers, projectors and presenters were at the right place at the right time. James Mott laboured at the graphs and verified the data behind them. We also thank the owners and managers of the Hotel Versailles for the flawless organization of the meals and conference facilities.
xiii
PREFACE
Finally, we wish to thank those members of the economic history profession who encouraged us in the belief that a small specialized conference might have a positive impact on the development of our field of study. We especially thank the participants who by their prepared and spontaneous comments revealed new meanings in old and new work: Lee Alston, Barry Eichengreen, Susan Carter, Alan Green, Bart Hamilton, Carol Heim, Michael Huberman, José Igartua, Carol Leonard, Joel Mokyr, Daniel Raff, Pierre Sicsic, Richard Sutch, David Weir, Jeffrey Williamson.
xiv
1 ECONOMIC HISTORY AND THE HISTORY OF LABOUR MARKETS George Grantham
What is the connection between the history of labour markets and the history of the economy? The answer to this question usually depends on whether it comes from an economist or an historian. When economists think about the economic history of the labour market, they are primarily concerned about how efficiently and equitably it allocated labour. To them the critical connections are those linking market outcomes to the evolution of technology, resource endowments, and individual preferences. Historians are inclined to consider labour markets sociologically, as one form of human organization among many. They blur, perhaps unconsciously, a line that in the economist’s way of thinking separates a given environment from events generated within it. The divergent visions of what the labour market means historically are rooted in the intellectual traditions of the two disciplines; they echo a fundamental ambiguity in how we understand historical causation. Economics was born under the sign of Natural Law: its vision of historical process is the working out in time of a permanent psychological propensity to optimize the employment of scarce resources. History grew out of the subjectivist revolution of the late eighteenth century, for which history represented the unfolding of latent propensities in the human spirit; each period of history had its own kind of humanity and its own kind of economy. Though rooted in the intellectual life of a time long past, these differences in vision survive to the present day. Economists’ interest in historical labour markets has crystallized around four issues that are clarified, though hardly resolved, by modern economic analysis. The first is Marx’s allegation that the extension of competitive labour markets during the early phases of industrialization reduced the range of economic opportunity for the working poor, in effect turning them into a proletariat with no alternative to working for wages.1 The second took the form of a debate about whether the extension of competitive labour markets lessened or exacerbated economic discrimination based on personal characteristics unrelated to economic performance. A peculiarly American variant of this issue has focused on the economics of slavery and racially based systems of coercive labour management that characterized tropical 1
GEORGE GRANTHAM
and subtropical commercial agriculture in the New World to the late nineteenth century (Engerman 1992). The third is whether the evolution of labour market contracts can be explained by a theory of ‘natural selection’ of efficient or incentive-compatible responses to the environment. The final issue has been stimulated by the persistent inability of contemporary competitive labour markets to achieve real wage levels consistent with full employment. Researchers seek to explain the reasons for and characteristics of unemployment and wage determination over time.2 To professionally trained economists these issues typically pose themselves in the analytical terms of market failure, which is conventionally measured relative to a benchmark of a perfect market, given specified technology, resources, and preferences. Because people can be expected to trade voluntarily with each other when they can gain from the exchange, a market failure implies the existence of impediments—it does not matter whether they are legal, technological, or strategic—to transactions that increase economic welfare. From the perspective of the perfect market, a given outcome that does not conform to the predicted outcome can be explained either as a non-equilibrium state in which the market has not completed its work of reconciling demand and supply, or as an equilibrium in which the optimal allocation has been blocked by institutional or other obstacles. Historical study of labour markets by economists mostly aims at distinguishing between these alternatives.3 In principle the analytical procedure involves defining a market, determining its competitive configuration of prices and quantities, comparing this configuration with the one that historically occurred, and explaining any differences. This approach requires, however, that the market to be analysed be identified in advance. To historians, on the other hand, the scope, as distinct from the existence, of labour markets is problematic. As in the concept of ‘liquidity’ in financial analysis, the ‘scope’ of a market is not easily defined. After all, everyone is ‘connected’ to the market in the trivial sense that there is a price at which it would pay someone to apply capital and labour to arbitrage away any conceivable difference in opportunity cost. Yet, in a non-trivial sense one can assert that the costs of such arbitrage, given the opportunity cost of capital and labour, are a function of the density and quality of information flows among potential participants to an exchange, and it is this density of information flows that historians have implicitly tried to monitor when they write or think about the ‘scope’ of the labour market (McCloskey 1993). It is precisely the scope of the market that is in question when the span of historical time becomes long enough to sustain a fully equilibrated response of institutions and expectations to the environment. In other words, does the labour market adjust to given conditions, or do some conditions adjust to the labour market?
2
THE HISTORY OF LABOUR MARKETS
Economists study how given market structures produce specific outcomes. To historians, however, the question is how that structure came to be what it is. How did labour market ‘space’ vary and in response to what panoply of opportunities and motivations? How and why has the mix of voluntary and compulsory elements in work organization changed over time? Who initiates changes in market scope? Did the integration of new social and geographical spheres come from efforts of employers to find workers or the struggle of workers to find jobs? Did pre-existing regional, occupational, and social structures affect the terms on which wage labour was supplied to industry and did they affect the subsequent evolution of the employment relation? At what dates and by what historical routes did wages and salaries come to constitute the core of family income? These are fundamentally evolutionary questions about the feedbacks from labour market structures to the direction of technological change and the other lines of psychological, social, and intellectual causation that define an economy’s path in time. The first sustained effort to explain the evolution of labour markets was proposed by early nineteenth century German historical economists who, inspired by romantic philosophy, invented a developmental morphology of organizational forms running from self-sufficient households to large firms and world-wide markets in a taxonomical sequence they believed corresponded to the historical sequence of stages in the division of labour. The inner force driving the sequences they called capitalism, by which protean term they meant not only the legal institutions that facilitated exchange and accumulation of private wealth, but also the markets that forced individuals to adopt economic strategies consistent with survival in competitive equilibrium. In a broad sense, the evolutionary theory proposed by the historical economists was an account of how the market came to be the central institution determining the economic welfare of families and individuals (Parker 1986). It was inevitable that an economic history based on a ‘stages’ vision of historical evolution should fasten on problems associated with the transition from self- to paid employment, as the major instance of an historical transformation of organizational forms. It was this transition to which the phrase ‘rise of the market’ usually applied. The dominant concern of western economic history to 1960 was, in a broad sense, labour history.4 Its narrative line commenced with the loss of individual economic autonomy during the Industrial Revolution and progressed through an eventful winning of universal suffrage and the right to organize free trade unions to the achievement of the welfare state. The linking of history to the problems of democratically adjusting social policy to the needs of early twentieth century industrial societies endowed it with urgency and contemporary relevance. As an historical justification of social democratic politics, it was adopted by reformminded intellectuals, many of whom had attended economic and social history seminars in Germany before the First World War, where they learned 3
GEORGE GRANTHAM
to apply the research tools and methods of interpretation of the German historical school to contemporary problems. In the long run, however, the primitive economics of German economic historiography and its inability to explore equilibrium states made it an easy target for economist-historians who began in the late 1950s to deploy formal economic theory to demolish its edifice, an effort subtended by the contemporaneous drift of ideology and opinion towards the virtues of unimpeded competition and the belief that early post-war America represented the end of history. It is nevertheless doubtful that a historiography based on a unilinear sequence of well-defined ideal types, which has implicitly been accepted as a stylized fact by the theorists of the ‘new institutional economies’, is a fruitful way of categorizing institutional change (North 1981). The discovery that western economic history did not logically end around 1960 in Sweden or the United States has encouraged economic historians to take another look at the way labour markets evolved to their present state (Wright 1986, 1987). The papers in this volume are part of the renewed effort by economic historians to recover the markets of our past, employing techniques pioneered by the new economic history to analyse statistical information generated by the great flash of statistical light generated between 1870 and 1930. Much of this information has been buried in the archives of governments and firms since its production. The papers are thus very much a voyage of discovery. We begin with a Mappamundi to this voyage. Like its original, the centre of the map of our historical world is Europe; its proportions are necessarily distended by what is known, and its southern and eastern borders are still empty spaces figured with fabulous beasts. Its cosmography is contemporary social physics; it summarizes ignorance as well as understanding. It guides by exaggeration, as did the charts that led Columbus to believe the ocean was narrow enough to cross. The study of twentieth century labour markets has been stimulated mainly by contemporary problems of economic and social policy, the urgency of which has tended to produce historically foreshortened accounts of the processes that underlie their development. The studies in this volume provide part of the material for a longer view, which, being informed by a sense of movement, will perhaps be less dogmatic. THE ROLE OF SPACE The central interpretative issue in the economic history of labour markets is whether labour market institutions are a natural consequence of adjustments to exogenously determined technology and individual preferences, or whether technology and individual preferences are themselves significantly affected by the structure of labour markets. Focused on factors that determine price, the analytical tradition in economics usually takes tastes and technology as given, ignoring feedbacks other than those directly connected to the formation 4
THE HISTORY OF LABOUR MARKETS
of market prices. This makes taste and technology the ultimate causes of observed patterns of resource allocation and the distribution of income and opportunity. This neat analytical distinction between exogenous and endogenous factors has long seemed to trained economists to be the only logically defensible way to analyse market processes, including processes that occur over long stretches of time. Its great value is that it permits a clean separation of cause from effect. Any other démarche would be considered by most economists to be muddled and even mystical. That the scope and form of labour markets may reflect influences besides technology and tastes, however, in no way means that a transcendental spirit animated their history; only that the individual decisions creating the history were not predicated on a full knowledge of all the possible consequences. When this happens, the feedbacks from decisions to environment create unforeseeable evolutionary paths, which had they been foreseeable might have led to a different set of decisions. The information flows that govern the most important decisions in the labour market—the kind of training to invest in, the kind of firm in which to apply for work, the kind of life cycle to plan for—are still most effectively transmitted by word of mouth and personal impressions and therefore operate across short distances. The externalities that have had the most influence on the scope of labour markets thus tended to present themselves at the level of the state or region rather than the whole intercommunicating world. This regionalization of labour markets is most obvious in the sphere of social regulation of the conditions of employment. Laws and social attitudes governing immigration, permitting or forbidding discrimination (see Chapter 13), regulating the employment of children and women, hours worked and other conditions of employment (see Chapter 8), as well as the macroeconomic policies pursued by government (see Chapter 11) all impose national or cultural definitions on labour markets. That people prefer to restrict job search to regions defined by linguistic and religious communities is another example of how labour market scope and behaviour reflect attitudes growing out of common values imposed by religious and social institutions and by the history that people have been taught to believe is their common heritage. From the standpoint of technology and innate preferences, however, there is nothing ‘natural’ about these communities. Different political histories give rise to different ‘natural’ groupings of people (e.g. Shiells 1990). Historically, the discontinuities and feedbacks that sustained the informational structure of labour markets have been rooted in geographical externalities. Among populations that are not continuously mobile, regional economic specialization gave rise to communities whose common history led to common preferences with respect to wages and working conditions. Such learned attitudes and expectations produce region-specific life cycle patterns of labour supply. For example, the kind of skills young people are 5
GEORGE GRANTHAM
likely to acquire reflects the quality of local schooling and local possibilities of learning by doing, which in turn reflect local opportunities for employment, which in turn is a function of the regional division of labour. Being intrinsically spatial, such facilities for learning and developing human character are fundamentally public goods (see Chapter 13). Everyone in the region is exposed to them, whether they like it or not. Although in theory the whole world with which a region communicates constitutes the opportunity set for young people when they are making up their minds on what they want to do in life, under normal circumstances the ties of place and family are too strong for working people to sort themselves geographically on the basis of individual preferences about the kind of public good they prefer. Most people have to take what is available in the region where they were raised.5 Consequently there is no reason to believe that the distribution of regional ‘environments’ reflects an optimal outcome with respect to exogenous preferences; and of course, children do not ‘sort’ themselves, but are sorted socially and educationally by their parents. Moreover, the kind of job people find is often determined by information they obtain from friends and relatives whose judgement they trust and share. Such privileged channels of job information are especially important for communicating non-local employment opportunities.6 Finally, employers form expectations about workers and discriminate among them on the basis of geographical and social markers. In geographically extensive labour markets, they recruit in places that are known to ‘produce’ certain types of worker. A similar dynamic governs the evolution of contractual forms. The labour market thus reflects strongly self-reinforcing learning processes on both sides of the market (Whatley 1990:43–56). Geographical externalities also determine an historical relationship between the evolution of labour markets and patterns of technological change. In the past and to a considerable degree even today innovations and industries displayed strong tendencies towards spatial clustering (Krugman 1991). The reason innovation clusters are initially strongly region specific seems to be that communicating complex pieces of information requires prolonged personal contact among the innovators. The subtle and idiosyncratic bits of technological information that constitute the raw material of invention are not easily transmitted by more formal means of communication. In this century the sensitivity of technological change to small variations in the economic environment has been surprisingly high. For example, the labour-using bias of Canadian industrial technology relative to that of the United States developed in response to a moderate gap in wages when the United States closed its frontiers to mass immigration after the First World War (Wylie 1989). This example of technological malleability causing a divergence in the technology of countries whose economies are very similar illustrates why one cannot assume that the technological constraint is exogenous. When the period of adjustment is long enough to allow for all 6
THE HISTORY OF LABOUR MARKETS
the equilibrating responses to work themselves out, regional differences in technology can be expected to emerge. The Canadian case illustrates the importance of public choices on the evolution of these constraints.7 The history of American manufacturing provides an important example of how the structure of labour markets influenced the development of a specific kind of technology, which in turn reinforced many distinctive features of the American labour market. American manufacturing from its birth was biased towards using unskilled labour. This was largely because the United States had virtually no indigenous tradition of high craftsmanship maintained by an effective system of long apprenticeships, nor a structure of demand capable of sustaining a small class of highly skilled workmen. Such a tradition could hardly be expected to take root in a newly settled agricultural colony that satisfied its limited demand for high-quality manufactures by importing them from Europe (Elbaum 1989; Rorabaugh 1986; Douglas 1921). In the absence of a domestic supply of highly skilled craftsmen, the young Republic could have continued to import manufactures from Britain, in line with the comparative advantage of the two nations. But politics intervened, as it had in the colonial period. Between 1800 and 1830 the federal government took two decisions that were to give rise to a uniquely American technology. The first was to subsidize a domestic small-arms industry; the second was to protect an infant cotton textile industry that had sprung up during the years when trade with Britain was interrupted by embargo and war. The resulting expanded demand for skilled and semiskilled labour to make large quantities of standardized items like spindles and gunstocks in a society with few craftsmen led to the extended development of templates and lathes that could be pre-set to reproduce shapes in wood and metal (Rosenberg 1972; Parker 1987; James and Skinner 1985). The first crude machine tools invented as a result of this effort permitted employers to substitute abundant raw materials and unskilled workers for skilled labour across a spectrum of mechanical operations that was narrow at first but progressively widened as the technology was perfected. The drive to economize scarce supplies of skilled labour is also evident in changes in the organization of the workshops. Where it was impossible to substitute a machine for a man, manufacturers divided the work by hiving off the simpler operations to unskilled workers.8 By imposing a division of labour along the lines of the level of skill rather than, as in the European craft tradition, along the lines of the type of material to be worked up, the United States took a precocious step in the direction of serial production of manufactured goods. The critical feature of all these changes is not that they saved labour per se in an economy with a high ratio of land to labour (Habakkuk 1962), but that they saved skilled labour in an economy that had never developed institutions to ensure the formation and transmission of the higher reaches of traditional craft skill. 7
GEORGE GRANTHAM
The initial bias towards mass production using semi-skilled labour imparted by the early development of machine tools had a decisive influence on the subsequent evolution of American industrial labour markets. It is not as well known as it should be that most industrial workers in the United States at the time of the First World War were either immigrants or children of recent immigrants. The overwhelming majority of native-born Americans who fought in the Civil War were farmers and craftsmen to whom the mind-and bodynumbing work of smelter, forge, and factory held little appeal. By the early twentieth century a taste for independence and the economic means to pursue it had produced an occupational segregation of the native white population—descendants of pre-Revolutionary immigrants—into the WASP fortresses of family farming, small business, white-collar work, and the liberal professions, leaving to outsiders—later immigrants, blacks and towards the end of the nineteenth century impoverished and illiterate Appalachian hill people—the work that was paid by piece or by hour.9 Indeed, much of the opposition to mass immigration from non-Anglo-Saxon Europe by early twentieth century progressive intellectuals stemmed from the belief that it was linked to poor—i.e. ‘un-American’—working conditions (Commons 1907). In the absence of a sufficient supply of industrial labourers from the original stock of white Americans, however, industrialists seeking to exploit the unprecedented opportunities for expansion turned to foreigners. The ‘American System of Manufacturing’ made it technologically feasible for them to expand by drawing on the huge reserves of undemanding labour from the small peasant holdings in French Canada, Ireland, Italy, and eastern Europe (Byington 1910). Skilled workers were recruited when needed from Germany and Great Britain; the others could be set to work with minimal training on machines designed to be used by semi-skilled labour. The tendency of nineteenth century American manufacturing technology to make maximum use of unskilled labour was confirmed in the last quarter of the century by the integration of the American industrial labour market with the labour markets of eastern and southern Europe. It is hard to imagine that the extreme subdivision of tasks resulting from Taylor’s time and motion studies could have been forced on native American workers the way it was imposed on the overwhelmingly immigrant industrial labour force of the early twentieth century United States.10 Ford’s assembly line was simply the apotheosis of a long line of technical and organizational development encouraging rapid insertion of unskilled workers into manufacturing processes. The ease with which workers could replace each other in unskilled operations—in economic jargon their high degree of fungibility—influenced the structure of the labour market in which they were recruited. Tasks that demanded little training were filled and refilled from the ever changing pool of workers in the spot labour markets, making it possible for firms to tailor hiring to seasonal and cyclical fluctuations in the demand for their output.11 The availability of short-term employment attracted workers with weak 8
THE HISTORY OF LABOUR MARKETS
attachments to regions or industries and dissuaded those who might have preferred a more settled existence.12 American industrial technology was probably uniquely adapted to a labour market that was characterized by high turnover and a predominantly immigrant workforce that had selected itself into a labour market offering high wages in return for intense effort, a trade-off that suited the majority of men who intended to stay only a few years in the New World before returning to their villages and farms at home. High turnover rates and immigrant labour facilitated and possibly necessitated the development of authoritarian styles of labour management and hierarchical chains of command.13 Once in place, however, machines and management traditions linked to this kind of labour market proved difficult to change in the face of a fundamental alteration in conditions of labour supply, and when free immigration from Europe ceased in 1923, American steel makers, rather than adapt new methods of recruiting and managing labour to the needs of a less mobile labour force, looked to Mexico and even contemplated importing East Asian labour to preserve a system of labour management and a production technology predicated on unlimited access to unskilled males who expected their connection to the industry to be temporary (Brody 1960:266–7). The lines of causation that sustained the American structure of hiring, firing, and promoting workers also led into recesses of consumer mentality that are usually taken to be an ultimate constraint on the performance of an economic system. Behind the protective wall of the tariff—a collective choice—American consumers were presented with standardized products that far from matched the quality of similar articles produced by the more highly skilled European industrial workforce. The failure of an indigenous system of craft apprenticeship to take root meant that quality differences between so-called mass-produced articles and those produced by local craftsmen were narrower in relation to their relative price than in Europe. As a consequence, American manufacturers found it easier to organize production by means of an extended division of labour at single locations.14 The limited range of goods produced by a highly centralized system of manufacturing reinforced social norms of consumption that placed a low value on differentiated personal consumption—a curious outcome considering that non-material marks of status were largely absent in a society that was, in its own eyes at least, born free. Relatively undifferentiated tastes in turn increased the scope for mass production technologies making use of unskilled labour (Rosenberg 1976: ch. 8). A Marxist might argue that this merely demonstrates the ability of capitalist employers to bend tastes and technology in the interests of exploiting an undifferentiated supply of unskilled labour; but equally capitalistic European manufacturers did not accommodate their societies to the economics of mass production until after the Second World War, although in the nineteenth century their technological prowess exceeded that of Americans and the profits of largescale production could have been no less. 9
GEORGE GRANTHAM
The ‘rise of the labour market’, then, was not a movement along a single axis from less to more perfect markets, but rather manifested itself in a sequence of changing configurations of workers’ characteristics, working conditions, modes of payment, types of job attachment, and varying extent of geographical and social job mobility, all of which were influenced by the feedbacks operating through public policy and geographical externalities. The evolution was the result of specific sequences of opportunities and responses. The most important opportunities were those making it profitable to organize enterprises whose demand for labour exceeded what could be supplied from within the households that owned or operated them. It was large-scale production—large from the perspective of economies that had traditionally produced most commodities in households—that created the economic basis for the formation of labour markets. The commercial opportunities underlying large-scale production manifested themselves as a sequence of impulses: population growth, market-induced specialization, geographical discovery, and technological change. The conditions governing the supply of labour reflected the peasant origins of the European population. Although vulnerable to war and demographic catastrophe, this population belonged to a social order that had an internal demographic logic, life cycles, and implicit contracts with their neighbours and governors. Potential employers of wage labour had to compete with an alternative structure of incentives—what one might call today a life-style. The relation between this original social structure and the expanding market for labour services is the subject of the remainder of this chapter. LAND HOLDING AND EUROPEAN LABOUR MARKETS 800–1800 In seeking to understand the dynamics of labour market evolution one cannot escape Marx’s question: under what circumstances did labour markets emerge? Apart from informal exchanges of favours that are the common heritage of mankind, trade in labour services during the formative period of modern European society mainly served to redistribute young people among farms and to accumulate workers for the harvest. Even in egalitarian societies the hazards of birth endow families with too few or too many hands in relation to the size of their holding. Local exchange of labour services was a way of compensating temporary farm-level imbalances that averaged out at the level of the community. Early European agrarian society was of course by no means egalitarian, and labour services therefore typically flowed from small to large holdings. The flows were modest because in the absence of outlets for agricultural produce there was little inducement to organize large operations. In principle, small adjustments in the terms of service for a renewable cohort of mobile young people could sustain a steady-state balance of demand and supply across a wide range of farm-size distributions.15 Stronger 10
THE HISTORY OF LABOUR MARKETS
impulses emanated from the demand for food in large early medieval establishments. Before the late eleventh century monastic and the more important seigneurial households constituted the only geographically stable concentrations of people capable of aggregating demand for food to a level that could support agricultural organization above the level of the individual peasant household (Russell 1972:35). Farms organized to victual early medieval households could have been worked by dependent cultivators, but the supply of labour secured by compulsion was never so perfectly tailored to the level of demand that the masters of land and men could avoid recourse to more flexible methods of recruitment. Or it may be that compulsion was an instrument to secure a flexible supply of labour on demand, as the one or two days a year demanded from the majority of dependents of the ninth century Abbey of St-Bertin seems to imply (Ganshoff 1975; Delatouche 1977). As early as the ninth century the Benedictine monks of Corbie held an inventory of sixty pence in coin to pay workers engaged to spade, weed, and harvest (Duby 1972:109). We know nothing about the civil status of the workers the monks hired, but they were probably serfs. The main point is that a farming system with no outlet for surplus agricultural product had little demand for large numbers of employed workers, whatever their civil status. The crucial factor causing the informal labour markets of a parish or seigneury to expand beyond the size needed to work demesnes organized to victual companies of monks and knights was the growth of towns. An impressive body of evidence now demonstrates that between 1150 and 1250 northern Europe’s agricultural space was radically reorganized in relation to new urban markets for farm produce (Campbell 1988, 1992). The growth of commercial outlets for food and fibre differentially favoured large farms because of the scale economies they enjoyed in the employment of draft animals and permanent hands.16 As the opportunities for large-scale farming enterprise expanded, so did the demand for agricultural labour. However, regional specialization also accentuated the seasonality of local labour demand. Always present but muted by the modest level of specialization occasioned by rural commercial demand for foodstuffs, the seasonality embedded in a technology that used animals to plough but required men and women to reap was to have a determining effect on the structure of European labour markets (Grantham 1993). The large seigniorial demesnes were well situated to profit from commercial opportunities because of their size and financial resources; faced with an opportunity to sell large amounts of grain, meat, and wool for cash, the seigneurs attempted to turn the inherited manorial framework to commercial advantage by reviving and extending the incidents of serfdom (Stacey 1986:926; Biddick 1985). By the end of the twelfth century, however, the lords’ ability to extract labour from a dependent peasantry was being limited by the codification of peasant status that had accompanied the development 11
GEORGE GRANTHAM
of official literacy and the professionalization of the different spheres of law (Berman 1983). Freezing the conditions of servile tenure inevitably limited the flexibility of labour supply obtained through manorial institutions. When cultivation of the demesne ceased to be guaranteed by the labour of serfs, manorial authorities lost the main reason for restricting the alienation and subdivision of tenements owing labour services. At the same time, the embryonic labour market allowed peasants owing labour services to hire substitutes, a practice that became more common as tenancies passed into the hands of families with no personal ties to their lord. Thus, even where there was no outright commutation of servile labour dues labour markets grew in the interstices of the old system, and as they grew, employers found it easier to obtain the labour they wanted through markets than through the cumbersome and old-fashioned mechanisms of serfdom. It was probably the system externalities of labour markets more than the relative efficiency of serf versus hired labour, that accounts for the rapidity of the transformation. In eastern Europe, where independent legal traditions protecting peasants from their lords failed to take root, the lords’ attempt to bend serfdom to commercial ends triumphed (Miskimin 1977:56–64). The economic history of the twelfth and thirteenth centuries is still too poorly understood to permit unqualified generalization about how fast a class of wage-dependent families emerged from the population of dependent manorial cultivators. The disintegration of coherent family holdings between the tenth and the thirteenth century is one of the murkiest regions of agrarian history. What is certain is that the bulk of the European labour force was recruited from the debris of the manorial tenantry, and that most labouring families in the thirteenth century possessed land. The extinction of feudal and family encumbrances to land transfer accelerated the subdivision of peasant holdings and multiplied the number of families who had to supplement the product of their fields with earnings from other work. It is misleading to call this process proletarianization, since so many farm labourers in the early Middle Ages had been slaves and so many workers in the later Middle Ages held land. The smallest holdings were not as a rule absorbed by larger ones. Instead the land market bifurcated on a fault line drawn at the minimum size capable of occupying a team of draft animals. Holdings above the line hired workers, holdings below the line supplied them. The larger holdings thus comprised the locus of labour demand and their possession was restricted to a narrow class of buyers and renters. The small plots, on the other hand, were bought and sold by smallholders and workers who lacked the means to stock farms. These holdings often turned over rapidly, reflecting the economic precariousness of their owners (Béaur 1984; Bottin 1983). The advantages to large farmers of an institutional arrangement permitting them to make intermittent calls on labour of nearby smallholders were so overwhelming that to the twentieth century they continued to sublet allotments 12
THE HISTORY OF LABOUR MARKETS
to their workers in exchange for the right to demand labour from them at will (Pinchemel 1957:87). Such arrangements perpetuated the medieval compromise between small-scale family enterprise and large-scale farms.17 As long as agricultural production involved intermittent and unpredictable moments of intense activity, agricultural labour was treated by the main employers as a ‘fixed’ input, and therefore one whose annual subsistence had to be maintained by one device or another. The problems this posed were greatest for crops that imposed great seasonal variations in the level of labour demand. When regions specialized, the range of crops grown narrowed, which in turn tended to accentuate seasonality and aggravate the problem of securing temporary supplies of labour.18 The technological obstacles to mechanizing the harvest do not, however, fully account for all the features of the labour market that issued from the medieval agrarian settlement. General conditions governing the nature of labour supply also mattered. The most important of these general conditions was the peasant origin of most of the European labour force. By the force of things, they had only small amounts of land.19 Owning land but not enough of it to cover subsistence, or at least to produce enough surpluses to pay for the minimum amount of tradable goods (including tax payments) they needed, smallholders had to engage in multiple activities—in French the generic term for this was Industrie, but a more telling English rendering is ‘makeshifts’—to make ends meet (Meuvret 1987). They inserted themselves into the economy providing products and services to be exchanged ultimately for food. The family holding was the fixed point that set the conditions for a long standing jointness of seasonal and family labour supply. Around its social and geographical kernel the family constructed a portfolio of complementary employments, supplying labour at different times of the year to different activities and pooling incomes. Participation in the labour market thus reflected a portfolio balancing act allocating labour across activities having different combinations of earnings and risk. It is probable that in this portfolio, wage income was especially risky. This proposition runs counter to a long tradition of contract analysis that views the wage as a fixed obligation of employers, in contrast to other possible arrangements for sharing risk between principals and agents. However, the constructions upon which such arguments are based usually abstract contracting arrangements from the markets in which they were typically concluded. In late medieval and early modern times, product markets were unstable in the short run and labour attachments of individual workers to individual employers were intermittent. A wage that depended on a continuously fluctuating demand for labour hired for short periods of time cannot be taken seriously as a guarantee. The jointness of labour supply created other problems. If the supply of labour by individuals or families to different market employments is truly joint, decline in demand for labour in any one employment can reduce the 13
GEORGE GRANTHAM
labour locally supplied to all the others, disrupting a whole chain of local labour markets.20 Disruptions to the normal flow of labour through seasonal and geographical markets must have occurred frequently, and did not conduce to labour market stability. Rural industry also tended to give smallholders unstable terms of employment, because an activity that depended for finance almost exclusively on private credit arrangements among merchants was highly vulnerable to financial panics and crises of confidence. Labour markets that had evolved as congeries of temporary employments involving spatial and sexual recombinations of the workforce did not adjust smoothly to shocks in demand. The chronic instability of wage incomes increased the value of holding land to smallholders, because it gave them direct control over the employment of their labour and, subject to climatic disturbances, of their income as well. Paradoxically, the possession of a holding often worked to extend shortterm labour mobility, as insufficient annual income periodically set family workers into motion to seek in distant labour markets the resources they needed to preserve and extend independence at home. For a long time, this kind of temporary migration dominated long-distance mobility, voluntary permanent migration being mainly confined to persons lacking the minimal wealth needed to establish a foothold in their community of origin.21 The most important consequence of the European agrarian system of large farms surrounded by a penumbra of smallholdings was that it continuously held out to the labouring classes the vision of family farming as an alternative to full-time participation in the labour market. This is the main reason why from the thirteenth to the twentieth century, the significant fact about competitive labour markets in Europe was that most workers did not want any part of them. The determination of ordinary people to ‘escape’ a life where their income and status was determined in a spot market may derive its psychology from the desire for autonomy and the uncertainty of wage income streams, but widespread landholding made the desire to escape a plausible goal. 22 Peasants in early industrial Europe hoped that by economizing they might acquire enough land to climb the agricultural ladder to a rung of self-sufficiency, and the fear that loss of this precious asset could put them back at the mercy of market forces beyond their control and understanding encouraged them to engage in multiple activities, including work for wages in labour markets to maintain and augment their holding. Wage labour was thus instrumental, not so much to maximizing income, as maximizing the chances of obtaining or keeping a way of life. Wage labour must thus be considered one of several omnipresent gainful alternatives to subsistence production that presented themselves in varying proportions to a class of poor people, who were not yet and in many places not ever a true proletariat, whose numbers and economic condition varied greatly from place to place but whose type was determined by their possession of a literal pied à terre. This class was the fixed point in the labour history of 14
THE HISTORY OF LABOUR MARKETS
western Europe to which other institutions had to adjust when technological and market opportunity expanded the scope for large-scale production. THE EXPANDING MARKET FOR LABOUR SERVICES There are three important consequences of a labour market in which workers retained strong links to the land. The first was the organizational and technological response to the seasonality in demand for labour; the second was the structure of labour markets that developed in rural industry; the third is reflected in solutions to the problems faced by early entrepreneurs in recruiting labour to mines and factories. We begin by considering the effects of seasonality. Seasonality and the Labour Market Until recently agricultural and industrial activity were characterized by significant seasonal variability. Some of it was caused by natural phenomena. Commercial food processing peaked after the harvest; the charcoal iron industry shut down in winter; grain mills were arrested by lower water and prolonged frost; shipping and fishing schedules responded to seasonal shifts in the direction of winds and the annual migrations of fish. The demand for coal and coal miners peaked in late autumn and early winter, that for ice and ice-men in summer.23 Social seasonality caused firms to schedule production to take advantage of moments when farmers and workers had cash in their pockets and to build up stocks of fashionable goods in anticipation of the autumn and spring buying seasons. Few sectors had such evenly distributed work throughout the year that seasonal bottlenecks in labour supply failed to appear when output increased rapidly (Douglas and Director 1931:48–50, 82–100). Seasonality was therefore a consequence of economic expansion as much as a by-product of a particular technology. Employers responded to increased seasonality in the demand for labour (not necessarily their own demand, since they had to compete for labour with any employer experiencing a seasonal rise in labour demand) in three ways. The first was to hoard labour; the second was to adopt labour-saving technology; the third was to hire labour on the spot market as needed. The three responses implied different labour market structures. The first response in effect made the labour force a fixed input, forcing employers to find ways to occupy or otherwise subsidize employees in the off season and creating conditions for a permanent employment relation and internal markets.24 Workers generally liked this strategy, which guaranteed their way of life. Innovation was rarely profitable at the level of the individual enterprise because of the low probability of success, unless it was a question of introducing an already tested technology, but extreme bottlenecks nevertheless brought out inventive effort. Workers, however, tended to oppose labour-saving innovation that made them into a variable rather than a fixed 15
GEORGE GRANTHAM
input or that reduced their control over the work process (Lazonick 1979, 1981; Robertson and Alston 1992). The option of the spot market was in principle the most efficient as long as employers could be assured that workers with the requisite skills and aptitudes would be available as demanded. This condition characterized both the high and low ends of the skill spectrum. But with the exception of the highly skilled and often highly mobile skilled craftsmen, workers resisted full-time commitment to the spot labour market. The agricultural solution was at first to hoard labour. Until demand got so large it could not be satisfied from local families, harvest hands were overwhelmingly drawn from smallholders and cottagers; but as regional specialization increased, the labour from local sources of seasonal supply became increasingly inadequate, giving rise to longer migrations of harvest workers. This is evident in the difference in the geographical scope of the markets for hired hands and harvest hands. Whereas as late as the seventeenth century, the average distance between successive jobs of permanent hands hired by the year was only 5 or 6 kilometres, by 1300 manors in Norfolk were attracting harvest workers from Wales 250 kilometres away (Kussmaul 1981; Dyer 1988; Armstrong 1989:661). Normal wage hierarchies were deformed at harvest time. Under the pressure of harvest demand the normal gap between male and female wages disappeared and the usual relation between industrial and agricultural wages was reversed (Penn 1987; PostelVinay, Chapter 4). Overlapping seasonal and non-seasonal labour markets with distinct patterns of spatial and social integration were to last a long time. It reached a climax at the start of the twentieth century when the hyperdevelopment of specialized grain-growing belts gave rise to international and intercontinental movements of harvest labour. Harvest hands were brought in by railway to northern France from Galicia; Italian peasants made regular round trips to Argentina to harvest wheat during the European winter. Such movements replicated on a world scale what had earlier been possible only on a regional one, and the underlying mechanisms on both the supply and demand side were much the same (Grantham 1993). In Europe the agricultural and industrial specialization that was set in motion by the growth of the market after 1600 can be seen as linked phenomena, in the sense that the sectoral demands for labour were often seasonally complementary. This complementarity is usually taken to be a ‘natural’ adjustment to technology, but it is hard to believe that the complex systems of labour migration that balanced demands and supplies throughout the early modern economy did not take considerable time to establish. The gradual extension of the spatial and occupational division of labour had allowed time for institutions and networks of labour exchange permitting the different sectors of the economy to take advantage of complementarities in the underlying seasonality of labour demand in cereals to develop (Grantham 1993:495–6). As Postel-Vinay demonstrates in Chapter 4, the 16
THE HISTORY OF LABOUR MARKETS
adjustment to seasonality of agricultural demand for labour affected many rural factories, which regularly shut down to accommodate the flow of workers into the fields.25 Even in a society as highly urbanized as early twentieth century England workers continued to make their way from garden plots and shops to the hop, grain, and fruit harvests (Sturt 1984). The variety and complexity of intersectoral and inter-regional labour flows before the twentieth century reflected mutual adjustments in technology and organization that could only have been made possible by a long history of commercial development. The analysis of these flows and their relation to macroeconomic adjustment is still in its infancy. Some of the problems of adjustment can be seen in the difficulties that emerged around the middle of the nineteenth century, when the separation of agriculture from industry accelerated in the industrializing nations. Seasonal pressure in continental labour markets peaked in the 1860s, just as the production of cereals was peaking and the railway boom was destabilizing the market for unskilled rural labour.26 The pressure in Europe was relieved in the following decade when grain began to pour in from abroad, with the result that the seasonal labour bottleneck shifted to the North American high plains (Engerman and Goldin 1991). Indeed, at the end of the 1860s the railway construction camps were having difficulty holding their immigrant labour force in the face of harvest wages twice the construction wage (Erickson 1957:133). The United States represents a situation in which agricultural specialization occurred too rapidly to allow time for the growth of institutions facilitating complex intersectoral flows of labour, and for this reason, farmers had a strong incentive to substitute machines for men at the harvest peak. Fortunately, the emergence of specialized grain belts in the United States coincided with the early industrial breakthrough in mechanical technology making possible the mass production of reapers, mowers, and other machines designed to reduce the peak demand for labour (Parker 1972:379–86). The rise of an agricultural machinery industry designing its product to meet the particular requirements of the American family farm is usually considered a triumph of nineteenth century American ingenuity, as indeed it was; but rapid specialization had its dark side. Hannon noticed that in western New York State the transition to specialized cereal production was accompanied by increased income instability among wage earners and a rise in the relief rolls (Hannon 1984). Specialization was associated with a variety of institutional forms. The delicacy of fruits and vegetables prevented the introduction of machines in picking. In the United States, the opportunities created by refrigerated railway transport provoked a search for stoop labour that was finally satisfied by teams of native Indian, black and Mexican migrant workers, moving like a pendulum in periodic swings across the fruit and vegetable belts of the nation, unhoused, uneducated, and by the standards of American life, virtually unpaid (Wright 1988; Fisher 1953). In Europe, a similar growth of commercial 17
GEORGE GRANTHAM
demand for fruit and vegetables was met from small farms using family labour. One technology, two vastly different social results within a common culture of market capitalism. A similar pattern of divergence between Europe and North America can be discerned in responses to industrial seasonality. The key variable was the nature of the labour force. Because American industrialists long enjoyed access to migrant workers whose primary goal was to make money as fast as possible and then return home, they could hire and fire most of their temporary labour in the spot market. The spot market for labour was important in the early part of this century, because many industries were still characterized by strong seasonal patterns of activity (Slichter 1919). Europe also possessed spot markets for industrial labour, often filled by Italians and in Germany by eastern Europeans, but the rapid growth of American industry after 1880 drew in so many migrants that firms were able to go much further in the direction of organizing production on the basis of shortterm hires. A labour force for whom the wage was the decisive element of a labour contract posed no obstacle to managerial policies that tailored labour demand to seasonal fluctuations in product demand. American firms began to hoard labour only when the market no longer provided an elastic supply of short-term hires. Thus, when the traditional supplies of casual labour dried up after the First World War, American firms found they had the technology to provide workers with more continuous work, and the seasonal variability of employment decreased sharply (Douglas and Director 1931; Brown and Phillips 1985; Woytinsky 1953). It is probably no coincidence that the end of mass migration to the United States occurred in the same decade that economists have detected a marked deterioration in the cyclical sensitivity of wages that appears to have been uniquely American (Sachs 1980; Gordon 1982). The migrant labour market was a true spot market; native-born workers placed a higher premium on stable employment relations. Seasonality, then, and its contractual fellow, temporary employment, were thus only partly technological in origin. To be sure, without seasonal bottlenecks there would have been no seasonal problem, but the technological constraints manifested themselves only when output was expanding rapidly or when production was becoming regionally concentrated. But there were many possible solutions to the problem of seasonality, and the ones chosen were not always optimal from a long-run or social standpoint. Did workers sort themselves into different employment patterns, or did employers sort regions and classes in order to find workers predisposed to loose employment affiliations? How binding were the technological constraints? What substitutions in output mix might have alleviated them? The answers to these often unanswerable questions are important, because there was a world of difference between societies that functioned on the basis of an ever mobile labour force with no ties to place-specific community 18
THE HISTORY OF LABOUR MARKETS
assets, and one that had an interest in developing and maintaining those assets. Home employment and rural industry For about 600 years—from the middle of the thirteenth to sometime after the middle of the nineteenth century—the expansion of European manufacturing usually meant expanding rural outwork. Rural industry originated when rising costs of production in the textile cities induced merchants to ‘source out’ simpler tasks to rural households who worked at them in their spare time. Two peculiarities of the market for rural outworkers call themselves to the attention of the modern economist. The first is that work was paid by the piece, which was the only feasible way of remunerating persons whose work could not be directly monitored (Milward 1981).27 The second is that the real income of domestic manufacturers was extremely variable because the price of foodstuffs in isolated rural markets was highly variable. Both characteristics had important consequences for the behaviour of workers and families in the market for this kind of employment. The implication of piece rate payment under the organizational conditions of rural industry was that fluctuations in the product market were transmitted back to workers in the form of wage fluctuations, as well as fluctuations in employment. This was in marked contrast to the short-term rigidity of nominal wages elsewhere in the economy (Bythell 1969:99; and see Chapter 3). Variable piece rates were an inherent feature of a form of industrial organization in which the supply of capital or stock in trade was more elastic than the supplies of labour it set to work. Rural industry was geographically highly specialized by product line, to the point where entire villages were occupied in manufacturing single objects (Engrand 1979). As a result, districts that had become hyperspecialized through concentration on cottage industry tended to experience ‘labour market congestion’, something usually associated with modern industry in times of shortfalls in demand. Rural labour market congestion was a predictable consequence of the spatial development of skill and credit lines tied to particular lines of manufacturing. Rural industry was an outgrowth of mercantile networks. Because most tradables prior to the middle of the nineteenth century were high in value relative to bulk and made little use of non-mobile factors they tended to locate in response to the dynamic economies of agglomeration rather than to the comparative advantages growing out of static resource endowments.28 Apart from seasonal work, agriculture could not absorb many unemployed workers during periods of industrial downturn. Had the supply price of rural outworkers been determined by the agricultural wage, fluctuations in demand for rural manufactures could have been absorbed by short-term movement of labour between the two sectors, which would have stabilized wages and real incomes in rural manufacturing. Instead, declining 19
GEORGE GRANTHAM
industrial demand generated high local unemployment among the population of domestic outworkers unless they possessed enough land of their own to cushion the shock. The problem of rural labour market congestion was greatest when contraction followed a long period of expansion, because rural industrial families had higher fertility than agricultural families. The congested linen-manufacturing districts of Flanders, which had grown up in the late seventeenth and early eighteenth centuries and were undermined by the expansion of cotton at the end of the eighteenth, are a good example of this. The Flemish cottagers constituted a pool of poverty and resentment that lasted a century (Mokyr 1976; Terrier and Toutain 1979). In this respect, the contraction of cotton handloom weaving in Lancashire in a decade and a half was an outstanding achievement, aided by the nearby expansion of weaving in Manchester (Brown 1990; Lyons 1989). Rural industrial households were vulnerable to harvest shortfalls. Because the districts of rural manufacturing were usually located away from the main grain-raising regions, they had to be provisioned by local producers who marketed their produce through small market towns. In periods of severe shortfalls, however, these local markets were often integrated into the larger metropolitan markets for grain, onto which the urban demand for foodstuffs was projected with all its buying power. The resulting outflow of grain from the towns shifted the burden onto the rural industrial families. Most food riots in the seventeenth and eighteenth centuries occurred in districts where farming was not the main occupation. In this respect, urban workers were far better placed than rural workers because urban provisioning networks were more secure (Grantham 1989). The real incomes of rural industrial workers were therefore often unstable, which raised the value of land as a hedging asset. Although it was geographically dispersed, rural manufacturing possessed characteristics that recall modern labour markets. One observes similar patterns of recruitment, although the organization of production remained decentralized in individual households. Merchant-entrepreneurs actively recruited workers in regions not yet colonized by this form of industrial organization in an effort to secure cheaper labour. They trained workers, extended them credit, and even stinted outwork among them during downturns in trade. Except for men with especially scarce skills, however, income guarantees were usually abandoned once a rural manufacturing population began to reproduce itself demographically, as they were no longer necessary to attract workers into the industry (Bythell 1969). Rural industry married highly mobile capital with immobile labour, an arrangement that shifted most of the risk onto workers. In the factory age, a greater share of the short-term risk was borne by owners. As a stage in the evolution of industrial organization, then, rural industry responded to the high cost of transporting food relative to the cost of moving intermediate and finished manufactures. It maintained the family economy and was flexible enough to 20
THE HISTORY OF LABOUR MARKETS
permit the seasonal intersectoral migrations necessitated by the rise of specialized grain growing. The instability of rural industrial income reinforced the importance to working families of maintaining multiple sources of income. This hedging solution was to give way in the face of the more permanent and profound division of the rural and urban industrial labour force that resulted from the rise of modern industry. The labour market and the factory Between 1780 and 1920 the thrust of technological change was to intensify the spatial concentration of manufacturing. Industrial regions contracted around ports and into coal-mining districts, while industrial production was increasingly concentrated into large plants.29 This massive alteration of economic geography had a major effect on labour markets: it permanently displaced working families from the countryside to the towns and led to the rise of working-class life cycles founded on expectations of permanent attachments to firms and industries. Physical migration of whole families was the crucial event in this transformation; when families moved, they transported with them a bundle of intergenerational expectations that had to adjust to the possibilities of their new environment. The rural-urban migration thus could not occur until the urban labour market offered the same degree of income and social security that membership in a rural community, for all its inequalities, still afforded. The paradoxical mixture of high temporary mobility of workers and the immobility of families in rural agrarian society was sustained by an equilibrium of mutual expectations about the obligations of family and community to provide aid in times of difficulty. Such expectations were place specific, as were the entitlements of industrial society prior to the twentieth century invention of the national welfare state. Down to the Second World War capitalists and entrepreneurs also tended to be regionally sedentary; they shared a community of interest with local businessmen, professionals, and their permanent employees. These ties constituted so many implicit contracts. Labour recruitment into urban factories, then, had at first to compete not only with the level of rural wages, but also with the implicit rural networks of social security and opportunity. In analysing the difficulties early factory masters had meeting the reservation price of the rural labourers they wanted to attract, one is led to ask why the difficulties of recruiting a stable labour force were not resolved by paying compensating wage differentials. This is a question badly in need of detailed research, but one reason why high wages failed to reduce turnover may be that exclusive reliance on the money wage to recruit workers induced adverse selection. The section of the rural population that was most sensitive to the money wage was the ‘footloose’ part, consisting of persons who as a rule had failed to establish themselves in their home community; such persons 21
GEORGE GRANTHAM
were least likely to possess the qualities needed by entrepreneurs who wanted their plants to operate continuously and efficiently. The ease with which the lees of rural society were recruited into factories made it more difficult to attract the better sort. High wages were also often counterproductive because workers accustomed to pursuing a multiplicity of casual occupations organized around the fixed point of an agricultural holding tended to revert to agricultural work when their cash income increased. The factory did not at first hold out the same promise of family establishment and a stable life cycle that rural occupations did, so that workers treated employment there as another marginal source of income among several, or a way of raising the sums needed to become established. The introduction of factories in rural areas placed unaccustomed stress on labour markets. The factory was a foreign implant in rural economies and for a long time struggled against rejection, not always successfully. Labour historians emphasize workers’ resistance to long hours and intense work; it is likely, however, that the main reason turnover rates were so high is that workers came from populations that had a tradition of multiple employments and short job attachments. The high turnover rates for individual firms meant that even modest establishments generated a high gross demand for workers, and as most rural people had some employment before the introduction of factories, the firms rapidly moved into inelastic range of local labour supply.30 In the face of locally inelastic supplies of labour, employers either had to innovate around the problem of high labour turnover, which was usually beyond the technological capacities of the early industrial period, or create environments that could attract and hold the kind of worker they wanted (McCaffray 1987). In attempting to explain the range of institutional solutions to the problem of labour recruitment in the early textile industry Saxonhouse and Wright (1984) identified two elements of rural society that had an especially important influence on the organizational outcome. The first was the prosperity of farm families in the regions where employers recruited labour; the second factor was the degree to which the rural labour market for female labour approximated the conditions of a competitive spot market. Rural society contained two reservoirs of recruitable labour: a contingent of young people from established families, and landless or underemployed families in regions of poor agricultural opportunity or declining rural industry. The first pool almost exclusively consisted of young women who worked in textile mills until they married, after which they reinserted themselves back into rural society. This age-specific supply of female labour was a feature of preindustrial society and spanned virtually the whole range of social situations, so that the potential supply included relatively well-off as well as poor persons (Ulrich 1991:81–2; Ardouin-Dumazet 1912). To persuade parents to allow their daughters to take up factory work, employers had to provide decent living arrangements and working conditions, which meant investing 22
THE HISTORY OF LABOUR MARKETS
in dormitories and clean factories. The second pool of labour was not so demanding; it was filled by the poor families for whom the traditional rural life cycle had broken down (Ware 1931). Such families usually migrated as a unit because they were unable to maintain themselves as viable rural households. To attract people of this class, employers had to invest in housing, minimal schooling, and find work for the men who came attached to the predominantly female factory labour force. On the other hand, they did not have to make big investments in working conditions. As the factory labour force ‘matured’—that is to say, as it reproduced itself—labour turnover declined and firms adjusted hiring practices and working conditions accordingly (Phillips 1985, 1986; Huberman 1991). Once families had rooted themselves in mill villages, employers gained considerable latitude to adjust wages and job assignments so as to minimize the cost of a given supply of family labour. This often meant new allocations of work by task and new job hierarchies (Saxonhouse and Wright 1984). It also meant adjustments in remuneration on margins other than the wage, as when night work was extended after the First World War in southern mills. This adjustment could occur only because the maturation of the mill village labour force had resulted in downward sticky wages, so that as the demand for labour in cotton textiles contracted in the 1920s, employers could impose night work as a condition of employment (Shiells and Wright 1983). A byproduct of the maturation of the labour force was that the mobility of workers tied to specific industries and regions declined, increasing the incidence of cyclical unemployment during industrial downturns (Wright 1981; Southall 1988). The change in worker life cycle expectations took about three generations, time enough for family life to organize itself around the work opportunities provided by expanding factory districts. This is why the prolonged expansion of factory-based industry after 1860—in Britain from the 1820s—was so critical to the timing of the rise of an urban working class. A man who entered a factory or mine as a young man in the late 1860s would have been joined by his sons in the early 1880s and by his grandsons at the turn of the century. By the third generation a life cycle pattern would have emerged. Historians, following Marx, have emphasized the ‘class consciousness’ that resulted from the rise of factory employment; but enterprises had no reason to dissuade the development of working-class society because, like ownership of a plot in farming districts, it stabilized labour supply and lowered its cost.31 It also provided a social environment in which implicit contracts, collective bargaining, and investments in specific human capital— whether it was by firms or workers is not important in this context—could be undertaken with a sufficient degree of confidence in their realization. This seems to have resulted in more rather than less security for workers. It gave rise to new points of reference for judging what was just and fair.32 Such community-based notions of a fair wage produced resistance to 23
GEORGE GRANTHAM
downward wage reductions and mechanisms of adjustment that left room for large amounts of unemployment in cyclical downturns. In retrospect, the epoch of western labour history immortalized in Modern Times can now be seen to have been a transitory form of organization induced by the conjuncture of a technology that involved large numbers of workers at specific sites and an industrial economy that was still confined (with the Japanese exception) to Europe and its outposts, which set a floor to wage and working conditions. This epoch, like the technology and society that subtended it, began to come to an end around 1960, as western technology and methods of organizing the manufacture of goods took root in the densely populated tropical world and the international mobility of goods and capital and finally of low-skill labour itself picked up again after the hiatus of the 1930s and 1940s. Life cycles have again broken down in many segments of the working classes and as the mobility of owners and managers, workers and taxpayers has risen, much of the geographical cement that previously bound them together has dissolved. Labour markets for ordinary people no longer provide the opportunities they did during the classic period of industrialization, which for all the intensity of its class conflict, nevertheless and in retrospect gave working people a degree of stability and certainty that has to be created again. NOTES 1 The British standard of living debate is the classic example. See Lindert (forthcoming) for an up-to-date survey. Hobsbawm (1964) is a representative of the ‘classical’ Marxist view. 2 Chapters 3, 11, and 12 deal with topics related to this issue. 3 For a sampling of this kind of study see Rothenberg (1988), Carter and Savoca (1991), and Hatton and Williamson (1991). 4 For a review of the history of economic history in Britain see Coleman (1987). 5 See Chapters 5, 6, and 7 for discussions of migration in the nineteenth and twentieth centuries. 6 For example, see Piore (1979). In the 1860s craftsmen in industrial districts in Britain were better informed about job prospects and working conditions in the United States than the American companies organized to recruit them (Erickson 1957:22–3). For a recent statistical demonstration of how strongly the lines of job market information for long-distance migration were tied to ethnicity, see Dunlevy and Saba (1992). 7 For a theoretical development of the importance of national boundaries with respect to Canada and the United States, see Krugman (1991). 8 Sokoloff (1984) and Parker (1987) suggest that there was rather more craft skill available in southern New England than what I am here asserting, but on any comparison with Europe the generalization offered here stands. 9 A tabulation made in 1893 of the occupations of farmers’ sons in one of the most industrialized sections of the United States showed that only one-twelfth had entered jobs classified as ‘labouring’ (Wright 1988:201–2).
24
THE HISTORY OF LABOUR MARKETS 10
11 12 13
14 15 16 17
18 19 20
21 22 23 24
25
‘Taylorism’ was strongly resisted by European workers and with considerable success until the mass unemployment of the 1930s. In the nineteenth century American industrialists had actively sought out persons with no previous industrial experience, which meant no pre-formed ideas about what a ‘just’ wage should be and no prior experience with trade unions. This is one reason why the early textile manufacturers recruited in French Canada rather than Britain (Erickson 1957:130). At Ford nearly 80 per cent of workers hired required less than one week’s training and of these more than half needed less than one day (Raff 1988:389)! See Chapters 9 and 10 for discussions of the extent of labour turnover in North America in the late nineteenth and early twentieth centuries. High turnover rates put the responsibility for labour performance in the hands of foremen, who commonly employed abusive language and threats to maintain work effort under what came to be known as the ‘drive system’ (Jacoby 1985:20– 4). This was especially marked in the production of shoes and boots. See Commons (1910:1–58) and Rorabaugh (1986:53). For an insightful technical analysis of the relation between initial income distributions and the evolution of labour markets see Banerjee and Newman (1993). The savings are documented in Chapter 5 of my forthcoming book The Long Agricultural Revolution. It was not the only possible compromise, although it probably was the one that released the most labour from farming because of the higher labour productivity of large-scale farms. Banerjee and Newman (1993) give an interesting taxonomy of the dynamics of labour systems that flow from initial differences in the distribution of wealth. This was as true of American cotton as it was of medieval corn (Whatley 1983). It is patently obvious that a society of medium-sized landowners would have had to have been a much less populous society, given the limited amount of land to distribute. For example, in Chapter 3 de Vries explains that in the seventeenth century when the Baltic carrying trade shifted from the smaller Dutch outports to Amsterdam, the coastal villages found themselves unable to maintain the seamen who normally took part in the fishing season. The situation was different for skilled craftsmen who had invested in apprenticeship for the very reason that the investment was itself predicated on the expectation of finding a place. See Patten (1988). As explained in the first section, this was spectacularly true in North America, where the availability of land was great enough to allow several generations of farmers to retire from labour market participation if they so desired. In the coal industry in the early part of this century, employment fluctuated from 35 to 40 per cent seasonally in both Britain and the the United States (Douglas and Director 1931:48–52, 82–3; Slichter 1919; Beveridge 1930). In the United States small-arms manufacturers balanced the seasonal hunting demand for arms and ammunition by developing new product lines, such as fishing tackle, skates, and carpenters’ tools. However, off-season diversification to hold high-skilled labour only worked for products requiring high skills. Otherwise the labour force ran into competition from producers employing lowskilled, low-wage labour (Douglas and Director 1931:97–100). For an example of the effects of this seasonality on one rural firm, see Rosenband (1985).
25
GEORGE GRANTHAM 26
27
28
29 30
31
32
Railway contractors were under great pressure to meet construction deadlines and could offer wages no agricultural employer could match. In the summer of 1858 workers left customary employers in the lurch when contractors doubled their wage (Hébert-Barrat 1980:99). Bythell notes that the outwork labour force in the Lancashire weaving sector was extremely heterogeneous with respect to productivity levels and that the average productivity of supervised workers in weaving sheds was higher than that of the average isolated weaver (Bythell 1969:135). This is why the models that have attempted to explain the distribution of rural industries before the nineteenth century have so many exceptions. It is not that comparative advantages, especially the advantage of cheap immobile labour, were unimportant, but the advantages of agglomeration were even more important. For clear accounts of how these economies generate trading opportunities and spatial patterns of specialization see Krugman (1987, 1991). On the long-term technological causes of these two concentrations, see Parker (1984: ch. 8). In New England, factory wages in adjacent towns in the 1820s and 1830s varied by as much as two to one. Since agricultural wages in the same period already exhibited strong tendencies towards regional equalization the gaps must have been due to the near simultaneous establishment of factories of different size in otherwise similar communities (Lebergott 1964:127–9). What they had reason to oppose was the emergence of working-class political parties made possible by the elimination of property qualifications for voting. It is only in this larger context that the notion of relatively homogeneous class interests makes sense. The reasoning in this section is based on ideas developed by Akerlof and Yellen (1990), who should not be held responsible for their possible misuse.
26
2 BRIDGING THE GAP BETWEEN LABOUR ECONOMICS AND ECONOMIC HISTORY A labour economist’s perspective David A.Green
A deep chasm appears to be developing between economic historians and other economists. Economic historians are placing greater emphasis on developing a competing economic paradigm which stresses path dependence and the importance of initial conditions. In contrast, theoretical and empirical models in the rest of economics are often built using a stationary environment with the underlying assumption that market forces will correct aberrations and render initial conditions unimportant. Not all papers can be so easily placed in one category or the other, of course. In labour economics, the past few years have seen everything from active participation by economic historians in debates over the extent and causes of labour market discrimination to cases, some of which are described below, where the historical context has been ignored or invoked incorrectly. Some of the chapters in this volume are direct contributions to current debates in labour economics. Nonetheless, the hyperbole present in the word ‘chasm’ is guided by a basic truth: the separation between those who study economic history and other economists is much greater today than in Simon Kuznets’ generation. This chapter is an attempt to reach across the gap and examine from the point of view of a labour economist the gains from increased interaction with economic historians. The discussion is based on the chapters in this volume. My conclusion, based on these papers, is that labour economists could gain much from increased interaction, but that the onus to close any gaps that exist does not rest solely on labour economists. Economic historians have to take a more active part in ongoing debates in labour economics. There appear to be three main lessons that labour economists can learn from economic historians. Primary among these are lessons concerning institutions related to labour markets. When one considers that such institutions include not only the family in its role as a labour-supplying unit 27
DAVID A.GREEN
and firms’ internal hiring practices, but the labour market itself (Solow 1990), it is clear that understanding the way institutions influence agents’ decisions in the short run1 is a necessary part of analysing labour market outcomes. Such an understanding requires an historical context, not only because the institutional ‘short run’ can actually be long in calendar time, implying the need for a long perspective to analyse it effectively, but also because knowing how institutions have evolved is a prerequisite to understanding how they function in today’s world. Thus, the first set of lessons concerns how labour market institutions function and evolve; the second is that one needs a long view to understand the functioning of institutions in the current context. The third lesson is perhaps the most obvious: economic history provides a wealth of examples for checking theories about labour markets and the stylized facts on which they are based. I will discuss each of these lessons in turn, beginning with those related to the evolution of institutions. Views on the value of studying the evolution of institutions stem from views about the basic process underlying that evolution. George Grantham (Chapter 1) discusses several models of this process. For our purposes we will delineate three such models. The first is based on the assumption that in modern economies the forms of institutions as well as economic outcomes are driven by rules of efficiency. Under this model, shocks and decisions made on incomplete information do not permanently affect economic outcomes, because arbitrage opportunities eventually lead back to the efficient solution. In the short run (where the short run is defined as the period within which institutions do not change), however, economic outcomes may be conditional on the form of related institutions and thus may fall short of efficiency. Economic history under this model serves mainly as a set of examples of how well the market works and what kind of impediments block its functioning. Further, the economic history of periods before the creation of a substantial number of markets would seem to hold little value to economists concerned with current problems, as such a history by definition provides little or no insight into how to fine-tune labour markets. Under the second model, the evolution of institutions follows a simple linear process. Thus, for example, the economic history of western economies might be viewed as a forward march towards an efficient form of organization that has not yet been realized; to appreciate the current state of an economy requires an understanding of where we are in the progression. There may, however, be more than one path to the final goal. Here, economic history has a role beyond providing examples, because one has to know the initial conditions that have determined the current path. One does not have to go too far back into history if the evolutionary process is simple enough, inasmuch as it takes only two points to define a line and the first does not have to be the starting point. Furthermore, one would not want to make use of history before the last major structural break because it would reflect an old trajectory no longer relevant. 28
A LABOUR ECONOMIST’S PERSPECTIVE
In the third model, new institutions evolve from old ones, but not in a simple linear manner. New markets and institutions that develop in response to new technologies or arbitrage possibilities may make parts of the economy more efficient but, by redirecting effort, make other parts less so. Thus, new institutions draw certain elements from existing institutions and abandon others. This is the model on which I believe Grantham’s introduction is based, and the one in which economic history is the most useful. Current institutions bear the thumbprint of institutions many iterations in the past. Further, since mutations are continuing and are not simple, one needs as long an historical perspective as possible to grasp the significance of current developments in the labour market. This point is brought out in the following passage from Wright: economic institutions arise in response to the pressures and conditions of a particular era, some of which may be eternal but many of which are fleeting…. Once in place, institutions change relatively slowly and incrementally as new pressures develop. An institution which evolves from decades of incremental change may be far from ideal. But in the meantime many kinds of behaviour, by employers and workers, will have adapted to the institution rather than vice versa…. The adaptations may approximate a kind of equilibrium which, at any point in time, may be rationalized any way you want. But…[u]nless you know the beginnings of the institution and have traced through adaptations…you will not understand the equilibrium you have observed. (Wright 1987:318) The evidence in this volume supports this kind of evolutionary model. Consider, for example, the studies of labour market integration (Chapters 3– 7). Postel-Vinay (Chapter 4) explicitly selects the third model of evolution in his study of regional and sectoral integration in the French labour market in the eighteenth and nineteenth centuries: The evolution of French labour markets in the course of the nineteenth century does not support a picture of linear progression towards complete integration. Instead we have a picture of one kind of labour market displacing another, each with its specific pattern of integration and segmentation. (Chapter 4:137) Postel-Vinay describes a mid-nineteenth century economy in which at harvest time there were large movements of labour from industry to agriculture and to some extent across regions, the size of the movements being related to the relevant crop technology. As the century wore on, more stable employment relations developed in both industry and agriculture, which 29
DAVID A.GREEN
resulted in smaller movements of labour across sectors and regions. Thus, while the dispersion of wages across regions narrowed from the 1840s to 1860, by 1892 it had recovered the levels existing at the start of the century. This is not to say the labour market had come full circle; rather, new patterns of market organization had emerged with less interaction between sectors and enlarged migration channels from outside the country. Similar conclusions are drawn in de Vries’ analysis of pre-industrial Dutch labour markets (Chapter 3). Again we see a transition from a pattern of flexible employment where work in more than one sector in a year for a given worker was common to employment portfolios with more stability but also more unemployment. Whether workers considered it an improvement to move from an environment of low job security to one with higher job security but greater difficulty finding a job is not clear. Boyer and Hatton (Chapter 5) find that regional wages in some occupations in England and Wales did not converge over the period 1850–1913 despite large labour movements. Indeed, they find evidence of increasing rather than decreasing agricultural wage dispersion. International comparisons lead them to conclude that economic development and labour market integration go hand in hand, but that the type of integration reached in more developed economies is not necessarily one with all wage differentials arbitraged away. The patterns of integration Boyer and Hatton find indicate the existence of established channels of migration between specific regions, similar to what Postel-Vinay describes for farmers preferring migrants from specific regions because of their experience with the relevant crop. Allen’s careful study of real wages in Britain, Canada, the United States, and Australia (Chapter 6) also leads him to conclude that for these labour markets just before the First World War: It is impossible to accept the view that labour markets were well integrated internationally….The example of Canada is very much to the point: despite massive migration both ways across the border with the US, Canada preserved a distinctive wage structure. (Chapter 6:221) On the basis of these observations he argues that one must study the development of these economies primarily at a national rather than an international level. Finally, O’Rourke’s contribution (Chapter 7) is motivated by the persistence of a paradoxical wage differential: higher nominal wages in Ireland than England. The paradox is that the main labour flows were from Ireland to Britain. Conversion to real wages does not reverse the differential. O’Rourke argues that the relevant comparisons are for expected, or unemploymentweighted, real wages, but even after this adjustment Irish expected real wages often exceeded those in England, at times by over 30 per cent (Table 7.6). 30
A LABOUR ECONOMIST’S PERSPECTIVE
The chapters by Postel-Vinay, de Vries, Boyer and Hatton, Allen, and O’Rourke, then, seem to depict labour markets in which large migrations did not ensure full integration. Like water working through sand, some channels linking regions or sectors became enlarged and persistent while others remained small. One does not witness expansion of all channels but rather a continual switching of the overall pattern over time as some channels expand at the expense of others. Whether such switching is optimal is not clear. Another example in favour of the third evolutionary model can be found in Whatley and Wright’s chapter (13) on racial discrimination in the United States. In the simplest and most powerful market model, employer discrimination will be driven out, or at least minimized, by market forces. Whatley and Wright point out that no large-scale elimination of the discrimination against blacks has occurred in United States history through market forces. Even the Ford Motor Company, which apparently played the role of the firm that begins the destruction of discrimination by hiring blacks at wages equal to those paid to whites, actually hired blacks into the toughest, dirtiest jobs and did not pay them a compensating differential. The history of black earnings does not fit a model of ongoing efficient markets or a progression towards equality through market forces. Rather, Whatley and Wright present a picture of racist attitudes and policies that were maintained by increasing segregation in all parts of life. Reduced opportunities for blacks led to reduced effort and educational investment and, along with the poor quality of black education, this confirmed white beliefs about blacks’ ability and increased segregation. Instead of breaking down discrimination, the labour market was part of a feedback loop that reinforced it. The segmented labour market can be traced back in part as a mutation of the slave economy. This is not the whole story, however. Whatley’s (1990) example of Atlanta firms switching back to segregationist policies despite wartime experience that blacks were able workers, suggests that a solely information-based model (in which the discriminatory equilibrium is not undone because segregation prevents firms from seeing other possible equilibria) does not fit perfectly. As in Donohue and Heckman (1991), Whatley and Wright suggest that societal prejudice and collusion may play a role. In any case, discrimination emerged as an institution from older institutions and was not broken down as part of an ongoing process towards efficiency. In the end government intervention was required to break the equilibrium and allow what we think of as efficient market forces to operate. Substantial numbers of Southern employers appear to have been willing to gain access to the supply of cheap black labour, but required the excuse of Federal pressure to defy long-standing community norms regarding employment of blacks. (Donohue and Heckman 1991:1605) 31
DAVID A.GREEN
If we accept the third model of evolution, what lessons can labour economists draw from economic history? One of the most important is that labour market institutions do not evolve according to a simple pattern but exhibit a high degree of persistence. Consider the example of labour market integration in the context of the relationship of Canada’s Atlantic provinces to other regions of the country. Allen (1986) shows that while real wages in Quebec, Ontario, the Prairies, and British Columbia are quite close to one another throughout the post-1945 period, real wages in the Atlantic region remain 5 to 10 per cent below those in the rest of the country. Regional inequality has been the target of considerable policy effort in Canada, but the question of what is the best approach to eliminate regional differentials remains unsolved. Under a simple linear model of progressive labour market integration, one looks for the main obstacle blocking arbitrage and removes it. The evidence of persistent differences between neighbouring regions presented in this volume might instead cause one to look for more complex explanations. If a different wage structure can develop in Canada relative to the United States in spite of massive migration (Chapter 6), why cannot a different structure develop in the Atlantic region relative to the rest of Canada, again in spite of massive migration? Perhaps a large part of the answer to the persistent wage differential results not from current impediments but from the virtual absence of in-migration to the Atlantic provinces during Canada’s surge of immigration in the early twentieth century (Green and Green 1993)? On a more concrete note, these persistent differences imply that in empirical studies of migration in Canada, the common practice of lumping together all interprovincial flows and treating them as observations on the same process is misplaced. Studies of the determinants of migration must respect the existence of established channels (i.e., the non-independence of the various flows) if they are to be successful in expanding our understanding. The practical side of this general lesson about the complexity of institutions is that labour economists should learn about the evolution of specific institutions, including evolution that occurred far in the past. This volume contains excellent discussions of the evolution of labour market institutions that make the point plain. Whatley and Wright (Chapter 13) demonstrate why understanding the evolution of discrimination is necessary to understand fully discrimination in its current context. Information on how a specific discriminatory equilibrium develops is useful in deciding whether, for example, employment equity policies (called affirmative action policies in the United States) would likely be effective. Rosenbloom and Sundstrom’s chapter (8) on the reduction in the work week provides a further example of how a study of labour market evolution could be used to inform current debate. They find a significant decline in the length of the working week in the United States at the turn of the century, which is shared by all occupations and regions they study. The timing of the hours changes shows a remarkable coincidence across 32
A LABOUR ECONOMIST’S PERSPECTIVE
occupations and regions, leading them to conclude ‘that broadly based social and economic forces must have been at work in this era’ (Chapter 8:307). They further find that hours adjusted through a sort of ratchet mechanism: reductions won in periods of labour market tightness were not reversed in cyclical downturns. These findings have implications for ongoing debates about work sharing. The fact that earlier adjustments to the work week seemed to involve a broadly based consensus indicates that significant moves towards work sharing as a response to unemployment may require active intervention to induce a critical mass of workers to take part in this option. The ratchet nature of earlier adjustments hints at rigidities that ought to be considered in implementing any policies affecting the work week. De Vries’ and Postel-Vinay’s chapters (3 and 4) on the importance of seasonal exchange of workers between sectors point to a further example of the usefulness of studying the evolution of labour markets. While I am unfamiliar with the details of the evolution of the French and Dutch labour markets, their descriptions bring to mind the Canadian labour market at the turn of the century where in February a worker might be ‘a lumber worker in Iroquois Falls, Ontario; in June a railroad navvy along the National Transcontinental; in August a harvester in Grenfell, Saskatchewan; in November a coal miner in Fernie, British Columbia’ (Avery 1979:8). In addition, the harsh winter imposed a strong seasonal pattern on all kinds of work that still exists to some extent today. This pattern had a strong influence on the form that unemployment insurance eventually took in Canada. Originally, seasonal workers were excluded from unemployment insurance on the grounds that being certain of their period of unemployment they did not need insurance. By the early 1970s, however, these workers were taken into the programme. The argument that existing employment patterns justified longer benefit entitlement periods in regions with high (and typically highly seasonally variable) unemployment rates, because it was more difficult to find new jobs in those regions, was accepted (Employment and Immigration Canada 1981:5–7). This policy change reinforced seasonal work patterns and influenced migration into and out of the economically disadvantaged Atlantic region. The original sectoral work patterns thus form the basis for the migration and unemployment insurance usage we witness today. The evolution of the Dutch economy presented in Chapter 3 also contains useful lessons for labour economists. By the early eighteenth century the Dutch economy showed signs of very modern problems: high wages combined with high unemployment and large migration. De Vries makes it clear that the transition to this state was by no means simple. Studying that transition promises potential dividends in understanding the sources of similar structural problems in modern economies. If institutions develop in complex ways, then a long view is needed to understand current events. This is the second main set of lessons that labour economists can learn from economic historians, and several examples can 33
DAVID A.GREEN
be found in this volume. Chapter 13 describes a society where discrimination is not quickly eradicated by market forces. Observation over a long time span is therefore required to understand how a discriminatory equilibrium can be reinforced and how it is propagated from generation to generation. Determining whether or not labour market integration exists, or even deciding what we mean by labour market integration, necessarily requires a long time frame to avoid confusing temporary shocks with lack of integration. O’Rourke’s study of the Irish and British labour markets (Chapter 7) shows that the combination of high real wages, high unemployment, and high emigration from Ireland is persistent, and thus needs to be examined as a clue to understanding the functioning of the Irish labour market. Chapters 11 and 12, on the behaviour and effects of inter-war wages in Britain and Germany, are also examples of situations where historical perspective is necessary and useful. This is partly because the inter-war years provide a unique opportunity to examine the workings of western economies in a time of crisis, and partly because a long-term perspective is necessary to unravel the links between real wage rigidity and the functioning of the economy. Finally, Postel-Vinay’s suggestion (Chapter 4) that rural industrial firms may have continued to use water rather than steam power into the late nineteenth century because it was a technology that was well matched to the seasonal labour demand from agriculture, points to research in economic history that focuses on the development of technology and its interaction with the labour and capital market. Again, an historical perspective is necessary to examine these issues. What labour economists gain from interacting with economic historians in discussions of issues that necessarily imply a long examination time frame is not just catalogues and descriptions, but also access to expertise in dealing with these issues. I can think of an example in modern policy research where the application of such expertise could provide useful insights (many more exist). In some areas in Canada a community response to the unemployment insurance programme has developed. In high-unemployment regions of the country it is possible to work for ten weeks and collect unemployment insurance for the rest of the year. In some regions, a form of collusion has developed between firms and community members in which part of the workforce works for ten weeks and then is laid off and takes unemployment insurance. The workers are then replaced by another group who in turn work just long enough to qualify for unemployment insurance, etc. (Green and Riddell 1993). Economic historians are ideally suited to discussing how such institutions develop and are maintained. The third set of lessons that labour economists can learn from economic historians comes in the form of specific examples from history; examples which, to paraphrase Grantham loosely, can serve to break down dogma. Given the view of evolution espoused here, relevant examples can come from historically quite remote time periods. However, one further assumption 34
A LABOUR ECONOMIST’S PERSPECTIVE
must be made to permit the use of historically remote examples: one must assume that our ancestors were not less rational and less able to process information than we are. With this assumption a reasonable model must be able to incorporate not only current observations but also the actions of our ancestors. Without it, any historical examples can be dismissed as irrelevant. It seems to me to be the height of arrogance to assume that our ancestors, controlling for the short-run institutional context in which they operated, were less able to spot what was in their best interests than we are. This point is made by James when he observes that the ability to discern the benefits of long-term employer-firm relations ‘should not be limited to large firms with bureaucratic structures after 1920, but should apply to nineteenth century employers as well’ (Chapter 9:351). Perhaps the most obvious contribution in the form of the third set of lessons in the chapters in this volume is in the chapters on worker-firm relations. James’ study of the samples of workers from Michigan and Kansas in the 1890s reveals that in some ways the labour market at that time was similar to the current labour market. In particular, the median length of attachment to firms in his samples is similar to that in 1978 and much of the short-term attachment can be accounted for by young workers. However, the percentage of lifetime jobs, controlling for differences in the age distribution of the workforce, was much lower in the 1890s than it is currently. His investigation of compositional changes in the workforce suggests that the shifts in the shares of skilled and manufacturing jobs do not account for the differences. James concludes that there must have been basic changes in employer-worker relations in the intervening years. But if such changes occurred, why did they occur? There was some long-term commitment in the 1890s in spite of the apparent lack of institutions reinforcing such commitments. What was the basis of this commitment? Modern theoretical attempts to understand internal labour markets, and institutions such as pensions that (at least in the days when they were not vested) seem to induce longer-term commitments, need to be informed by an understanding of the alternatives. Answers to the questions just listed would provide important grist for that mill. Similarly, an understanding of early pension schemes, such as the one at the Canadian Pacific Railway that MacKinnon studies (in a companion paper to Chapter 10), is interesting in its implications for modern debates over mandatory retirement and internal labour markets (MacKinnon 1993b). Lazear (1979) pointed out the seeming paradox of mandatory retirement: why should firms want to commit themselves to releasing a worker who apparently the day before the pre-set retirement day had been quite satisfactory? One suggested answer is that workers and firms form implicit contracts in which a bonus is given to workers at the end of the contract provided they are never caught shirking. The difficulty is that firms have an incentive to claim there has been shirking even when it has not occurred, in order to avoid 35
DAVID A.GREEN
paying the bonus. Firm reputation ought to discipline this type of behaviour since a firm will have difficulty persuading future workers to take part in such a contract if it is perceived as having cheated in earlier instances. Mandatory retirement may make it easier for workers to tell whether such cheating has occurred (Carmichael 1989). MacKinnon’s findings raise several questions about the implicit contract framework and mandatory retirement. On the most obvious level, Lazear (1979) suggests that 65 is a common age for mandatory retirement because the initiation of entitlement to public pension benefits at that age makes it focal. However, 65 was the mandatory retirement age at the CPR long before public pensions were created, and early public pensions in Canada had a minimum pension age of 70. Uncovering why 65 was chosen may bring out interesting information on the purpose of mandatory retirement. Perhaps more importantly, MacKinnon’s evidence indicates that the CPR may at times have reneged on promised pensions, denying pensions to those who appear to have been eligible. Under the implicit contract model, workers should not enter a contract involving a bonus at the end unless they believe the cost to cheating on the part of the firm is greater than the benefit. This cost can be created by a combination of firm reputation and quasi-rents from operation that would be lost if the firm could not hire replacement workers. The CPR seems a likely candidate for possessing such quasi-rents; it certainly could not disappear after cheating and reappear the next day under a different name in order to fool potential employees. Why then did it appear to ignore the effect its behaviour could have on its reputation? For those workers who were told their right to a pension was being denied for disciplinary reasons but who continued to work at the CPR, is there any evidence of a change in the nature of their work? Was the CPR able to give the pensions out virtually randomly to a subset of long-service employees (providing an incentive at reduced cost) because so few other companies had any pension provisions at all? Answers to these questions will provide interesting data with which to evaluate theories of implicit contracts. It seems possible that interactions which theory assumes must be self-enforcing actually require a tighter legal framework to be established by the government than we had imagined. Alternatively, the pattern of lay-offs and pensions might be rationalized in a model in which workers continually compete for better jobs (or the bonus). In either case, historical evidence provides a unique perspective not found in data sets generated in the last few decades. The interaction between historical evidence and theory should be fascinating to watch. It must be evident to any reader by this point that I believe that labour economists could benefit greatly from increased interaction with economic historians. Greater attention to economic history could reduce the tendency mentioned by Grantham to ‘produce historically foreshortened accounts of the processes that underlie’ the development of contemporary problems of economic and social policy (Chapter 1:7). Too often labour economists look 36
A LABOUR ECONOMIST’S PERSPECTIVE
back no further than a few decades or the last main structural break, whichever is closer. Nowhere is this more evident than in studies of policies central to the modern welfare state. Even historical accounts in this area typically go back no further than the passage of the first bill introducing the policies, leaving the reader with the impression that what came before was starvation and chaos.2 Based on the arguments made in this paper, it appears that such an approach may lead to imperfect policy evaluations and certainly constitutes a throwing away of useful evidence. If many labour economists appear to need to read more history, however, economic historians, in my opinion, ought to pay closer attention to contemporary debates in labour economics. Economic historians bring a unique perspective to issues along with an understanding of a wealth of information which labour economists dabbling in the past could not hope to command. It is not sufficient for economic historians to append cursory statements to their work claiming that their results ought to have implications for how we view modern labour markets. Direct linkages to contemporary issues should be made. If, as this collection of papers suggests, economic historians as a group tend to think the wedges in the economic system are of paramount importance, they need to spend more time not just cataloguing the wedges but explaining to other economists why those wedges are not arbitraged away and how the wedges develop and are reinforced. Economic historians are needed to inform contemporary debates in other fields. There is, of course, a strong precedence for this. Simon Kuznets, whom economic historians claim as one of their own, comes quickly to mind as someone who in part used an historical perspective to inform contributions to contemporary debate.3 Whatley and Wright’s work, and indeed the whole ongoing debate over the evolution of discrimination, clearly fits into this tradition. If I might be excused the conceit of referring to the work of my own father, A.G.Green’s (1976) book on post-war immigration policy in Canada used a long-run perspective on immigration policy to inform his contributions to the contemporary debate on that policy. In general, however, the distinction between economic historians and other economists is more sharply drawn than in the past. In the absence of a blurring of that line, more interaction is needed, and economic historians ought to be held in part responsible for ensuring that interaction takes place. This is not to deny the importance of basic research. Clearly, contributions such as the creation of accurate price series do not need to be justified with a paragraph linking them directly to some modern policy debate. But it would be of great usefulness if from time to time economic historians who have the inclination could spin off papers from their main research linking their results to contemporary issues. Alternatively, contem-porary issues could serve as a fertile ground for finding topics in economic history. In case anyone has missed the point, I do not believe that economic historians should enter ongoing debates in labour economics like librarians providing 37
DAVID A.GREEN
the appropriate reference at the appropriate point. My belief is that economic historians bring a unique perspective to economic issues and that the debate will be enriched by their increased presence. NOTES I would like to thank Bob Allen for interesting discussions. Most of all, I would like to acknowledge years of learning from Alan Green, my father: any interesting ideas in this paper are his; any errors or shortcomings in the paper are my own. 1 The short run is defined below as a time period within which a given institution does not undergo a significant structural change. 2 Emery (1993) discusses this point with respect to health insurance. He studies fraternal association sickness insurance before the introduction of nationalized health care in Canada. 3 Probably the clearest example of this is his work on the national accounts. In the ultimate policy-affecting application, Keynes used Kuznets’ estimates in The General Theory (1936:102).
38
3 HOW DID PRE-INDUSTRIAL LABOUR MARKETS FUNCTION? Jan de Vries The question posed in this chapter’s title is neither small nor easily answered. After all, economists find the behaviour of contemporary labour markets continually eluding their efforts to capture it in models, and, in comparison to modern society, the pre-industrial societies of Europe collected and preserved much less in the way of wage data and contemporary observations about the behaviour of labour. Thus, to economists the title question poses a challenge. But to some readers it will surely stand as a provocation. Many historians deny that economic forces dominated the determination of what labourers were paid for their time, skill, and effort and are sceptical that economic objectives suffice to account for the behaviour of labourers. To some, culture and custom were far more influential than the market; to others, the market power of wage earners was simply too small, rendering the markets one sided and tending to keep wages always at or near subsistence. In either case, economics is not the discipline of choice to approach the issue of labour conditions in pre-industrial society. This chapter proposes to take the pre-industrial labour market seriously. It is worth doing, despite the difficulties that stand in our way, because current issues might be illuminated by the study of labour markets existing in a very different institutional setting, and because we now know that wage labour was no marginal phenomenon in sixteenth through eighteenth century Europe. This is particularly true for the Dutch Republic, whose experience will be drawn upon in this study. No one could pretend to offer a comprehensive analysis of the Dutch economy in these centuries without confronting the issue of labour market behaviour, since by the midseventeenth century easily half the labour force worked mostly for wages and salaries. THE PRE-INDUSTRIAL LABOUR MARKET Such insights as we now have about pre-industrial labour markets are mostly inferred from the patterns of nominal and real wages that emerged from projects of price and wage data collection sponsored by the Price History Commission beginning in the 1930s and supplemented by many 39
JAN DE VRIES
more recent local and regional studies. I will not attempt to survey all these studies here, but will confine myself to a single, exemplary study, that of E.H.Phelps Brown and Sheila Hopkins. In 1955–6 they made use of existing data to construct a time series of daily wages for southern English construction labour stretching from 1264 to 1954—for seven centuries. They went on to construct a cost of living index, the cost of a ‘basket of consumables’, which spanned the same period. Deflating the first by the second time series, they arrived at an index of the purchasing power of builders’ wages—a time series of real wages. Their articles (1955, 1956) are cited with great frequency, even now after nearly four decades, as the classic statement of wage behaviour not only for England but for much of Europe. Indeed, their own later work (1957, 1959) confirmed that the patterns they uncovered for England were broadly consistent with those found in France and Germany.1 What they found, briefly stated, was this: 1 Nominal wages were extraordinarily ‘sticky’, remaining fixed for very long stretches of time, sometimes over a century, before being changed. They abundantly confirmed Adam Smith’s (1904:83) observation that there was no tendency for wages to vary with short-term fluctuations in the cost of living. 2 Nominal wages never fell (almost never; they fell briefly in 1921–3, 1930–3, and back in the 1330s). Phelps Brown and Hopkins conceded that their method and the type of labour they were studying biased their findings in this direction, but the absence of any sustained decline of wages, even during long periods of falling prices, amply demonstrated the truth of John Maynard Keynes’ foundational statement in The General Theory (1936:9), that ‘whether logical or illogical, experience shows that this [i.e., the total rejection of nominal wage reductions] is how labour in fact behaves’. 3 Skill premia (the ratio of craftsmen’s wages to those of unskilled construction labourers) remained fixed for centuries on end. After a rise in the relative position of the unskilled in the early fifteenth century (a period of great labour scarcity), the ratio remained fixed at 3:2 until the outbreak of the First World War. This suggests that wage earners have long been very sensitive to their relative position, a powerful factor, according to Keynes (1936:14), once again, for workers to resist nominal wage reductions and for a given structure of pay, once achieved, to be perpetuated by employers if only in the interest of labour peace. 4 Throughout the pre-industrial era trends in the real wage were determined overwhelmingly by changes in the cost of living—by prices—rather than by changes in the nominal wage. Those price trends, in turn, followed faithfully the broad trends of population change. This suggested to Phelps Brown and Hopkins, and to many other scholars before (such as Wilhelm Abel (1935)) and since (such as B.H.Slicher van Bath (1963) and E.Le 40
PRE-INDUSTRIAL LABOUR MARKETS
Roy Ladurie (1966)), that the real compensation of labour was determined chiefly by Malthusian forces. Labour—indeed, everyone—stood powerless before these fundamental factors. The familiar graphs prepared by Phelps Brown and Hopkins documented a great rise in labour’s purchasing power occurring after the Black Death (the so-called ‘golden age of labour’ of the fifteenth century) reaching its apogee in the third quarter of the century (see Figure 3.1). There followed, during the price inflation of the sixteenth century, a disastrous long-term decline of purchasing power. When it ended, in the first quarter of the seventeenth century, the real wage stood at its lowest point in the seven-century time series (if we ignore for now the considerable index number problem inherent in any exercise of this type). The restoration of real wages that followed was slow and far from complete when it was reversed by the renewal of price inflation after 1750. The nineteenth century brought a fundamental change in the pattern of real wages: for the first time real wages rose in the face of rapid population growth, but this is not our immediate concern. 5 One final observation can be made about the behaviour of wages. Nominal wages almost never kept pace with trends in the prices of foodstuffs (which 41
JAN DE VRIES
dominate any cost of living index), but they did more or less track trends in the prices of industrial products. Phelps Brown and Hopkins call attention to this and conclude that wage earners during the sixteenth century price inflation (to the extent they were industrial workers) were not denied the customary share of their product. Their unenviable position was the result of deteriorating terms of trade between the agricultural and industrial sectors of the economy. By these accounts, wage labour before the Industrial Revolution was truly imprisoned in an iron cage, or, to use Le Roy Ladurie’s (1974, 1977) phrase, caught in an ‘histoire immobile’—a motionless history. In a world without much productivity growth and with a mentalité that could not conceive of incentives to labour other than poverty, attachment to custom must have seemed the best defence a labourer could deploy to protect his or her interests (Coleman 1956). THE DUTCH ECONOMY What Phelps Brown and Hopkins found for England has been shown to be broadly true for most of Europe.2 This finding has not only lent authority to their analysis but has also reinforced the view that such a pan-European phenomenon must be explained by very deep-seated, fundamental causes. In this historiographical context I have sought to analyse the workings of labour markets in the Dutch economy from the sixteenth through the eighteenth centuries, as the newly formed United Republic experienced its ‘golden age’ and subsequent decline. Among the questions that need to be addressed are these: 1 Did wages in Holland follow the general European trends? Specifically, did labour benefit from the ‘golden age’ economy, or did it experience the same dismal fate of historically low real wages observed in England, France, and Germany? 2 Did labour markets function as Phelps Brown and Hopkins say, or did the more highly urbanized, more proletarianized, and more rapidly expanding Dutch economy develop more flexible labour market behaviour? 3 How did Dutch labour markets respond to the transition from vigorous economic expansion to stagnation and decline? On the first of these questions there is much disagreement. My survey of the literature reveals an ‘optimist’ and a ‘pessimist’ camp (if I may borrow the polarizing terminology of the British standard of living debate), and, remarkably, both camps base their views to a large extent on the same prominent characteristics of the seventeenth century Dutch labour market. The optimist position can be traced back to the turn-of-the-century German scholar Otto Pringsheim, who was the first to advance the proposition that ‘the seventeenth century was a time of economic brilliance 42
PRE-INDUSTRIAL LABOUR MARKETS
for Holland that has never returned, and this forceful upswing also worked to the benefit of the labourers’ (Pringsheim 1890:52, my translation). J.A.Van Houtte expressed, without elaboration, a similar view in his Economic History of the Low Countries (1977:232) when he wrote ‘Rich or poor, the Dutch could hardly have found better conditions anywhere else.’ One must admit to a certain unease with these statements, for they seem to be based on the notion that a period called a golden age simply could not have passed the workers by. For Pringsheim a telling fact in support of his belief was the large migration of rural dwellers and foreigners to the Dutch cities. Surely, he reasoned, these multitudes were attracted by a standard of living that exceeded that available where they had previously lived. The pessimist position was launched by no less an advocate than Karl Marx. If to Pringsheim and Van Houtte a vigorous economic expansion such as we call the golden age seemed on the face of it to be good for the wage earner, to Marx it was, again, on the face of it, sure to be a disaster. Marx was no stranger to the Netherlands (he visited relatives there, who included the founder of Philips), and in Das Kapital he recites the economic achievements of the Republic around 1648, observing that ‘the total capital of the Republic was probably more important than that of all the rest of Europe put together’. He then adds that ‘by 1648 the people of Holland were more overworked, poorer, and more brutally oppressed than those of the rest of Europe put together’ (Marx 1887:754). Marx does not elaborate on this obvious exercise in hyperbole, but later pessimists have been consistent in arguing that labour was denied the economic fruits of the golden age because of a massive inflow of labour. Van Ravesteyn (1906:251), Charles Boxer (1965: ch. 3) and J.G.van Dillen in his textbook (1970:294) all held that the vigorously expanding demand for labour elicited an even greater supply, causing the labour market to be chronically oversupplied. This, in turn, kept wages low. A third group of economic historians, including all the most recent commentators, might best be called agnostics. Van Deursen (1978:11–52), in a book devoted to the study of the common man’s economic and social situation in the golden age, writes matter-of-factly ‘We do not generally know how high the workers’ wages were.’ Caught between the evidence of poverty and the knowledge that foreigners eagerly took jobs in the Republic, he was unwilling to draw a conclusion. Peter Klein (1977:89, 111) insists that ‘the data are too scarce to form judgments’, while the most recent study, by Leo Noordegraaf (1985), adds no small amount of wage data to that previously available, but still manages to avoid coming to any firm conclusion. I have reviewed the literature in some detail because it demonstrates clearly that progress in answering the question has been beyond the reach of historians for two reasons. There is, first of all, an obvious need for broadly-based time 43
JAN DE VRIES
series of wages and salaries that can be compared with the data for other countries. I have tried to respond to that need, about which more follows below. But that, by itself, surely will not be enough. In the case at hand, the pessimists seem to believe that the growing Dutch economy could call upon unlimited supplies of surplus labour available at a constant, subsistence wage. The optimists implicitly hold that only the offer of a superior standard of living can induce labour to migrate to a new location. In the absence of an appropriate model of labour market behaviour, ambiguity will continue to adhere to this discussion, and preconceptions will reign unchecked. THE DATA The wage data displayed in the following graphs are drawn from my Dutch Labor Market Project.3 I have focused on the wages of construction labour— masters, journeymen, and unskilled labourers—because of their relative homogeneity over time and space, and because construction labour is nearly universally used in other studies of historical wages. These data are supplemented, however, with wage data for sawyers, shipyard workers, and rope works employees, as well as with salary data for clergymen, teachers, municipal employees, and sailors. Real wages are computed with a price deflator modelled after the Phelps Brown and Hopkins ‘basket of consumables’ index. I have constructed a basket of consumables for Holland with weights broadly similar to the English series, but with a selection of commodities guided by Dutch consumption habits and, of course, with Dutch prices drawn from the price history of N.W.Posthumus (1908–39, 1944–64).4 Nominal wages for construction labour have been collected for ten locations in the western Netherlands and another seven in the eastern and southern provinces. (Here, we are chiefly concerned with the western provinces of Holland, Utrecht, and Zeeland.) Individually, they show most of the same characteristics as the English series, particularly the rarity of wage reductions and the general stickiness. Indeed, some series show no change at all for a full two centuries, from 1650 to 1850 (see Figure 3.2). In order to acquire an overall indicator of wage trends for the region I have taken averages of the prevailing wage at each location in each five-year period from 1500 to 1815. This indicator is primitive, in that the observations are unweighted and most of the component time series are not complete for the full period. Consequently, the short-term fluctuations are unlikely to be particularly meaningful. The long-term trends, on the other hand, represent the experience of labour in city and country, large cities and small. Figure 3.2 reveals a pattern of wage behaviour rather different from that summarized by Phelps Brown and Hopkins. To begin, nominal wage adjustments in the late sixteenth and early seventeenth centuries were very fre44
PRE-INDUSTRIAL LABOUR MARKETS
quent. At one location, a drainage board workplace in Spaarndam, the prevailing wage for common labour rose eight times between 1570 and 1610. Yet, at other times, wages at this same location exhibited a truly phenomenal rigidity, remaining unchanged for over a century after 1680. Second, the structure of pay—the pay differentials between skilled and unskilled labour, etc.—was by no means constant. From 1500 to 1570 the 45
JAN DE VRIES
relative pay of unskilled labourers deteriorated, falling from a peak of 70 to no more than 50 per cent of the master craftsmen’s wage (see Figure 3.3). This is consistent with our knowledge of urban industrial decline in this period. The era of the revolt against Spain brought sudden disruption to these established relationships, and the rapid economic expansion that followed consolidated important relative gains for the lower ranks of wage workers. By the 1640s unskilled labour earned 70 per cent of masters’ wages. Yet, these movements in the skill premium were as nothing compared with the behaviour of the wages of sawyers. The demand for sawyers sensitively reflected investment levels, since urban expansion, farm improvements, and shipbuilding all required sawn lumber. As beneficiaries of the investment accelerator, sawyers’ wages rose from some 90 per cent of the masters’ wage in the first half of the sixteenth century to a peak of 130 per cent by 1610. It should not surprise us to hear that a certain Cornelis Cornelisz. rigged up a windmill to saw timber mechanically in 1596. It took time to perfect his sawing mill, and even more to overcome municipal restrictions to its use, but the relative wages of sawyers fell after 1610, and after 1665 there are not enough references to their employment to sustain the time series. If the sawyers became early victims of technological unemployment, the unskilled labourers who benefited so much from the expansive growth of 46
PRE-INDUSTRIAL LABOUR MARKETS
the post-Revolt economy found their relative position under attack in the more stagnant economic environment of the eighteenth century. From 70 per cent of the masters’ wage in 1690 common labourers’ relative wages drifted gradually downward to 60 per cent by the 1750s. While casual and unskilled labour suffered some deterioration of their relative position in the eighteenth century, salaried employees of many types gained. The heterogeneous character of salaried workers is an impediment to the construction of a meaningful index, one I have sought to overcome by incorporating evidence from the broadest possible range of salaried employees. The index presented in Figure 3.4 incorporates salaried occupations spread across the western Netherlands: Reformed clergymen in Amsterdam, smaller cities, and rural areas, school teachers in Amsterdam, Leiden, and Middelburg, a broad range of municipal employees in Amsterdam, Middelburg, and Goes, orphanage employees in Amsterdam and Utrecht, and, finally, windmill operators employed by the Hoogheemraadschap van Rijnland. The twelve separate indices have been combined into a single unweighted index stretching from 1540 to the early nineteenth century. The divergent course of salary compensation is readily apparent when it is related to the course of wages in the building trades (Figure 3.5). Salaried employees fared poorly during the sixteenth century price inflation: their salaries were adjusted even less frequently than were day wages. After the 1580s salaries rose even faster than wages, more than restoring the relative position of salaried employees by the mid-seventeenth century. After midcentury, when wages began to assume their notorious rigidity, many salary earners enjoyed further raises. By 1800 the salary index stood 25 per cent higher than in 1664. This brief glance at salary data reminds us that wage quotations are not the same thing as earnings. The number of days worked, or the incidence of unemployment, most obviously intervened between wage rates and earnings. This topic will be discussed in some detail below; here we consider other, often ignored, factors that strongly influenced actual earnings. Wage rigidity exists in part because there is a cost to publicly changing wage rates that employers, and sometimes employees, are eager to avoid. Most obviously, a raised wage is nearly impossible to lower again. In response to this fact employers often respond to demand-induced pressures on wages— in either direction—by altering the effective compensation through the manipulation of terms of employment other than the wage rate itself. This ‘wage drift’ is hard for the historian to detect for the obvious reason that it was not intended that it be generally known. I have stumbled upon a sufficient number of examples of wage drift to suggest that it was a widely practised way of responding to changed labour market conditions. Consider these three examples.
47
JAN DE VRIES
1 Throughout our period wage labourers received a lower wage in winter than in summer, usually about 75 per cent as much. Sometimes intermediate wages were paid in the spring and autumn. This was obviously related to the changing length of the work day in a world where sundown brought work to a halt. Yet, the details of the seasonal pay pattern were by no means as predictable as the course of the sun. Both the size of the winter pay reduction and the length of time the lower wages were in effect could vary by job and over time. Consider the case of day labourers employed by the city of Kampen (Van der Vlis 1981:79–81). In 1644, when we can first observe their wages, they earned 12 stuivers per day, year round. There was no separate winter wage. This was introduced in 1675, whereafter the labourers earned 12 st. during the 32 summer weeks and 11 st. during the 20 winter weeks. In 1690 the winter wage was reduced to 10 st., but by way of compensation the 12 st. summer wage now prevailed for a longer period, 35 weeks, and an autumn and spring wage of 11 st. was introduced. Finally, at a date not known exactly, but after 1715, the city shortened the summer period to 27 weeks and reduced 48
PRE-INDUSTRIAL LABOUR MARKETS
the winter wage to 8 St., which now prevailed for a 17 week period. The net effect of all these shifts was to reduce the maximum annual earnings of Kampen’s day labourers by nearly 20 per cent (from f. 187.20 in 1644– 74 to f. 152.00 in the post-1715 period) even though the summer wage rate had never been altered. 2 The payment of substantial bonuses was periodically used to attract recruits to the Navy and the East India Company, but it became a common supplement to the earnings of many permanent employees towards the end of the eighteenth century. The unskilled labourers of Amsterdam’s forestry service(!) earned 18 stuivers per day in the summer and 15 st. in the winter until 1783. Their maximum annual earnings then stood at f. 263.70. Thereafter the winter wage was raised by 1 stuiver, but, more importantly, they now received an annual ‘New Year’s Gratification’ of 25 guilders, raising their annual earnings by 11 per cent.5 Beginning in 1796 the city restricted this annual bonus to selected workers and in 1811 the practice disappeared, returning the maximum annual pay to its 1783 level (Van Zanden 1983:7–8). 3 Finally, we come to a probably widespread practice, but one difficult to detect: the transformation of the ‘day’ for which the daily wage rate was paid from a real to a notional concept. Throughout the period under discussion the work day was divided not into hours but into schoften, or work periods between breaks. Pauses in the day were typically at midmorning, a dinner break, and a final one in late afternoon, dividing the 49
JAN DE VRIES
work day into four schoften. The shorter winter work day could be divided into three periods, providing a ready justification for the lower winter wage. The detailed pay records of the Hoogheemraadsckap van Rijnland, a large drainage authority with responsibility for the maintenance of dikes, locks, pumps, etc., reveal that most of their permanent labour force worked about 260–70 days per year in the eighteenth century (they were usually laid off in the dead of winter). In 1784 the bookkeepers recorded a sudden jump in the number of days the crews worked, to 280, and in the following decade the days worked rose steadily so that by 1795 it reached a level in excess of the total number of available days, 313, and by 1799 reached 342. It hovered around 325 for the next 20 years before declining again.6 What had happened? In fact, the employees did not labour more after 1784 than before, but were being paid for five schoften (1.25 days) for summer days that previously had been compensated at the rate of four schoften. The drainage board raised the annual earnings of its labourers, in stages, by 22 per cent between the 1780s and 1800 without ever changing the posted pay for a day’s work.7 It is not easy to summarize all these scraps of information, but it seems fair to say that Dutch labour markets were by no means as captive to customary arrangements, nor were wages quite as sticky, as Phelps Brown and Hopkins found for England. Around an impression of impressive wage rigidity we see numerous accommodations being made to market forces. These accommodations tended to compress the structure of pay during the period of rapid growth and to extend it when the demand for labour slackened. REAL WAGES The course of nominal wages deserves attention in its own right; after all, this is what employers paid and this is what directly affected the cost of production and the demand for labour. It can be shown that the level of nominal wages in the maritime provinces of the Netherlands was approximately at the southern English level in the first quarter of the sixteenth century (in both areas building craftsmen’s daily wages translated to about 4.5 grams of silver), rose beyond the English level, but were still under the Belgian (Brabant) level in the last two decades of the sixteenth century (see Table 3.1). They surpassed that level after 1600, to reign as the highest wages in northern Europe until the end of the eighteenth century, when English wages assumed that position (de Vries 1984a). Dutch employers paid higher wages than their competitors in neighbouring countries, but did the wage earners enjoy higher living standards? Here we must turn to my consumer price index. It reveals the familiar price trends that were common throughout Europe (see Figure 3.6): after a slow build50
PRE-INDUSTRIAL LABOUR MARKETS
up in the early decades of the sixteenth century, consumer prices rose irregularly but rapidly until 1600. Then, after a brief respite, the inflationary process resumes, to reach its apogee in the 1650s. Thereafter prices tend to fall, again very irregularly, to the 1730s, whereupon an upward trend is reasserted which becomes sharply inflationary after the French Revolution. When the nominal wage series are deflated by this consumer price index they reveal short-term patterns similar to those found for England and other neighbouring countries (see Figures 3.7 and 3.8). This stands to reason, since when wages are sticky, price movements dominate the series. 51
JAN DE VRIES
The trends, however, are very different. The best way to appreciate this is to compare directly the Phelps Brown and Hopkins real wage index, cast in the form of a thirteen-year moving average, with the same index for Dutch master carpenters and masons (see Figure 3.9). Both time series are indexed to 1451–75, the era of peak real earnings in England. I cannot prove that a given index number represents equal purchasing power in the two countries, but the similar construction of the baskets of consumables, and the approximately equal exchange values of nominal wages in both the early sixteenth and early nineteenth centuries, lend support to my construction of Figure 3.9. Figure 3.9 makes four main points. In both countries labour suffered a decline in purchasing power from the late medieval high points into the mid-sixteenth century. Between 1580 and 1620 the Dutch economy followed a separate path from England and most other countries. Instead of further decline, real wages staged a recovery and this occurred because of large and frequent increases in the nominal wage rather than because of divergent price behaviour. The gap that opened up between Dutch real wage levels and 52
PRE-INDUSTRIAL LABOUR MARKETS
those of her neighbours in the early seventeenth century was maintained for more than a century. Indeed the highest real wages were reached in the 1680s and 1730s. Thereafter, the Dutch lead wasted away as English nominal wages rose periodically while those in the Netherlands remained at their mid-seventeenth century levels. Based on this comparison of craftsmen’s wages, the exceptionality of the Dutch ‘golden age’ resides not so much in the rise of real wages as in the avoidance of erosion. This, I am sure, is too modest a conclusion, for the comparison wage used, that of master craftsmen, experienced a slower growth of nominal wages than any other category of labour investigated in this study. Unskilled labour did significantly better, and salaried workers better still. But the optimists’ case is still not really made, for we are focusing our attention on daily wage rates and not on what workers actually earned. To pursue this issue further we must penetrate the veil of wage observations, the intersections of the supply and demand curves, to the working of the labour market itself.
53
JAN DE VRIES
THE LABOUR MARKET IN THE GOLDEN AGE The first matter demanding attention is the supply of labour, a topic that both optimists and pessimists regard as central to understanding the experience of labour in the Republic’s numerous urban centres. There can be no doubt that the supply curve of labour shifted sharply to the right, particularly in the half century after the 1570s. This relatively brief period repays careful attention because in it was concentrated the most intense immigration to the Republic, the most rapid urbanization within the Republic (and, hence, a probable increase in the percentage of the labour force working for wages), and a major increase in the length of the working year (the consequence of the reform of religion). In this period Holland’s population nearly doubled, the urban population rose from about 40 to nearly 60 per cent, and the length of the work year rose by 18 per cent, from about 260 to 307 days. Natural increase, immigration, proletarianization, and reformation all combined to flood the Netherlands with wage labour: the effective supply of non-farm labour grew by nearly 3 per cent per year in the period 1570– 1620. If the demand for labour had not also increased as least as fast, the equilibrium wage certainly would have fallen as we know that it did in surrounding countries. But, we know that in this very period Dutch money wages were raised repeatedly, far outstripping the rise in prices. Moreover, 54
PRE-INDUSTRIAL LABOUR MARKETS
we know that the pay differentials narrowed in these decades, suggesting a relative shortage of unskilled labour. There is no hint of Malthusianism here. On the contrary, the wage evidence is consistent with an economy in which heavy investment is expanding employment opportunities and increasing productivity. Sawyers’ rapidly increasing wages reflect the accelerator effect of capital investment (and their disappearance reflects the further consequences of that investment process), while the rise of productivity can be inferred from the tendency for nominal wages to rise faster than industrial prices between the 1580s and 1650s. The imperfect industrial price data before the 1580s show a rough constancy of the wage to industrial price ratio, but by the 1650s labour productivity would have had to exceed that of 1580–1600 by a third in order to hold the unit cost of labour constant. A labour force growing rapidly, attaining higher levels of productivity, and earning higher real wages—this is the attractive picture we have of the Dutch labour markets in the period 1570 to the mid-seventeenth century. But we cannot stop here. A large portion of the total demand for labour in this period offered casual and temporary employment. The herring fishery, Baltic shipping, peat digging, and the many infrastructural projects such as 55
JAN DE VRIES
polder drainage, canal construction, and urban expansion all offered employment on a seasonal or highly temporary basis. Specific conditions varied from place to place, of course, but a remarkable feature of the labour markets in much of Holland was the way in which these varied employments were integrated by a large, flexible labour force, both urban and rural. Rural smallholders took work as fishermen and construction labourers on polder projects; Enkhuizen herring fishermen also signed on, before and after the fishing season, with merchant skippers sailing for the Baltic. Indeed, the sailing dates for the main Baltic fleets, mid-April and the end of August, seem calculated to avoid the peak period of the herring fishery, July and August. We know little about just how these activities were integrated by individual families, but we have many examples of hundreds, even thousands of workers suddenly appearing to dig a canal, repair dikes, or work on polder drainage. And we know from contemporaries that the dates of the harvests and sailing of fleets would cause their work crews to melt away like butter left out in the summer sun (Baars 1973; Van Dam 1992; Lucassen 1984). There were, of course, other, more permanent forms of employment available. But, the highly flexible supply of casual labour was both conspicuous and strategic to the most rapidly growing sectors of the economy. Few formal institutions had much influence over the organization of employee-employer relations in these sectors. Even in the organized crafts and industries of the cities, the guilds were usually creatures of the municipal government (rather than the other way around) and usually forced to be accommodating to new entrants and market forces. The institutional position of the worker in the golden-age economy was not particularly secure. The one thing labourers had going for them was a vigorously expanding demand for their services. What would happen to this flexible, open labour market when the demand stopped growing? THE POST-GOLDEN-AGE LABOUR MARKET The very expansion of the Dutch economy increased the opportunities for full-time, specialized employment. The drainage projects, among other factors, increased the number of large, commercialized farms. Rural society offered progressively less scope for the smallholder with by-employments, but it offered an increased demand for permanent farm servants. The expansion of commerce to other continents increased the demand for sailors prepared to work for years at a time rather than for a voyage of two or three months. New, capital-intensive industries such as industrial windmills, refineries, and ship wharves offered relatively regular employment in comparison to textiles. The growing cities created many positions in what we would call the service sector, and the magistrates tended to encourage guild formation in previously
56
PRE-INDUSTRIAL LABOUR MARKETS
unorganized occupations in order to facilitate the maintenance of an orderly and peaceful urban society. All of these fruits of economic growth increased the number of specialized and secure positions, but they could hardly compensate for the unravelling of the seasonal employment system that occurred with greatest intensity in the 1670s and 1680s. In these decades that inaugurated a longer period of population decline and falling prices, the labour-intensive industries contracted in the face of intensified foreign competition, and the investmentrelated sectors collapsed as urban expansion, land reclamation, and infrastructural improvements came to a standstill. At the same time, but for reasons that seem to be coincidental, the herring fishery began to decline. All of these sectors had offered seasonal employment that fed into a single labour system (Lucassen 1984). With their decline it became increasingly difficult for Holland’s workers—accustomed to high wages, and by now largely separated from agriculture—to offer their labour to the surviving remnants of the seasonal and casual labour markets. The case of Enkhuizen, a port on the Zuider Zee, is instructive in this regard. Enkhuizen had grown rapidly to a population in excess of 20,000 by about 1650, on the strength of the herring fishery and merchant shipping to Norway and the Baltic. As noted earlier, the seasonal character of these activities made it possible for many workers to unite them. But the demand for seamen began to decline in the 1630s as the Baltic trade became ever more concentrated in Amsterdam, and even more after 1650 as the skippers sought their crew members from small communities in Friesland, the North Sea islands, and abroad. It has long been thought that this displacement was motivated by the search for lower-cost labour, but a decline in the number of Hollanders offering themselves for this work may also have played a role. At any event, by the early eighteenth century there were very few seamen in either Enkhuizen or the rural communities in its region. The decline of employment in the merchant marine had a major impact on the organization of the herring fishery. This began to decline after 1670, although it remained the major economic activity of the city throughout the eighteenth century. But, the fishing firms soon came to depend on rural communities for their labour force as fishermen disappeared from Enkhuizen. In the absence of other employments they could not survive on the seasonal work of the herring fishery. Such fishermen could continue to exist only in small coastal communities where other forms of inland and ocean fishing could supplement the seasonal herring activity. Ironically, this declining industry found its labour force shrinking even more rapidly, so that wages had to be increased around 1700. Indeed, by 1750 the fishing firms resorted to recruiting foreign labour, despite laws prohibiting this (Willemsen 1988:54–60). Many seasonal employments remained after the mid-seventeenth century: hay making, peat digging, and the rapidly growing whaling industry being 57
JAN DE VRIES
the most important. But, instead of complementing each other, as had been the case while the economy grew, they now tended to compete with each other, all requiring labour in a relatively short summer period. This explains why, in the face of a sharp decline in domestic employment opportunities, the maritime provinces of the Netherlands began to attract migrant labour from the eastern provinces and, particularly, from Germany, beginning in the 1650s and 1660s (Lucassen 1984). The wrenching changes occurring in the Dutch labour market in the second half of the seventeenth century did not generally result in permanently high unemployment because of a third and final new element: the expansion of the Dutch East India Company (the Verenigde Oostindische Compagnie or VOC). Although the VOC had grown rapidly from small beginnings early in the century, the doubling of its size between 1680 and 1730 had an enormous impact on the labour markets of the Dutch port cities, and of the vast hinterlands that supplied them with labour. Working for the VOC as a sailor had long been regarded as highly undesirable: the pay was low, the dangers high, and the separation from home lasted for years. Most—in fact, two-thirds—of those who embarked for Asia aboard a VOC vessel never returned. Even in the early days of the company foreign employees had to be recruited in large numbers. Already in the 1630s, 40 per cent of those setting sail for Asia were not Dutch, and by the 1660s the foreign share had risen to half. Signing on as a VOC soldier was even less attractive: in the 1660s only 25 per cent of them came from the Republic’s maritime provinces (see Figure 3.10). The diminished employment opportunities of the later seventeenth century, particularly the crisis in the casual labour sector, generated a new interest in VOC employment. The Dutch share of all VOC employees rose from 50 per cent around 1660 to 73 per cent around 1700 (see Figure 3.11). The maritime provinces’ share of newly recruited VOC soldiers rose from 25 to 50 per cent in the same interval. Since the Dutch share was rising when total company employment was growing, the VOC’s overall impact on Dutch labour markets grew enormously. One historian speaks of the VOC sweeping clean the city streets of beggars and the unemployed (Van der Woude 1980). Specifically, he calculates that in the first half of the eighteenth century an average of 20 per cent of each cohort of 20 year old men in the province of Holland died abroad while in the service of the VOC.8 If we survey the character of Dutch labour markets in the early eighteenth century, we find a very different situation from that which obtained during the golden age. The wage data show a remarkable stability throughout the period of transition, but it reflects not tranquillity or stagnation, but a transformation from a rather fluid, flexible regime to one characterized by pronounced labour market segmentation. The segmented labour market featured a sector composed of large public and semi-public employers—the shipyards, the drainage authorities, and the 58
PRE-INDUSTRIAL LABOUR MARKETS
59
JAN DE VRIES
guild-organized activities—which paid the nominal wages that had been attained by the mid-seventeenth century, even though the intervening decline of prices had substantially increased the purchasing power of those wages. Perhaps the ‘efficiency wage’ hypothesis can help explain the persistence of these high wages. By honouring the previously attained wage levels employers could choose the best workers, reduce turnover, and elicit more intensive effort from their employees. On the other hand, this policy ensures that markets will not clear, that supply will exceed demand.9 There is little room for doubt that the considerable advantages of employment in this sector created queuing effects, whereby the unemployed were prepared to engage in extended searches—or waits—for one of the privileged jobs. The knowledge that the patronage of urban regents figured in the disposition of many of these choice jobs is consistent with this scenario (Kooijmans 1985:36).10 And, where the rents formed by these ‘above market’ wages were controlled by guilds, entrants faced rising fees (for the right to produce the required ‘masterpiece’), payments to widows (for the right to succeed the deceased husband), guild sickness fund contributions, etc.11 Those workers unable to acquire a permanent job in the high wage sector could participate in the casual labour market. Wages in this sector were not necessarily lower—although they tended to be more sensitive to immediate supply and demand conditions—but the intense seasonal concentration of demand in this sector sentenced these workers to extended unemployment. Unless non-wage income or charity offered sufficient alternative support, this sector was not viable, and, as we have seen, Dutch labour tended to abandon it to foreign migrant labour—labour that did have access to complementary household resources to secure its reproduction. Every year, throughout the eighteenth and into the nineteenth century, thousands of Scandinavians and, especially, Germans entered the Netherlands to man merchant ships and whaling vessels, dig peat in the moors of Groningen and Drenthe, mow hay on farms in Friesland and Holland, and work in the brickyards, bleaching fields, and any number of other seasonal employments. Finally, the ranks of the casually employed were thinned by the existence of a third segment, an employer of last resort offering long-term employment to very nearly all comers: the VOC. This three-part segmented labour market persisted throughout the eighteenth century. The one major change, occurring gradually after 1740, was the declining interest of Dutch labour in embarking on the ships of the VOC. Both the VOC and the Admiralties faced pressing labour shortages which they sought to overcome with foreign recruitment and the payment of signing bonuses. One can only speculate as to the reasons for this reduced supply of labour, for no increased demand in other sectors acted to bid this labour away. Consequently, in the second half of the eighteenth century, 60
PRE-INDUSTRIAL LABOUR MARKETS
workers experienced higher levels of unemployment, particularly chronic urban unemployment, than they had in earlier periods. The eighteenth century Dutch economy exhibited several strikingly modern features: high wages, probably the highest in Europe, together with substantial unemployment, and, at the same time, sectoral labour shortages and a dependence on seasonal migrant labour from a neighbouring country with a much lower standard of living. This occurred in an institutional context very different than what we now know. There was no unemployment insurance, minimum wage, or unionization—to name only the usual suspects for the crime of labour market disequilibrium. There were guilds, to be sure, and a system of public and church charity that was extensive and generous by contemporary standards. Contemporary observers can be found who did not doubt that much of the unemployment they witnessed was ‘voluntary’, and caused by the availability of relief payments.12 Historians are inclined to dismiss charitable institutions as a factor in labour market behaviour because of the very low levels of support they offered the able-bodied poor. Moreover, most recipients of relief were not participants in the labour force.13 However, it is fair to say that the Dutch labour market of the eighteenth century was one capable of producing a ‘modern’ form of unemployment, one in which reservation wages, queuing for employment in the high wage sector, and the temporary availability of public support all played some role. With the benefit of historical hindsight the maintenance of the high nominal wages achieved by 1650 beyond the 1670s looks like a fatal mis-step. The sharply falling price levels that drove real wages to their peaks after 1670 were not a product of cost-reducing investment and increased production in the Republic. The rising real wages were the product of an exogenous shock in the form of a falling international price level. Under these circumstances the failure to adjust nominal wage rates led to a redistribution of income, away from profits and rents and towards the wage bill. The high real wage was compensated for by depressed land values and low returns on capital, which generated, in short order, greater unemployment. This, in turn, speeded the construction of the segmented labour market regime that I described above, a regime that tended to protect the high wage levels by reducing the scope for labour market competition. Of course, wage reductions are unpopular—and almost as rare—today as they were then. We tend to reduce real wages, when this is necessary, through inflation and/or exchange rate policy. These measures were not available in the seventeenth century monetary regime, certainly not to a nation that functioned as the financial nerve centre of Europe. The only measure that seemed available to respond to the reduced demand for labour at the high wage levels of the post-1670 economy was a reduction of population. This demographic response did, indeed, occur, both through 61
JAN DE VRIES
fertility behaviour at home and the VOC abroad. But it was, to say the least, a slow and imperfect policy tool. Without effective measures to adjust wages to maintain international competitiveness, the Dutch Republic endured a long, debilitating period of labour market disequilibrium. At the safe distance of three centuries, we can look back and identify beneficial reforms. But contemporaries gave little indication of unhappiness or unease. People like security, and the post1670 labour regime offered more of this, in the unfavourable economic environment of the time, than had the more flexible regime of the preceding decades of economic expansiveness. That is, it offered this security to the ‘insiders’, those with the permanent high-wage jobs. The beauty of this system is that it also commands the loyalty of many outsiders, those who hope to get one of the secure jobs. CONCLUSIONS Early in this chapter I posed three questions about Dutch labour markets, and it would seem that the best way to conclude is to see how those questions might now be answered. The first question was the easy one. Real wage trends in the early modern Netherlands did not, after the 1570s, follow the English trends, or those of the other neighbouring countries. Dutch exceptionality seems to be confirmed at least until the late eighteenth century, when English wage levels catch up. But, so large was the early advantage of the Dutch that even well into the nineteenth century wages remained high in comparison with most countries. Did the labour market behave as Phelps Brown and Hopkins say? Here the answer depends on how literally they are to be taken. The stickiness of wage rates is generally confirmed, but that is a nearly universal phenomenon. It makes more sense to stress the relative modernity of the Dutch Republic’s labour markets. Such contemporary aspects as labour market segmentation, reservation wage, efficiency wage, and job queuing seem to have a place in explaining Dutch behaviour, and the features that puzzle us are not really different from those features of modern labour behaviour that continue to challenge the labour economist. Finally, concerning the nature of the Dutch transition from a dynamic to a more stagnant economic environment, the stress should be placed on the radical changes that were made in order to defend the pay structure attained early in the seventeenth century. The response took the form of shifting employment risks to ‘outsiders’ through labour market segmentation. Concurrently, the economy became structurally unable to supply—to reproduce, in Marxist terminology—the labour for many seasonal employments. It became dependent on labour from zones of low living costs at the same time that it could no longer fully employ the domestic labour force. 62
PRE-INDUSTRIAL LABOUR MARKETS
The pre-modern character of this society is revealed by the absence of any policy measures to restore labour market equilibrium; its modern character, by the broad allegiance of the population to a system that gave security—at a price, to some. NOTES 1 2 3
4
5 6 7
8 9 10 11 12
13
All four papers are reprinted in Phelps Brown and Hopkins (1981). For sixteenth and seventeenth century wage data, see Phelps Brown and Hopkins (1959); for eighteenth century data see Söderberg (1987) and Braudel and Spooner (1967). This project has been supported by the National Science Foundation grant SOC78201788, for which the author here expresses his gratitude. Pending full publication of the assembled data, the reader is referred to de Vries (1984a, 1993). The basket of consumables consists of the following items: rye (1050 kg), yellow peas (143.5 kg), beef (100 kg), cheese (50 kg), butter (50 kg), beer (621 litres)— for which coffee, tea, and gin are substituted after 1750, peat (100 turf tons), and a composite of industrial product prices weighted to equal 25 per cent of the food subtotal in the period 1575–99. The basket of consumables is for a household of average size. There are 20 stuivers to a guilder. These records are in the Oud Archief van het Hoogheemraadschap van Rijnland, no. 10905–10920, werklijsten. The payment for extra, fictive schoften occurred elsewhere, especially in the 1770–80s. Master carpenters and masons in Alkmaar used this device in the 1770s when they faced labour shortages, the Admiralty ropeworks in Amsterdam introduced the ‘fifth schoft’ in 1774, and in distant Groningen the same practice is observed beginning in 1782 among craftsmen employed by a village church. In the two last cases the fifth schoft was later abolished when formal wage increases were introduced. These wage increases thus simply ratified a practice that had begun earlier (Noordegraaf 1980:38–9; Algemeen Rijksarchief Den Haag, Archief Admiralititen, Archief van de Lijnbaan te Amsterdam, no. 8–20, Betalingsboek van spinders en draijers; Paping 1991:38–62). For the reasoning and evidence behind this statement see de Vries (1984b: 210). On this issue see Blinder (1988), Akerlof and Yellen (1988), Dickens and Lang (1988), and Hall (1975). Note also that the four Burgemeesters of Amsterdam carefully rotated their rights to appoint persons to no less than 3200 offices. Rising fees are documented in Nusteling (1985:150–1). A spectacular example is provided for the southern Netherlands port of Antwerp in Deceulaer (1992). On widows’ rights, see de Vries (1981:121–2). Laurens van de Spiegel, who would later become Raadspensionaris (the rough equivalent of prime minister), was convinced that native Dutchmen, particularly those adhering to the public church, were especially reluctant to do heavy and dirty work. ‘Why should they take up such work or walk so far to become employed? They can find a much easier source of income in the payment of the poor relief agencies, and if these do not suffice, they can call upon the magnanimity of their fellow citizens’ (van de Spiegel 1772:15). See Van Zanden (1983:19). A more nuanced analysis of this issue is found in Van Leeuwen and Smits (1987).
63
4 THE DIS-INTEGRATION OF TRADITIONAL LABOUR MARKETS IN FRANCE From agriculture and industry to agriculture or industry Gilles Postel-Vinay
]Labour markets in the industrializing economies of seventeenth and eighteenth century Europe were highly segmented, but the degree and significance of this segmentation is often misunderstood. Large wage differentials between sectors and regions were typical, but the labour markets of that era did not simply involve separate groups of immobile workers. Rather, the supply of labour to individual sectors of the economy was a complex mixture of immobile permanent and mobile temporary workers. Such combinations are still with us, but they were more critical to the functioning of the economies of earlier centuries, because seasonal bottlenecks in production and transportation were extremely widespread and could only be overcome by temporarily hiring extra hands. The grain harvest was the most important bottleneck; every summer thousands of industrial workers left their jobs to work in the grain fields. This annual great migration between sectors imparts a special character to the labour markets of the advanced economies of Europe before the twentieth century. Without it, pre-industrial economies probably could not have attained such high levels of specialization in production (Grantham 1993). However, intersectoral mobility put stress on pre-industrial labour markets, especially when it involved the exchange of labour between industry and agriculture. The familiar categories of sectoral labour market analysis are not well suited to the study of the kind of short-term intersectoral labour mobility that was common in most European industrial economies before the twentieth century. In France, estimates of the sectoral distribution of the labour force in the nineteenth century have been worked up from the occupational designations of the census of population (Toutain 1963; O’Brien and Keyder 1978; Marchand and Thélot 1991). Anyone who declared himself to be an agricultural (or industrial) worker 64
TRADITIONAL LABOUR MARKETS IN FRANCE
was considered to be permanently attached to his professed occupation by the census-takers, even though a significant part of the population floated seasonally between agriculture and manufacturing. For example, Grantham finds that at the end of the eighteenth century, between 25 and 40 per cent of the rural labour force in France were jointly employed in the agricultural and manufacturing sectors (Grantham 1993:440). This makes it impossible to measure sectoral labour productivity using census occupational designations alone. Estimating implicit wage gaps from intersectoral labour productivity differentials is therefore not much help in understanding the economics of intersectoral labour mobility (O’Brien and Keyder 1978; Sicsic 1992). Similarly, analyses of urban-rural wage gaps typically focus on the annual wage and thus ignore the seasonal variations that are critical for understanding temporary movements of workers between sectors (Williamson 1985:49).1 The problem of interpreting wage gaps is greatest in precisely those countries where statistical monitoring of economic growth has been carried back past the middle of the nineteenth century, when temporary intersectoral migration was probably at its peak. This paper attempts to measure the extent of such movements and considers their implications for the evolution of French labour markets from the eighteenth to the early twentieth century. TRADITIONAL LABOUR MARKETS In the eighteenth century agricultural labourers worked about 200 days a year, and their working year increased only slightly to the middle of the nineteenth century. A relatively short working year was also typical of industrial workers. According to the industrial census of 1861 (Statistique générale 1873), they worked on average 254 days. If 300 days is taken as a full working year, both figures indicate significant levels of temporary unemployment. This reading of the data, however, is untenable. For one thing, farm labourers had for centuries been finding part-time employment in rural industry, while rural artisans and urban workers commonly participated in field work in summer and autumn (Lequin 1977). Large farmers often employed more workers than they needed for their average daily labour input in order to meet the frequent though unpredictable episodes of intense demand for labour produced by day to day changes in the weather, but the harvest peak in labour demand was of a wholly different order of magnitude. Eighteenth century farms in grain-growing regions required four times as many workers for the harvest as for the rest of the year (Grantham 1993:484). The substitution of the scythe for the sickle and the advent of mechanical reaping caused the peak to subside in the second half of the nineteenth century, but down to the 1870s the efforts of individual farmers to innovate around seasonal labour shortages were offset by continued growth in regional specialization. In the middle of the nineteenth century, the harvest still required a huge force of temporary workers who engaged in 65
GILLES POSTEL-VINAY
other occupations the rest of the year (Collins 1976; Grantham 1993). A consequence of the massive summer emigration of industrial workers was that many plants had to shut down for lack of sufficient labour. The industrial census of 1861–5 (Statistique generate 1873) provides evidence that temporary spells of industrial unemployment were linked to quits by workers who went into the agricultural sector at harvest time. Every industrial firm had to report at what time of the year and for how long it stopped production, whether it laid workers off during the shutdown, and whether laid-off workers were able find work and if so what kind. Most of the original questionnaires have disappeared, so one cannot directly reconstruct answers to all these questions. The published statistics allow one to distinguish three possible responses to lay-offs: (a) workers remained unemployed for the duration of the shutdown; (b) they found work in other firms in the same sector; (c) they found work in other sectors. The first option seems to have been infrequent. The second is also ruled out because it is hard to see why some firms in an industry would have increased hiring when others were shutting down for the season. The third alternative was thus the most probable. Most plant closures took place in summer, which strongly suggests that many firms geared their production schedule to the predictable voluntary departure of part of the labour force into field work (see Figure 4.1).
66
TRADITIONAL LABOUR MARKETS IN FRANCE
SEASONAL COMPETITION FOR LABOUR IN LOCAL MARKETS The seasonal pattern of sectoral wages provides a first indicator of how strong was the incentive for workers who lost their jobs between July and September to look for temporary work in agriculture. Table 4.1 shows that the usual differential between industrial and agricultural wages was reversed in the summer. Normally, the male industrial wage outside of Paris was 20 per cent higher than the agricultural wage, but at harvest time it was only 80 per cent as high. The shift was even more pronounced for women. Most of the year, agricultural and industrial wages were about the same; during the harvest women’s earnings in agriculture were more than a third higher than in industry. Industrial employers clearly faced extremely strong competition for labour at harvest time. How many workers were affected by the seasonal inversion of agricultural and industrial wages? It is difficult to answer this question directly from Table 4.1, because inter-regional wage differentials were large—up to three to one—and there is no obvious way to determine the geographical limits defining relevant employment opportunities for workers in each region. If we take the département—the fundamental administrative unit in France— to be a local labour market, we can measure the number of industrial workers for whom the harvest agricultural wage exceeded the industrial wage. Table 4.2 breaks down the industrial labour force of 1860 according as the departmental agricultural wage in summer was higher or lower than industrial wages. Industries that did not match the harvest wage were highly vulnerable to summer closure. This pattern suggests that shutdowns were the consequence of a rational choice by employers not to compete for labour at the harvest peak (Bompard et al. 1990a). We can analyse the links between the seasonal inversion in the relation of industrial and agricultural day wages and the seasonal spells of temporary industrial unemployment, by assuming that labour demand and supply in 67
GILLES POSTEL-VINAY
each département was affected not only by the relation of the agricultural to the industrial wage, but also by the share of agriculture and industry in total employment, the level of urbanization, and the productivity of agricultural labour. The Industrial Census of 1861–5 reports output, intermediate inputs, the value of capital, the employment of men, women, and children along with their wages, and the shutdown periods for 800 industries in a regional grid of more than 350 arrondissements.2 The geographical and sectoral breakdown of the variables along with the number of firms reported in each can be used to construct representative firms in each industry in every arrondissement as units of observation.3 Suppose that industrial firms responded to the summer rise in agricultural wages according to a schedule of marginal productivity determined by their capital intensity, labour productivity, the size of the firm as measured by its number of workers, and its rate of profit or rate of value added relative to reported capital value.4 Factor analysis of the relationship between these variables brings out two main conclusions: the geographical distribution of firms did not depend on strictly local conditions of employment and wage levels—i.e. low-wage districts did not attract more industry than high-wage districts (Bompard et al. 1990a, 1990b); despite this spatial independence agricultural and industrial wages were highly correlated, especially for female workers at the level of the arrondissement. The tight connection between agricultural and industrial wages meant that competition between the two sectors for labour was especially intense in summer. The seasonal competition for labour was strongest in the market for female workers, but it also operated in the market for male labour, especially in the numerically important lowwage textile sector, whose male employees were not immune to the financial attractions of field work. Did manufacturing plants shut down because their employees quit in order to participate in the harvest? To explore this question we employ principal component analysis to analyse the relation between the 68
TRADITIONAL LABOUR MARKETS IN FRANCE
abovementioned variables and firms’ decisions to cease operating for a few months in summer. If workers quit when expected harvest earnings exceeded expected industrial earnings, the difference between the specific industrial subsector wages and the local harvest wage will be positively correlated with the level of voluntary departures from the subsector. Workers’ decisions to quit depended on the wage they expected to earn when the harvest was over. Since they did not necessarily re-enter their old industry or at least did not rejoin their old firm, we take the mean autumn departmental industrial wage as the measure of the expected wage for workers who quit for a time to engage in agricultural work. The difference between the autumn wage in the industry they quit and the mean industrial wage measures the expected gain or loss to workers from quitting and returning to industrial employment (Bompard et al. 1990a). Figure 4.2 shows the main results of a principal components analysis of the characteristics of our representative industrial firms. The projections of the principal variables on the two main factorial axes classify the mean firm in each département and industry according to the proportion of summer closures (rising from left to right), and according to the ratio of the industrial wage to the summer agricultural wage and profit rates (higher profit rates towards the bottom, relatively high industrial wages towards the top). The plane of the axes, which respectively explain 21 and 18 per cent of the total variance, supports the argument that shutdowns were triggered by the loss of workers to agriculture. The first axis distinguishes firms paying wages that exceeded the summer agricultural wage and that also exceeded the mean departmental industrial wage. Industries that paid exceptionally high wages did not shut down in summer. Notable among them were heavy transport equipment, machinery, iron and other metallurgical industries, and mining. The high wages they offered dissuaded employees from quitting. In contrast, quits were frequent in low-wage and low-profit firms that were highly concentrated in textiles and clothing. The second factorial axis distributes firms by mean wage and profit level. One expects high wages to discourage seasonal quits, but the prior on profit level is not obvious. Summer mobility scarcely touched the high-wage metallurgical and mining industries, whereas seasonal agriculture demand readily tapped the ill-paid and predominantly female workforce in textiles and clothing. The food processing industries were another reservoir of summer harvest hands. The two industries released workers for different reasons: textile firms usually had low capital-labour ratios and could quickly stop ordering raw materials in anticipation of the summer shutdown; in contrast firms in the food processing industries were small, capital intensive, and had high profit rates and high labour productivity, but paid low wages. Low wages were the main source of their vulnerability to seasonal competition from agriculture. Firms that regularly lost workers to agriculture in summer employed about a third of the industrial labour force.5 69
GILLES POSTEL-VINAY
The seasonal supplies of labour from industry affected patterns of agricultural specialization. There was a strong local connection between the number of workers released by firms in summer and the extent of wheat cultivation in the same or adjacent districts. Wheat production expanded most in districts where industrial workers were temporarily available for harvest work. Seasonal migration of industrial workers, however, had important consequences for industry as well. Many new technologies involved costly investments that needed to be continuously operated to cover their cost. Firms intending to introduce new techniques thus had to raise wages to hold workers through the peak summer season in order to develop an experienced workforce. This was particularly important for technologies 70
TRADITIONAL LABOUR MARKETS IN FRANCE
demanding continuous production, for example when steam replaced water as the source of industrial power. The period of low water coincided with the harvest. This is possibly a major reason why France’s main source of energy through the 1860s continued to be water power. In any event, capitalintensive firms tended to hold their workers through the summer more successfully than labour-intensive firms. THE GEOGRAPHICAL SCOPE OF TEMPORARY LABOUR MOBILITY Industry and industrial employment were unevenly distributed across the country, and firms that shut down in the summer were still more unevenly distributed. In contrast, since all regions maintained large agricultural sectors, seasonal peaks in demand for agricultural labour occurred throughout the country. As a result, there were systematic variations in local wages. Where the number of workers released from industry exceeded local agricultural demand, farm wages were depressed, and vice versa. This relation is demonstrated for the early 1860s by a regression of the departmental male agricultural wage in summer on a measure of excess local demand for seasonal farm workers;
where W is the male summer agricultural wage, XLA the excess seasonal demand for agricultural labour as measured by the number of persons reported to have immigrated into the département to work in agriculture, and XLI is the number of workers locally unemployed by seasonal plant shutdowns.6 The results of a test of the effect of the differential exposure of male and female workers to plant shutdowns on the male-female agricultural wage ratio are qualitatively similar.
where WM/WF is the ratio of male to female harvest wage, MXLA and FXLA are the number of male and female temporary agricultural migrants from outside the department, and MXLI and FXLI the number of seasonally unemployed of the two sexes in industry. The regressions show that agricultural and industrial labour markets in the two sectors were locally connected. But is the département the right level of labour market aggregation? Workers migrated beyond the boundaries of départements in response to inter-regional wage differentials. The differentials reflected imbalances in local labour markets resulting from the combined effect of temporary industrial shutdowns in July and August and the surging 71
GILLES POSTEL-VINAY
local demand for harvest labour. Workers in regions where there was a temporary excess supply of labour spread out into neighbouring départements. This movement was especially important between départements on the periphery of the Paris Basin, many of which had rural textile industries, and its grain-growing core, and between the southern fringe of the Massif Central and the cereal-growing and wine-making districts of Languedoc. At harvest time these flows temporarily linked otherwise distinct regional labour markets. Although seasonal migrants were highly responsive to inter-regional wage differentials, choosing a region in which to look for work was risky, and information about alternative employment conditions in competing regions was usually incomplete. In particular, the expected real harvest wage in any region was unpredictable. In terms of real seasonal wage differentials, the advantage of temporary migration could be reversed from one year to the next. As a result, workers tended to look for work in regions where they had worked before. For their part, farmers tended to restrict hiring to workers from regions that they knew supplied workers skilled in specific techniques of harvesting. They often dealt year after year with the same labour contractors. For all these reasons the massive migrations of the mid-nineteenth century failed to eliminate regional differences in agricultural wage levels, and local pockets of seasonal unemployment continued to co-exist with tight agricultural labour markets. Seasonal labour migration into agriculture was widespread and long standing, but was not universal. According to Robert Allen, seasonal intersectoral migrations did not characterize English labour markets in the eighteenth century, at least not in the southern Midlands, where rural unemployment in non-peak seasons was high (Allen 1992; Thirsk 1989:678–81). On the other hand, France experienced high intersectoral mobility through the 1860s. The decade marked the high point of intersectoral occupational mobility, which began to decline shortly after and had all but disappeared by the end of the century (Bompard et al. 1990b). SEASONAL MIGRATIONS AND THE EARLY INDUSTRIAL ECONOMY Agriculture A long history of temporary flows of labour from industry and the crafts into agriculture reached its apogee between 1830 and 1870. In some regions one can observe by 1860 a gradual shift towards steadier patterns of agricultural employment. Where cattle raising was becoming important, the proportion of annually employed farm servants increased. In areas that were becoming more specialized in wheat growing, however, the demand for labour became even more seasonal. This in turn caused larger seasonal wage differentials between agriculture and other activities. The trend towards greater regional 72
TRADITIONAL LABOUR MARKETS IN FRANCE
specialization in wheat production occurred slowly, since most regions had originally been self-sufficient in grain. The precise timing of the increase in the premium is difficult to trace, but the wage gap had become quite large by the 1850s. It is clear from the spatial pattern of the wage gap in 1852 that it was positively linked to regional specialization in small grains (Crebouw 1986:644–7). The summer/winter wage gap continued to grow in the 1850s and early 1860s. In 1852 the average summer wage had exceeded the winter wage by 40 per cent; by 1862 the difference was 50 per cent, and was even greater in regions specializing in cereals (Statistique générale 1858–60, 1870). The link between regional specialization and seasonal wage differentials was especially close for male wages.7 The rapid expansion in the cultivation of wheat, which made France the world’s largest wheat producer in the 1850s,8 had its counterpart in the increasingly pronounced seasonality of agricultural wages. Little by little the rise in the summer wage of temporary day labourers influenced other wages in the agricultural economy and eventually other sectors of the economy (Crebouw 1986:646). Industry To judge from the census of industry that was conducted in the 1840s (Statistique générale 1847–52) approximately 45 per cent of industrial employees worked in communes with fewer than 2000 inhabitants.9 Towards the middle of the century when agricultural wages rose twice as fast as industrial wages, firms in rural areas had particular difficulty holding onto their workforce during the summer. Many enterprises were located in rural communes because of the availability of water power, and therefore had large investments in fixed capital. In the early 1840s the amount of installed capital per worker was twice as high in rural communes as it was in urban communes.10 Table 4.3 shows the distribution of rural firms according to their capital intensity and the wages they paid relative to the agricultural wage. It is clear that the industrial wage was strongly influenced by the degree of capital intensity in the industry concerned. This was especially true of textiles, which was the most important rural industry and displayed the widest range of organizational forms, production being carried out in both cottages and factories.11 The positive correlation between capital intensity and wage levels had important consequences. Rural industrial wages frequently exceeded urban industrial wages because the capital intensity of rural firms was higher (see Table 4.4). To maintain continuous operation rural firms had to pay summer wages that at least equalled the summer agricultural wage. If they failed to give workers a summer premium, they had to accept the likelihood of closing down for the season and losing experienced hands, which impeded the introduction of sophisticated equipment. Thus, firms that adopted the 73
GILLES POSTEL-VINAY
strategy of summer shutdown seem to have had lower than average capital intensity, which in the long run placed them at a serious competitive disadvantage vis-a-vis urban firms, because industrial innovation tended to be associated with rising capital intensiveness. The disadvantage for firms that had to shut down every year is perceptible in the industrial census of 1861–5. Firms operating at a loss on the basis of reported sales and the cost of variable inputs had the lowest amounts of installed capital per employee (see Table 4.5). They also tended to be the most exposed to quits by their employees in summer. By freezing or at least discouraging industrial investment in rural firms, an industrial labour market that was subjected to strong seasonal flows of labour into agriculture was tending to marginalize many sectors of the rural economy that had long been of great importance. THE DISCONNECTING OF AGRICULTURAL AND INDUSTRIAL LABOUR MARKETS The closely knit fabric of seasonal industrial and agricultural labour markets, which had been so marked a feature of the French economy at mid-century, unravelled after 1870. The long depression affecting both industry and agriculture and the subsequent restructuring of both sectors caused the two main sectoral markets for unskilled labour to become less closely connected. The harvest wage ceased to exceed the industrial wage and seasonal 74
TRADITIONAL LABOUR MARKETS IN FRANCE
migration fell dramatically. A new set of circumstances emerged, with regular spells of underemployment for agricultural workers in the off season, and reasonably permanent (although cyclically interrupted) jobs for industrial workers. In the industrial sector, internal labour markets developed. Industry The transformation of the industrial labour market between 1860 and 1890 was particularly striking. The earlier practice of combining agricultural and industrial employment over the year had largely vanished by 1890. During the Second Empire the difference between maximum and minimum monthly employment levels in industry had been very considerable. Around 1860 at least 500,000 and more probably 800,000 workers quit their jobs every year to work temporarily in agriculture. Thirty years later the number of seasonal quits was barely 100,000. A seasonal immigration that in 1860 amounted to more than a tenth of the agricultural labour force had by 1890 been reduced to an insignificant addition of 2 per cent (Marchand and Thélot 1991). To judge from the inquiry into employment and working conditions conducted by the Office du Travail (1893–7) in the early 1890s, short-term intersectoral migrations within France had almost ceased. The length of the working year in industry had risen close to the maximum 300 days and the number of workers employed in industry no longer fluctuated seasonally, or fluctuated to a minor degree only (see Figure 4.3).12 Taken at face value the survey indicates that employment behaviour had undergone major change during the previous three decades. However, the difference between the labour markets of 1860 and 1890 is not quite so straightforward. Some industries continued the traditional pattern of a working year of less than 250 days and occasionally even less than 150 days. While the summer trough of industrial employment had disappeared, marked seasonality remained in a subset of industries (Figure 4.3). The industries that maintained older traditions of seasonally flexible employment 75
GILLES POSTEL-VINAY
were concentrated in a few departments and their seasonal shutdowns were not synchronized with harvest labour demand.13 Moreover, although half the firms reporting indicated they had a stable workforce, the others reported significant fluctuations in the level of employment over the course of the year, in the range of plus or minus 10 per cent of mean employment. Short-term variation in activity forced some firms to rely on temporary hires and accept high turnover rates. On the basis of fragile calculations from the information it received from reporting firms, the Office du Travail concluded that on average firms hired 115 workers to fill every 100 positions. Of these hires ‘perhaps 75…were stable, and the rest unstable. The stable labour force works on average 295 days a year, while the unstable workers are likely to work no more than 170’ (1893–7 vol. 4:16, my translation). In addition, many firms employed as part of their flexible workforce ‘external’ help working at home. (These workers are not included in Figure 4.3.) In effect, firms were reviving the older system of domestic industry. There was about one ‘external’ worker for every four employees. 14 The employment patterns that had emerged by the 1890s were thus characterized by a stable core of employees surrounded by a cloud of temporary workers. According to the survey, about two-thirds of industrial workers were stable employees, so permanent and continuous employment was becoming the rule for the majority of workers.
76
TRADITIONAL LABOUR MARKETS IN FRANCE
Agriculture Employment patterns in agriculture followed a similar path. We know that for several decades following 1870 this sector experienced a difficult reorganization. The wheat-growing regions faced the greatest problems as grain imports pushed prices down, and wherever it was possible to redirect production, cattle raising expanded at the expense of cereals. The slow diffusion of mechanical harvesting attenuated the seasonal peak in labour demand (Chatelain 1976, vol. 1:126–9, 245–6). The contraction of commercial wheat growing also affected the viability of rural industry. Industrial activities that had developed by using labour available during the agriculturally slack times of the year were undermined as agricultural employment for day labourers contracted. Nevertheless, rural industry resisted total elimination for a long time, despite low labour productivity. In the long run it could not survive, because the changed condition of the agricultural labour market left cottage industry no room within which to operate.15 The contraction of seasonal migration, the decline of cottage industries, and the increasingly stable pattern of industrial employment reduced the seasonal flow of labour across branches of agriculture and between agriculture and industry. As day labourers in agriculture lost complementary incomes from rural industry, their numbers shrank. Between 1862 and 1892 the number of agricultural day labourers fell most rapidly in those regions where the working year had traditionally been shortest and where, by extension, rural industry had been most important to the income of day-labouring families. In contrast, servants in husbandry hired on annual or seasonal contracts were almost as numerous as ever, despite the contraction of agricultural output. Indeed, if we consider the whole active agricultural population, including farmers as well as employees, the number of persons permanently engaged in farming actually increased in this period (see Table 4.6). 77
GILLES POSTEL-VINAY
As the seasonal pattern of labour demand progressively smoothed out, local seasonal migrations of labour abated. To the extent that a temporary force of agricultural workers continued to be required, as it did in the beet and grain fields of northern France and the vineyards of Languedoc, the migrants were now drawn not from a domestic population of rural industrialists and small farmers, but from the peasant population of Spain, Italy, and Poland.16 Thus after a long period during which they were intimately connected by the seasonal sharing of the labour force, the agricultural and industrial sectors of the economy parted ways. They henceforth constituted two separate ensembles, in each of which employment had become more regular, continuous, and stable, but between which the connections were effected by permanent rather than temporary migration. To be sure, on the margins of both activities there remained persons who were excluded from this system of employment relations. Relegated to the temporary tasks and suffering frequent bouts of underemployment, they represented a kind of mobile labour reserve. But for all practical purposes the reciprocal flow of labourers from one sector to the other had ceased by the First World War. By the end of the nineteenth century the pattern of relative wages had developed to the point where even at the peak season agricultural remuneration no longer exceeded industrial wages. In fact the gap between farm and industrial wages was widening (Sicsic 1992). As a general rule, the agricultural day wage did not come close to the level of earnings of local unskilled factory workers and the rise in agricultural wages lagged far behind increases in industrial wages. At mid-century, nine out of ten men and women working in industry could have increased their income by taking part in the harvest. In the 1890s, this was true for only one out of three women and one out of seven men, and the bonus earned from quitting and returning to industrial work was much lower than it had been.17 THE CHANGING PATTERNS OF SPATIAL INTEGRATION AND JOB ATTACHMENT Wages paid to servants in husbandry had always varied considerably from region to region. The spatial disparities in the wages of hired hands were largest for female servants, but were high for men as well. This suggests that the markets for permanent hired hands were more localized than other labour markets. (See Table 4.7). The mobility of servants in husbandry was low because most were young people undergoing what was in effect an agricultural apprenticeship in the region where they had been raised. The premium farmers placed on their employees’ knowledge of local farming practice also encouraged geographical segmentation of the market for men whose principal tasks were handling animals and ploughing. Towards the middle of the century, as labour markets tightened, the regional differentials 78
TRADITIONAL LABOUR MARKETS IN FRANCE
narrowed even for this local and fairly immobile component of the agricultural labour force. The trend towards greater regional integration was reversed in the decades after 1870 and, by the end of the century, the regional dispersion of agricultural wages was of the same order of magnitude as in the first half of the century. Agricultural wages of permanent employees remained low, and the regional hierarchy changed little over time. Nevertheless, the hired hands remained the most attached to farming, to the point where by the end of the century they constituted the majority of all agricultural employees (see Table 4.6). They were especially numerous in the regions specializing in rearing cattle and dairy farming. The habitually mobile day labourers had long enjoyed the smallest interregional wage differentials. In the course of a single generation, their number fell by half. The contraction of the most mobile segment of the agricultural labour force helped to increase spatial segmentation of agricultural labour markets. In contrast, industrial labour markets in the same period were becoming more integrated. Industrial wages, which were rising rapidly throughout the period, continued to converge throughout France (Table 4.7). As a consequence, the regional correlation of agricultural wages between 1862 and 1892 is much higher than the regional correlation of industrial wages over the same period.18 This pattern shows up in the way the industrial wage varied by town size and rural or urban location in 1840 and 1892 (Table 4.8). In 1840 wage levels varied by town size, and the rural industrial wage significantly exceeded the wage paid in towns of 5–20,000. By 1892 the industrial 79
GILLES POSTEL-VINAY
wage in rural areas was lower than the urban industrial wage, and wages rose as town size increased. A large part of the differential across towns of different sizes, however, was due to differences in the cost of living between the large cities (Marseille, Lyon, Paris) and the rest. As employment relations became more permanent, or at least as sectoral attachments of workers became more stable, views as to what constituted the problem of unemployment began to change. By the 1890s administrators defined the unemployed as a statistical class (Salais et al. 1986). People out of work, now defined and counted, were becoming the object of a new institutionalized politics. It was at this time that internal labour markets began to emerge in many industries.19 In the railroads, for example, the wage structure was reorganized by collective bargaining to facilitate the creation of career paths within firms (Caron 1973). The change can be illustrated by comparing the new wage structure with the one described by Villermé at the birth of France’s industrial growth in the 1830s. He observed that for workers in the textile industry wages rose to about the age of 30, rapidly at first, and then more gently. From the age of 35 or 40 they declined. Moreover, the upward-sloping part of the wage curve was steeper than the downwardsloping part (Villermé 1971 edn: 203). At the end of the nineteenth century the age-wage curve had become more ‘modern’. Wages rose with age and experience over a long period of time, and did not begin to decline until workers were much older. The new pattern of lifetime wages reduced short-term seasonal mobility between the sectors and it affected long-term mobility as well. It had long been common for farmers’ children to work in the mines or factories until the age of 35 to 45 and then return to take up the family farm (Trempé 1971) (Figure 4.4). But this traditional movement was becoming much less common at the turn of the twentieth century as the returns to remaining permanently with an industrial firm rose. The change in the employment relation is consistent with the change in the nature of industry and the growing importance to firms of long-term job attachments. Workers who expected a career in firms that held out the promise of increasing (or at least not decreasing) wages over most of their 80
TRADITIONAL LABOUR MARKETS IN FRANCE
life cycle no longer had to return to the village to live out the years of semiretirement on the family holding (Hatzfeld 1971:318–22). The forms and destination of workers’ savings had also altered. In this respect, too, a link between industry and agriculture that had operated through workers reinvesting their savings in land was also being severed so that part of peasant agriculture was becoming increasingly isolated from the rest of the economy (Postel-Vinay 1991). Nevertheless, one should not jump too quickly to the conclusion that the new systems of permanent employment represent a total break from more flexible, if unstable, employment relations of a prior era. Historians who study European labour markets in the early phase of industrialization and the pre-industrial period have described the market environment of skilled workers whose employment relations were highly structured by urban guilds and other corporations (Kaplan 1989). These were no doubt a minority of all workers, but they were the critical reservoir of skill and their labour market, at least for some of them, was truly national, certified by the age-old tradition of the ‘Tour de France’. This brief sketch has been concerned with the structure of the markets for the large group of low-skilled workers and has ignored the more institutionalized markets for highly qualified craftsmen which are better known but accounted for a small share of all labour market transactions. Nevertheless, the market for unskilled labour cannot be considered to be non-institutionalized. Indeed it is worth noting that it was precisely where employee-employer transactions presented the features of 81
GILLES POSTEL-VINAY
an auction market that we find the earliest evidence of collective bargaining (Moriceau 1985). CONCLUSIONS The evolution of French labour markets in the course of the nineteenth century does not support a picture of linear progression towards complete integration. Instead we have a picture of one kind of labour market displacing another, each with its specific pattern of integration and segmentation. In the first, partially integrated labour markets were embedded in complex intersectoral arrangements that for centuries had successfully coped with the seasonal variations in demand for agricultural labour by permitting the seasonal migration of substantial numbers of workers. Seasonality was an inherent characteristic of agriculture in general and of arable agriculture in particular, and whilst it posed major problems to farmers, it also concerned many industries, especially those located in the countryside. The normal competition for available labour between industry and agriculture during the peak harvest season became increasingly intense as agriculture became more commercialized and specialized between 1750 and 1870. For a time this competition generated fairly high levels of spatial and intersectoral integration. However, the seasonal labour market integration broke down during the subsequent agricultural depression and the rise of industrial and mining activities that employed a full-time workforce. As a consequence sectoral labour markets became more separate than before. NOTES 1 It probably makes more sense to think of a separation of agriculture and industry in Britain than in any other European country. 2 An arrondissement is an administrative subdivision of the département. 3 The relevant information about seasonal employment in agriculture is taken from the agricultural Enquêtes of 1852 and 1862 (Statistique générale 1858–60, 1870). Finally, the population census of 1866 (Statistique générale 1869) allows us to estimate the share of industrial, agricultural, and service employment in every départementy along with an important residual class of unemployed persons declaring no occupation. 4 The profit rate is set equal to gross operating surplus/value added. 5 The 1861–5 census probably exaggerates true quit rates, because it reports the quits by industry, so that in our construction all firms in an industry reporting some closure are considered to have been closed. On the other hand, industries where shutdowns were most frequent are underrepresented because they were too small to be reported. Thus, despite gaps in coverage, the census provides a reasonable estimate of the importance of temporary labour migration between sectors to the economy of mid-nineteenth century France. 6 The t-statistic is given in parentheses. XLA and XLI are measured in thousands. 7 At the departmental level, the percentage of wheat production in total agricultural production is highly correlated with the increase in male summer wages between 1852 and 1862.
82
TRADITIONAL LABOUR MARKETS IN FRANCE 8 9
10
11 12
13
14
15
16 17
18 19
The level of French cereal production increased by 1.3 per cent per annum between 1816–20 and 1858–62 (Heffer 1986:187). The proportion is lower if one includes the manufacturing population of Paris, which is not included in the Industrial Census. A survey carried out by the Paris Chambre de Commerce in 1847–8 gives an estimate of the manufacturing workforce that implies that around the middle of the 1840s, about 35 per cent of the industrial workforce lived in communes of less than 2000. These calculations, however, do not include workers in cottage industries. See Statistique générale (1847–52) and Chambre de Commerce (1851). The first census reported for each firm the value of the tax paid (patente) and the basis of this tax (valeur locative). The latter can be taken as a good proxy for capital. We estimated the capital value of a firm at twenty-five times its valeur locative. This is admittedly a crude measure since the ratio capital/valeur locative differs from one industry to another, but the resulting error is not likely to alter our results. On this basis, the mean capital per capita is 584 and 1338 francs for urban and rural firms, respectively. For each firm the ratio of its wage to the farm wage is strongly correlated with the capital of the firm and capital per capita. For rural textile firms, for instance, a 10 per cent increase in capital per worker raises the wage ratio by 1 per cent. The inquiry made by the Office du Travail recorded two pieces of information on days worked. Firms had to state how many days a year they worked (the average was 303). Since they also reported annual and daily wages for each category of worker, an implicit duration can be derived. On this basis, men worked an average of 293 days a year, women 285. Seasonally flexible employment remained a significant phenomenon in departments where the food industry was important. In Aisne, for instance, the firms in the sample of the Office du Travail reported an average working year of 251 days. The firms that employed a fixed number of workers throughout the year hired most of the external workers. About a third of employees worked at firms that reported they had a stable workforce. However, these were the firms that employed most of the ‘external’ workers—roughly three-quarters of the total. Since this survey was based on a fairly small sample, these results must be treated with caution. There are similarities with the Netherlands in the later seventeenth and eighteenth centuries, where workers could live on a combination of jobs as fishermen and sailors, but with the decline of one source of employment, the other also gradually declined (see Chapter 3). Again, see Chapter 3 for a discussion of the growing role of seasonal immigrants to the Netherlands in the late seventeenth and eighteenth centuries. This result is based on a comparison of farm wages (Ministère de 1’agriculture 1897) and those collected by the Office du Travail in the early 1890s. While the Industrial Census of 1860 gave only average wages per firm or group of firms, the Office du Travail inquiry was more detailed and published wages by category of worker. The greater precision of the source indicates that at the end of the century even unskilled industrial workers usually earned higher wages than those offered by farmers for harvest work. This seems broadly consistent with evidence on lack of convergence of wages in English agriculture at this time. See Chapter 5. See Chapters 9 and 10 for discussions of the emergence of internal labour markets and long-term job attachments in North America.
83
5 REGIONAL LABOUR MARKET INTEGRATION IN ENGLAND AND WALES, 1850–1913 George R.Boyer and Timothy J.Hatton
It often is maintained that as economies mature, their labour markets become better integrated both geographically and between the rural and urban sectors. Declining transport costs, improved information flows, and declining institutional impediments to migration lead to an increase in labour mobility. As a result, regionally or locally segmented labour markets become increasingly integrated at the national or even the international level. This chapter examines the integration of labour markets within the rural and urban sectors of England and Wales during the second half of the nineteenth century. Although there is a large literature on internal migration and emigration in Victorian Britain, historians typically have focused on the direction and causes of migration rather than on its consequences for the labour market.1 Broadly speaking, the literature has found that workers did indeed migrate towards better wage-earning opportunities, that most moves were short-distance moves, and that once certain patterns of migration were established they often persisted. The studies leave the strong impression, if only implicitly, that although there was considerable migration, opportunities for arbitrage were not fully exploited. However, analyses of the pattern and extent of migration movements shed little light on the issue of integration. Markets could be perfectly integrated but exhibit little migration or they could exhibit high rates of migration but be poorly integrated. A better measure of labour market integration can be obtained by examining wage rates. There is a large literature on the history of wages during the nineteenth century. However, with the exception of Arthur Bowley (1898, 1900a, 1900b, 1901), A.Wilson Fox (1903), and most importantly E.H. Hunt (1973, 1986), historians have not examined the changes over time in local or regional variations in wages within occupations. This chapter extends the work of Bowley, Fox, and Hunt, by offering several tests for the degree of labour market integration and its trend from 1850 to 1913. We construct annual regional wage series for agricultural 84
REGIONAL LABOUR MARKET INTEGRATION
labourers and carpenters, and use these new wage series to test for regional wage convergence and to estimate structural models to assess the degree of labour market integration between regions. PREVIOUS STUDIES OF LABOUR MARKET INTEGRATION E.H.Hunt’s (1973, 1986) recent conclusion that the British labour market became better integrated from 1850 to 1914 represents somewhat of a break from the historical literature. The few contemporaries and historians who previously had examined wage variations within occupations before 1914 generally were struck by the apparent lack of labour market integration. The existence of large local and regional wage differentials was often commented on, especially in studies of the agricultural labour market. We begin by reviewing the conclusions reached by these earlier studies of labour market integration. Perhaps the best known comment on labour market integration in nineteenth century Britain was made by the agricultural journalist James Caird (1852:517), who concluded from his tour of rural England in 1850–1 that a ‘marked inequality in wages…bisects the kingdom by unmistakable lines into two great geographical divisions’, the high-wage north and the low-wage south. Caird blamed the regional wage differential on low labour mobility caused by the Poor Law and the Settlement Law (1852:517). Frederick Purdy (1861), the author of a study of agricultural labourers’ wages in 1860, reached a conclusion similar to that of Caird. He wrote that ‘no commodity in this country presents so great a variation in price, at one time, as agricultural labour…. A labourer’s wages in Dorset, or Devon, are barely half the sum given for similar services in the northern parts of England’ (1861:344). Purdy (1861:345) also noted that ‘considerable differences in wages obtain in small areas’. He cited several instances in which wages varied by 2s. to 4s. across parishes within a Poor Law union. Thirty years later, Joseph Ashby and Bolton King (1893), the authors of a study of the living conditions of agricultural labourers in south Warwickshire in 1892, also found significant local variations in labourers’ wages. They wrote that ‘in one village the rate of wages has been throughout, from 1872 at least, 2s. below the rate in the surrounding villages. It is not at all uncommon to find a difference of 1s. 6d. to 2s. in adjacent villages’ (1893:5). They attributed these differences partly to differences in the demand for and supply of labour, and partly to custom (1893:5–8). Economic historian J.H.Clapham, discussing the extent of wage variations in agriculture in 1902, wrote that: Variations in earnings from county to county remained astonishingly great. The broad divisions corresponded roughly with differences in labour efficiency….Yet differences in efficiency can hardly explain such gaps as those between the 14s. 6d. of Oxfordshire,…the 16s. 4d. 85
GEORGE R.BOYER & TIMOTHY J.HATTON
of Buckingham, the 16s. 11d. of Essex and the 20s. of Surrey. Surrey no doubt runs into London; but so does Essex, and the Buckingham and Oxford boundaries are not very far away. Some explanation connected with mobility is more probable. (Clapham 1938:97–8) The literature on wage variations within urban occupations is much smaller than that on agriculture. The first person seriously to examine wage variations across cities was F.W.Lawrence (1899). He found (1899:18) that ‘marked changes [in wage rates in the same trade] are noticeable as we pass from one part of the country to another’. Rather than view these regional wage differentials as evidence of labour immobility, Lawrence (1899:52) concluded that they largely were a result of regional differences in the ‘character of the people…. So that where the higher wages are paid they are merely a higher price for a better article.’ J.W.F.Rowe (1928:67) concluded from his study of wages in the building trades that, before 1914, ‘even amongst towns of the same size in the same county, or group of counties, the variations [in wage rates] were appreciable’. He maintained that these variations could only partly be explained by differences in the cost of living or in the demand for labour, and attributed the remaining differentials to ‘that unsatisfactory agency, custom’ (1926:68). Concerning the trend in regional integration, Rowe (1928:70) concluded that the ‘variations in wage rates [in 1914] presented a picture hardly more uniform or symmetrical than that of 1893’. E.H.Hunt (1973, 1986) is the only individual to examine explicitly the trends in labour market integration in nineteenth century Britain. His analysis focused on wage rates of agricultural labourers in 1867–70, 1898, and 1907, and, to a lesser degree, wage rates of carpenters and building labourers in 1886 and 1906. In his 1973 book, Hunt divided Britain into thirteen regions and examined trends in the regional wage structure, whereas in his 1986 article he considered agricultural wages at the county level and carpenters’ wages in six cities. Hunt (1973) concluded that the regional wage structure was relatively stable over the period 1850 to 1914. He wrote that in broad terms there were two high-wage areas in 1850: the London area; and the counties of the north of England together with parts of the Midlands as far south as Birmingham. In 1914 the position was similar except that these two areas had been joined by South Wales and much of southern and central Scotland. (Hunt 1973:4) The only other significant change in the regional wage structure was that the rural southeast of England replaced the southwest as the lowest wage region in Britain (Hunt 1973:17, 64). 86
REGIONAL LABOUR MARKET INTEGRATION
Hunt (1973:58) maintained that ‘there is one generalization which may be made with confidence about most [wage] differentials—they were less wide at the end of the period than in 1850’. To determine the trend in regional labour market integration, Hunt calculated the coefficient of variation of agricultural labourers’ weekly earnings across eighty-six British counties in 1867–70, 1898, and 1907, and of bricklayers’ weekly wages across twentyone English and Welsh towns in 1886, 1900, and 1913.2 For agricultural labourers, the coefficient of variation declined from 0.14 in 1867–70 to 0.10 in 1898, and then to 0.09 in 1907. For bricklayers, it declined from 0.145 in 1886 to 0.107 in 1906, and then remained roughly constant at 0.108 in 1913 (1973:59). Hunt (1973:60) concluded that ‘what is remarkable about [the wage] differentials is not that they were reduced but that they persisted so tenaciously’. He maintained that this was not caused by labour immobility. Rather, ‘between 1850 and 1914 there were strong forces working to reinforce the existing wage differentials, forces so strong that it would be rash to assume that the slow erosion of regional differentials necessarily implies a high degree of immobility’ (1973:242). Hunt’s finding that regional wage variations declined from 1850 to 1914 is suspect, for two reasons. First, it is based on an examination of wage rates of agricultural labourers and bricklayers at only three points in time. For the purposes of establishing whether there was a discernible downward trend in the variation of wages it clearly would be desirable to examine annual time series of wages rather than rely on three benchmark years. Second, Hunt’s earnings data for agricultural labourers are drawn from a separate source for each benchmark year, which raises issues of comparability. The problem is made worse by Hunt’s comparison of estimated weekly earnings rather than cash wages. The earnings data include estimates of the average weekly value of extra wages at harvest time, occasional piece work, and perquisites ranging from a rent-free cottage to beer or cider at harvest time. Although the value of these items should be included in the total wage, it is not clear that the items were adequately measured in the surveys or that when included they were appropriately valued. Most importantly, there is no reason to believe that the valuation of the items was comparable across the three surveys. Because of the problems associated with using estimates of agricultural labourers’ earnings, we focus on weekly cash wage rates, although we recognize that this is not the ideal measure. Like the previous studies of labour market integration, in this chapter we focus on regional variations in wages in two occupations: agriculture and the building trades. Wage rates in these occupations generally are considered to be representative of rural and urban unskilled and skilled wage rates. Bowley (1914:617) maintained that ‘agricultural wages are in close sympathy with wages that can be obtained for unskilled labour in the same neighbourhood’. Similarly, Hunt (1973:4) contended that ‘farm labourers’ earnings are probably the most useful single guide to wage levels in different 87
GEORGE R.BOYER & TIMOTHY J.HATTON
parts of Britain’. Bowley (1900a: 59, 63) used bricklayers’ and carpenters’ wages as a proxy for general wage trends. Hunt (1986:962) maintained that in most respects the wages of carpenters and other building workers can be regarded as a reasonably reliable guide to town wages. There were building workers in every town; the nature of their work did not vary much from place to place, nor was it much affected by technical change; and within towns a single rate of pay prevailed in each of the building occupations. We use two previously neglected data sources to construct annual time series of average regional wage rates for agricultural labourers from 1855 to 1903 and for carpenters from 1865 to 1913. Agricultural wage data were obtained from the Second Report…on the Wages, Earnings, and Conditions of Employment of Agricultural Labourers in the United Kingdom (1905a), written by Wilson Fox for the Board of Trade. Appendix V reports annual observations of ‘weekly cash wages paid to ordinary agricultural labourers’ on individual farms scattered throughout England and Wales for the years 1850 to 1904. Continuous wage series for the period 1855–1903 are available for seventy-nine farms. Carpenters’ wage data were obtained from an unpublished Board of Trade report on Rates of Wages and Hours of Labour in Various Industries in the United Kingdom (1908a), supplemented for the years 1907–13 by the Abstracts of Labour Statistics (1907–1912–13) and the Reports on Standard Time Rates of Wages (1909–14) published by the Labour Department of the Board of Trade. Continuous wage series for the period 1865 to 1913 are available for twenty-nine towns. The published wage rates typically are the standard trade union rates, which might create questions of representativeness if non-unionists worked at lower rates. However, Rowe (1928:65) maintained that even before the First World War ‘in each town or country district the building operatives of each grade were almost invariably paid the trade union standard rates of wages’.3 We grouped the farms and towns in our samples into six regions: the north, midlands, east, south, southwest, and Wales. The regions correspond broadly to the regional breakdown used by Hunt (1973), although we distinguish fewer regions.4 For each region we construct time series of the average weekly wage of agricultural labourers and carpenters. For agricultural labourers the regional wage is calculated as the simple average of the wages of the farms in the region, while for carpenters the regional wage is calculated as a weighted average of the wages of the towns in the region; the weights used were the population of the city in 1891. Because most of the towns for which we have carpenters’ wage data are in the industrial north and midlands, the other regions are represented by very few towns. For Wales we have wage data for Cardiff and Newport, for the southwest we have data for Bristol and Plymouth, and for the south and east we have data only for London. 88
REGIONAL LABOUR MARKET INTEGRATION
MODELLING WAGE RATES AND TESTING FOR MARKET INTEGRATION In this chapter we pursue a number of different approaches for examining labour market integration. These go somewhat beyond the methods used in the existing literature on regional labour markets in Britain. The usual approach, reflected in the discussion above, assumes that the wage, suitably adjusted, should be the same in each location. In principle, this involves adjusting for differences in the cost of living, non-cash payments, hours, working conditions, and wage-earning opportunities for other family members, as well as measuring (and appropriately valuing) all the nonwage characteristics of location. Any gap remaining is a measure of market failure. This approach has been used formally in recent studies of late nineteenth century US wage differentials between urban centres (Rosenbloom 1990) and between the urban and rural sectors in a single region (Hatton and Williamson 1992). It is also one of the principal methods used by Hunt (1973) in his analysis of late nineteenth century Britain. There are several drawbacks with this approach as a test or measure of labour market integration. First, we cannot expect to measure exactly all the advantages and disadvantages of location. Hence, we would probably observe some wage differential even in the unlikely event that the labour market was perfectly arbitraged. If, as is more likely, we only observe cash wages, there is no way of knowing whether a given wage gap represents equilibrium or disequilibrium. Second, even if we could measure and correctly value all the wage and non-wage attributes of a region, we have no criterion to determine how large the gap could be before we conclude that labour markets were not integrated. To demand that the wage gap be equal to zero, requiring as it does that labour markets are instantly arbitraged, is surely too harsh a test. Third, under imperfect arbitrage, shocks to labour supply and demand are likely to drive wage differentials away from equilibrium. Cross-sectional comparisons will not, in general, distinguish between the impact of shocks to labour demand or supply and lack of labour market mobility. One approach to testing for changes in the degree of market integration is to search for trends over time in a measure of the dispersion of wage rates, such as the coefficient of variation. This technique has been used by Rothenberg (1988), who finds that farm wage rates within the state of Massachusetts converged during the first half of the nineteenth century. She interprets this as evidence that a state-wide labour market had emerged by 1850. For this to be the case, it must be assumed that the measured wage effectively summarizes the advantages or disadvantages of each location. Otherwise a narrowing of the ‘true’ wage differentials could be accompanied by a divergence of the measured differentials or vice versa. However, this approach has the advantage that, provided a sufficiently long time series of 89
GEORGE R.BOYER & TIMOTHY J.HATTON
the chosen measure of dispersion is available, the underlying trend can be discerned even in the presence of temporary shocks.5 Unfortunately, the analysis of trends in wage dispersion says little about the degree of integration of labour markets. An absence of any distinct trend in the measure of dispersion could be consistent with very well integrated markets or very poorly integrated markets. For these reasons it is worth pursuing time series methods. Several different time series methods have been used to determine the degree of integration between two markets. One method calculates the correlation coefficient of wage changes between the two labour markets. The higher the correlation coefficient, the more closely integrated the two markets are assumed to be.6 However, this test for integration is flawed, because it does not enable one to distinguish between a strong tendency for instantaneous arbitrage and common shocks to labour demand or supply. A high correlation in changes does not necessarily mean that the two wage series move together in levels in the long run as market integration implies. An alternative method for determining the extent of integration between two labour markets consists of testing for a unit root in the ratio of the two wage series. A number of widely used unit root tests have been developed by Engle and Granger (1987). Although these tests enable one to examine the long-run integration of labour markets, they also have drawbacks. In empirical applications, the tests are often found to be weak against the null hypothesis of no cointegration (Engle and Granger 1987:269). A major disadvantage of simple correlations or cointegration (unit root) tests is that they generally are not based on a structural model of the forces that determine the relationship between two wage rates. We develop a structural model based on labour demand and supply in two labour markets linked together by migration. The model is developed in full in the Appendix and will only be sketched out here. The chief force towards the convergence of wages in two labour markets is migration. Where there are only two labour markets i and j, migration from i to j can be represented as follows:
where mij is the rate of net migration from i to j (negative for migration from j to i), c is a parameter measuring the responsiveness of migration to a given wage differential, and k represents the non-wage advantages of market i relative to market j. The rate of migration depends on the wage ratio and on the mobility parameter, c. The greater is the wage ratio, and therefore the incentive to move, the greater will be the flow of migrants. The parameter c determines the degree of integration between the two labour markets. As c approaches infinity labour becomes perfectly mobile and the two labour 90
REGIONAL LABOUR MARKET INTEGRATION
markets are perfectly integrated. If c was zero then the markets would not be integrated at all. The migration relation which links the labour markets together has implications for the wage ratio. If log(Wj/Wi)t-1 >k then labour would flow from i to j (assuming c > 0). The increase in labour supply in market j would tend to lower Wj while the fall in labour supply in i would raise Wi. The ratio log(Wj/Wi) would, over time, tend to fall and this process would continue (at a decreasing rate) until log(Wj/Wz)=k. Hence, there would be a long-run tendency towards the wage ratio Wj/Wi=ek and if k=0 then in the long run Wi/Wj=1. In the Appendix we have used a variant of this migration equation which allows for migration to and from third markets. We also introduce a simple model of labour demand in each market which allows us to eliminate migration and to express the relation between the two markets in terms of the wage alone. This yields the following relationship which we can estimate:
The parameter d1 reflects the degree to which there are common forces affecting both labour markets while d2 is a measure of the degree of integration of the two markets. The size of d2 is determined by the mobility parameter in the migration equation; if the mobility parameter was equal to zero then d2 also would be equal to zero. This type of model is sometimes called an error correction model. The two wage rates are related in changes but the error correction term d2log(Wi/ Wj)t-1 prevents them from wandering apart over time in levels (provided d2 is negative).7 If we imagine a long-run stationary equilibrium where ∆logWit =∆log Wjt=∆t=0, then log(Wi/Wj)t-1=-d0/d2. That is, -d0/d2 represents the longrun equilibrium wage ratio. We now turn to the empirical analysis of regional labour market integration. We begin in the next section by testing for trends in wage dispersion, and then move on in the following section to time series tests for market integration. COEFFICIENT OF VARIATION TESTS FOR MARKET INTEGRATION E.H.Hunt’s (1973:58–9) conclusion that labour markets became better integrated from 1850 to 1914 was based on estimates of the coefficient of variation of wages for agricultural labourers and bricklayers at a few points in time. In this section we re-examine the trend in labour market integration for these two occupations, using the same methodology as Hunt. We construct annual wage series for seventy-nine farms from 1855 to 1903 and for carpenters in twenty-nine towns from 1865 to 1913, and then calculate the annual coefficient of variation of wages for both occupations. Our analysis offers a more accurate measure of the trend in labour market 91
GEORGE R.BOYER & TIMOTHY J.HATTON
integration, for three reasons. First, an analysis based on an annual time series of wages is preferable to one based on three benchmark years. Second, Hunt’s agricultural wage data are drawn from a different source for each benchmark year, while our wage data are for the same set of farms for each year. Thus, Hunt’s data might not be comparable across years, while our data almost certainly are. Third, Hunt’s estimates of the coefficient of variation are based on nominal wage data. The use of nominal wages is valid for the analysis of the agricultural labour market but not for the analysis of urban occupations. Hunt (1973:80–7) examined differences in rents and food and fuel prices across rural Britain, and concluded that ‘the cost of living of rural workers did not vary significantly in different parts of the country: differences in real wages paralleled differences in money wages’.8 We agree with his conclusion and therefore also use nominal wages in agriculture. On the other hand, there were significant differences in the cost of living across English towns. A 1908(b) Board of Trade Enquiry into Working Class Rents, Housing and Retail Prices estimated the cost of housing, food, and fuel for working-class families in seventy-seven English and Welsh towns in October, 1905. The study found that, although the cost of living differed by less than 10 per cent among the majority of cities there were outliers, the most important of which was London, whose cost of living was significantly higher than that of most other cities.9 To take account of these differences in the cost of living, we deflated the nominal wage of carpenters in each of our twenty-nine towns using the Board of Trade’s cost of living index for 1905. Because town-level cost of living estimates do not exist for any earlier year, we were forced to assume that the differences in the cost of living across towns in 1905 held for the entire period 1865–1913. To determine whether labour markets became more integrated over time, for both occupations we regressed the time series of coefficients of variation on a constant and a time trend (Table 5.1). If there was wage convergence over the period the coefficient of the time trend will be negative and significantly different from zero. Consider first the regressions for agricultural labourers’ wages in part (A) of Table 5.1. The dependent variable in equation (1) is the coefficient of variation of wages for the seventy-nine farms. The regression results show that the coefficient on the time trend is positive and significantly different from zero. Its magnitude suggests that the coefficient of variation increased by 0.023 from 1855 to 1903. Our results therefore do not support Hunt’s conclusion that the variation in agricultural wages declined during the late nineteenth century.10 Rather, they suggest that wage dispersion increased from 1855 to 1903. The upward trend in the coefficient of variation might have been the result of either increasing inter-regional wage variation or increasing intra-regional wage variation or both. To determine whether intra-regional wage variation 92
REGIONAL LABOUR MARKET INTEGRATION
increased over time we grouped the seventy-nine farms into six regions, calculated the coefficient of variation for each region, and then calculated the weighted average (by number of farms) of the coefficients of variation of the regions. The resulting series provides a measure of the average intra-regional wage variation for the six regions. In equation (2) we regress this series on a constant and a time trend. The coefficient on the time trend is positive but not significantly different from zero, suggesting that, on average, intraregional wage variation did not increase from 1855 to 1903. To examine the trend in inter-regional wage variations, we calculated regional average wages, then calculated the coefficient of variation of wages across the six regions and regressed it on a time trend. Equation (3) shows a significant trend towards increasing inter-regional wage dispersion. Together, equations (2) and (3) reveal that the increase in the coefficient of variation of agricultural wages from 1855 to 1903 was caused almost completely by increases in wage variations across regions. The above results suggest that the agricultural labour market was regionally segmented in the late nineteenth century. However, the increasing regional wage differentials might have occurred despite a high degree of inter-regional labour mobility, if the ‘strong forces working to reinforce existing wage differentials’ noted by Hunt (1973:242) could not be overcome even by relatively high migration rates. Unfortunately, we cannot distinguish between these two possibilities by looking at trends in the coefficient of variation. Part (B) of Table 5.1 analyses carpenters’ wages. It is generally maintained that the labour markets for skilled occupations were better integrated than those for unskilled 93
GEORGE R.BOYER & TIMOTHY J.HATTON
occupations, because of the higher mobility of skilled workers. According to Hunt (1981:155), ‘urban skilled workers were… better educated, better informed, and more likely to move over long distances. Their trade unions, in particular, were an important source of market information.’ Table 5.1 supports this hypothesis; the mean level of the coefficient of variation of carpenters’ wages across twenty-nine towns was approximately one-half of the mean level of the coefficient of variation of labourers’ wages across seventy-nine farms. The regressions to examine trends in the coefficient of variation of carpenters’ wages were run using both nominal and real wages. Because wage data are available for only a small number of towns within certain regions, we were able to estimate only the trends in overall and inter-regional wage variations. The variation in nominal and real carpenters’ wages declined sharply over time both at the town and at the regional level. The coefficients on the time trend in equations (6) and (7) suggest that at the town level the coefficient of variation of real wages declined by 0.037 from 1867 to 1913; at the regional level by 0.044. In sum, the labour market for carpenters became significantly better integrated in the half century before the First World War. In this section we have examined changes over time in labour market integration for farm labourers and carpenters by statistically testing for trends in the coefficient of variation of wages. Despite using the same measure of labour market integration that Hunt used, we have reached different conclusions than he did. Regional wage differentials for carpenters declined in the late nineteenth century, but regional differentials for agricultural labourers increased. This result raises several questions about the extent of regional labour market integration, questions that cannot be answered by the analysis of trends in the coefficient of variation. We now turn to time series models of market integration in order to determine the degree of labour market integration between individual regions. TIME SERIES TESTS FOR REGIONAL INTEGRATION In this section we examine the time series relationships between pairs of regions for agricultural labourers’ and carpenters’ wage rates, to see which regional labour markets were most and which were least integrated with the others. In order to allow for changes over time in the equilibrium wage ratio for each pair of regions, we introduce time trends in the error correction model. We first examine the agricultural wage rates. Figures 5.1 and 5.2 present the average weekly wage rate of agricultural labourers for each region from 1855 to 1903. In the north, midlands, and south, wages increased sharply from the late 1850s to the mid-1870s, then declined slightly in the late 1870s to a plateau that was maintained from
94
REGIONAL LABOUR MARKET INTEGRATION
about 1880 to the late 1890s, and then increased again around 1900. The wage series for the east follows a somewhat similar path from 1855 to 1880, although it is much more volatile than the other series. Wages declined during the 1880s, and then increased sharply after 1896. By contrast, the southwest experienced a slow and steady increase in wages with few distinct fluctuations. As in the other regions, wages maintained a plateau in the 1880s and increased sharply around 1900. Finally, the wage series for Wales followed a pattern similar to that of the southwest, but with fluctuations in the 1860s and 1870s similar to the midlands and the south. The results obtained from estimating the error correction model for the fifteen pairs of regions are presented in Table 5.2. In each case the coefficient on the ?logWjt term (d1 in equation (5.2)) is positive and significantly different from zero at the 5 per cent level. This suggests that common shocks arising from changes in labour supply or demand were important in agricultural labour markets. It is likely that these arose chiefly on the demand side due to fluctuations in agricultural prices and variations in weather conditions. However, there is little evidence that the mainly arable areas of the midlands, south, and east show higher short-run correlations and larger coefficients between them than they do with the more predominantly pastoral areas.11 A negative and statistically significant coefficient on the log(W i/W j) t-1 term (d 2 in equation (5.2)) is evidence of convergence to a long run equilibrium wage ratio. The coefficient is negative in each of the regionalpair regressions in Table 5.2. However, seven of the fifteen coefficients are not
95
GEORGE R.BOYER & TIMOTHY J.HATTON
significantly different from zero, suggesting that the equilibrating supply side forces were weak. The regional pairs exhibiting the strongest evidence of integration are: the midlands and south, north and midlands, north and south, Wales and south, and midlands and Wales. Wales was integrated with every region except the north, while the south, north, and midlands each were integrated with three of the five other regions. All but one of the insignificant coefficients were associated with regressions involving either the southwest or the east. The results suggest that the southwest was integrated only with Wales. Oddly, the east was integrated with the distant north and Wales, but not with the midlands, south, or southwest. The rate of adjustment of the wage in region i to a shock to the equilibrium wage ratio, holding Wj fixed, can be estimated from the coefficient on the log(Wi/Wj)t-1 term, d2. The predicted lag between an initial shock and the return to equilibrium is equal to (1-d2)/d2. In the regressions in Table 5.2 the lag varies from one year and 1.4 years for the midlands/south and north/midlands pairs to 7.3 years and 8.1 years for the Wales/east and north/east pairs. Before considering the implications of these results we turn to the carpenters’ wage rates for comparison. Figures 5.3 and 5.4 present the average weekly wage rate of carpenters for each region from 1864 to 1913. It should be recalled that the south and east are represented only by London. To take account of differences in living costs across cities the series have been adjusted using the Board of Trade’s relative cost of living index for 1905. As was observed for the agricultural wage rates, there is a strong similarity in
96
REGIONAL LABOUR MARKET INTEGRATION
the profiles of the north and midlands. Both exhibit a steep ascent from the mid1860s to the late 1870s followed by a short decline and a plateau in the 1880s. After another sharp increase in the 1890s the profiles are flat until 1912. The other series exhibit fewer changes, reflecting the fact that Wales and the southwest are represented by only two towns each and the south and east only by London. The few changes in these series illustrate the stickiness of trade union wage rates. Despite this there is a close resemblance between the broad pattern exhibited in Wales and the southwest and that of the north and midlands. The clear outlier is London. Only at the very beginning and in the last fifteen years or so of the period does the London wage follow a pattern similar to that of the other regions. Overall the increase in the London wage is less marked than the increase in the wage for the other regions.
97
GEORGE R.BOYER & TIMOTHY J.HATTON
The results obtained from estimating the error correction model for the ten pairs of regions are presented in Table 5.3. The coefficient on the ?log Wjt term is positive and significantly different from zero for all but two of the regional pairs, both of which involved London. The coefficient on the error correction term log(Wi/Wj)t-1 i is negative in each of the regional-pair regressions, and is significantly different from zero in seven of the ten 98
REGIONAL LABOUR MARKET INTEGRATION
regressions. The regional pairs exhibiting the strongest evidence of integration were Wales and southwest, and midlands and Wales. London, the midlands, Wales, and the southwest each were integrated with three of the four other regions. Surprisingly, the north and midlands were found not to be integrated. The predicted lag between an initial shock to the equilibrium wage ratio and the adjustment back to equilibrium varied from 1.8 years and 1.9 years for the Wales/southwest and midlands/Wales pairs to 7.3 years for the midlands/London pair. How do the labour markets for carpenters and agricultural labourers compare in terms of regional integration? Although agricultural wage rates appear to have been more volatile than carpenters’ wage rates, the tests for regional integration produce results that overall are similar for the two sectors.12 The error correction model produces coefficients with similar orders of magnitude in the two sectors, although a larger share of the urban regional pairs were found to be integrated. It is tempting to conclude that urban labour markets were better integrated than rural markets, but the evidence is not compelling.
99
GEORGE R.BOYER & TIMOTHY J.HATTON
IMPLICATIONS OF THE TEST RESULTS In this section we further analyse the results obtained in the preceding sections. First, we compare the extent of labour market integration in England and Wales with that of other industrializing countries in the late nineteenth century. Second, we use the results of the error correction model to determine the extent of labour market integration between regions within England and Wales. We then compare our conclusions with the results of the literature on inter-regional migration. In Table 5.4 our estimates of the coefficient of variation of agricultural labourers’ wages from 1860 to 1900 are compared with estimates for the United States, France, Prussia, and Sweden. The reported coefficients of variation for France, Prussia, and Sweden are for wages of agricultural labourers and general unskilled labourers at the departmental/county level of aggregation, while those for the United States are for wages of agricultural labourers at the regional level. Thus the estimates for the United States correspond to the regional estimates for England and Wales, while the estimates for France, Prussia, and Sweden correspond to the county estimates for England and Wales. Three results stand out in Table 5.4. First, from 1860 to 1880 the market for unskilled labour was better integrated in England and Wales than in the other economies.13 Given that Britain began to industrialize several decades before the other countries and was by any measure the most developed economy at least until the 1890s, this result supports the hypothesis that improvements in transportation and information networks during economic development play a major role in increasing labour market 100
REGIONAL LABOUR MARKET INTEGRATION
integration. Second, the two most rapidly industrializing economies during the period, the United States and Prussia, experienced wage convergence in their unskilled labour markets in the late nineteenth century, while England and Wales did not. Again, this suggests that economic development leads to increased labour market integration. Conversely, it is possible that regional wage convergence was a cause of rapid economic growth in the United States and Prussia. Third, wage dispersion in England and Wales was significantly higher in 1890 than in the other years. The sharp increase in wage dispersion from 1880 to 1890 and its sharp decline from 1890 to 1900 were largely the result of movements in wages in the lowest wage region, the east, and the highest wage region, the north. From 1880 to 1890, nominal wages declined by 9.1 per cent in the east and increased by 2.3 per cent in the north, while from 1890 to 1900 wages increased by 20 per cent in the east and by 5.3 per cent in the north.14 The ratio of wages in the north to wages in the east increased from 1.39 in 1880 to 1.56 in 1890, and then declined to 1.37 in 1900. Turning to regional integration within England and Wales, the results obtained from the error correction model indicate that many regional pairs’ labour markets at best were only weakly integrated. For each regional pair there was some tendency for wage ratios, once disturbed, to revert to a longrun stable path. However, the small size and lack of statistical significance of many of the coefficients on the log(Wi/Wj)t-1 term show that this often was a slow process. Some regions seem to have been better integrated than others. In agriculture, the north, midlands, south, and Wales were well integrated with other regions, while the southwest and east were poorly integrated. The results of the error correction model suggest that the southwest was integrated only with Wales. This finding is consistent with the literature on migration. Baines (1985:234) found that the southwestern counties of Devon, Cornwall, and Somerset had relatively low rates of internal migration.15 The major destination of out-migrants, especially in the 1870s and 1880s, was South Wales (Friedlander and Roshier 1966:254–6, 263). The Royal Commission on Labour (1893) reported that [farm] labourers from Wiltshire, Somerset and Devon, from Hereford and the Cirencester district of Gloucestershire have almost everywhere superseded the indigenous Welsh labourer in the Vale of Glamorgan. …But even the newcomers do not remain very long on the land,… so that other labourers have to be continually drafted from the same English counties to replace them. (quoted in Thomas 1969:45–6) The east’s lack of integration with other regions is harder to explain, although Baines (1985:234–5) found that Norfolk and Suffolk had relatively low out101
GEORGE R.BOYER & TIMOTHY J.HATTON
migration rates, and Frederick Clifford (1875:124–5) wrote that East Anglian farm labourers ‘are far too slow in migrating…even when there is a reasonable assurance of higher wages’. Easterners who did migrate overwhelmingly moved in the direction of London, which for most of them could be reached without passing through another region (Friedlander and Roshier 1966:252– 8).16 Turning to the carpenters, the results of the error correction model suggest that each of the regions was reasonably well integrated with the others. London’s wage profile looks different from the wage profiles of the other regions. However, although the short-run correlations between wage changes in London and other regions are low, in the long run London appears to be integrated with every region except the southwest. The regressions of the coefficients of variation on a time trend reported in Table 5.1 show that regional wage dispersion for agricultural labourers increased during the second half of the nineteenth century. Is this result consistent with the migration literature? Recent estimates by Baines (1985:238) suggest that the rate of male internal migration from thirty-four rural British counties declined from 10.9 per cent of the ‘total native population’ in 1871– 80 to 9.3 per cent in 1881–90 and further to 8.9 per cent in 1891–1900. The rate of male internal migration declined sharply in the 1880s in all five eastern counties and in the southwestern counties of Devon, Cornwall, and Somerset (Baines 1985:287–90). Baines (1985:241) concluded that the decline in rural outmigration in the 1880s and 1890s ‘must have been caused by changes in the relative attractiveness of towns… [perhaps] by fluctuations in the demand for building labour’. However, rural-urban wage differentials were larger in the 1880s and 1890s than in the 1870s (Boyer and Hatton 1994:21–2; Williamson 1985:49). Moreover, Boyer’s (1986) analysis of migration from twenty rural southern counties to six major destinations in each decade 1851–60 to 1891–1900 found that the rate of migration declined significantly in the 1880s and 1890s even when one controlled for origin and destination wage rates and the proportion of the rural population who were young adults. On the other hand, migration rates in the second half of the nineteenth century were higher than English migration rates in the 1920s and 1950s or migration rates in current Third World countries (Hunt 1973:262; Williamson 1988:291–3). From 1841 to 1911, 4.5 million persons migrated from rural registration districts in England and Wales (Cairncross 1949:85). Even Norfolk and Suffolk, which had relatively low out-migration rates, experienced a loss of 373,000 migrants from 1851 to 1900 (Cairncross 1949:84). In sum, England and Wales experienced a high rate of labour migration in the second half of the nineteenth century, but our results show that it was not high enough to ensure that labour markets were arbitraged in the short run. 102
REGIONAL LABOUR MARKET INTEGRATION
CONCLUSIONS The only previous empirical analysis of market integration in England and Wales, by E.H.Hunt (1973, 1986), concluded that both rural and urban regional wage differentials declined significantly in the second half of the nineteenth century. Our results indicate that the issue of market integration is much more complex than Hunt’s analysis suggests. Statistical tests for trends in the coefficient of variation of wages show that regional wage differentials for carpenters declined in the second half of the nineteenth century, but regional differentials for agricultural labourers increased. We also divided England and Wales into six regions and examined the degree of market integration between each pair of regions using a simple error correction model. The results of these tests suggest that some regions were better integrated than others. In particular, the agricultural labour markets in the southwest and the east were very poorly integrated with labour markets in other regions. Although there is some evidence that the labour market for carpenters was better integrated than that for farm labourers, some regional pairs’ labour markets were weakly integrated at best. Overall, the results support the conclusion of Clapham (1938), Rowe (1926), and many contemporaries that the integration of labour markets at the national level was far from complete in 1914. APPENDIX In this appendix we develop a simple model of labour demand, labour supply, and wage rates where two labour markets are linked together and also to third markets by migration. From this analysis we obtain the error correction model used in estimation and which appears as equation (5.2) in the text. In the two labour markets, i and j respectively, labour demands are
where Wi, Wj are nominal wages and all other factors (e.g. output prices) are in the demand shift terms Di, Dj. Changes in labour supply are simply the difference between rates of natural increase, n, and rates of net outmigration, m, so that
Rates of out-migration are determined by the (geometric) weighted average of the wage rate in the other region (i or j) and in the third market, z, relative to that in the own market:
103
GEORGE R.BOYER & TIMOTHY J.HATTON
where si, sj represent the ‘weight’ of the third market immigration decisions in i and j, and ki and kj represent the non-wage advantages of location i and j relative to the weighted average of the others.
Equation (5A.6) is essentially the equation we use for estimation in the text. The terms in ∆D) reflect demand conditions which we do not have in our estimating equation. If the variables driving labour demand are integrated of order 1 (as is frequently found with price series) then ∆D will be integrated of order 0 and will form a random disturbance. In the estimating equation, the variables n and k are treated as constants, although any secular changes will be captured in the time trend which is added to the model. The coefficient on ∆logWjt is expected to be positive and would be close to one for two symmetric regions. The coefficient on log(Wi/Wj)t-1 will be negative since ai, the labour demand elasticity, is expected to be negative. It is worth noting that the lower the migration elasticity, ßi, which reflects the degree of labour mobility, the smaller will be the coefficient on the error correction term in the equation. NOTES We would like to thank George Grantham, Mary MacKinnon, Joel Mokyr, Jeffrey Williamson, and conference participants for helpful comments on an earlier version of this paper. 1 The literature on migration in Victorian Britain includes Redford (1926), Cairncross (1949), Thomas (1954, 1972), Friedlander and Roshier (1966), Baines (1985), and Boyer (1986).
104
REGIONAL LABOUR MARKET INTEGRATION 2
3
4
5 6 7 8
9
10
The coefficient of variation of wages is defined as the standard deviation of wages divided by the mean wage. Hunt refers to the coefficient of variation as the ‘standard deviation of county earnings from their mean’ (1973:59). In his 1986 paper, Hunt (1986:951) calculates the coefficient of variation of agricultural labourers’ weekly earnings across forty-two English counties in 1767–70, 1794–5, 1833–45, 1867–70, and 1898. According to Rowe (1928:65–6), ‘all employers [in the building trades] had to pay the same rate, not mainly as a result of trade union organisation, but owing to the peculiar economic conditions of employment in their industry…[Because] all ordinary building operations are much alike, there is therefore a great potential fluidity of labour between different firms in any district.’ Hunt (1973:8) divided England and Wales into ten districts. The major differences between our regional breakdown and Hunt’s is that Hunt divides Wales into two regions and the north into four regions, while we consider Wales and the north each to be one region. In addition, Hunt includes a region, London and Home Counties, that we do not, and we include a region, east, that Hunt does not. Because Hunt (1973, 1986) examines the coefficient of variation of wages for only a few benchmark years, he is unable to separate trends in wage dispersion from temporary shocks. For instance, Hodne and Gjolberg (1981) use correlation coefficients to analyse labour market integration in nineteenth century Norway. Thus, this approach combines short-run correlation of changes in wages and long-run cointegration. Hunt (1973:80–7) found that food prices were similar throughout rural England, that rural rents were slightly higher in the north than in the south, and that coal prices were higher in the south than in the north. Hunt’s findings are supported for an earlier period by Crafts (1982:68), who found that in 1843 regional price variations were relatively small even when urban areas were included. The major reason for the high cost of living in London was its relatively high cost of housing. The Board of Trade report (1908b: xvi) found that ‘the average rents in the towns included in the enquiry are from 50 to 60 per cent of those prevailing in London’. Overall, the report found that the cost of living was higher in London than in any other British city. Examples of the relative cost of living in several of the cities and towns for which we have wage data are: London=100; Newcastle=94; Cardiff=91; Leeds=86; Manchester=86; Birmingham=85; Stockport=81. Hunt’s conclusion was based on a comparison of the coefficient of variation of weekly earnings across counties in 1867–70 and 1898. For our sample of seventynine farms, the coefficient of variation of weekly wages increased from 0.157 to 0.165 from 1867–70 to 1898. Hunt obtained earnings data for 1867–70 from Bowley (1900a: end table) and for 1898 from Board of Trade (1900:28–9, 58, 73–4). These sources report weekly wage rates of agricultural labourers as well as weekly earnings. The coefficient of variation of weekly wages across counties calculated using these data is equal to 0.163 in 1867–70 and 0.136 in 1898. Bowley (1898:704–7, 1900a: end table) also reports county-level wage data for earlier years, including ‘the returns of agricultural wages collected officially for Michaelmas and Christmas, 1860, and Lady Day, 1861’. From these data we calculate that the coefficient of variation of weekly wages across counties in 1860–1 was equal to 0.131. Thus, the county-level wage data reported by Bowley and Fox suggest that wage dispersion did not decline, and might have increased slightly, from 1860–1 to 1898.
105
GEORGE R.BOYER & TIMOTHY J.HATTON 11 12 13
14 15 16
The one exception is the east, the most arable region in Britain. The coefficients on the ?log Wjt term are significantly larger in the south/east and midlands/east regressions than in the other regressions involving the east. The apparent differences in volatility may be due in part to differences in the nature of the wage data, although it seems likely that urban wages were indeed more sticky than rural wages. The unskilled labour market also might have been better integrated in England and Wales than in the other countries in 1890 and 1900. Weir’s estimates suggest that the county-level variation in agricultural labourers’ wages in 1890 was lower in France than in England and Wales, but the estimates by Sicsic suggest that wage dispersion in France was equal to that in England and Wales. Moreover, the coefficient of variation of wages for French unskilled labour was higher than that for English and Welsh agricultural labour in 1890. The estimates in Table 5.4 also suggest that wage dispersion was lower in Prussia than in England and Wales in 1890, and that wage dispersion was about the same in the two economies in 1900. However, Prussia is only one region within Germany, and there is reason to believe that the Prussian estimates understate the extent of wage dispersion across German regions. Bry (1960:106) maintained that there were ‘striking differences in wages paid in the several geographic regions of Germany’ in 1885 and 1905. For England and Wales as a whole, nominal wages of agricultural labourers declined by 0.4 per cent from 1880 to 1890, and increased by 9.6 per cent from 1890 to 1900. On the other hand, each of these counties had ‘exceptional’ emigration rates (Baines 1985:234). The fact that the intervening counties between the migrants’ place of origin and London were in the east suggests that easterners might have had a high rate of intra-regional migration but a low rate of inter-regional migration. For evidence that this was the case, see Ravenstein (1885:206–11).
106
6 REAL INCOMES IN THE ENGLISH-SPEAKING WORLD, 1879–1913 Robert C.Allen
In the late nineteenth and early twentieth centuries, millions of Brits were on the move. Migration within the British Isles was substantial, as explained in Chapter 5. Moreover, between 1879 and 1913, 5 million people left the United Kingdom altogether: 2.7 million of them went to the United States, 1.3 million to Canada, and 700,000 to Australasia.1 The British migrants to the United States were joined by many other Europeans, and, of course, Americans and Canadians joined the foreigners in the march across the North American continent. These population movements were an integral part of the development of the world economy. The industrialization of Britain, the northeastern United States, and central Canada increased the demand for primary products. When prices were high enough to attract labour and capital from Britain and the east coast of North America, western North America and Australia developed by exporting grain, minerals, lumber, wool, and meat to the metropolitan regions that supplied the periphery with capital and labour. The slowdown of growth in Victorian Britain abetted the movement of population by adding a ‘push’ factor to the ‘pull’ factor of economic opportunities in the developing west. The westward migration was substantial and tended to equalize wages in the sending and receiving regions. A question, however, which is receiving increasing attention (cf.Williamson 1992) is how tightly integrated labour markets were between sending and receiving regions. If they were perfectly integrated, real wages would have been the same in Britain, Canada, Australia, and the United States except for the amortized cost of moving. The allowance for moving costs implies that British wages would have been less than those elsewhere, but differences between the booming regions of North America and Australia would have been small. With real wages equalized by migration (and real profits equalized by parallel flows of capital), real per capita GDP would have been the same in the four countries, except for differences in capital—labour ratios and differences in the value of land and natural 107
ROBERT C.ALLEN
resources. In this simplest model of the world economy, differences in economic dynamism and growth would have affected the flows of labour and capital. Different rates of economic growth would have led to real estate fortunes in the growing regions but not to different wage or profit rates. The main purpose of this chapter is to test this simple model by comparing real wages in major cities in the English-speaking world. I will focus on Manchester, Chicago, San Francisco, Toronto, Vancouver, and Sydney. Manchester and Sydney were leading industrial cities. Chicago was at the centre of the American industrial revolution, and San Francisco was the largest city in the American west. Toronto and Vancouver occupied analogous positions in Canada. In addition to real wage comparisons involving these cities, I will compare real GNP in the countries concerned. AN INTERURBAN-INTERTEMPORAL CONSUMER PRICE INDEX In order to compare real wages, it is necessary to adjust nominal wage rates for differences in consumer prices. Constructing an interurban-intertemporal consumer price index is consequently the first task of the chapter. A consumer price index should average retail prices and reflect actual spending patterns. Table 6.1 summarizes the weights used in some important studies of the cost of living in Britain, Canada, and the United States. Generally, the weights are derived from contemporary expenditure surveys of workingclass budgets. In all cases, there was some ‘miscellaneous expenditure’ omitted in computing weights. These omissions do not create serious distortions in most cases since the items included in the index account for most spending. For instance, omitted items in the 1914 Birmingham survey (4) discussed by Shergold (1982:211) amounted to only 7.4 per cent of spending. Table 6.1 also reports the shares implied by Prest’s (1954) reconstruction of UK consumer spending in 1900 (5) and the results of Knibbs’ (1911) 1910–11 survey of Australian spending patterns (8). There are significant differences in the weights implied by Prest’s work and the surveys of British working-class spending. A much bigger share of working-class income was spent on food, which is not surprising, and much less was spent on drink. Either the middle and upper classes drank vastly more than the working classes or the working-class surveys underreported expenditures on alcohol. Knibbs’ Australian results—with their very small share of income spent on food and correspondingly large share spent on ‘other’ things—are of importance in showing the extent of the difference in spending patterns across the English-speaking world. Since the prices of consumer goods did not vary equiproportionately over time and space, the choice of weights has a bearing on the measurement of aggregate price differences.
108
INCOMES IN THE ENGLISH-SPEAKING WORLD
In constructing my consumer price index, I have adopted a simple weighting procedure: my weights, which are detailed in Table 6.2, are patterned after Bowley’s (1937) British weights ((1) in Table 6.1) and Shergold’s 1914 Birmingham survey.2 I made this choice since the British worker was the typical migrant. An index using weights reflecting his spending pattern interprets the world as he saw it—and is thus an appropriate set of weights when the effectiveness of migration to equalize wages is an issue. It should be noted, however, that an index using other weights or a combination of weights might give a different ranking of real incomes. The food, fuel and light, tobacco, and beer components are the most securely established. For most commodities in many years, it was possible to find retail prices in the city concerned or in a nearby city where one would expect prices to have been similar. Collecting these prices was facilitated by the work of the major statistical agencies early in the twentieth century. The Board of Trade’s Report on Wholesale and Retail Prices in the United Kingdom (1903) and the New South Wales Year Book, particularly for 1905–6, supplied numerous price series running back to the 1870s, and the Eighteenth Annual Report of the United States Bureau of Labor (1904), did the same thing back to 1890. The inflation at the end of the First World War brought more inquiries including an important collection of time series in the New South Wales Statistical Register (1919–20). These sources provide the majority of price quotations for food, lighting, etc., although some series were extracted from other sources, and later publications were used to carry sources forward to 1913. For the UK and the United States, these later data 109
ROBERT C.ALLEN
were inferior to the data for the 1890s, while for Canada there is almost no information before 1900. Full details of sources are provided in the Appendix. It was not always possible to find the retail price of every commodity in every city in every year. Cattle, for instance, were slaughtered in one way in Britain and Australia and another in Canada and the United States, so meat cuts did not correspond exactly. I matched them on the advice of my butcher. I found the prices of raisins and currants in Britain and Australia, but the price of prunes in the United States and Canada. In the end, I decided that dried fruit was dried fruit and made no distinction between the various kinds. I could not find the prices of lard and dried beans in Australia, so I presumed they bore the same relation to English prices as did the prices of butter and potatoes. I could not find Canadian prices for beer and tobacco and substituted American prices in their stead. I could not find Sydney prices for a number of commodities, so I used Brisbane and Hobart prices. Likewise, I applied London prices to Manchester.3 110
INCOMES IN THE ENGLISH-SPEAKING WORLD
Rent was not difficult to find, although comparisons of rent are bedevilled by differences in the quality of the housing and by climatic differences which affect the desirable characteristics. I measured rent as the average rent per room of a four- or five-room working-class house. My sources and procedures are explained in detail in the Appendix. I made no attempt to correct the rent series for quality differences or environmental variation. The least reliable component of the consumer price index is clothing. American and Canadian studies of the cost of living have made great headway by using Sears, Montgomery Ward, and Eaton’s catalogues to track clothing prices (Rees 1961:79–96, 148–53; Bertram and Percy 1979:304–5, 311–12). Someday these sources will provide a good measure of the relative cost of clothing in the two countries, but no results are yet available. It is not possible to use comparable sources to measure clothing prices in Australia and Britain. There were department stores in Australia, but the population has always been highly concentrated in the major cities, so a mail order business was not important as it was in North America. Consequently, there are no catalogues to consult. In Britain, department stores were not important nor indeed was factoryproduced clothing—until a late date most clothing was made at home or custom made by tailors. Other sources must be used to measure clothing prices. I measured clothing prices as follows. I used Bowley’s (1937:121) clothing price index as the price in the UK.4 Shergold (1982:172) has sifted through the available information on British and American clothing and footwear prices. Following his lead, I set the Chicago price in 1905 at twice the British price and then extended the Chicago price forward and backward using Rees’ index. Quotations in the Aldrich Report (1892) suggest that San Francisco prices in 1890 were 10 per cent higher than Chicago prices, so my San Francisco clothing index equals my Chicago index plus 10 per cent.5 Various comparisons of Toronto and eastern American prices indicate little difference between Canada and the United States, so I applied my Chicago and San Francisco price series to Toronto and Vancouver, respectively. Finally, on the basis of the New South Wales Report on Census and Industrial Returns Act of 1891, I set the price of clothing in Sydney 20 per cent above the British price. In computing the consumer price index, Australian, American, and Canadian prices were converted to sterling at the par exchange rates, and then price relatives for all six cities were computed with Manchester prices in 1896 as the common base. The consumer price index was then calculated as a weighted average of these price relatives with the weights shown in Table 6.2. Full series of prices for Manchester and Sydney spanned the period 1879–1913, but the American series began in 1890 and the Canadian in 1900. I extrapolated the American price indices back to 1879 using the Hoover (1960) and Long (1960) indices. Coats (1915) reported Canadian data only for 1900, 1905, and 1910–13. I interpolated the values for the missing years, 111
ROBERT C.ALLEN
which does not introduce serious distortions, and extrapolated the series back to 1879 using the Barnett retail price index for Kingston, Ontario (Urquhart 1986:88). The extrapolated values are highly suspect. Figure 6.1 (and Table 6A.I) show the six consumer price indices. All indices show the Kondratieff pattern of falling prices before 1896 and rising prices thereafter. Prices were lowest in Britain and highest on the west coast of North America. Australian prices were only as high as Toronto prices. There were big differences across the cities in relative prices. Manchester had the lowest rent by far. Rents were highest in San Francisco and Sydney and usually two and a half times Manchester levels. Rents in Chicago, Toronto, and Vancouver were one and a half to two times Manchester levels. Rents were more stable than other prices especially in Britain and the United States. Sydney rents exhibit a mild Kondratieff pattern, and the rate of rent inflation was highest in Canada after 1900. Since housing was not a traded good, considerable variation in the levels and rates of change of rents is not surprising. On the other hand, textiles were widely traded. North American prices were double British prices. Tariffs were the cause of the high level—in Sydney, where there was no tariff, textile prices were only 20 per cent above Manchester levels. While the Manchester entrepreneurs may not have adopted the latest techniques in spinning and weaving, they did, nonetheless, produce cloth more cheaply than anyone else. Food was also widely traded, and tariffs were not a factor. Food, especially 112
INCOMES IN THE ENGLISH-SPEAKING WORLD
meat, was cheapest in Australia. When shipping rates fell enough to allow trade, Australia always exported wheat and meat to the UK. In the 1890s, food prices were also lower in Chicago than in Britain, and at that time American meat invaded the UK. After 1900, food prices inflated more rapidly in North America than in Britain. By 1913, food was more expensive in Chicago and Toronto than in Manchester, and American exports had dropped considerably. Food was always more expensive in San Francisco and Vancouver than it was in the UK. Fuel and light costs were highest on the west coast of North America, mainly due to the prices of coal and illuminating gas. Coal was particularly expensive in San Francisco and gas in Vancouver. The prices of these items were similar in the other cities. There was no important variation in the prices of candles or kerosene. One test of my consumer price index is to compare it with the national price indices for Britain, the United States, Canada, and Australia. (See Table 6.3.) Bowley’s index for the UK is the most venerable. Recently, Gazeley (1989) has subjected it to careful scrutiny and decided it is reliable, while Feinstein (1991) has revised it principally by shifting the weights away from food and rent and onto tobacco, alcohol, and services. My index closely parallels Bowley’s and is within four percentage points of Feinstein’s, which is the most stable of the three. Rees (1961) has revised Douglas’ (1930) American index. My San Francisco index follows Rees’ closely, and my Chicago index is much closer to his than to Douglas’. Likewise, my Toronto index, in particular, follows Bertram and Percy’s (1979) Canadian index closely. My Vancouver index is also closer to theirs than to the discredited Department of Labour index. My Sydney index tells the same general story as the Australian ‘A’ series consumer price index, but there are considerable differences in detail. In this case, my index is superior to the official Australian index, which was based only on food prices and house rents in this period. My consumer price index mimics the behaviour of the better national price indices reasonably well. The main contribution of my index is its comparisons of price levels across North America and between countries. Several comparisons have been made of prices in Chicago and San Francisco in 1890. I found that San Francisco prices exceeded Chicago prices by 22 per cent. Haines (1989) computed the differential as 17 per cent, Rosenbloom (1990:91) as 23 per cent, and Coelho and Shepherd (1979:73) found a differential of 18 per cent between the Pacific and East North Central regions. None of these indices is as complete as mine, but it is encouraging that they yield similar results. Phelps Brown and Browne (1968:46) estimated the difference in American and British prices to have been 45–52 per cent in 1905–9, and Shergold (1982:210–11) estimated that Pittsburgh prices exceeded Birmingham prices by 65–7 per cent in the same period. I found that the differential increased between 1900 and 1913, i.e. inflation was higher in the United States than 113
ROBERT C.ALLEN
114
INCOMES IN THE ENGLISH-SPEAKING WORLD
in the UK. Chicago prices exceeded Manchester prices by 29–42 per cent and San Francisco prices exceeded Manchester prices by 58–66 per cent. My estimates are in rough agreement with these previous attempts and extend the comparisons over time and to more countries. REAL GNP PER CAPITA I will examine the history of real GNP before discussing levels of real wages. There are three important reasons for comparing GNP. First, real gross national product per capita is the most widely used measure of economic development. Second, a major issue in the history of all four countries is whether the benefits of economic growth ‘trickled down’ to the working class. Answers to this question often turn on whether real wages were growing as rapidly as real per capita GNP. Third, migration—and thus labour market integration— may have depended on real GNP instead of real wages. Many migrants bought property in their new countries and many established businesses, so they received income attributable to land and capital as well as labour. If those markets were perfectly integrated, such investments could be made without migrating. It is probable, however, that decisions about investing in land and capital could not be separated from one’s residence and job. In that case, migration may have depended on all factor incomes rather than just labour income. Accurate estimates of GNP are now at hand. Butlin’s (1962) reconstruction of Australia’s national accounts is one of the earliest attempts using modern concepts. Feinstein’s (1972) estimates for Britain are regarded as definitive. Recently, Firestone’s (1958) estimates of Canadian national income have been superseded through a massive research project (Urquhart 1986, 1993). Kuznets’ American estimates have been revised several times—originally by Kendrick (1961) and Gallman (1966) and most recently by Romer (1989) and Balke and Gordon (1989). I use the latter as providing the best estimate of nominal GNP. (There is little difference in their estimates of real GNP.) With the revisions of the Canadian and American accounts, reliable comparisons of per capita income are possible for the first time.6 The problem remains of converting Canadian dollars, for example, into British pounds. The simplest procedure— converting currencies at market, parity, or other official exchange rates—is poor since exchange rates did not reflect price differences. Figure 6.1 makes the point dramatically so far as consumer prices are concerned. The solution is to calculate socalled ‘purchasing power parities’ by indexing prices in the various countries.7 The prices and weights needed to construct such indices for all components of GNP are not yet at hand.8 Instead, I report the results of two limited PPP conversions. The first uses my consumer price index.9 (See Figure 6.2 and Table 6A.2.) This procedure can be rationalized in two ways. First, deflation by the CPI compares
115
ROBERT C.ALLEN
incomes in terms of wage goods per person. Second, the CPI might be interpreted as a consumer spending deflator. Since consumer spending is the largest component of gross national expenditure, the CPI might be a reasonable proxy for a full-scale GNP deflator. The second PPP conversion relies on backward extrapolation. Following Bairoch (1976), Maddison (1982), and Crafts (1983), I begin with recent measurements of per capita GDP based on PPP calculations—in this case the estimates prepared by the OECD (1987) for 1985 and carried back by them to 1970. Then I extrapolate these 1970 values backwards using the indices of real national income per capita developed by the national statistical agencies and then by Feinstein (1972), Butlin (1962), Urquhart (1986), and Balke and Gordon (1989). Finally, I multiply the extrapolated values of GDP by the ratio of GNP to GDP in each year in order to estimate GNP in 1970 US dollars.10 The strength of this procedure is that the PPP exchange rates are based on the full range of prices. The obvious weakness is that all errors are accumulated over a century of backcasting. The two methods give very similar results for the relative real incomes of Australia, Canada, and the United Kingdom. It is only in the American case that there are discrepancies between the two series: while the trend of American per capita income growth remains unchanged, the backward extrapolation produces a level of real income (in comparison with the other countries) that is about 20 per cent higher than the level implied by conversion using consumer prices. Perhaps the American real GDP series is not as 116
INCOMES IN THE ENGLISH-SPEAKING WORLD
sound as the recent work suggests. Nonetheless, the two methods give similar answers to the main historical questions. Australia In the 1880s, real Australian per capita GNP was much higher than any other country’s—about one-third higher than average income in the UK, over double that in Canada, and two-thirds to twice as high as American real income (depending on the exchange rate used). This is dramatic confirmation of the traditional view of the continent’s prosperity. The depression of the 1890s was exceptionally severe and ended Australia’s pre-eminence. Real per capita GNP fell about one-third from the peak in the late 1880s to the trough in the mid-1890s. Furthermore, Australian income did not rebound to its 1880s level until the eve of the First World War.11 By then, the rest of the English-speaking world had caught up with Australia, and the country was no longer exceptionally rich. Britain British retardation is also apparent in Figure 6.2. Income did grow from 1879 to 1900 and, indeed, average British real income generally exceeded American before the early twentieth century. However, there was stagnation from 1900 to the First World War. This stagnation is particularly marked when GNP is deflated by the CPI.12 American per capita income had reached or surpassed British income by the First World War, depending on the deflator used. United States The United States was a prosperous country at the end of the nineteenth century but was not the world’s richest. As noted, Australian income was much higher in the 1880s as was British income before 1900. American growth accelerated in the 1890s, and Figure 6.2 puts the United States just behind Australia and Britain on the eve of the First World War.13 While the roots of American mid-twentieth century prosperity go back to the 1890s, the high relative income in, say, 1970 was mainly a post-First World War phenomenon. Canada Canada, with an income only half of Britain’s in the 1880s and 1890s, was the poorest country of the four. The mid-1890s, however, marked a decisive and dramatic turning point. By the First World War, Canada had closed much of the gap that had existed with respect to Britain and Australia. The gains, of course, reflected stagnation in those countries as well as the acceleration of growth in Canada. 117
ROBERT C.ALLEN
My results reaffirm the importance of the 1890s and the wheat boom as turning points in Canadian economic history. Their role was regarded as central in older historical accounts but was questioned by Chambers and Gordon (1966) and Bertram (1965). The former developed a general equilibrium model of the Canadian economy which implied that the wheat boom made little contribution to economic growth, while the latter rejected the traditional view that manufacturing growth accelerated at the same time. Recently, Urquhart (1986) and Altman (1987), using the revised Canadian national accounts, have argued that the 1890s marked an important break. Figure 6.2 supports that reassessment by showing that the growth of Canadian per capita income after 1896 was important in closing the gap with other countries. REAL WAGES If the British, American, Canadian, and Australian labour markets were well integrated, then workers would have chosen the best jobs wherever they were, with the result that the standard of living (net of moving costs) would have been equalized across the four countries. To test whether labour markets were integrated, I will compare real wages in Manchester, Chicago, San Francisco, Toronto, Vancouver, and Sydney. This is an imperfect test for two reasons. First, differences in unemployment and in the length of the working year also affected annual labour income. Second, comparisons of material income do not capture the standard of living to the degree that people valued a nice climate, the right to vote, or the ease with which a house could be purchased (Offer 1989:132–4). Despite these caveats, real wages were an important factor in defining the standard of living, so they are worth a close examination before the more elusive factors are integrated into the analysis. I have computed real wages for three categories of workers—bricklayers, labourers, and the average factory worker. Bricklaying was an important skilled trade well represented in all six cities. Moreover, the wages of bricklayers were similar to those of carpenters, plasterers, etc., so the experience of bricklayers was representative of skilled building workers in general. Generally, the labourer’s wage that I have analysed is that of a building labourer, although I have used some quotations for labourers in foundries and machine shops in Chicago and San Francisco.14 The wage of labourers was similar in the building and metal trades. I computed the average wage of manufacturing workers by dividing total wage payments shown in a census of manufactures (or similar source) by the total number of wage earners. The calculation lumps together different industries and makes no distinction between men, women, and children, or the skilled and the
118
INCOMES IN THE ENGLISH-SPEAKING WORLD
unskilled. Any of these factors could influence the comparisons of average manufacturing earnings. The three sets of real wages are graphed in Figures 6.3–6.5 and tabulated in Tables 6A.3–6A.5. The salient features can be summarized by considering the countries in turn. Britain In all cases, British wages were at the bottom. In one sense, this is not unexpected in view of the millions who emigrated from the UK. In another sense, however, the pattern is a surprise since British real wages performed much worse than British GNP. Until 1900, per capita income in the UK was at least as high as in the United States, yet the wages of most American workers exceeded British wages. The divergence is an aspect of the greater inequality of British society. Another remarkable feature of the real wage series are the small gains made by British workers between 1879 and 1913. The net gain of bricklayers was only 19 per cent. Labourers did better with a 35 per cent gain, and the average factory worker realized a 36 per cent advance. All of these gains were made in the 1880s and 1890s when retail prices were falling. In the Edwardian period, real wages declined for all three groups. The magnitude of the gains from 1879 to 1913, and the difference in experience before and after 1900, are consistent with the standard accounts of British labour history (e.g. Bowley 1937). Feinstein (1990) has recently offered a more refined analysis of the wage
119
ROBERT C.ALLEN
120
INCOMES IN THE ENGLISH-SPEAKING WORLD
data in the context of his national income accounts. His figures point to a higher rate of growth in real income per employee (primarily since he makes allowance for the shift from low- to high-wage jobs), but he still finds a deceleration of real wage growth after 1900. The question of why real wages stopped growing after 1900 has received attention from British historians. The main possibilities include a deterioration in the terms of trade (e.g. Lewis 1978), a redistribution of income in favour of capital (e.g. Gourvish 1979), and a decline in the rate of growth of productivity (e.g. Phelps Brown and Browne 1968:157–74). The deterioration of the terms of trade would have worked across all industrial nations and so is hard to reconcile with the rise in American real wages. Feinstein (1990) has produced a comprehensive reconciliation of the real wage history with his national accounts for the period. He shows that labour’s share of the national income did not decline and concludes that the retardation of productivity growth explains the check to real wages. An important lesson of the international comparisons is that distributional considerations played a much more active role than this analysis indicates. Changes in distribution may not have been responsible for changes in real wage growth in Britain, but differences in distribution were critical in explaining why wages in Victorian Britain were less than in the United States. Australia The most impressive feature of the Australian wage series is the high level prevailing in the 1880s. Labourers in Sydney received a wage double that prevailing in Manchester, Chicago, or San Francisco. Bricklayers did well but not as well—their wage was almost twice that of their counterparts in Manchester, but bricklayers in Chicago and San Francisco did even better. The average earnings of manufacturing workers in New South Wales, however, exceeded the level in Chicago, San Francisco, Ontario, British Columbia, and the United Kingdom. If labour and capital markets were perfectly integrated, then real wages and profit rates in Australia would have equalled those in other countries, and the high average income generated by the Australian boom of the 1870s and 1880s would have accrued solely to the owners of land. The lack of labour market integration meant that the benefits of growth were shared among the working class, especially among the least skilled. This period was remembered nostalgically thirty years later. The period from 1880 to 1887 was, perhaps, the brightest in the State [of New South Wales] from the industrial standpoint, as at no period, except in the five golden years, 1853–57, were wages so high, and at no period was the purchasing power of money so great. (John R.Trivett in NSW Year Book 1911:516) 121
ROBERT C.ALLEN
This view is regarded as panglossian by some Australian historians— dubbed the ‘dominion pessimists’ by Offer (1989:132). Fitzgerald (1987:226), for instance, writes cynically of ‘the much heralded “prosperous” period 1870–90’. She concentrates on social indicators of well-being—her book is called Rising Damp—instead of real wage indices. She laments the rise in the relative amount of unskilled work in Sydney without recognizing that labourers made wages twice those in Britain or North America. Lee and Fahey (1986:1) also deny that ‘the working class [received] a substantial share in the fruits of economic growth’ during the 1870s and 1880s. Much of their case consists in showing that seasonality was important and casual labour common. This was undoubtedly true, but there is no reason to believe that the situation was different elsewhere: in Britain and North America, building workers were also hired job by job. Buckley and Wheelwright (1988:142) note the high ‘wages paid to construction workers and builders’ labourers’ but dismiss that fact as unimportant since their wages ‘were high compared with those paid to unskilled workers in manufacturing and other industries’. Building labourers’ wages were certainly exceptional, but, as Figure 6.5 shows, manufacturing wages were also very high by international standards. The benefits of growth spread widely across the working class. In the 1880s Australia was an exceptionally rich country and her workers might have done better, but it is a mistake to jump from that possibility to the conclusion that the Australian worker did not receive any of the gains of the great boom. Australian workers maintained their wages despite the collapse of the 1890s and the slow recovery. The real wage of labourers dropped in 1892 and 1893 and then recovered to the level prevailing in the preceding decade. The wage of bricklayers and the average earnings of factory workers show no abrupt change. The big falls in national income occurred first in urban real estate and finance and then in agriculture, which was devastated by drought. Canada There are no Canadian series of bricklayers’ or labourers’ wages before 1900, so one must rely on average annual earnings in British Columbia and Ontario manufacturing to span the period. In 1880 and 1890, real factory earnings in Canada were lower than in the United States and Australia. In part, this is a reflection of the low Canadian income in this period, but there is more to the explanation since Canadian earnings were higher than British earnings even though Canadians were poorer, on average. High Canadian manufacturing wages may reflect a more favourable industrial mix, but, more likely, the greater equality of Canadian society compared with British. Whatever the explanation, neither the low GNP per capita nor the low average earnings in factories provided an economic incentive for anyone to 122
INCOMES IN THE ENGLISH-SPEAKING WORLD
emigrate to Canada rather than Australia or the United States in the 1880s. While many people left Britain for Canada, more left Canada for the United States, so that net immigration into Canada was negative between 1861 and 1901 (Urquhart and Buckley 1965: series A245, A250). After 1900, the Canadian wage data are broader, and they show Canadian living standards to have risen substantially. Toronto and Vancouver displaced Sydney as the cities in which labourers earned the highest wage—in Vancouver the rate is almost twice the rate in Manchester.15 Bricklayers also earned much more than in Manchester or Sydney but not as much as in San Francisco or Chicago. Real earnings of manufacturing workers increased 56 per cent in British Columbia between 1890 and 1900 and 61 per cent in Ontario between 1880 and 1905. Thereafter, earnings were constant (Ontario) or declined (BC). Nonetheless, they remained well above the 1880s level. Canadian factory earnings exceeded British and Australian but were not as high as American. High earnings outside of agriculture played an important role in the surge of immigration after 1896. As with Australia, there are ‘dominion pessimists’ in Canada who argue that the working class did not share in the expansion. Piva (1979:27–59) has computed real wage indices for many groups of Ontario workers between 1900 and 1921. For the period 1900–13, he found that the standard of living declined for most workers. In contrast, my figures show that the standard of living of Toronto workers changed little. In Vancouver the results were mixed with construction wages rising and factory earnings falling. (The national real wage index of Bertram and Percy (1979:307) rose 11 per cent between 1900 and 1913.) In part Piva’s more pessimistic results for Toronto stem from his use of the Department of Labour consumer price index, which rises more rapidly than mine or Bertram and Percy’s. Whatever the exact movement of real wages after 1900, the important point is that Canada had become a high-wage country by the early twentieth century, even if wages were no longer rising. United States The history of American wages has several distinctive characteristics. First, the rate of growth of real wages was unusually high in the United States. Between 1879 and 1913, the real wage of bricklayers increased 69 per cent in Chicago and 124 per cent in San Francisco, the wage of labourers grew 77 per cent and 42 per cent in the same cities, and the average annual earnings of factory workers grew 53 per cent and 68 per cent, respectively. These measurements are consistent with Rees’ (1961) view that real wages increased in this period rather than Douglas’ (1930) view that they were constant.16 American workers did share in the rise in average income between 1879 and 1913. Over the same period, the real wages of most categories of workers
123
ROBERT C.ALLEN
in Manchester and Sydney grew little. Only in Canada did real wage gains approach American levels. Second, some categories of American workers received the highest real wages in the world. American bricklayers earned the highest rate almost without exception. In 1879, the real earnings of manufacturing employees were less in Chicago and San Francisco than in New South Wales. By 1889, Chicago factory workers had outstripped their Australian counterparts, and by 1899 the same was true of San Francisco workers. Third, one category of American worker received the lowest real wage in the world—labourers. In the 1880s, it was a dead heat for last place between building workers in Chicago, San Francisco, and Manchester. In the 1890s, British labourers, in fact, did better than American. It is only after 1905 that the real wages of American labourers jumped above British levels.17 Phelps Brown and Browne’s conclusion that ‘not until after 1900…did the real wage in the U.S.A. draw ahead of the British’ (1968:163) applies to unskilled workers—not all workers. Fourth, the preceding two characteristics imply that wage inequality was much greater in the United States than in the other three countries. On average, the ratio of a bricklayer’s wage to a labourer’s wage was 2.95 in Chicago and 3.11 in San Francisco. In contrast, the ratio was 1.50 in Manchester, 1.48 in Sydney, 1.87 in Toronto, and 1.68 in Vancouver. Phelps Brown and Browne (1968) remarked on the exceptional wage inequality in the United States. Shergold (1982:214–15) argued that it characterized industrial wages as well as construction wages in Pittsburgh and Birmingham. The evidence of San Francisco shows that the phenomenon was not confined to the northeastern states. Several factors might explain the relatively high wage of skilled labour in the United States. Rapid industrialization may have increased the demand for skills, while the heavy migration of unskilled labour to the United States may have lowered its price. In contrast, the modest rate of industrial expansion in Britain may have moderated the growth in demand for skilled labour, and the high (in comparison with the United States) proportion of skilled labour in the workforce may have forestalled a rise in the skill premium (Harley 1974). Williamson has developed general equilibrium models to assess these possibilities (Williamson 1985; Williamson and Lindert 1980). His results are particularly important since they place Anglo-American wage history into the context of a general analysis of inequality and economic development. Williamson believes that in both countries the ratio of the wage of skilled to unskilled labour (the pay ratio) rose through the phase of early industrialization and declined thereafter. The rise in the pay ratio was due to a high rate of productivity growth in the industries that used skilled labour intensively (Williamson and Lindert 1980:287–8). In Britain, the pay ratio was reduced as skills per worker increased over the nineteenth century 124
INCOMES IN THE ENGLISH-SPEAKING WORLD
(Williamson 1985:158). In both the United States and Britain, the pay ratio traced out a ‘Kuznets curve’ of rising then falling inequality. The difference between the countries is in the timing of the industrial revolution. In Britain the wage premium peaked in the 1860s. By the end of the nineteenth century, the ratio of skilled to unskilled wages had declined. In the United States, however, the wage premium peaked in the early twentieth century. The difference in Anglo-American wage inequality in Figures 6.3 and 6.4 is due to the later industrialization of the United States. Williamson’s analysis elegantly accounts for the difference in British and American wage inequality. It has, however, been challenged. Jackson (1987) and Feinstein (1988) have argued that the rise and fall in the pay ratio in Britain is due to the rise and fall in the incomes of lawyers and doctors. Those incomes, in their turn, are mis-measured, so the British ‘Kuznets curve’ is an illusion. The histories of Australia and Canada provide a different sort of challenge to Williamson’s model. Why were their wage structures like Britain’s? Were Canada and Australia pre-industrial economies that had not yet entered the phase when the premium of skilled labour rose? Were they frontier economies with a peculiarly high demand for unskilled labour? Even if that were true, why did migration between Canada and the United States fail to equalize the wage structure in the two countries? Are supply and demand factors sufficient to explain why the pay ratio was the same in Canada, Australia, and Britain or does a common cultural heritage provide the key? CONCLUSIONS It is impossible to accept the view that labour markets were well integrated internationally in view of the evidence reviewed here. This finding is at variance with the results of recent research like Williamson (1992). As expected, wage rates were lower in Britain (the main sending country) than in Australia, Canada, and the United States (the main receiving countries). More to the point, wages differed widely between the receiving countries. It is hard to imagine that transportation costs could have caused these differences. Further, the differences changed over time—Australia had high wages in the 1880s and low wages in the 1900s—in a way that is not readily explicable in terms of changes in transportation costs. The example of Canada is very much to the point: despite massive migration both ways across the border with the United States, Canada preserved a distinctive wage structure. The international comparisons of real incomes lead then to contradictory lessons for the historian. First, an international perspective changes the interpretation of many national questions: the Dominion pessimists are right that Australian workers did not receive rising real wages between 1870 and 1890, and Canadian workers may have been in the same straits in 1900–13. Nevertheless, international comparisons show that Canadian and Australian real wages were exceptionally high at those times. The 1880s and 1890s 125
ROBERT C.ALLEN
were a period of rising real wages for British workers, but the gains were small compared with advances in other countries and the level ultimately reached was low. The United States was the land of opportunity for some workers but certainly not for all: skilled American workers received the highest incomes in the world, but wage inequality was greater there than anywhere else, and unskilled American workers were at the bottom of the wage heap. Second, an international perspective reaffirms the importance of national factors in explaining the level and growth in income. Since there was not a closely integrated labour market across the English-speaking countries, wages in each country depended on national supply, demand, institutions, conflicts, and customs. This view is implicitly recognized in the historiography of all of the countries. For instance, Feinstein explains the slowdown of real wage growth in Britain after 1900 by a slowdown in the growth of GDP. Likewise, Williamson explains the levels of wage inequality in Britain and the United States by the rates of productivity growth and skill accumulation in the two countries. Neither of these explanations would make sense if labour markets were perfectly integrated, for, if they were, the wage of each grade of labour in each country would depend on the total demand and supply across the integrated economies. The wage in Manchester would have been just as responsive to an increase in labour demand in Sydney as it would have been to the same increase in Manchester. Differences between countries in the growth of GDP, productivity, or skill accumulation would have affected the volume of migration and the composition of the workforce in each country but not the level or rate of change of wages. Therefore, while an international perspective can help frame the right questions, it is not enough to explain the evolution of incomes in Britain, Australia, Canada, and the United States.
126
INCOMES IN THE ENGLISH-SPEAKING WORLD
127
ROBERT C.ALLEN
128
INCOMES IN THE ENGLISH-SPEAKING WORLD
129
ROBERT C.ALLEN
130
INCOMES IN THE ENGLISH-SPEAKING WORLD
131
ROBERT C.ALLEN
Sources for the interurban-intertemporal CPI, 1879–1913 Australia 1879–1913: vinegar: contract price of Tasmanian government purchases delivered in Hobart (Tasmania, Statistics 1879–1913). beer: colonial ale, Hobart market price (Tasmania, Statistics 1879–1913). coal: Brisbane retail price (1905 interpolated) (Queensland, Statistics 1879– 1913). lard, beans: These prices were estimated by multiplying the prices of butter and potatoes, respectively, by the ratios of the average (over 1879–1913) British prices of lard relative to butter and beans relative to potatoes. tobacco, imported: (NSW Year Book 1905–6:467; 1913:995). gas: 1893–1913 Sydney price (NSW Year Book, various years). Gaps are filled by interpolation; 1879–92 price is extrapolation based on the London series. rent (average rent per room of a five-room house): (NSW Statistical Register 1919–20:411). clothing: 1.2 times the price index of British clothing (Bowley 1937:121). For the 20 per cent mark-up, see NSW Legislative Assembly (1891–2:1081). 1901– 13: eggs, milk, cheese, butter, potatoes, bread, flour, tea, coffee, sugar, oatmeal, rice, gravy beef, sirloin beef, corned beef, bacon, mutton leg, pork chop, raisins, currants, Australian jam, starch, soap, candles, kerosene: (NSW Statistical Register 1919–20:413). 1879–1900: eggs, cheese, butter, potatoes, bread, tea, coffee, sugar, oatmeal, rice, starch, soap, candles: (NSW Year Book 1905–6:466–7). There are some minor discrepancies between series at the overlap. gravy beef, sirloin beef, bacon: prices extrapolated back to 1879 using the price series for beef and bacon (NSW Year Book 1905–6:466–7). currants, raisins: Tasmanian government contract price (Tasmania, Statistics 1879–1900). milk: Brisbane market price (Queensland, Statistics 1879–1900). jam: 1879–81 prices are extrapolations of succeeding data; 1882–92 Hobart price extrapolated back to 1882 based on the value per pound of jam produced in Tasmanian jam factories; 1893–1900 Hobart market price (Tasmania, Statistics 1893–1900). kerosene: New South Wales price for 1901–13 extrapolated backwards using Brisbane price (Queensland, Statistics 1879–1900). candles (tallow): 1879–80 from Brisbane, 1881–1900 from Hobart (current price, not contract price). (Queensland and Tasmania, Statistics, 1879–1900). flour, pork, mutton: Hobart market price. For mutton, the low value in the
132
INCOMES IN THE ENGLISH-SPEAKING WORLD
range was used as it links with the New South Wales price in 1901. For flour and pork, the average values were used (Tasmania, Statistics, 1879–1900). Great Britain 1879–1913: bread: Ministry of Labour (1926:143). rent (average rent per room of a five-room house in Manchester), 1905: Board of Trade (1913:40). Extended to 1879 and 1913 by the rent index in Bowley (1937:121). clothing: price index of British clothing from Bowley (1937:121); 1879 assumed to be 108, the same value as 1880. 1903–14: beef (imported), mutton (imported), bacon, flour, tea, sugar, milk, potatoes, butter, cheese, eggs, pork, rice, oatmeal, coffee, jam, currants, raisins: prices extrapolated from 1900 using the index numbers of retail prices in London given in Board of Trade (1925:21). lard, haricot beans (import price used), raisins, currants, candles, kerosene, beer, tobacco, coal, gas: prices extrapolated to 1913 on the basis of the prices or price indices developed by Prest (1954:48, 57, 64, 102, 116, 76, 60, 107, 115). 1879–1902 (except as noted): eggs (1886–1902), milk, cheese (1886–1902), butter (1886–1902), potatoes (1894–1902), flour, tea, coffee, sugar, lard, oatmeal (1886–1902), rice, haricot beans, American gravy beef (1889–1902), American sirloin, wing end (1889– 1902), corned beef, bacon (1886–1902), New Zealand mutton leg (1889– 1902), pork chop (1886–1902), raisins, currants, raspberry jam, starch, soap, vinegar, tea (China, Kaisow Congou), coffee (mocha, roasted, not ground), bitter ale (East India), tobacco (English cut Cavendish): Board of Trade (1903). gravy beef, sirloin (wing end), mutton (1886–8): prices of British beef used, Board of Trade (1903). gravy beef, sirloin (wing end), mutton, bacon, butter, oatmeal: 1879–85 prices extrapolated back using indices in Ministry of Labour (1926:136); flour index used for oatmeal. potatoes: 1879–1893 price extrapolated back using index in Ministry of Labour (1926:136). pork, cheese: 1879–86 prices extrapolated back using average value per cwt of imported pork and average price paid for cheese by Bethlem Royal Hospital. The latter is an open market, not a contract, price and is slightly less than the price of American and Canadian cheese (Board of Trade 1903:131, 150). candles, kerosene, coal, gas: 1879–1903 Board of Trade (1905b). Rough extrapolations and interpolations:
133
ROBERT C.ALLEN
eggs: 1879–85 set at 14d.; starch: 1903–14 set at 6d.; vinegar: 1903–14 set at 19d.; soap: 1903–14 was interpolated incorporating the 1913 value of 3.5d. given in the Report on the Soap Industry (1921:19). Canada 1900, 1905, 1909–13: new laid eggs, fresh milk, new Canadian cheese, creamery butter, bread, flour, black tea, mocha coffee, granulated sugar, lard, rolled oats, rice, handpicked beans, chuck beef, sirloin steak, pork, best smoked bacon, mutton hind quarters, coal oil, vinegar, bituminous coal, prunes (used for raisins and currants): Coats (1915: vol.II, 18–19, 128–9). rent per room: Coats (1915: vol.II, 405). Missing values interpolated, gas (illuminating): Coats (1915: vol.II, 331–2). beer and tobacco: the San Francisco prices were used for Vancouver, and the Chicago prices for Toronto. jam: estimated as the price of sugar multiplied by the ratio of the average (over 1879–1913) British price of jam relative to the average British price of sugar. soap and candles: Toronto: estimated by multiplying the Toronto price of starch by the ratios of the prices of soap and candles to starch in Chicago. Vancouver: price of soap estimated by multiplying the Vancouver price of starch by the ratio of the price of soap to starch in San Francisco. The Toronto price of candles was applied to Vancouver. Chicago and San Francisco candle prices were the same. clothing: the prices for Chicago and San Francisco were used respectively for Toronto and Vancouver. The similarity between Chicago and Toronto prices is born out by comparisons of Toronto and Pittsburgh prices (Coats 1915:130; Shergold 1982:169, 172; US Senate 1911:1483–9). Vancouver prices exceeded Toronto prices for the same reasons that San Francisco prices exceeded Chicago prices. 1901–4, 1906–8: indices were interpolated. 1879–1900: indices were extrapolated using the Barnett retail price index for Kingston, Ontario (Urquhart 1986:88). United States 1890–1913: jam: estimated as the price of sugar multiplied by the ratio of the average (over 1879–1913) British price of jam relative to the average British price of sugar. starch, soap, candles, kerosene, beer, tobacco: 1890 Aldrich (1892). No kerosene price was reported for San Francisco, so the Los Angeles price was used instead. Starch, soap, candles, and tobacco: for 1891–1913 the 1890 134
INCOMES IN THE ENGLISH-SPEAKING WORLD
retail prices were extrapolated forward using the corresponding wholesale prices in Bureau of Labor Statistics (hereafter BLS) Bulletin, no. 181. For kerosene, Rees’ (1961:110) index of the retail price of kerosene was used. For beer, the Canadian wholesale price given in Coats (1915:68–9) was used. The American wholesale price of draft beer was first collected in 1913. In that year, the Canadian price was $6.25 per barrel while the American price was $5.5025. gas: 1913, BLS Bulletin, no. 138; 1912, BLS Bulletin, no. 132. 1907–11, BLS Bulletin, no. 105; 1889–1906 extrapolated backwards from 1909 using the values per ton of manufactured gas in California and Illinois reported in the US Census of Manufacturing (Census Office 1895:715; Census Office 1902b:712; Bureau of the Census 1913b: 649). Intervening years interpolated. coal: 1912–13 prices for October from BLS Bulletin, no. 138; 1907–11 annual averages from BLS Bulletin, no. 105 (firm A used for Chicago); 1890–1906 extrapolated backwards from 1907 using the wholesale price of bituminous coal at Cincinnati from BLS Bulletin, no. 181. We can compare the extrapolated values for 1890 with the values reported in Aldrich (1892). For Chicago the former was $3.12 while the latter was $3.00. For San Francisco the former was $12.09 while the latter was $11.00. rent: Chicago: The rent in 1909 was taken to be the average rent (as shown by Rees (1961:102)) recorded by the (UK) Board of Trade (1911). Rees’ (1961:97) index of Chicago rents derived from newspapers was used to extrapolate the 1909 value to the other years. San Francisco: following Rees (1961:96–105), I computed the average rent per room in houses, flats, and apartments rented by people making less than $1500 per year in San Francisco and Oakland from BLS Bulletin, no. 357. That rent was $46.06 per year. Rees’ national rent average shows no change between 1913 and 1914 (Rees 1961:74), so I set the San Francisco 1913 rent at $46.06 per room per year, and extrapolated back to 1890 using the national rent average. The San Francisco rent series is the least reliable. Nonetheless, the resulting index agrees with other, scattered information about Illinois and California rents. According to the US Commissioner of Labor (1905:370), the average rent per room in California was 41 per cent higher than the average in Illinois while the differential was only 23 per cent in 1914. This greater differential suggests that rents in Chicago were rising faster than rents in San Francisco, and, indeed, Rees (1961:97) found that Chicago rents rose the fastest of any city he studied. Moreover, the difference in the growth rates between the national (taken for San Francisco) and Chicago series is exactly consistent with the measured differences in rents in the two cities in 1903 and 1914. clothing: roughly following Shergold (1982:172) I set the index of Chicago clothing, footware, and furnishings at twice the British level in 1905. Then I extended it between 1890 and 1913 using the NBER retail, all items, clothing index in Rees (1961:91). I set the San Francisco index 10 per cent higher 135
ROBERT C.ALLEN
than the Chicago index. This is supported by price comparisons in Aldrich (1892). 1908–13: chuck beef, sirloin steak, creamery butter, corn meal (instead of oatmeal), fresh eggs, wheat flour, leaf lard, fresh milk, mutton leg (sometimes yearling lamb), pork chops, bacon, potatoes, granulated sugar: BLS Bulletin, no. 138. The 1912 and 1913 values are prices for September. beans, bread, cheese, coffee, prunes, rice, tea, vinegar: extrapolated using wholesale prices from BLS Bulletin, no. 181. 1890–1907: beans, chuck beef, sirloin steak, bread, creamery butter, cheese, coffee, corn meal (instead of oatmeal), fresh eggs, wheat flour, leaf lard, fresh milk, mutton leg (sometimes yearling lamb), pork chops, bacon, potatoes, prunes (instead of raisins and currants), rice, granulated sugar, tea, vinegar: US Commissioner of Labor (1904) and BLS Bulletin, no. 77. 1879–1889: Both the Chicago and San Francisco indices were extrapolated using the Long and Hoover indices for the United States given in Long (1960:156–7). Incomes and wages Nominal gross national product per capita, 1879–1913 All series are GNP at market prices. Australia: Butlin (1962:6–7). Canada: GNP from Urquhart (1986:11–14). Population from Urquhart and Buckley (1965: series Al). UK: Feinstein (1972: T10, T120, T121). USA: GNP from Balke and Gordon (1989:84–5). Population from US Bureau of the Census (1975: series A7). Wages of bricklayers and labourers Australia: Wealth and Progress of New South Wales, various years. Canada: Coats (1915, vol.11:484–5, 488–9). Some Vancouver numbers were interpolated. UK: 1879–1900: Bowley (1900b:311) with some interpolation. Extended to 1913 on the basis of the constant index shown in Bowley (1937:8, 10). USA: 1879–89: daily wages for bricklayers and ‘labourers, other’ in BLS Bulletin, no. 18(672, 678), divided by the number of hours worked per day. For bricklayers in San Francisco and for labourers in Chicago and San Francisco, number of hours taken from Aldrich (1893, pt 1:178) for the building trades. This gave close agreement with hourly wage rates used for 1890–1903 in the overlap. According to the US Commissioner of Labor (1905:447), bricklayers in Chicago, however, were working an eight-hour 136
INCOMES IN THE ENGLISH-SPEAKING WORLD
day in 1890 (instead of the 9.6 hour day shown in Aldrich (1893)). I assumed they worked an eight-hour day throughout the whole period 1879–89. This may overstate their hourly rate at the beginning of the period. 1890–1902: US Commissioner of Labor (1905). The wages are for bricklayers in Chicago and San Francisco, building labourers in San Francisco and Chicago. 1903–13: again bricklayers and building labourers, except in San Francisco for 1904 no return was made for building labourers, so labourers in foundries and machine shops were used instead, and in Chicago labourers in foundry and machine shops (1904–6) and blacksmiths’ helpers in railroad shops (1907–13). In San Francisco, the wages of building labourers and labourers in foundries and machine shops were similar. There is a Chicago series for building labourers, but it rises abruptly in 1907 with the change to reporting union wage scales (BLS Bulletin, nos 65, 71, 131, 143). Earnings of manufacturing workers Australia: Butlin (1962:161). Canada: Dominion Bureau of Statistics (1926:378). UK: Deane and Cole (1962:143, 152). USA: 1879 and 1889 from Census Office (1902a: 994, 1002); 1899, 1904, and 1909 from Bureau of the Census (1913a: 93, 95). NOTES 1 2
3 4 5 6
Mitchell and Deane (1962:50). These are net figures, i.e. the difference between inward and outward movements. Feinstein (1991) has argued that more weight should be placed on tobacco, alcohol, and services than Bowley allowed and less weight should be placed on food and rent. Whatever the virtues of this approach, it introduces practical difficulties since it puts more weight on services for which prices are hard to find. Even Feinstein was forced to assume a constant price for over half of his service expenditure. International comparisons accentuate this problem, so Feinstein’s lead has not been followed. Little hinges on this decision: Feinstein’s index differs from other British indices by at most four percentage points in any year and that is a small discrepancy compared with many of the real wage gaps I measure. Consequently, following Feinstein’s approach at the international level would cause only minor variations in the results. Gazeley (1989:211) concludes ‘that, except for the cost of accommodation, regional variations in prices [across Britain] were modest’. This index is criticized by Gourvish (1979:16) as ‘grossly inadequate’ (with a second from Gazeley (1989:213)), but no one has produced a replacement. Haines (1989:101) computes a 6 per cent premium. Coelho and Shepherd (1979:73) compute a 9 per cent premium between the Pacific and East North Central regions. Maddison (1982:8) offers some comparisons of real GDP per capita for 1870. His figures are derived by backward extrapolation of 1970 income levels with growth rates taken from national studies. While he had to rely on less reliable sources, his estimated income levels are not far different from those presented here.
137
ROBERT C.ALLEN 7 O’Brien and Keyder (1978) are one of the few to have attempted the exercise with nineteenth century data, and the sweep of their conclusions shows the importance of the method. 8 Haig (1989) has computed PPP exchange rates to compare per capita GDP in Britain and New South Wales. He found that New South Wales GDP per capita was 36 per cent higher than British in 1891 and 3 per cent higher in 1900. These results agree with my estimates. 9 I use the Manchester price index to deflate British GDP, the Chicago series for the United States, the Toronto series for Canada, and the Sydney series for Australia. 10 I did not make this adjustment for the United States since the difference between GNP and GDP has always been minute. The same has not been true of the other countries where GNP and GDP have sometimes diverged by as much as 6 per cent. 11 McLean and Pincus (1983) have posed the question ‘Did Australian living standards stagnate between 1890 and 1940?’ They answer ‘no’ in part because per capita GDP deflated by the CPI grew faster than Butlin’s real per capita GDP figures indicate. Figure 6.2 shows that any growth in per capita GDP deflated by the CPI occurred after 1913. The period from 1890 to the First World War was one of stagnation. 12 Matthews et al. (1982:606–7) also emphasize that 1900 marked the slowdown in the growth of output. 13 The backward extrapolation method puts the United States slightly ahead. 14 In 1907 the basis of wage reporting by the US Bureau of Labor Statistics changed. Beginning in that year, they report union wage scales for building workers, whereas earlier reports included non-union workers (Rees 1961:6–9). The reported wage of building labourers rose markedly in Chicago after this change. The wage of unionized building labourers in Chicago was much higher than that in many other cities and much higher relative to the skilled building wages. 15 However, MacKinnon (1993a) has argued that these figures overstate the earnings of unskilled Canadians. 16 Douglas and Rees came to different conclusions mainly because of differences in their consumer prices indices. My index is patterned after Rees (although weighted differently) and incorporates some of his components. 17 Shergold (1982:218) notes that if Douglas’ cost of living index is used instead of Rees’, then unskilled workers in Pittsburgh maintained a consistent 15 per cent real wage advantage over Birmingham workers in the period 1890–1913.
138
7 DID LABOUR FLOW UPHILL? International migration and wage rates in twentieth century Ireland Kevin O’Rourke
Two issues dominate the discussion of Irish labour markets. First, the Irish labour market is clearly exceptionally well integrated into world labour markets, with high levels of emigration being a fact of Irish economic life. This chapter examines migration from Ireland to Britain in the twentieth century. Its paradoxical finding is that for much of the century Irish wages in at least some important industries have been as high as, or higher than, British wage levels. The second major issue facing Irish labour economists is the Irish unemployment level, which has been chronically high throughout the century. This chapter considers whether the high rate of unemployment can explain why the Irish labour market was simultaneously characterized by high wages and high emigration. Table 7.1 gives estimated emigration figures for the period 1911–71.1 How can this be related to Ireland’s economic development in this century? The economic history of the period 1926–83 can be broken down into several distinct phases. Until 1932, the Irish Free State pursued a free trade policy based on agricultural exports; the success of this policy is disputed, with
139
KEVIN O’ROURKE
national income figures being of insufficient quality to resolve the debate. In 1932 Eamonn de Valera came to power and introduced a policy of industrial protection; industrial growth was rapid during the decade, while agriculture suffered as a result of British protection (O’Rourke 1991). Trade was further restricted during the war, which led to fuel and many raw materials being in short supply. Industrial output fell by 27 per cent between 1938 and 1943, and industrial employment also fell. A post-war boom followed but was succeeded by severe depression throughout the 1950s, the worst decade for emigration since the 1880s. The autarchic experiment was brought to an end at the end of the decade, with protection gradually being reduced (a process culminating in Ireland’s entry to the EEC in 1973), and multinationals being actively encouraged to invest in Ireland. The 1960s was a period of rapid growth for the economy; EC membership (and irresponsible fiscal policies towards the end of the decade) meant that the 1970s were also good, and the economy actually experienced net immigration. However, the 1980s were stagnant economically, and emigration resumed.2 It is possible to view international labour market integration in two ways: it can lead to wages in different countries co-varying over time, and it can lead to wage levels in different countries gradually converging over a number of years. Several recent studies have convincingly shown that Irish emigration is positively linked to British wages and the Irish unemployment rate, and negatively linked to Irish wages and the British unemployment rate; such studies are consistent with the hypothesis that the Irish and British labour markets are well integrated in the first sense (Walsh 1974; Geary and McCarthy 1976; Keenan 1978; Geary and Ó Gráda 1989). However, formal cointegration tests of the sort found in Chapter 5 have not been carried out here. The purpose of this chapter is rather to focus on the second aspect of integration, namely, to what extent Irish wage levels have converged to British wage levels over time. By the twentieth century, Irish wages had not only converged to British levels in several industries, but had actually overtaken them. I try to reconcile this finding with the fact that Irish labour continued to migrate to Britain throughout the period under investigation; hence the title of this chapter. High Irish unemployment can explain why the Irish emigrated despite high wages at home; in the language of the Harris-Todaro model, Irish expected wages may not have been higher than British expected wages. An important question remains, however: given an exogenous expected wage in Britain, what determines whether the Irish equilibrium involves high wages and high unemployment, or low wages and low unemployment? Are high Irish wages the exogenous factor, or high unemployment, or are both simultaneously determined by other factors? What are the policy implications of all this? This chapter speculates about these issues but does not attempt to resolve them. Its contribution lies principally in presenting previously unused wage data, and documenting that Irish wages were indeed higher than British 140
DID LABOUR FLOW UPHILL?
wages in several industries at various stages of this century. The first section provides a thumbnail sketch of what is known about the wage convergence that took place between the Irish and British labour markets in the years between the Great Famine of 1845–9 and the twentieth century.3 The second section presents skilled and unskilled wage data for Dublin and London for three industries: building and construction, engineering, and printing and publishing. The third section adjusts the data for cost of living differences between the two countries. The fourth section discusses unemployment in Ireland and Britain in the context of the Harris-Todaro model. Finally, the fifth section discusses some alternatives to the Harris-Todaro hypothesis that could also explain why wage levels in the labour-exporting country (Ireland) exceeded wage levels in the labour-importing country (Britain) so often during this century. WAGE CONVERGENCE AFTER THE FAMINE It is clear that Irish wages were much lower than British wages on the eve of the Famine; according to Ó Gráda Irish living standards were only one-third of British living standards at that time.4 It is equally clear that Irish wages increased rapidly in the decades after the Famine; for example, real Irish agricultural wages rose more rapidly than American real wages during the period 1856–76 (O’Rourke 1989: ch. 5).5 One would expect the Famine to have triggered a process of wage convergence for several reasons. First, of course, the Famine directly reduced the population of the country through deaths and forced migration. Second, the Famine greatly increased the integration of the Irish and world labour markets. Once large Irish population centres had been established abroad through the forced migration of the Famine years, networks of informational and financial flows made it much easier for the Irish to emigrate (Geary 1935–6). In the context of a long-run story of wage convergence triggered by the Famine, the question is to what extent convergence had been achieved by the twentieth century. Irish real wages have never caught up with American real wages, and, since 1930, when American immigration restrictions effectively stopped Irish emigration to the United States, there has been no direct labour market pressure to bring that about. What about the relationship between wages in Ireland and Britain? In 1905, the Board of Trade surveyed wages in various UK cities.6 The results can be compared with those of its 1893–4 General Report on the Wages of the Manual Labour Classes in the United Kingdom which gave wages in Ireland and Britain for 1886 and 1891. As Table 7.2 shows, by 1905 integration seems to have greatly narrowed the gap between Dublin and London skilled wages, although a substantial gap remained for unskilled wages. The remainder of this chapter asks whether wage convergence continued
141
KEVIN O’ROURKE
during the twentieth century. Let us begin by considering part (E) of Table 7.2, which gives nominal wage ratios for 1844 and 1944. Three features should be noted. While the unskilled wage ratio in 1844 was not unlike that in 1886, the skilled wage in 1844 was actually higher in Dublin than in London!7 Consistent with the evidence presented in the next section, Dublin unskilled building wages caught up completely with their London counterparts over this ninety-year period, and in the 1930s both skilled and unskilled building workers were better paid in Dublin than in London. NOMINAL WAGES IN THREE INDUSTRIES, 1926–79 Most studies of labour market integration have worked with average hourly earnings data, which are easily available. These data show lower Irish hourly 142
DID LABOUR FLOW UPHILL?
earnings in manufacturing and agriculture than in Britain, with the gap narrowing in the 1960s and 1970s.8 One problem with the earnings data is that even at their most disaggregated, they are still only available at the industry level, and thus ignore possible differences in the skill composition of the labour forces within an industry in the two countries. For example, bricklayers and construction labourers could be paid the same in Ireland as in Britain; yet if there were a higher ratio of bricklayers to construction labourers in Britain than in Ireland (and this was quite possibly so), average hourly earnings would appear higher in the British construction industry than in the Irish construction industry.9 Fortunately, the International Labour Organization began to collect data on wage rates, by occupation, in the 1920s.10 These data have formed the basis for the average industrial wage rate indices that have been calculated for Britain and Ireland; otherwise, the data have been surprisingly neglected. Irish and British wage rates are available over the period 1926–83, for both skilled and unskilled occupations in three private sector industries: building and construction, engineering, and printing and publishing. The data are reproduced in Appendix B. Figures 7.1–7.3 give the ratio of Irish to British nominal wages, for both skilled and unskilled labour for 1926–79. For construction, bricklayers are the skilled workers; for engineering, fitters; and for printing and publishing, hand compositors. In all three industries, wages specified for labourers or general labourers are taken to represent unskilled wages.11 Figures 7.1–7.3 show movements in relative wages which are intuitive given the thumbnail sketch of post-independence Irish economic history provided in the introduction. Irish relative wages rose until the mid-1930s (the exception is engineering wages) and then declined as the war approached; they rose during the post-war boom but fell during the early 1950s. From the late 1950s on, they began a steep rise as the Irish economy began to develop rapidly. The picture in the 1970s was more mixed. The principal result to emerge from Figures 7.1–7.3 does not, however, have to do with trends but with levels: surprisingly, Irish wage rates often exceeded British wage rates, when the reverse might have been expected. An apparent paradox thus emerges. With the exception of the period 1954– 63, wage rates in Ireland (the labour-exporting country) were generally greater than in Britain (the labour-importing country).12 If inferences about the Irish economy as a whole can be drawn from the experiences of these industries, then by the mid-twentieth century, international labour market integration had indeed exerted a significant impact on the Irish economy. Irish wages had done more than catch up with British wages, with obvious implications for the Irish economy. The remainder of the chapter will attempt to explain the co-existence of emigration from Ireland and high Irish wages. Note, however, that any satisfactory explanation (one which leaves us satisfied as to the rationality of 143
KEVIN O’ROURKE
the international labour market) will not alter the basic fact that Irish wages were not in general lower than in Britain. To the entrepreneur deciding where to invest, it is this basic fact which matters, rather than the explanation for the fact. ADJUSTING FOR DIFFERENCES IN THE COST OF LIVING The most obvious question which arises is whether the gap between Irish and British wages documented above is due to measurement error, and in particular whether it is legitimate to compare Irish and British nominal wages. The Irish and British currencies were linked one for one throughout this period, until Ireland joined the European Monetary System in 1979. Costs of living may, however, have differed in the two countries; if they were higher in Ireland, then real Irish wages may have been the same as or lower than real British wages throughout the period. The necessary cost of living adjustment factor has been calculated for 1975 by Kravis et al. (1982)(KHS). They find that the Irish pound’s purchasing power was in fact greater than that of the British pound in 1975; in real terms Irish wages were therefore higher relative to British wages (in 1975) than is indicated by Figures 7.1–7.3. Using consumer price indices it is possible to
144
DID LABOUR FLOW UPHILL?
calculate relative real wages in the two countries for the post-war period; results for selected years are given in Table 7.3.13 Since the Irish cost of living was less than the British cost of living in 1975, it is clear that adjusting for cost of living differences in the post-war era is only going to increase Irish wages further relative to British wages. Indeed, once the cost of living adjustment has been made, real wages in Ireland exceed British real wages throughout the post-war period for all categories of labour, even during the 1950s (the sole exception being unskilled typographical labour). This adjustment has clearly made our puzzle more puzzling. For the pre-war period, I make use of purchasing power parity calculations based on ILO data for 1938. These ILO data only contain food prices, and so the PPPs calculated for different countries only take account of differences in food price levels. The ILO data suggest that in 1938 food costs in Ireland were 9.3 per cent higher than in Britain.14 Adjusting relative wages to take account of differences in food prices therefore involves adjusting Irish relative wages downwards. On the other hand, rents, an important item in the household’s consumption basket, were lower in Ireland. It is possible to adjust for rent differences in the two countries, in an ad hoc but reasonable manner (details are given in Appendix C). Adjusting for rent differences obviously boosts relative real Irish wages. Table 7.4 shows that when pre-war wages are adjusted for 145
KEVIN O’ROURKE
differences in food costs only, Irish wages are no longer systematically higher than British wages; but that rent differences counterbalance this. The net effect of making both adjustments is minimal, as a comparison of Figures 7.1–7.3 and Table 7.4 makes clear. The conclusion to be drawn from Tables 7.3 and 7.4 is that the surprisingly high Irish nominal wages experienced during this century cannot be 146
DID LABOUR FLOW UPHILL?
explained by relatively higher Irish living costs. If anything, adjusting for differences in the cost of living makes Irish workers seem relatively better off, rather than relatively worse off (this is especially true for the post-war period). UNEMPLOYMENT IN THE THREE INDUSTRIES If unemployment is a feature of labour markets which are linked by migration flows, it may be worth while to focus on expected rather than current wage gaps. This is the simple insight of the Todaro model, which has been extensively used to examine wage gaps in the Third World, as well as in developed countries (Todaro 1969; Hatton and Williamson 1992). This model argues that if unemployment in a labour market increases, then the expected wage in that market is correspondingly reduced; actual wages will have to rise if the expected wage in the market is to remain equal to expected wages elsewhere. Given the strong intellectual and empirical tradition associated with the Todaro model, it seems worth while to compare unemployment in these three industries in Ireland and Britain.15 Table 7.5 gives unemployment rates in the three industries in both countries. As can be seen, Irish unemployment rates are at all times, and in all industries, higher than in Britain. Especially striking is the continually high level of unemployment in the Irish building industry. We can use these unemployment rates to calculate simple Todaro-adjusted wage ratios in the three industries, by multiplying the ratio of Irish to British wages in industry i, in year t, by a factor of (1-UiIt)/(1-UiBt) where Uijt is the unemployment rate in industry i in country j in year t. Adjusting real wage ratios in this way we get Table 7.6, which shows results for selected years. Comparing Table 7.6 with Table 7.3, we see that expected wages are much closer in the two countries than are actual wages, 147
KEVIN O’ROURKE
real expected Irish wages clearly and consistently higher than expected real British wages. Without engaging in any formal testing, it seems reasonable to conclude that the Todaro model may be relevant in analysing flows of labour between Ireland and Britain during this period. While no one has to my knowledge used the term ‘Todaro model’ in analysing urban-urban labour flows between Ireland and Britain, there are hints of such an analysis in work by Roy Geary and J.G.Hughes (1970) on 148
DID LABOUR FLOW UPHILL?
Irish post-war unemployment. In a section on ‘the labour surplus’, they state that over the years 1961–1966, to repeat, a period of unprecedented economic expansion, the average rate of non-agricultural unemployment remained steady (at a high rate in the absolute, if comparatively low for Ireland), showing that the associated labour pool increased pro rata with employment…. In our view what has happened during the last ten years or so is that, for the first time in living memory, there was a great increase, solidly based on the economic advance, in the numbers of young people who decided to ‘give Ireland a trial’ and who, in less optimistic times, would have emigrated. By far the most heartening feature in recent Irish demography is the sensational decline of net emigration from an annual average of 40,000 in 1956–1961 to 16,000 in 1961–1966. It resulted, however, in the formation of a rather too large labour pool. (Geary and Hughes 1970:14) The Harris—Todaro hypothesis was explicitly stated in the context of Irish rural-urban migration as long ago as 1926, by a Danish observer, Jorgen Pedersen. He wrote that wages in the towns are still kept high both in comparison to the wage level before the War and in particular in comparison with the agricultural wage. Despite the unemployment it is therefore tempting to move to the towns in the hope that one will be among the lucky ones who get jobs at the comparatively high wage that is paid. (Pedersen 1926:429, my translation) This opinion is not incompatible with the hypothesis outlined above, namely that some workers also preferred to remain unemployed in Dublin, rather than move to Britain where the probability of finding work was higher, but wages were lower. The many econometric works which relate emigration not only to real wage gaps but to differences in unemployment rates are clearly compatible with a Todaro view of migration. In addition to the general level of unemployment in these industries, it may also be the case that seasonal fluctuations in employment were worse in Ireland than in Britain. Geary and Hughes single out the seasonality of unemployment in the construction industry for special consideration, and complain that ‘in these days when technology is so advanced one would think that means might be found to enable the construction worker in Ireland to work the whole year round as in some other countries’ (1970:5). What would be interesting about a Todaro story in this context is that instead of a low but certain wage in the labour-exporting sector interacting with a higher but uncertain wage in the labour-importing sector, we 149
KEVIN O’ROURKE
have a combination of high wages and high unemployment in the labourexporting sector. As mentioned in the introduction, this raises an interesting set of issues: in particular, what determines whether Ireland will experience high wages and high unemployment (as is the case), or low wages and low unemployment (given an exogenous expected wage in Britain)? This is not an issue on which the Harris-Todaro model can itself shed light; for example, the original model focusing on rural-urban migration has to assume fixed nominal wages in the urban sector (where the unemployment occurs). One possible explanation for high Irish wages and unemployment hinges on the different sizes of the Irish and British labour markets. In thicker labour markets one might well expect to see less variability of employment, and hence to see lower wages and lower unemployment rates. In thin labour markets, by contrast, it will take a longer time on average to find another job once one has been laid off. Thus one will observe higher wages and higher unemployment rates in thinner labour markets. This sort of explanation would rely on the Irish and British labour markets being sufficiently well integrated that a Todaro relationship between expected real wages would hold, but not so well integrated that the labour markets were simply one big market. If there is some small fixed cost of moving across the Irish Sea (in either direction) then it will presumably not be worthwhile to shuttle between Dublin and London every few months in response to changing job opportunities. It might, however, be worthwhile to move to Britain (or move back to Ireland) if expected real wages were persistently higher in one country than another. One prediction of this analysis would then be that unemployment rates should be higher among older or married construction workers in Dublin, for whom the fixed cost of moving will be greater than for young unattached males, and who will consequently be more likely to stay in Dublin and wait for the next job opening that comes along. This prediction of the model is at least in theory testable. ALTERNATIVE HYPOTHESES There are, of course, other explanations of the co-existence of relatively high Irish wages and Irish emigration. This section speculates about several of them. Trade union behaviour It is possible that powerful Irish unions kept wages high for those fortunate enough to have work; those without work were forced to go to Britain. Unions were important in both countries: for example, in 1951, 47.9 per cent of the British non-agricultural workforce was unionized. The comparable figure for Ireland in 1952–3 was 53 per cent (O’Mahony 1964:9). The 150
DID LABOUR FLOW UPHILL?
overall unionization rate in Irish construction was quite low, which was ‘attributable in the main to the high proportion of unskilled casual workers employed. Skilled workers, especially in the urban areas, are highly organised’ (O’Mahony 1964:13). In the Irish engineering industry, there were many unions, mostly craft unions. These sometimes placed restrictions on the numbers of apprentices allowed to enter certain engineering trades, leading at times to ‘a shortage of some categories of skilled men’ (O’Mahony 1964:47). The Irish printing industry was highly organized, with relations between labour and management being characterized by O’Mahony (1964:49) as generally good. It is difficult to think of a theoretically sound metric of union strength: for example, do strikes imply strength or weakness? It is nevertheless of interest to examine the number of man-days lost, per person employed, in the two industries (building and engineering) for which data are readily available for both countries. These data are displayed in Table 7.7. The table gives some support to the view that Irish trade unions in these industries were more strike-prone than their British counterparts, certainly for the period up to the mid-1950s. Migration patterns What if there were no migration from Ireland to Britain of workers in the three industries studied in this chapter? Then relatively high Irish wages in 151
KEVIN O’ROURKE
these industries would not seem so puzzling. During the Second World War travel and work permits were required for Irish citizens emigrating to Britain. For the rest of our period, however, travel was unrestricted in both directions and no work permits were required. For this reason, information regarding the composition of the emigrant stream leaving Ireland and going to Britain during the twentieth century is depressingly scarce. The hypothesis that workers in the building, engineering, and typographical industries did not emigrate during this period does not seem plausible. Anecdotal evidence suggests that labour employed in these sectors did indeed flow across the Irish Sea. For example, in 1964 it was stated that management also knows, particularly in those industries such as building and the metal and engineering trades where there is a good deal of coming and going between Ireland and Britain, that it simply has to pay something very close to the wage demanded if it is to retain its workers with the result that the whole bargaining process is very largely a matter of shadow boxing. O’Mahony (1964:36) Furthermore, there is some information in the British Census on the occupations of foreign-born workers in Britain. Before the Second World War, information is not given for Irish natives (an unfortunate statistical consequence of Ireland’s insufficiently foreign status of the time); the 1951 Census gives information on the occupational group of Irish workers resident in England and Wales (Table 7.8). For the sake of comparison, Table 7.8 also gives the occupational breakdown of the labour force in Ireland in the Census years 1946 and 1956. It appears that the engineering and building industries were well represented among emigrants, while relatively few Irish workers, both at home and in Britain, were in the typographical industry. Note, however, that the British census only gives information about the jobs emigrants went into, not the ones they left in Ireland. Some information on the occupations of emigrants before and after leaving Ireland can be had for the period 1939–52, when travel permits were required for travel between Ireland and Britain (Commission on Emigration 1955:267). Between 1940 and 1951, three-quarters of males applying for travel documents ‘described their last occupation as that of unskilled labourer, builder’s labourer or agricultural worker’.16 Between 1948 and 1951, 31 per cent of emigrants gave agriculture as their intended occupation, 37 per cent unskilled labour, 10 per cent industry (including building), and 22 per cent were other skilled workers and clerks. It seems clear that at the very least unskilled workers, both inside and outside the building trades, were extremely mobile between the two countries. There is one issue, however, which needs to be brought up here. Most emigration from Ireland to Britain has traditionally been from rural areas, 152
DID LABOUR FLOW UPHILL?
whereas the wage data we are looking at are for Dublin.17 What if Dublin were segmented from both the rural Irish and British labour markets, with country people being more likely to emigrate than to move to Dublin, and Dubliners being comparatively unlikely to emigrate? Meenan (1970:210) once wrote that ‘the contrast between the extraordinary mobility of the Irish people throughout the world and their relative immobility within their own country is striking’. While it is beyond the scope of this chapter to go into these issues thoroughly, it seems somewhat fanciful to assert that Dublin could have remained insulated from an international labour market embracing the rest of Ireland and Britain. Dubliners did emigrate, and in 1946 one out of every three residents of County Dublin had been born outside the county (Commission on Emigration 1955:132). In conclusion, it is implausible to assert that the links between the Irish and British labour markets caused by emigration did not affect the three Dublin industries considered in this chapter. Relatively high Dublin wages thus remain somewhat of a mystery. 153
KEVIN O’ROURKE
The Reder hypothesis M.W.Reder (1955) advanced a series of hypotheses about how skill differentials might vary over the course of the business cycle, and why they might differ between geographical regions. Reder envisaged a world where workers could be ranked in terms of their skills, and where jobs also occupied different positions on a skill ladder. During a boom, employers at the top of the ladder would find themselves having to hire workers with slightly lower qualifications than those they had previously employed; employers lower down the ladder would have to replace these workers with still less-skilled workers; and so on. Thus workers in general would find themselves shunted up the skill ladder during booms; the result would be an excess demand for workers at the bottom of the ladder. Their wages would thus increase. Reder thus predicted that skill differentials would narrow during booms and widen during recessions. In a geographical context, Reder predicted that in areas where the capital labour ratio was higher, and where competition between skilled and unskilled workers was more intense (due to the existence of jobs on the middle rungs of the ladder, in particular semi-skilled manufacturing jobs), wage differentials would be narrower. His hypothesis thus predicts that skill differentials should be higher in Ireland than in Britain—which, as Figures 7.1–7.3 show, they indeed were before (but not after) the mid1960s. If the market for unskilled labour was well integrated (and as we saw in the last section, the unskilled were particularly numerous among the ranks of emigrants) then skilled wages should be higher in Ireland than in Britain. This all seems to fit reasonably well with the stylized facts as presented in the preceding two sections, but the hypothesis seems difficult to test; nor does the hypothesis explain the fact that skill differentials in construction and engineering were lower in Ireland than in Britain after the mid-1960s. The age distribution of emigrants Irish emigration has been predominantly of young people. For example, between 1911 and 1921, 41.6 per cent of male emigrants and 39.5 per cent of female emigrants were between 20 and 24 years old; between 1943 and 1951, 20–24 year olds received 33.4 per cent of travel permits granted to men, and 37.9 per cent of permits granted to women (Commission on Emigration 1955:122, 129). The fact that emigrants were on average younger than workers staying at home could help explain the paradox, if wages increase with the age of workers: higher Irish wages would in part reflect the fact that workers in Ireland were on average older than workers in Britain.
154
DID LABOUR FLOW UPHILL?
CONCLUSIONS In the three industries studied here, Irish wages have clearly caught up with real wages in Britain, and had indeed already largely done so by the interwar period. Moreover, in these industries Irish wage rates have often been higher than in Britain. This chapter has argued that a Todaro framework can help explain the co-existence of relatively high wages in Ireland and continuing migration from Ireland to Britain. There are some lessons in this for contemporary Irish policy makers. In particular, Irish industrial policy has often tried to sell Ireland as a low-wage location. Wages in Ireland may indeed be low for some grades of labour (eg clerical labour, or certain types of professionals) but the evidence of this chapter suggests that, in general, Ireland can hardly be viewed as a low-wage country. One obvious implication may be that Irish industries should attempt to compete internationally on the basis of quality rather than on cost. This chapter has only begun to scratch the surface of what needs to be done if Anglo-Irish migration in this century is to be fully understood. To name but a few of the more obvious examples, we would like to have information on a host of variables: unemployment by skill category within the two countries; regional data for Ireland; data on what sectors the Irish migrants came from and what sectors they moved to; data on how the stock of Irish migrants, by occupation, within Britain, evolved over time; and more information on trade union behaviour and apprenticeship practices. If the challenge is somewhat daunting, however, the potential benefits would seem to be great: an enhanced understanding of migration between rich and poor countries, and maybe some policy lessons for contemporary Ireland. APPENDIX A: WAGE RATE DATA AND EARNINGS DATA What can explain the contrast between the data on average hourly earnings and on wage rates? O’Mahony summarizes the possibilities succinctly: …basic rates in Ireland are sometimes considerably higher than in Britain and …they are scarcely ever more than marginally lower. Average earnings in Britain, at least in industry, tend to be higher than in Ireland. This is a reflection of differences in the structure of particular industries, in the composition of the labour force and in the incidence of overtime and other opportunities of earning more than the basic rates. All these factors result in a relatively greater proportion of higher paid workers in Britain than in Ireland… (O’Mahony 1965:10) It is sometimes argued that while earnings were closely tied to rates in Ireland, they were less so in Britain. Kaim-Caudle stated that ‘earnings in Ireland are often close to minimum rates while in Britain earnings are well 155
KEVIN O’ROURKE
above the agreed rates’ (Kaim-Caudle 1964:6). Similarly, O’Mahony noted that in the construction industry ‘basic official rates are very similar to those paid in Britain but there is not much overtime worked here and there is no jointly accepted bonus scheme so that earnings tend to be higher there’ (O’Mahony 1965:52). In opposition to this view, several writers have pointed out that earnings in Ireland have been rising more rapidly than wage rates since 1953 (Nevin 1963:2; Cowling 1966:16; Mooney 1978:238–9), and Dicks-Mireaux and Shepherd (1962) find that wage rate changes do affect earnings in Britain. There are thus grounds for arguing that earnings and wage rates were not necessarily more mechanically linked in Ireland than in Britain, and that wage rate comparisons do not necessarily produce a bias in favour of finding higher Irish living standards. At the least we should require some hard quantitative evidence that wage rates were more binding in Ireland than in Britain before dismissing the wage rate data as irrelevant. More importantly, the earnings data are given at the industry or even the economy-wide level, rather than at the level of particular occupations and skill levels. Clearly if unskilled workers form a higher proportion of the Irish labour force than of the British labour force, Irish earnings will appear to be lower than British earnings, even if workers of particular skill categories are paid the same in both countries (or, possibly, even if Irish workers are better paid than their British counterparts). The Census of Population for the two countries in the inter-war and early postwar period shows that unskilled workers formed a much greater proportion of the labour force in Ireland than in Britain (England and Wales, Registrar-General 1934, 1957; Irish Free State, Department of Industry and Commerce 1928–34, 1938–41; Ireland, Central Statistics Office 1949, 1952– 4). For example, about three-quarters of Irish workers in building and construction were unskilled, compared with at most half in England and Wales. It is evident that industry-level average earnings are completely inappropriate, at least for this industry, if the focus of interest is relative wages of particular skill classes, labour market integration, or the relative cost of labour in the two countries.
156
DID LABOUR FLOW UPHILL?
157
KEVIN O’ROURKE
APPENDIX C: CALCULATING COST OF LIVING ADJUSTMENTS Kravis et al. (1982) (KHS) state that in 1975, £0.372 sterling was equivalent in purchasing power terms to IR£0.352, which was equivalent to $1.00 (KHS: 198). For 1975, the ratio of Irish to British wage rates should therefore be adjusted upwards by a factor of (0.372/ 0.352). The factor of adjustment will be different for different years, if inflation rates in the two countries differ. Specifically, if CPIit is the consumer price index in country i at year t (with a base of 1975=100), then the appropriate adjustment factor for year t is (0.372/ 0.352)X(CPIBt/CPIIt).18 For the pre-war period I use PPP calculations based on ILO data for 1938 due to Taylor and Williamson. These PPPs only take account of differences in food price levels; in 1938 Irish food costs were 9.3 per cent higher than food costs in Britain. Adjusting relative wages to take account of differences in food prices therefore involves adjusting Irish relative wages downwards (dividing by 1.093 for 1938, and by 1.093 X CPIIt/CPIBt in year t). An important item in the household’s consumption basket left out of these calculations is rent, which will presumably be lower in Ireland than in Britain. Some indication of how important this factor may be can be gained in a rather ad hoc way by comparing PPPs based on food costs alone with PPPs taking rent into account, calculated for the two countries for earlier years. Based on Board of Trade data, Taylor and Williamson have calculated PPPs for Ireland and Britain for the years 1905 and 1912. These show 1905 food costs 4.8 per cent, and 1912 3.7 per cent, higher in Ireland than in Britain. Incorporating both food and rent, however, the Irish cost of living was 93.8 in 1905 and 93.3 in 1912 (UK= 100). The ratio of the Irish to the British cost of living was therefore boosted by about 11 per cent when rent was left out. Similarly, in 1926 the ILO calculated an index of relative food costs in several countries, including Ireland and Britain. In 1926 they found that a standard basket of provisions cost $1.86 in Dublin, as opposed to $1.80 in London (International Labour Review 15, 1:126–31). The ILO went on to include some measure of rent in their cost of living calculations. Adding rent raised Irish relative to British real wages estimates by 9 per cent. This 9 per cent difference corresponds fairly closely with the discrepancy found for the years before the First World War.
158
DID LABOUR FLOW UPHILL? A very rough and ready way of taking into account the impact on relative living standards of lower Irish rents might thus be to multiply the ratio of Irish to British wages (deflated by food costs) by a factor of 1.09 (or 1.111, or 1.117).19 Since I am trying to argue that price adjustments do not really solve the paradox posed by Figures 7.1–7.3, I chose the lowest of these numbers, 1.09.
NOTES I am grateful to Jeremy Atack, George Grantham, Mary MacKinnon, Larry Neal, Cormac Ó Gráda, Sungsup Ra, Alan Taylor, Jeffrey Williamson, and the staff at the National Library of Ireland, as well as to seminar participants at Montreal, the University of Michigan, and the University of Illinois. 1
2 3 4 5 6 7
8
9 10 11
12 13 14 15
Since labour has been free to move between Ireland and Britain throughout the twentieth century (with the exception of a brief period during and immediately after the Second World War) we need to rely on census data for estimates of emigration. For a succinct summary see Cullen (1987), the source of the output and employment figures given above. A task of major importance will be to document this process more fully, but this is outside the scope of the present chapter. Ó Gráda (1988a: 7) is comparing Mokyr’s and Feinstein’s national income data. However, in the mid-1870s American real wages were still nearly twice the level of real Irish wages (Ó Gráda 1988b: 161). This survey is also used by Boyer and Hatton in Chapter 5. Although it is beyond the scope of this chapter, it would clearly be of interest to try to explain why Dublin skilled building wages were so high before the Famine, and why this collapse occurred in the relative skilled Dublin wage between 1844 and the 1880s. In industry average nominal hourly earnings were 4/6 in Ireland and 6/9.5 in Britain in 1960. This gap had disappeared by 1977 (ILO Year-book of Labour Statistics, various years). Irish agricultural earnings remained lower than British earnings into the 1980s. This seems an obvious point, and yet I have only seen it made in print by O’Mahony (1967:16, fn. 19). See also O’Mahony (1965:10). Wage rate data can be subjected to many criticisms. Appendix A considers the relative merits of the available wage rate and earnings data. The data refer in all instances to hourly wage rates in a week in October, except for 1929 and 1930, when July wages are given, and 1941, when November wages are given for Ireland. Whenever possible the Irish data represent Dublin wages, and the British data London wages. The data can be found, in the years up to 1963, in the ILO International Labour Review (or its Statistical Supplement), and in subsequent years, in the ILO Bulletin of Labour Statistics (in the quarterly issues devoted to the results of the annual October survey). Data are not available for all years; hence the gaps in the figures below. The exception is unskilled labour in the printing and publishing industry. Appendix C gives details of how all the cost of living adjustments in this section were made. These calculations are the work of Alan Taylor and Jeffrey Williamson. I am grateful to them for making their unpublished findings available to me. Unfortunately, unemployment data by skill category are not available.
159
KEVIN O’ROURKE 16
17
18 19
Commission on Emigration (1955:127). On the same page the Commission cites an earlier study of ‘overseas’ emigrants during the period 1924–39:68 per cent of male emigrants were accounted for by the occupational group ‘agricultural labourers and labourers not in transport and communications’. For example, over the period 1943–51, 13 per cent of travel permits issued went to residents of County Dublin, 19 per cent to other urban districts, and 68 per cent to rural districts (Commission on Emigration 1955). In 1946 County Dublin accounted for 21.5 per cent of the country’s population (Statistical Abstract 1946:20); the county was thus underrepresented among emigrants. This will be theoretically legitimate as long as tastes in the two countries remain unchanged. This is not theoretically justifiable but seems a reasonable strategy given the data constraints I am working with.
160
8 THE DECLINE IN HOURS OF WORK IN US LABOUR MARKETS, 1890–1903 Joshua L.Rosenbloom and William A.Sundstrom
Reduction of hours has been one of the two major demands of labor in the United States…. The movement for shorter hours has been…a constant, dogged, frequently hopeless insistence by one great part of society on the need for a shorter day in the face of an equally determined resistance by a smaller part to every suggestion for immediate change. (Cahill 1932:11)
In 1880 the standard work day in the United States was roughly ten hours. By the 1940s, it had declined to about eight hours—where it remains today. Meanwhile, elimination of Saturday work resulted in an even greater reduction in weekly hours.1 The decline in hours of work was partly a reflection of the rising standard of living: as workers became more productive and their incomes rose, they opted to take some of this increase in the form of leisure. In other words, the decline in hours constituted a movement along a backward-bending labour supply curve caused by the secular increase in real wages.2 Such a simple story, however, does not explain the process by which reductions in hours were actually achieved. While productivity and per capita income grew at a relatively steady pace, progress towards shorter hours was uneven both over time and across industries. Employers bitterly resisted workers’ demands for hours reduction. Although demands for shorter hours had surfaced occasionally before the Civil War, it was during the economic expansion of the late 1860s and early 1870s that the issue first rose to national prominence. The issue largely disappeared during the subsequent contraction, but agitation for the eight-hour day re-emerged during the 1880s, culminating in a general strike called by the newly formed American Federation of Labor (AFL) in 1886, in which nearly 190,000 workers participated.3 Employers adamantly resisted workers’ desires, arguing both that shorter hours would raise production costs and that workers 161
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
would use additional leisure time for drinking or other activities that would impinge on their health and reliability at work (Atack and Bateman 1989:31– 2; Cahill 1932:225–6). In contrast to more workplace-specific issues such as wages, accident risks, and unpleasant working conditions, the desire for shorter hours was widely shared, and provided workers with an issue around which it was possible to organize a broad coalition of interests. During the latter part of the nineteenth century, the movement for shorter hours was international in scope. The periods of most intense agitation for the eight-hour day coincided across the United States, Britain, France, and other industrializing countries (Cross 1988:155). Hours of work in the United States declined only gradually until the late 1890s. Thereafter, progress towards the eight-hour day quickened (see Figure 8.1). By 1920 a forty-eight-hour work week—eight hours a day, six days a week—was the norm in most industries. Weekly hours remained stable throughout the 1920s, resuming their fall only with the onset of the Great Depression (Jones 1963). The timing of hours reduction varied considerably by industry. According to Douglas (1930), hours began to decline first in the building trades, turning down before 1900. While the decline in hours in unionized manufacturing began soon thereafter, hours reductions in socalled payroll (largely non-union) manufacturing began to drop significantly only around the time of the US entry into the First World War. There have been many attempts to explain these patterns of hours reduction.4 Cahill’s study of the shorter hours movement concentrated on 162
HOURS OF WORK IN US LABOUR MARKETS
the role played by legislative limitations on hours and the pressure of organized labour. Recognizing that only a minority of workers were represented by unions or directly affected by legal restrictions, however, Cahill was obliged to conclude that the bulk of hours reductions occurred through the Voluntary’ actions of employers (1932:222–3), and suggested that these actions might be explained by the pressure of labour scarcity and the ‘force of example and competition of the better hours standards established by legislation, and trade-unions’ (1932:223). In his Theory of Wages (1957), Douglas interpreted the decline in hours during this period as a direct consequence of the backward-bending labour supply curve, using the inverse relationship between wages and hours exhibited in his time series to estimate the short-run labour supply elasticity. An implicit assumption of his analysis was that most fluctuations in wages and hours reflected shifts in the labour demand schedule along a relatively stable supply curve. More recently, Owen (1970) and Kneisner (1976) have advanced a similar interpretation. It is unclear whether this approach can adequately account for cross-sectional differences in hours worked or for the unevenness of hours decline across time and occupations. For instance, in the 1890s pattern makers and carpenters had comparable hourly wages, but the former worked about five hours longer per week. Recently the topic of hours reduction has received renewed attention. Whaples’ (1990a) quantitative analysis of changes in weekly hours in manufacturing from 1914 to 1919 provides support for both the backwardbending labour supply hypothesis and the labour scarcity hypothesis proposed initially by Cahill. Exploiting cross-section data by city and industry, he finds that hours reductions were attributable largely to short-run labour market tightness caused by the combination of the strong wartime demand for labour and the cut-off of European immigration. If the long-run decline in hours of work is to be explained by temporary market tightness, however, some sort of ratchet effect is necessary, so that once achieved, hours reductions are not reversed as demand pressures ease. Shiells’ (1990) study of the steel industry, where the transition from twelve to eight hours per day did not occur until the 1920s, suggests that the role of hours as a ‘workplace public good’ posed a coordination problem that slowed responses to changing labour market conditions. While Whaples and Shiells have examined the reductions in hours that took place in the (largely non-union) manufacturing sector of the economy after 1910, relatively little is known about the decline in hours around the turn of the century. This chapter examines the nature and causes of the decline in hours during the period 1890 to 1903. Like Douglas’, our examination rests on data from the 19th Annual Report of the Commissioner of Labor (US Department of Commerce and Labor 1905). Douglas concentrated on the aggregate behaviour of hours and wages, but we place special emphasis on variations in the relationship between hours and wages across locations 163
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
and occupations. By focusing on disaggregated patterns we are able to examine the dynamics of hours reduction at a level approximating that of the individual establishment. Our examination proceeds in two parts. First, we examine the extent of heterogeneity in the behaviour of hours and wages across occupations and locations. The national trend in hours documented by Douglas is also apparent in the disaggregated data, but there are important regional and occupational variations in the magnitude and timing of hours reductions. While the similarities across cities and occupations suggest that hours reductions were the consequence of broadly felt social and economic forces, the variation across occupations and regions suggests that more localized forces were also important in determining labour market outcomes. Second, we consider whether the time series behaviour of hours and wages can shed light on the dynamics and causes of hours reduction around the turn of the century. Consistent with the hypothesis that shorter hours were achieved during episodes of rising wages, we find a negative relationship between changes in hours and changes in wages. This relationship was asymmetric: while increases in wages were associated with reductions in hours, there was no systematic relationship when wages fell. Adjustment costs or worker norms may help explain the asymmetry. This upward ‘stickiness’ in hours is evidence of the sort of irreversibility of hours reductions that would have to exist if the long-run downward trend in the length of the work day were to be explained by temporary periods of labour scarcity. It also poses a challenge to naive microeconomic interpretations of changes in hours of work as movements along a backward-bending labour supply curve, since such movements should be fully reversible. The timing of hours changes across certain related groups of occupations supports both Shiells’ suggestion that hours may have been a workplace public good and Cahill’s suggestion of a spillover effect from trade union pressure or legal restrictions on hours (the ‘force of example’). Among foundry and machine shop workers, there appears to have been a need to coordinate changes in hours across all workers at a particular establishment. Hours changes were not generally correlated within the building trades, but hours of bricklayers and hod carriers did tend to move together, which again appears to reflect the need for coordination. There is some evidence that the success of bricklayers, the ‘aristocracy’ of the building trades, led in subsequent years to reductions in hours in other building trades occupations. Hours of municipal street and sewer workers and the skilled building trades are strongly correlated, suggesting that political pressure may have been an important source of leverage for some occupations in winning hours reductions.5
164
HOURS OF WORK IN US LABOUR MARKETS
WAGE AND HOURS DATA The 19th Annual Report of the Commissioner of Labor (US Department of Commerce and Labor 1905; hereafter called the 19th Annual Report) contains annual data on wages, hours, and employment for twenty-one occupations for a large number of US cities from 1890 to 1903. We selected ten occupations drawn from the building trades, foundry and machine shop occupations (an important component of Douglas’ category of union manufacturing industries), and street and sewer construction. Five occupations are highly skilled (bricklayers, carpenters, machinists, iron moulders, and pattern makers), four unskilled (building trades labourers, foundry and machine shop labourers, municipal street and sewer workers, and contract street and sewer workers), and one semi-skilled (hod carriers). To ensure a reasonably uniform geographic representation across occupations we limited our sample to the thirty-eight cities with observations for five or more of the selected occupations.6 All of the data in the 19th Annual Report were derived from the inspection of payroll and other records from a constant sample of establishments in each city, and reflect conditions during ‘that portion of each year…most nearly representing the normal conditions of the establishments during each year’ (US Department of Commerce and Labor 1905:11). Typically the data for each city represent from two to six establishments, so the data are probably best interpreted as reflecting these employers’ decisions about what employment terms to offer in each year. Weekly hours were described as ‘normal hours’ for full-time employees, which likely removed most shortrun seasonal fluctuations. Wages were computed as the average hourly earnings of all workers in an occupation at the sampled establishments. REGIONAL AND OCCUPATIONAL TRENDS IN HOURS AND WAGES As several chapters in this volume demonstrate, despite substantial geographic mobility in this era, labour markets remained imperfectly integrated. Within the United States, both the west and the south remained at least partially distinct regional labour markets, and there were substantial and persistent differences in labour market conditions between cities within the northeastern and midwestern regions of the country (Rosenbloom 1990; Sundstrom and Rosenbloom 1993). It is not clear from Douglas’ series (Figure 8.1) whether the reduction in hours around the turn of the century was also characterized by regional differences or was relatively uniform. As we show below, there was in fact a considerable degree of uniformity in hours reductions across cities and occupations, suggesting that broadly based social and economic factors must have been at work in this era. Significant regional and occupational diversity 165
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
is also evident around these broad trends, suggesting that more localized forces also require attention. Average weekly hours varied across occupations, locations, and years. A convenient way of separating each of these effects is to regress hours (h) in a particular region (i), occupation (j), and year (t) on a series of dummy variables representing region, occupation, and year effects. This is an analysis of variance regression with no interaction, where we are interested in the parameter estimates as well as the analysis of variance. Using the estimated coefficients weekly hours can be represented as:
where α is the intercept, β is the region effect, δ is the occupation effect, π is the year effect, and ε is an unexplained residual. The intercept measures hours in a base occupation, region, and year, and each of the other coefficients is the incremental effect of being in another occupation, region, and/or year. Here we take as the basis of comparison building trades labourers in the northeast in 1890. In Table 8.1, column (1) reports the regression coefficients from an analysis of variance regression run on the entire sample. The intercept shows that building trades labourers in the northeast worked nearly fifty-nine hours a week in 1890. Hours were considerably shorter in the other building trades, but both skilled and unskilled foundry and machine shop workers had a slightly longer work week. Whereas the work week of street and sewer contract workers did not differ appreciably from that of building trades labourers, street and sewer workers employed by municipalities enjoyed work weeks almost as short as those of the skilled or semi-skilled building tradesmen. Thus skill level by itself does not appear to account for differences in the length of the work week. The region effects show that hours of work throughout the northern part of the country were roughly equalized. Hours in the north central and west regions were only slightly, and not significantly, lower than in the northeast. In contrast, there was a significant north-south difference. Work weeks were from one to three hours longer in the southern regions than in the north. Hours declined continuously after 1890, as is indicated by the negative sign and increasing magnitude of each of the year effects. By 1903 the work week had declined by just over four hours, or 6.9 per cent of the work week for building trades labourers in 1890. Most of this decline was achieved after 1899. Up to 1899 the average rate of decline was only 0.3 per cent per year, but in the next four years, the pace of reductions accelerated to an average annual rate of slightly more than 1 per cent. The coefficients reported in column (1) of Table 8.1 implicitly assume that occupation effects are identical in every region and year and that the year effects are the same in every region and occupation. Columns (2)–(6) report the results of estimating separate analysis of variance regressions for each 166
HOURS OF WORK IN US LABOUR MARKETS
167
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
region. These regressions still impose the assumption that occupation effects remain constant across years and that year effects are constant across occupations, but occupation and year effects are now allowed to vary between regions. Two patterns of cross-occupational variation in hours of work are apparent. In the northern and western parts of the country bricklayers, hod carriers, carpenters, and municipal street and sewer workers enjoyed a much shorter work week than any of the other occupations. In the extreme, municipal workers in the west worked about eleven hours less than building trades labourers, and nearly ten hours less than any of the foundry and machine shop workers. In the southern cities, however, the range of occupational variation was much more compressed. Hours were still shorter for bricklayers, hod carriers, and carpenters, but the differences between them and building trades labourers were considerably smaller. Similarly, municipal employees in the south worked a much longer week than their northern counterparts. In every region, foundry and machine shop workers worked more hours than skilled or semi-skilled building trades workers, and in most cases their work week was longer than that of building trades labourers. Regional differences in the year effects are also apparent. In the west, year effects varied substantially around a weakly declining trend. In the rest of the country, a stronger downward trend is more clearly visible, closely paralleling the pattern found in the first column. The decline was smallest in the south Atlantic region, where hours declines began somewhat later and hours fell more abruptly once they did begin to fall. All of the regressions reported in Table 8.1 assume that year effects were identical across occupations. Distinct occupational trends in hours are revealed by estimating separate analysis of variance regressions for each occupation, combining the observations from the different regions.7 The results are depicted in Figures 8.2 and 8.3, which show weekly hours for each occupation in the northeast.8 The overall time pattern of declines in hours worked appears quite similar across occupations, with acceleration in the downward trend coming slightly before 1900 for the building trades and the unskilled workers, and shortly after 1900 in the foundry and machine shop occupations. The most significant departure from this pattern is the municipal street and sewer workers, whose work week fell sharply between 1897 and 1900 and then levelled off. The overall picture is one of a gradual decline before the late 1890s, followed by a perceptible acceleration around the turn of the century. Although the timing of hours declines was quite similar across occupations, the magnitude of hours reductions varied substantially. Hours fell most in those occupations that had enjoyed the shortest hours in 1890—skilled building trades and municipal workers—thus increasing the occupational dispersion in hours of work. The framework developed to analyse variations in hours can also be applied to wage variations. Table 8.2 reports the results of analysis of 168
HOURS OF WORK IN US LABOUR MARKETS
169
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
170
HOURS OF WORK IN US LABOUR MARKETS
The occupational structure of wages was quite similar throughout the country. Bricklayers commanded an exceptionally large margin relative to building trades labourers in every region of the country, but skill premia were also rather large for all of the skilled and semi-skilled occupations. Among the unskilled occupations, the wages of building trades labourers and street and sewer contract labourers were quite close, with the wages of foundry and machine shop labourers slightly lower and those of street and sewer municipal workers appreciably higher. Generally, wages increased from year to year. As was true with weekly hours, real hourly wage growth accelerated around the turn of the century, though the extent of the change was less pronounced for wages than it was for hours. This acceleration is also visible in Figures 8.4 and 8.5, which show the wage rate for each occupation implied by separate analysis of variance regressions for each occupation. Wages in most of the occupations rose throughout the 1890s, accelerating slightly around 1899. The south Atlantic and the west seem to have diverged the most from national patterns in both hours and wages, suggesting a persistent isolation from the forces operating in other regions. Occupational differences were more pronounced than regional differences. Bricklayers, hod carriers, carpenters, and municipal street and sewer workers enjoyed the shortest hours at the beginning of the period and they also won the greatest reductions in hours over the period considered.9 In the next section we seek to exploit these differences more fully to illuminate the mechanisms through which hours reductions were achieved. 171
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
UNDERSTANDING THE PACE AND PATTERN OF DECLINING WEEKLY HOURS This section describes the observable implications of the various explanations of hours decline that have been proposed in the literature. The next section assesses the ability of the various theories to account for the pace and pattern of hours decline observed in the data from the 19th Annual Report. The regression results confirm the empirical link between rising pay and falling hours, but they are also consistent with the influence of other factors that shaped the dynamic of hours reductions in specific occupations and locations. As workers’ wages rose, it is plausible that they would have chosen to consume some of their rising income in the form of increased leisure time. The decline in hours would represent one facet of the broad upward trend of American workers’ standard of living. If such a movement along a backward-bending labour supply curve were a principal cause of declining hours, one would expect to observe an inverse relationship between changes in hours worked and changes in hourly wages. Workers’ ability to achieve hours reductions in the short run, however, may have been constrained by the availability of only a limited number of employment packages in the labour market, and adjustment costs (caused for example by the public goods character of workplace hours decisions) that restricted the flexibility of hours. Local political and trade union pressures may have further affected the rate of hours decline in specific locations and occupations. According to the textbook microeconomic model of a worker’s labour172
HOURS OF WORK IN US LABOUR MARKETS
leisure decision, rising hourly wages would lead to shorter desired hours if the income effect outweighed the substitution effect. A number of studies have interpreted aggregate wage and hours data as evidence of such a backward-bending labour supply curve (e.g. Kneisner 1976; Douglas 1957). In theory, because employers can offer any combination of hours and wages, we would expect them to adjust their terms of employment smoothly and continuously to minimize labour costs in response to the shifting labour supply behaviour of workers. A related implication of the textbook model is that the wage-hours relationship should be symmetrical: any reduction in hours that occurs as wages rise should be fully reversed if wages return to their original level. When workers acted collectively to achieve shorter hours, they usually linked hours reductions to wage increases, often with the goal of maintaining their prior daily or weekly earnings levels. Evidence on strikes collected by the US Commissioner of Labor for the year 1886, for example, shows that when strikes ‘for a reduction of hours’ fully succeeded (from the strikers’ point of view), workers almost always won hours reductions without any change in their daily pay. Moreover, workers in about half the strikes that were considered only ‘partly successful’ were able to maintain their daily earnings, while in the other half they took a pay cut.10 Whaples (1990a) modifies the conventional labour supply approach to take account of the possibility that employers are not indifferent about the number of hours their employees work. Under such circumstances supply and demand forces interact to determine hours of work and wages simultaneously.11 Applying this analytical framework to the period around the First World War, Whaples argues that hours reductions at this time can be explained as a consequence of a shift in employer demand caused by tight wartime labour markets. He finds that movements in the exogenous determinants of the real wage have considerable explanatory power in a crosssection regression analysis of hours changes. Because the wartime escalation of real wages was coincident with a period of temporary labour shortage, Whaples’ findings are consistent with two distinct if related hypotheses: first, that hours and real wages were generally inversely related due to labour supply behaviour; and second, that periods of extreme labour shortage provided workers with bargaining power to press for permanent hours reductions. The prediction that full-time or ‘normal’ working hours would decline during business-cycle expansions must be distinguished from the expectation that actual hours worked would be procyclical, reflecting the historical practice of work sharing as a partial alternative to employment reductions (see Bernanke and Powell 1986; Carter and Sutch 1991). Because the hours data reported in the 19th Annual Report represent standard fulltime hours, this hypothesis implies that a proxy for demand conditions will be inversely 173
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
related to the measure of hours in our data. If real wages move procyclically, changes in wages may in part be a proxy for cyclical demand conditions. Whereas the textbook theory permits continuous and symmetrical adjustments in hours of work, in historical practice employment packages were standardized around only a small number of possible choices; changes in hours usually took the form of discrete and discontinuous shifts between the offered packages. In 1890, for example, carpenters in about 80 per cent of establishments in northern cities were working either a nine- or ten-hour day, six days a week; other occupations exhibited similar degrees of standardization.12 This restricted range of hours is an instance of the standardization of labour contracts that has been observed in many historical contexts. Wright (1987) argues that the costs of obtaining and evaluating information on the complex set of components that make up an employment ‘contract’ encourage both sides of the market to simplify choices by standardizing a limited number of packages. Aside from informational considerations, another potential source of adjustment costs in the choice of hours was the need to coordinate work effort by different groups of workers at a particular establishment (Shiells 1990). Because of potentially heterogeneous tastes across workers, and the difficulty of establishing mechanisms of collective choice in such situations, this coordination problem probably discouraged continuous hours adjustments and may have slowed progress towards shorter hours. If workplace coordination issues played a significant role in the process of hours adjustment during our period, then changes in hours would be correlated across different occupations employed in any given establishment. For example, the hours of machinists, moulders, pattern makers, and foundry labourers would all tend to move together in each establishment. Although the observations drawn from the 19th Annual Report are not establishment specific, each represents only a small number of establishments in each city, and strong cross-occupational correlations within establishments would probably persist in the city-level data. Another observable implication of adjustment costs might be asymmetries in the relationship between wage and hours changes. Suppose, as appears to have been the case by the late 1890s, that the general trend of the labour market equilibrium was towards higher wages and shorter hours, but that adjustment costs prevented the continuous movement of hours to their equilibrium level. Real wages might then appear to fluctuate around an upward trend, with hours of work occasionally ratcheting downwards. The data might then on average exhibit downward flexibility but upward ‘stickiness’ of hours in response to wage changes. Upward stickiness of hours might also be the product of fairness norms or other behavioural factors, as we note below. Looking beyond the microeconomist’s income effect and public goods, historians have cited legislation, city politics, and trade union pressures as important factors in the historical decline of working hours. Although these 174
HOURS OF WORK IN US LABOUR MARKETS
factors are not directly quantifiable using existing data, they often imply predictable and measurable spillover effects of hours reductions between occupations. With the exception of some specific industries (such as the railroads), special categories of workers (women and children), and brief periods of federal intervention (the First World War), state and federal regulation of working hours was very limited before the Great Depression. Nevertheless, it may be the case, as Cahill (1932) has suggested, that in some cities working hours were effectively regulated, perhaps informally by way of public pressure on employers and the threat of actual legislation. Recently Friedman (1992) has argued that the large wage premium (and shorter hours) enjoyed by the skilled building trades in big cities around the turn of the century can best be understood as a consequence of pressure placed on employers by city government. During the late nineteenth century, big-city political machines successfully diffused potential class conflicts by forging a coalition encompassing both property holders and important elements of organized labour. Because of the localized nature of the construction market, which prevented employers from moving to more hospitable locations while protecting them from product-market competition, such alliances were most common and had the greatest influence on conditions in the building trades. Political machines sought accommodation with the building trades both through extensive public works spending that raised demand for their services and through implicit and explicit pressure on employers to concede to labour’s demands.13 In Chicago, for example, where twenty-two building trades leaders were on the city payroll by 1900, the Democratic machine that dominated city politics had enacted eight-hour legislation for workers on city construction projects in 1889, and openly supported favoured unions in labour disputes by denying employers and their strikebreakers police protection (Friedman 1992:461–2). If local legislation or other more subtle political influences were an important factor in the reduction of hours, one might observe reductions in the hours of municipal employees leading or strongly correlated with reductions in the hours of private sector workers, especially those in the politically powerful building trades. Cahill (1932) has also suggested that successful efforts by unions to reduce their hours may—through the ‘force of example’—have led to hours reductions for non-union trades. Although the national and international unions seem to have given a low priority to hours reductions after the early 1890s, union locals and the rank and file may have continued to press hard for the eighthour day well into the twentieth century. Disaggregated data on unionization are poor or non-existent for this period, but some occupations were much more highly organized than others. The skilled building trades, for example, were relatively well organized into various AFL affiliates, while unskilled labourers were largely unorganized. As we have noted, the hours reductions of our period were not restricted to the highly organized building trades, 175
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
although the rate of decline was more rapid in those trades. Perhaps successes in winning hours reductions by unionized workers spilled over to less wellorganized occupations in the same city, for example from carpenters to building labourers. More generally, organized campaigns for hours reductions, especially at the height of the eight-hour movement, were often conducted on a city- wide basis (Cahill 1932). If these movements benefited from organizational economies and demands for equitable treatment, hours reductions in any one occupation could have had spillover effects on any others in the city. REGRESSION ANALYSIS OF CHANGING HOURS, 1890–1903 It is possible to shed considerable light on the mechanisms generating hours declines in the late nineteenth century by examining the relationship between hours and wage movements within occupations as well as the relationship of hours movements across occupations within cities. Because of the extreme difficulty of identifying structural equations in a realistic model of hours of work, our regressions should be regarded as descriptive equations designed to identify important correlations and patterns in the data, rather than as estimates of structural models. We begin with a baseline model of the relationship between year-to-year percentage changes in hours and the corresponding percentage changes in real hourly wages. Let hit and wit be the weekly hours and real hourly wages of workers in some occupation in city i and year t. Our baseline model can be written
where ∆Xt=Xt- Xt-1, pt is a year-specific effect captured by dummy variables, and uit is an error term assumed to be identically and independently distributed across observations. The coefficient ß1 can be interpreted as the elasticity of hours changes with respect to wage changes, and is predicted to have negative sign under either the labour supply or market tightness hypotheses. The model is easily modified to take account of asymmetries in the effect of changes in the wage or other explanatory variables. If hours of labour were reduced during periods of rising real wages but rose only slightly when real wages fell, then the elasticity of hours with respect to wages would be more negative for increases in wages than it would be for decreases. We include a dummy variable for decreases in wages and interact it with the wage-change variable, which allows a variable’s effect to have separate slopes and intercepts for increases and decreases. The model can also be augmented to include other explanatory variables intended to
176
HOURS OF WORK IN US LABOUR MARKETS
capture aspects of market tightness not measured by wages or to reflect cross-occupation influences on hours of work. One potential problem with the implementation of the model is that our wage data are nominal, and good cost of living indices are not available disaggregated by cities. We have therefore deflated the wages by the national cost of living index proposed by David and Solar (1977). Although this procedure does not allow us to measure intercity differences in real wage levels, so long as price movements within individual cities paralleled national trends it will provide an accurate representation of changes in real wages. Because our model is specified in terms of changes in hours and wages rather than levels, any locational price differences that persist throughout the period will not affect the results, nor will any other persistent cityspecific effects such as local amenities or differences in growth rates. Column (1) of Table 8.3 reports the hours-wage elasticities estimated using ordinary least-squares regressions for equation (8.2), with separate equations for each occupation. The wage-hours relationship varied considerably across occupations. An F test rejects at the 5 per cent level the null hypothesis that the separate occupational equations can be combined into a single equation with separate intercepts for each occupation but a common wage elasticity. The wage elasticity (ß1) is negative and statistically significant in most cases. For carpenters, for example, the estimated coefficient implies that a 10 per cent increase in the real hourly wage was associated with a 1.9 per cent reduction in weekly hours. At the extremes, the wage elasticity of hours is not distinguishable from zero for pattern makers, and is nearly -0.5 for the street and sewer municipal workers. The conventional microeconomic interpretation of the negative elasticities would be that workers were cashing in some of their increases in real hourly earnings for increased leisure time. An alternative interpretation, analogous to Whaples’ findings for the First World War, is that wage increases are proxying for temporary local labour shortages. If the negative coefficients reported in column (1) of Table 8.3 were simply a reflection of movements along a stable backward-bending labour supply schedule, then we would expect the wage elasticities to be identical whether wages were increasing or decreasing. As the separate upward and downward hours-wage elasticities in columns (2) and (3) show, however, this was not the case. Wage increases were associated with hours decreases for most workers, but the converse does not generally hold: hours were ‘upward sticky’ when the real wage fell.14 The elasticity of hours changes with respect to wage increases is significantly negative in almost all the occupations. But for no occupation is the elasticity with respect to wage decreases distinguishable from zero. Because the standard errors are rather large on the elasticities for wage decreases, the null hypothesis that the difference between the elasticities is non-zero can be rejected for only five of 177
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
178
HOURS OF WORK IN US LABOUR MARKETS
the ten occupations (the results of such a test are summarized in column (4) of Table 8.3). The asymmetry is weakest in the foundry and machine shop trades. Comparing columns (1) and (3) shows that the estimated elasticity with respect to wage increases tends to be larger than the elasticity estimated when the effects of wage increases and decreases are not separated out. This asymmetrical relationship between wages and hours is difficult to square with conventional economic theory but may reflect adjustment costs. Behavioural or cognitive biases might also have contributed to this pattern. For example, it is possible that workers expected monotonic improvements in such basic working conditions as regular hours while accepting with minimal protest occasional real wage cuts if they took the form of price increases without nominal reductions (see Kahneman et al. (1986) on the role of ‘money illusion’ in fairness norms). Indeed, there is an increasing body of evidence showing that workers at the turn of the century tended to oppose nominal wage reductions, with the result that wage adjustments also exhibited asymmetry over the cycle (Carter and Sutch 1991; Hanes 1990; Sundstrom 1992b). The association between wage increases and hours decreases is strong enough to account for a large proportion of the actual decline in average hours nationally. This is apparent from Table 8.4, which answers the following counterfactual question: by how much would the national average of weekly hours have changed between 1890 and 1903 as a result of the observed wage changes alone, given the estimated wage elasticities from columns (2) and (3) of Table 8.3? It should be emphasized that because the regression from which the elasticities were estimated includes separate year effects for each year after 1890, any trend in hours that is unexplained by wage movements has been removed. In the calculation underlying Table 8.4, only the wage-change coefficients and the actual wage changes were used. The last column of Table 8.4 indicates that the wage effect alone predicts more than 70 per cent of the decline in average hours for most of the occupations, and in fact overpredicts the decline in hours in several of them. The role of the asymmetry is worth noting: if a variable affecting hours, such as the wage, has irreversible effects, hours will tend to ratchet downwards over time even if the explanatory variable fluctuates randomly with no overall trend.15 If wages responded only sluggishly to local or national demand conditions, they would be an imperfect proxy for the temporary demand pressures that Whaples hypothesizes caused hours reductions during the tight wartime markets. Unfortunately, there are few alternative measures of demand pressure available. Using employment data from the 19th Annual Report, however, we were able to form two alternative proxies for market demand conditions at the city level based on deviations of employment from trend: one occupation specific and the other encompassing all the sampled occupations in the city. The occupation-specific proxy is the first difference 179
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
of the deviation of log employment from its trend for the specific occupation and city. Because the employment data for single occupations are based on reports from as few as two establishments in many cities, however, they are almost surely subject to considerable random noise reflecting the good or ill fortune of the businesses surveyed. Therefore we also calculated a second measure of market tightness that averages out the establishment-specific noise by taking the deviation from the trend in total city employment for the occupations in our data set.16 To examine the effects of our demand proxies and wage changes simultaneously, we added the demand proxies (each in a separate regression) to equation (8.2), allowing the wage effect to differ for increases and decreases. The alternative demand proxies add little information to the baseline model presented in Table 8.3.17 The coefficients on the demand proxies were usually negative, as expected, but they are quite small in magnitude. The hours-wage elasticities were not substantially different from those reported in columns (2) and (3) of Table 8.3. The small magnitude and weak statistical significance of our demand proxies may be a consequence of measurement error; indeed, it is plausible that the wage variable is a better proxy for local labour market tightness than the employment changes. 180
HOURS OF WORK IN US LABOUR MARKETS
While the predictions of the backward-bending labour supply and market pressure hypotheses focus on the relationship between wage and hours movements, the observable consequences of other explanations of the pace of hours decline involve the cross-occupational pattern of hours movements. One would expect hours movements to be correlated across occupations if only because many occupations experienced broadly similar movements in real wages, and as we have seen, real wages were inversely related to hours. A stronger test for spillover effects is to look for correlations in the residual changes in log hours, after controlling for the (asymmetric) wage effects and year dummies as in Table 8.3. The cross-occupational correlations in the residual are generally largest among the foundry and machine shop occupations (about 0.4). There is also a very high correlation (0.67) between the unexplained component of hours changes of bricklayers and hod carriers, which is hardly surprising given the complementarity of these tasks in construction work. In these cases it would appear that hours of work did indeed display the sort of public goods character that Shiells has suggested was important in the steel industry. It is possible that the smaller and more gradual reduction in hours of work among foundry and machine shop workers may reflect the greater difficulties confronted in coordinating hours choices among a heterogeneous workforce. On the other hand, it is unclear why the same coordination problems did not impede reductions in hours for hod carriers and bricklayers. There were fairly large (0.28–0.45) correlations between the residual hours changes of municipal labourers and the skilled building trades, a relationship consistent with a model of political pressures of the sort Friedman (1992) has suggested were important at this time. As Friedman has pointed out, the need for political machines to win support from segments of the working class was greatest in big cities, where wage labourers comprised a larger share of the electorate. It is thus possible that the relatively slower rate of urbanization in the south explains the smaller and more gradual reduction in hours in the building trades in that region. Looking only at contemporaneous correlations we cannot conclusively identify the direction of influence between occupations. To investigate this issue we used the regression framework of equation (8.2) to test for leading or lagging cross-occupational effects. If occupation B has a ‘causal’ influence on A, then the hours change in B should help predict the hours change in A when the lagged hours change in A is also a regressor. We ran 100 crossoccupational regressions of this type, including the asymmetric own-wage effects and year dummies as regressors. The results, not reported here, generally reveal few significant lagged cross-occupational effects. The one occupation that had a fairly strong spillover effect on several other occupations was the bricklayers. Decreases in bricklayers’ hours tended to lead decreases in the hours of carpenters, hod carriers, and foundry and machine shop labourers. Determining whether or not this pattern reflects a high rate of 181
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
union representation among bricklayers or some other factor requires further research. CONCLUSIONS Over the last century there has been a substantial decline in daily and weekly hours of work in the United States. While this decline is undoubtedly a reflection of the rising standard of living of American workers, this interpretation does not by itself explain the process by which hours reductions were achieved. Reductions in hours were the subject of acrimonious conflict between workers and employers. Understanding the process by which hours reductions were actually won holds the promise of new insights into the workings of the labour market and the ways in which changing economic conditions are reflected in the terms of the labour bargain. Economic historians have only recently begun to look carefully at the process of hours reduction, and most previous efforts have focused on the sharp decline in hours for primarily non-union workers that occurred during or after the First World War. Hours of work began to decline considerably earlier in a variety of unionized trades. Studying the disaggregated behaviour of wages and hours sheds considerable light on the validity of various explanations for the process of hours reductions around the turn of the century. The strongest and most interesting result obtained from our examination is the asymmetric relationship between changes in real wages and changes in hours. Wage increases were associated with hours reductions, but real wage cuts did not result in hours increases. This behaviour casts doubt on the view that short-run movements in working hours can be interpreted simply as movements along a backwardbending labour supply curve. However, it does appear consistent with the hypothesis that workers were able to use periods of temporary market tightness to win irreversible reductions in hours. We also found significant contemporaneous cross-occupational effects of hours changes among workers likely to have been employed together. Such coordinated hours changes appear to confirm that hours were a workplace public good, though it is not clear whether this gave rise to the significant problems of collective action that inhibited hours reductions in the steel industry. We also found evidence consistent with the view that political pressures may have influenced local labour market conditions in the building trades. Combined with the relatively high degree of organization among skilled building trades workers, such political pressures enabled workers in these occupations to win substantial reductions in working hours well in advance of most other workers.
182
HOURS OF WORK IN US LABOUR MARKETS
Notes The authors thank Jeremy Atack, David Green, Tom Weiss, Robert Whaples, participants in the McGill conference, and especially the editors of this volume for helpful comments on previous drafts. Rosenbloom’s research was supported by University of Kansas General Research Allocation #3282–20–0038. 1 2
3
4 5 6 7 8 9
10
11 12
Evidence on hours of work in the nineteenth century is surveyed in Atack and Bateman (1992). Kneisner (1976) traces the path of hours from 1900 to 1970. Such an explanation obviously must deal with the fact that the decline in hours of work ceased circa 1940 even though per capita incomes have continued to rise. Kneisner (1976) and Owen (1970) suggest a variety of other variables whose incorporation into the standard labour supply model help to account for the Lshaped pattern of hours decline from 1900 to the present. On the history of the shorter hours movement see Cahill (1932), and Roediger and Foner (1989). While labour leaders generally rationalized shorter hours as a means to winning higher wages, rank and file workers seem to have been more interested in escaping the drudgery and toil imposed by long hours (see Rodgers 1978:159). Whaples (1990b) provides an exhaustive review of the literature on this subject. This is consistent with Friedman’s (1992) argument that political machines in a number of big cities sought to accommodate building trades unions as a means of dividing organized labour and diffusing potential class conflicts. Several other industries (printing, saw and planing mill products) were represented in the data, but with observations for only a few cities. A much greater range of industries is represented in the 19th Annual Report at the regional level. Little information is lost by combining the regions because the year effects appear to be quite similar across regions. By assumption the time trend in the other regions is identical to the Northeast, though the level will differ. A similar pattern of occupational variation in hours of work is apparent in British data for the nineteenth century. There, hours reductions were confined to skilled building trades and engineering workers (Bienefeld 1972: ch. 4). Data for municipal workers in Britain are unfortunately not available. These conclusions come from an examination of all strikes ‘for a reduction of hours’ reported in the Third Annual Report of the Commissioner of Labor for the states of Illinois and New York in 1886 (US Department of the Interior 1888: Table I). Over half of all establishments in the United States affected by strikes in 1886 were located in one of these two states. For a reduction of hours’ was the single most common cause of strikes reported during that strike-prone year. The report does not generally state whether the strikers sought compensating hourly wage increases with their hours reductions, but such demands can be inferred by examining the recorded wages and hours before and after the strikes that were considered successful. Sundstrom and Rosenbloom (1993) use restrictions implied by a similar model of simultaneous wage and hours determination to test for evidence of labour market integration using data from the 19th Annual Report. The 19th Annual Report lists distributions of establishments working different hours patterns, by region. The decline in hours of work after 1897 tended to take the form of a changing mix of the basic standards; the building trades, for example, witnessed a rapid increase in establishments working eight-hour days, and a sharp decline in the number working nine-hour days. In some occupations— such as those in the foundries and machine shops—fractional daily hours and
183
JOSHUA L.ROSENBLOOM & WILLIAM A.SUNDSTROM
13
14
15
16
17
short time on Saturdays provided greater flexibility in weekly hours, but such patterns characterized a minority of establishments. That the benefits of such political support were not quickly dissipated through in-migration reflects the high degree of unionization in the building trades, which enabled big-city locals to establish artificial barriers to labour mobility. Sundstrom and Rosenbloom (1993) provide evidence that the markets for skilled building trades were less geographically integrated than those for other occupations. Of course, the strength of union organization was in turn influenced by the support of city hall in disputes with employers seeking to import lower-cost labour. The equations included a dummy variable equal to 1 if the change in the log of the wage was negative and 0 otherwise, which allowed the intercept to differ for wage decreases and non-decreases. The effect of this dummy variable was not statistically significant in any of the occupations. Results quite similar to those reported in columns (2) and (3) of Table 8.3 are obtained if the effects of positive wage changes are compared with those of non-positive wage changes. Workers’ preferences over hours of work might have depended as much on weekly earnings as hourly wages. Indeed, since wages at this time were often paid on a daily, not hourly, basis, some workers viewed hours reductions holding daily wages fixed as a means of increasing hourly wage rates. Unfortunately, the 19th Annual Report does not provide direct data on weekly earnings, so as a measure of earnings the best we can do is calculate the product of the average hourly wage and average weekly hours (which generally will not equal the average weekly earnings). The results of replacing the hourly wage rate in equation (8.2) with our measure of earnings generally reveal positive elasticities; that is, decreases in earnings were usually associated with cuts in hours. As hours were reduced, daily and weekly earnings generally fell, implying that workers did have to give up some potential earnings to win hours reductions. The relationship is again asymmetrical for many of the occupations, with rising weekly earnings having a smaller effect on hours than falling earnings. These results suggest that rising earnings were relatively seldom the consequence of rising hours, whereas falling earnings were more often associated with declining hours of work. If a positive deviation of employment growth from its trend is a reasonable proxy for labour market tightness, then our demand proxies should be correlated with wage growth, a hunch that is confirmed for several of the occupations by a simple regression of the change in log wages on the demand proxies. The results are therefore not reported here. We also estimated regressions substituting our demand proxies (separately for each proxy) for wage changes, both without and with year dummies. These generally show an inverse relationship between hours and demand movements for the building trades and foundry labourers, and a weakly positive or no relationship for the other occupations. In all cases the fit was quite poor, but the inverse relationship between demand and hours for at least some of the occupations reaffirms the hypothesis that during periods of market tightness (above-trend increases in employment) some workers achieved reductions in full-time hours. Interestingly, allowing the elasticity of hours changes with respect to demand changes to differ for increases and decreases in employment reveals no systematic evidence of asymmetries of the kind we found for the wage effect.
184
9 JOB TENURE IN THE GILDED AGE John A.James
The long-term relationship between many workers and their employers is one of the (perhaps the major) defining characteristics of the modern labour market. Because both parties recognize the lasting nature of the relationship, the time path of wages is influenced primarily by factors of mutual convenience (e.g. income smoothing or insurance) rather than the state of the labour market. Wages therefore are not particularly responsive to current economic conditions. The cyclical insensitivity of wages—their failure to adjust rapidly to restore labour market equilibrium—follows from these widespread permanent relationships between employers and workers. The modern labour market therefore represents a marked departure from the textbook auction or spot market. By the same token, these long-term arrangements also have important implications for the nature, structure, and distribution of unemployment.1 A major area of research in labour economics today involves investigation of the factors behind these employer-worker bonds and exploration of their effects. Enduring relationships between employers and workers are not necessarily intrinsic in labour markets. Indeed, they have been argued to be relatively recent developments in labour relations. Jacoby (1984, 1985) dates the development of bureaucratic employment practices and internal labour markets, policies designed to retain workers with the firm for the long term, to the period following the First World War. Even if it is fairly generally acknowledged that over the twentieth century the incidence of long-term employment relationships has increased, the question remains as to how large a change this was. Was the pre-First World War labour market essentially a spot market, implying that the changes in worker-employer relationships, and thus in the behaviour of wages and unemployment as well, were large and dramatic? Or was there a substantial component of long-term employment relationships even in nineteenth century labour markets, so that the changes over this century were by and large changes in degree rather than in kind? One needs, therefore, to examine the distribution of job tenure before the 185
JOHN A.JAMES
First World War, when the institutional and economic environment was presumably quite different, both for some insight into the operation of ‘premodern’ labour markets, and to be able to gauge the extent of the transformations in labour market practices over the twentieth century. Jacoby, for one, argues that before the First World War the employment relationship was one of only weak attachment on both sides. While he allows that there may have been a relatively small component of stable, quasi-permanent jobs among ‘skilled workers, typically employed in mediumto-large sized firms that operated on a year-round basis producing standardized commodities’ (Jacoby and Sharma 1992:171), the jobs of the bulk of manufacturing workers were quite unstable. Employment for most industrial workers was ‘unstable, unpredictable, and frequently unjust’ (Jacoby 1985:23)—as also were wages. The pre-First World War labour market therefore must have been one of substantial fluidity and flexibility, one in which wage and employment relationships were significantly more mutable than those in the contemporary labour market, and hence more responsive to current economic conditions. Although this is certainly a plausible and, indeed, persuasive characterization, virtually no direct systematic evidence on job tenure lengths in the period before the First World War has been offered in support of it. One piece of empirical evidence that might be consistent with such a picture is the high rates of labour turnover in the early 1900s.2 A substantial literature developed in the years after the First World War on the problem of labour mobility or ‘instability’ (e.g. Slichter 1919; Brissenden and Frankel 1922; Feldman 1925). Drawing on surveys of various firms and particular industries over the 1910s, these investigators found annual separation rates in excess of 100 per cent to be not uncommon. For the nineteenth century direct evidence on turnover is even more fragmented and anecdotal, although the degree of geographic mobility has been used as indirect evidence. Less than half of adult men living in Boston in a census year were still there ten years later, for example (Thernstrom 1973), and Boston does not appear to be atypical. The extraordinary level of geographic mobility throughout the nineteenth century has been taken as an indication that labour mobility and industrial turnover remained continuously high before the First World War (Jacoby 1983, 1985). Turnover, however, is an ambiguous indicator of the state of the labour market. An annual turnover rate of 100 per cent might imply that every job is vacated every year, but it is also consistent with stability in 90 per cent of the positions and the remaining 10 per cent being turned over ten times a year. Turnover statistics in themselves do not convey a clear picture of the degree of labour market fluidity. Indeed the contemporary American labour market is one with rapid labour turnover, but at the same time a substantial proportion of the workforce are in long-term relationships with employers or ‘lifetime jobs’ (Hall 1972, 1980, 1982; Baily 1983:31). 186
JOB TENURE IN THE GILDED AGE
Even if it is true that the labour market in the late nineteenth and early twentieth century experienced very high levels of labour turnover—and that is not completely obvious3—it does not necessarily follow that it was one of great fluidity and flexibility with rapid adjustments to disturbances. Just as high turnover rates today do not preclude widespread long-term employer-worker relationships, so also it might have been during the Gilded Age. Rapid turnover today is concentrated primarily among younger workers, many of whom eventually settle into lifetime or near lifetime jobs. The population in the nineteenth century was on average younger than today— in 1970 52.6 per cent was under 30 years of age, while in 1900 it was 62.6 per cent. It is possible that the greater proportion of younger workers with high turnover rates masked the experience of middle-aged and older workers with relatively long job tenures. Similarly, the inflow of substantial numbers of unskilled immigrants between the Civil War and the First World War may have boosted perceived turnover rates. About two-thirds of immigrants arriving between 1860 and 1910 were unskilled, many of whom were single or married with families still in the old country. Jacoby (1983:265–6) argues that they represented a particularly unstable component of the labour force because of their high emigration rates as well as their ‘pre-industrial work ethic’. Very high mobility rates may have been limited to these segments of the labour force, obscuring the more common stable relations. There are other reasons one might doubt the pervasiveness of high turnover rates and short job tenures in this period. Labour is not a homogeneous product. It would always seem that it would be wise (and profitable) for employers to try to hold onto good workers or valuable employees. The logic of this observation should not be limited to large firms with bureaucratic employment structures after 1920, but should apply to nineteenth century employers as well. Moreover, James (1992) has found industrial wages between 1840 and 1890 to have been remarkably cyclically unresponsive, a result that does not seem to fit well with the picture of a generally fluid and flexible labour market. Even if there were good data on labour turnover rates before the First World War, those alone might be misleading indicators of the nature and functioning of the labour market. The length and distribution of job tenure, in particular the prevalence of long-term employer-worker relationships, is an important key to understanding the character and operation of the labour market. Worker surveys conducted by state Bureaux of Labor Statistics in a number of states over the last two decades of the nineteenth century collected a wealth of individual data on worker characteristics, employment experience, and expenditure patterns.4 Several states collected information on years spent with present employer which can be used to calculate job tenure length. Carter has used a California study of 3493 workers done in 1891–2 to compute actual and expected lengths of job tenure and found a pattern of job tenure rather different than that implied by essentially a spot or auction market. 187
JOHN A.JAMES
While in an earlier paper she concludes that her results support the view that lengthy employment relations have become a more important feature of the labour market since 1892 (1988:297–8), in a later paper based on the same data she observes ‘brief job attachment does not describe the typical employment experience of workers in the late nineteenth century’ (Carter and Savoca 1990:14). Thus, the only direct evidence on job tenure in the period before the First World War appears to undermine, if not directly contradict, the picture of a very fluid labour market with primarily brief or transitory jobs. This paper, however, in turn has been severely criticized by Jacoby and Sharma (1992) for both methodological reasons and on the grounds that the observations on San Francisco workers were not representative of non-farm US labour markets in the 1890s. The distribution of job tenures, and by implication the nature of workeremployer relationships in the United States before the First World War, are empirical issues. Thus far, little convincing evidence on tenure lengths has been marshalled to examine this question. In this chapter I shall use ‘new’ data drawn from worker surveys in two other states to calculate patterns of actual and eventual job tenure lengths. The first sample is that of Michigan furniture workers collected in 1889, representing a major manufacturing industry in an important industrial state. The other is of Kansas workers collected in 1899 and 1900, representing a labour market in which manufacturing was not particularly important. This evidence on the job tenure experience of American workers offers us a perhaps unique insight into the functioning of the pre-war labour market. There are certainly no comparable British data, for example, for this period. JOB TENURE DATA The Michigan Bureau of Labor Statistics surveyed workers in the furniture industry in 1889 to gather ‘reliable information relating to the social conditions surrounding employees’. Special agents conducted personal interviews at the factory in seventy-eight firms in twenty-four cities.5 Usable responses were reported by 5419 individuals, 78 per cent of total employment in the canvassed firms and 64 per cent of total Michigan furniture industry employment as reported in the 1890 Census of Manufactures. The survey apparently covered most or all of the leading firms. Although the sample is certainly a sizeable slice of industry employment we do not know how it was chosen—i.e. if it was random, composed of those who showed up at work on a given day, or what. The sex of the worker is not reported, but it seems reasonable to assume virtually all the respondents were male. Define job tenure as the length of time an employee has worked steadily for the same employer, although not necessarily in the same occupation. The Bureau of Labor Statistics today considers employment to be terminated if the worker quits, is laid off for thirty days or more, is discharged, or enters 188
JOB TENURE IN THE GILDED AGE
the Armed Forces. In the Michigan sample mean job tenure is 2.98 years, with median tenure of 2.0 years. In 1978 median job tenure among American workers was 3.6 years (US Bureau of Labor Statistics 1980:49). The median job tenure length of Michigan furniture workers is at least of the same order of magnitude (low) as that of contemporary workers, and these figures overstate the true difference. The Michigan furniture industry, the third largest industry in the state by 1890, expanded rapidly after the Civil War and over the decade of the 1880s in particular. Between 1879 and 1889 alone Michigan furniture industry employment grew by 138 per cent. In such a fast growing industry, it is not surprising to find a large number of workers who have been on the job only relatively briefly—indeed, it follows necessarily. The raw data then understate the job tenure experience of manufacturing workers in general. I need therefore to adjust the number of workers in each length of job tenure category to offset the influence of rapid industry employment growth. In other words, I create a (crypto) steady state in 1889 assuming that furniture industry employment after 1869 grew at the average rate of increase of the US population over the mid- and late nineteenth century; 1869 is taken as the benchmark, the year in which workers with twenty years of experience in 1889 started. To adjust for the subsequent growth in employment I calculate the rate of industry employment growth in excess of population growth (from 1869) and then use the reciprocal of the compounded total to deflate the appropriate length of job tenure category—e.g. those with 15–20 years experience (category average—17.5 years) are multiplied by 0.9791, and so on. The assumption of a steady state also underlies Hall’s method of calculating eventual job tenure discussed in the next section. All of the Michigan results presented in this chapter are based on the adjusted data reflecting steady-state growth. The Kansas data are a pooling of worker survey results from 1899 and 1900 (Kansas Bureau of Labor and Industry 1900, 1901), with 1187 usable responses. The survey was done by mail, so that in spite of a $25-$ 100 fine for failure to respond, there is certainly a possibility of some self-selection bias, understating the more transient, less skilled, and less literate in the results. A separate tabulation of female workers was reported, so unless noted otherwise the Kansas results represent males. Kansas was chosen as a contrast to the Michigan furniture workers. By using a state in which manufacturing was not a major activity, I hope to capture some aspects of ‘pre-industrial’ employment relationships and hence get some sense of how they may have evolved over time in response to structural change. Occupations reported in the Kansas sample include a large representation from trade and services—teachers, barbers, butchers, office help, retail clerks, and so on—and little from manufacturing (leather workers, mill workers). In addition, there are 463 observations on railroad workers (of varying skill levels) and 179 on workers in the building trades. These data should give us some sense of employer-worker relations outside the factory. The growth of 189
JOHN A.JAMES
the Kansas male population did not differ too much from that for the United States as a whole over the postbellum period, so adjustments to reproduce steady-state growth (as described for Michigan) had little effect in this case. EVENTUAL JOB TENURE In the modern American labour market, as we have noted, the data on job tenure do not obviously suggest long-term employment is an important factor. In 1978, median job tenure among all workers was just 3.6 years, and only 9.5 per cent of workers had been in the same job for twenty or more years (Hall 1982:717). Such an inference, however, fails to take into account the fact that the observed workers have not yet completed their term of employment. Just counting the workers who as of now have twenty or more years on the job ignores the large numbers of middle-aged workers who are now working in jobs that will last twenty or more years. To get a more accurate picture of the prevalence of long-term employment, one must project the additional time a worker will spend on his current job. Then I may compute what Hall calls ‘eventual tenure’, the sum of actual tenure to date and the projected additional time in the job. The cross-sections of workers reporting age and time with present employer can be used for this. To calculate retention probabilities, (the probability that a worker will remain in his current job for five, ten, or whatever, years longer) I compare the number of workers in the different age-tenure categories. Thus, to calculate the probability that a worker aged 30–34 with five years on the job will remain there for ten more years we divide the number of workers aged 40–44 with fifteen years of job tenure by the number of workers aged 30–34 with five years of tenure (Hall 1982:718). Failure to take into account the non-uniformity of the age distribution will distort these results. Since there are fewer people in the 40–44 year cohort than in the 30–34 year one, the calculation described above in some sense understates the probability of remaining on the job ten years longer. To correct for this we multiply the calculated age-tenure ratios by the relative shares of the older and younger age categories in the US male population above age 15.6 The Michigan results are adjusted by figures from the 1890 US male age distribution, while the Kansas calculations are adjusted by the 1900 distribution.7 This adjustment implicitly corrects or allows for two major forms of job termination, death and emigration. It seems reasonable to exclude death as a reason of job leaving from our calculations—there is clearly an involuntary aspect for both parties! Emigration is a voluntary separation, but unfortunately it is difficult to separate these effects. All other forms of movements from jobs—movements into other jobs, retirement, or otherwise out of the labour force—are taken into account in the calculation of retention rates (Hall 1982:718–9). Carter (1988) makes several additional corrections to the calculated age-tenure ratios. If in-migration is positive, then age-tenure ratios understate the likelihood of remaining with an 190
JOB TENURE IN THE GILDED AGE
employer. While this may be a problem in principle, it is not probably a difficulty here because rates of population growth in Michigan and Kansas did not differ a lot from the US average over this period.8 She also goes on to correct (the foreign-born in-migrants) for out-migration (1988:294). Similarly, she corrects for those leaving wage labour for self-employment, and also for exits into retirement (by adjusting the labour force participation rates) (1988:293). Job termination by death is not so interesting here, so I allow for differences in age cohort sizes. But job terminations due to outmigration, a switch to self-employment, or retirement are clearly relevant to a study of the length of job tenure. We should be interested in job departures for whatever reason (except for death). By excluding these possibilities, Carter must overstate eventual job tenure lengths and hence the incidence of lengthy or ‘lifetime’ jobs in her sample of California workers. Table 9.1 presents estimates for Michigan furniture workers of the percentage of each age category that has or will achieve eventual job tenure of more than 1, 2, 5, 10, or 20 years.9 We see a number of points of similarity with contemporary data. As today, expected tenure rises as we move from younger classes to middle-aged ones. Only about 6 per cent of 15–19 year old workers, for example, could expect to be in their present job for more than 10 years, while almost 50 per cent of 50–54 year old workers could do so. Turnover is more concentrated among younger workers—about a third of 15–19 year olds have a job expected to last less than a year. At the same time, however, lengthy periods of employment appear fairly widespread. By middle age, almost half of the workers in the cohort achieved jobs with eventual tenure greater than ten years and about two-thirds or more were in jobs lasting at least five years. I do not have data on the causes of turnover in this period, much less on 191
JOHN A.JAMES
causes of separations by age. Appealing to figures from the 1910s, however, one sees that quits (voluntary separations) represented the great bulk of all separations, as they do today (Slichter 1919:85–9; Brissenden and Frankel 1922:78–81; Clark and Summers 1979:46–51). If this holds most strongly for younger workers, we see a pattern of labour market mobility in the late nineteenth century similar to that which Hall (1982) describes for the contemporary labour market. After trying a succession of jobs while young, many workers eventually find one that is a ‘good match’ between employer and worker and lasts for a reasonable period of time. Consider Table 9.2, showing the percentage of workers in each age group with jobs that have lasted under one year. The percentage of new workers in each age category clearly declines with age, as it does in 1978.10 Given the similarities between the patterns of eventual job tenure in the Gilded Age and today, we might ask whether high turnover rates among the young (who were a larger relative share of the population in the Gilded Age) then simply obscured widespread lengthy jobs among middle-aged and older workers similar in extent to today. The answer seems to be no. Comparing the percentage of workers in lifetime (twenty or more years) jobs by specific age cohorts in Table 9.1 with figures for 1978 (Hall 1982:723), we find the incidence of lifetime jobs among the 1889 Michigan workers to be substantially lower than today. By ages 25–29, while only 4 per cent of the 1889 workers had achieved lifetime jobs, 27 per cent of males in 1978 had. By ages 45–49 almost 15 per cent of the Michigan workers were in lifetime jobs, compared with 51 per cent in 1978.11 Similarly, turnover rates were higher for middle-aged and older workers in the 1889 sample than today. The percentages of new workers in all categories above age 30 were higher in the Gilded Age (from Table 9.2) than in 1978. 192
JOB TENURE IN THE GILDED AGE
One striking difference in the age-tenure profiles is in the experience of older workers with respect to long-term jobs. Hall finds the proportion of men in 1978 with eventual tenure of 20+ years to remain fairly steady (around one-half) from the early 40s to the late 60s. In contrast, consider Table 9.1 again. The higher eventual tenure categories peak around ages 45–54 and then decline significantly at higher ages. The prospect of being in a lifetime (20+ year) job, for example, about halves between ages 45–49 and ages 60– 64. The decline is less pronounced but still there for workers with 10+ years of eventual tenure. Ransom and Sutch (1986), noting the under-representation of older workers in industrial occupations (using various data sets including this one), attributed it to early retirement and also to downward occupational mobility. Their estimates of labour force participation rates for males in 1900 start to decline by the mid- to late 40s. After age 60 the participation rate falls to less than two-thirds (1986:12). Workers who lacked the resources to support full retirement may have continued to work, but often in a less skilled position. Can we attribute the rise in lengthy employment relationships between the Gilded Age and the present simply to structural or compositional shifts in the labour force? Although Hall focuses on ‘lifetime’ jobs, lasting twenty years or more, given the scarcity of such positions in the Michigan data, it seems reasonable to concentrate instead on the incidence of jobs lasting a fairly, rather than very, long time, say ten or more years. Table 9.3 presents the percentage of workers in each age category with eventual job tenures of ten or more years, classed by skill level, nativity, and size of firm. As a standard of comparison, in 1978, 68.8 per cent of males aged 40–44 and 75.2 per cent aged 50–54 would achieve or had achieved ten or more years of job tenure (US Bureau of Labor Statistics 1980). 193
JOHN A.JAMES
In Table 9.3 we see that while the prospects for eventual tenure of 10+ years increase with skill level—skilled workers aged 45–54 were about twice as likely to be in a job lasting ten or more years than were unskilled workers12— a shift in the composition of the workforce towards more skilled or semiskilled workers alone cannot account for the increased incidence of lengthy employment over the twentieth century. The proportion of skilled workers in 1889 with prospective tenure equal to or greater than ten years was substantially lower than that for all males in 1978. Similarly a decline in the share of unskilled, foreign-born workers in the labour force in itself cannot explain the increasing prevalence of lengthy employment relations. Again, in Table 9.3 the percentage of native-born workers with eventual tenures of 10+ years in 1889 was much below the 1978 benchmark figures. Finally, consider the influence of changing firm size. Suppose over time firms simply grew larger while retaining the structure of employment relations that existed in 1889. Could that account for the observed increased incidence of long-term jobs? The last column of Table 9.3 reports the percentage in each age class for firms with at least 200 people. Again we see the incidence of lengthy tenure to be substantially below our 1978 benchmarks. Structural or compositional changes in the workforce or among firms therefore cannot alone explain the increased importance of long-term employment. There must have been a change in employer-worker relations. Comparing the job tenure experience of Kansas workers in 1899 and 1900 (Table 9.4) with those for the Michigan furniture workers (Table 9.1), we see the proportion of Kansas workers in lengthy jobs (10+ or 20+ years eventual tenure) is substantially higher at every age. The percentage of workers in lifetime (20+ years eventual tenure) jobs was at least twice as high for the Kansas workers as for the Michigan ones. Table 9.5 reports the proportion of workers in each age class in jobs with eventual tenures of 10+ years by skill level, nativity, industry, and sex.13 194
JOB TENURE IN THE GILDED AGE
The longer tenures of many Kansas workers were not due to a higher proportion of skilled workers in the sample (curiously, probably due to the relatively small number of observations for the unskilled, a larger proportion of unskilled workers than skilled workers were in jobs lasting 10+ years). The presence of union members is not the reason either—for ages 30–60, a higher proportion of non-union than union workers were in jobs lasting more than ten years.14 Rather, the higher incidence of lengthy jobs in the Kansas data was due to the presence of railroad workers. With one exception, at least around 80 per cent of railroad workers between the ages of 25 and 65 were in jobs lasting over ten years—figures half again as high than those for all Kansas workers, which in turn were substantially greater than those for the Michigan furniture workers. Such results support Jacoby’s claim (1984) that the railroads were in the vanguard of bureaucratizing employment practices, rewarding seniority, and encouraging long-term tenure. In contrast, workers in the building trades appear to have experienced much shorter eventual tenures than average. Now compare the incidence of lengthy jobs among other workers in Kansas, primarily employed in trade and services, with a smattering in manufacturing, with that among the Michigan furniture workers (Table 9.3). Note that at the two ends of the age distribution, jobs lasting 10+ years were more widespread among the Kansas, primarily non-industrial, workers. Fewer workers in their 20s were in lengthy jobs in the Michigan furniture industry, implying a higher turnover rate among younger workers in manufacturing. While the percentage of workers in 10+ year jobs falls off after ages 50–54 in (Michigan) manufacturing it does not among the Kansas workers. Between the ages of 30 and 55, roughly middle age, the proportion of ‘other’ Kansas workers in lengthy jobs appears to have been in the same ball park as that among Michigan furniture workers. 195
JOHN A.JAMES
Taking the Michigan and Kansas ‘other’ workers as stylized representations of industrial and non-industrial workers, then we might infer (cautiously) that the growth of manufacturing over the last part of the nineteenth century should have had little impact on the prevalence of lengthy jobs among the non-agricultural workforce. Because employment patterns and practices might be influenced by current social conventions or standards of fairness, one should be cautious in using this cross-section (across states) relationship to infer how employment patterns may (or may not) have evolved over time. But the inference that the change in labour force composition may have had little influence on the prevalence of lengthy jobs, is at least consistent with the observed stability of the Phillips curve relationship over the mid- and late nineteenth century (James 1989). In the Kansas data there were only ninety-six female responses reported, so the problem of small sample sizes in some cells is particularly acute. Nevertheless, the pattern in the last column of Table 9.5 is a consistent one. Employed women were relatively young (none over 44), and thus their chance of being in a lengthy (10+ years) job was quite small. However, if we look at jobs that last at least five years, we find a surprisingly high proportion of younger women: 40 per cent of women aged 20–24, and 50 per cent of women 25–29 were in jobs lasting at least five years. This pattern of few women in lengthy jobs but substantial proportions of younger women in jobs of moderate length is also found by Carter in the 1892 California data (1988:296–8). The retention probabilities for 5–10 year jobs for women in their 20s were, as she points out, actually higher in the Gilded Age than in 1978. Our calculations indicate that a significant number of workers in the Gilded Age were in reasonably long-lasting jobs. The labour market does not seem to have been a simple spot or auction market based on constant and continual turnover. There must have been some ‘glue’ holding workers and employers together. Can we say anything about what the factors behind this bond were and how strong they were? Can this evidence on lengthy job tenure in the Gilded Age be explained by the presence of internal labour markets, i.e. systems of promotion and training of workers for skilled positions within the firm? After all, contrary to Jacoby (1985) who dates the development of internal labour markets in the 1920s, Sundstrom (1988) presents evidence that such schemes existed in a number of firms even by the turn of the century. It seems unlikely that internal labour markets (or promotion ladders) were responsible for the lengthy tenure observed among Michigan furniture workers. For one thing, while internal labour markets are usually associated with large firms, as Table 9.3 shows, the proportion of workers employed by large firms in relatively lengthy jobs was not dramatically higher than the average. 196
JOB TENURE IN THE GILDED AGE
Second, if there was an internal promotion ladder, we should observe workers, especially skilled and semi-skilled ones, with longer tenure with the firm than in their present occupation. Table 9.6 reports average years in occupation and with present employer by skill class and age category. In no case do average years with the firm exceed years in occupation. Similarly, Table 9.7 shows how few workers in each age-skill category had been longer with the firm than in their occupation. There seems to be little support for the idea that lengthy job tenure among Michigan furniture workers was due to internal promotion ladders encouraging the worker to stick with the firm.15 Indeed, the case for internal labour markets is overstated here because Tables 9.6 and 9.7 capture skill downgrading as well as upgrading. Was firm-specific human capital, often suggested as the important glue in twentieth century worker-employer relationships, the bond promoting lengthy employment relationships and reducing labour mobility? The results in Table 9.5 might suggest this. The building trades are probably the classic example of occupations in which skills are readily transferable from firm to firm or from job to job. In such trades human capital is general rather 197
JOHN A.JAMES
than firm specific. The proportions of Kansas building trade workers in lengthy jobs are generally quite low, compared with those for ‘other’ Kansas or Michigan furniture workers.16 Firm-specific human capital therefore may have been a factor promoting relatively long-term employment relationships in the late nineteenth century labour market. But how strong was the attachment between employer and worker? In the furniture industry at least, I would suspect not that strong.17 First of all, if firm-specific knowledge made workers particularly valuable to the firm, we might expect workers with longer tenure (and hence more firm-specific capital) to be more insulated from lay-offs. Michigan furniture workers reported the number of days of lost time during the year and the reason. I interpret days lost due to ‘no work’ as including time spent in unemployment between jobs, and as a result of selective lay-offs. 18 Table 9.8 presents the average number of work days lost over the year due to ‘no work’ or ‘laid off’ (a relatively small entry) classed by length of actual job tenure (years with present employer).19 The figures are somewhat higher for those with less than one year with their present employer, reflecting some mixture of time spent in unemployment between jobs in the past year and that which was the result of lay-offs. Among unskilled workers there seems to have been no consistent relationship between days lost per year and length of tenure with present employer. In contrast, among skilled and semiskilled workers, days lost due to lay-offs appear to decline as years with present employer rise. The lack of a relationship between job tenure and days laid off for unskilled workers, and an inverse one for the skilled and semi-skilled, is consistent with the existence of firmspecific capital. Robert Whaples (1990c) has estimated a wage equation based on these same data which shows years with present employer to have a positive effect on the daily wage. While there appears to have been a premium to seniority (presumably due to the acquisition of firm-specific capital), it seems to me to be quite small—the estimated coefficient is less than one-quarter than that 198
JOB TENURE IN THE GILDED AGE
for age. Table 9.8 is also not very persuasive evidence for firm-specific capital being a pervasive factor in influencing the length of employment relationships. Among skilled and semi-skilled workers, the pronounced decline in annual number of days lost due to no work sets in only after ten or fifteen years of tenure. For workers with less than ten years on the job, the influence of job tenure on lay-offs is negligible. Some workers must have been particularly valuable to the firm, as evidenced in their long tenures and low lay-off rates, but they were by no means typical. Consider Table 9.1 again, and note the decline in the proportion of older workers in jobs lasting 10+ years. We do not know for sure where these people went, but it seems likely that many of them were fired. Over the 50– 59 year old age span, the proportion of workers in 10+ year jobs seems to drop more sharply than the labour force participation rate (Ransom and Sutch 1986:12), suggesting that not all of them went into voluntary retirement. If firms made provision for their older workers by moving them into less strenuous jobs, there would not have been a decline in the proportion of older workers in lengthy jobs. Additional evidence against paternalistic downgrading but continuing employment of older workers may be seen in Table 9.7. The small percentage of unskilled older workers who had been with the firm longer than in their present job also suggests that downgrading within the firm was not widespread. The evidence suggests that while many middle-aged workers may have been in reasonably lengthy jobs, firm-specific capital was probably not generally important enough to provide a long-term bond resulting in a ‘lifetime’ job. Moreover, the implicit understanding between worker and employer usually did not go so far as to promise or provide continued employment in old age, even in a downgraded position. To the extent that 199
JOHN A.JAMES
older workers moved into less skilled positions (Ransom and Sutch 1986:26), it must have been most often after being discharged from a more skilled job.20 Note also that among Kansas ‘other’ (non-agricultural and primarily non-manufacturing) workers in Table 9.5, there is no sign of a fall-off in the prevalence of lengthy jobs in the older age classes. Perhaps in manufacturing, where jobs were more physically demanding, the decline in worker productivity past middle age overwhelmed whatever importance firm-specific capital or on-the-job training had. The accumulation of substantial firm-specific capital resulting in lifetime jobs seemed to apply to relatively few workers. The weak relationship between days lost due to lay-offs and length of job tenure for workers with less than ten or fifteen years on the job (Table 9.8) and the apparent shedding of older workers (Table 9.7) suggest that the accumulation of firm-specific capital was modest for many or most workers. CONCLUSIONS The pattern of job tenure in the United States before the First World War suggested by the surveys discussed here is squarely in the middle of the spectrum. In the middle, that is, between the views that there has been relatively little change in the incidence of long-term worker-employer relationships over the twentieth century (so that brief job attachments were atypical in the Gilded Age) and the belief that there essentially was a spot market in labour in this era, characterized by predominantly short job tenures and weak attachment on both sides. Relatively long-term relationships between employer and worker were not uncommon in the Gilded Age. By middle age, between a third and a half of Michigan furniture workers sampled in 1889 were in jobs lasting ten years or more and roughly two-thirds were in jobs lasting at least five years. In addition, the pattern of job tenure lengths by age is similar to that in the contemporary labour market. Turnover rates were higher and job tenures shorter for younger workers and then turnover decreased through middle age. Younger workers probably tried a succession of jobs before settling into one lasting a reasonable length of time. Just as high turnover masks the widespread presence of long-term employment relationships today, so also in the late nineteenth century, rapid mobility among young workers was consistent with many middle-aged workers being in lengthy jobs. Notwithstanding such parallels, one should be careful not to exaggerate the similarities between the labour market of the late nineteenth century and the modern one. In the Gilded Age, the labour market was in fact much more fluid and employment relationships more tenuous than today. Although a significant number of middle-aged workers made it into reasonably lengthy jobs, few achieved the ‘lifetime’ job status (20+ years) that is now quite widespread. Rather surprisingly, a larger percentage of young workers were 200
JOB TENURE IN THE GILDED AGE
in new jobs in 1978 than in 1889 (at least for Michigan furniture workers), but the comparison reverses after the early thirties. Many middle-aged workers in the Gilded Age never seemed to settle into stable employment.21 Even in their 30s and 40s, about one-quarter of the Michigan furniture workers were in jobs lasting less than two years. Akerlof and Main (1981:1007) calculate the average mean completed tenure of male workers in durable manufacturing on the job in 1968 as 18.3 years, while the mean eventual tenure for the 1889 Michigan furniture workers is estimated at 6.35 years.22 Moreover, the figures derived here based on a sample of workers are upper bounds to the incidence of lengthy jobs in the general labour force. If short-term jobs were filled by a succession of different workers, then I understate the proportion of short-tenure workers in the labour force by just observing who was employed at one time. The incidences of reasonably lengthy jobs for Michigan manufacturing (furniture) workers and Kansas ‘other’ workers, most of whom were not in manufacturing, were roughly similar. This suggests that the growing share of manufacturing in total employment over the later nineteenth century would have had little influence on the prevalence of long-term employment relationships in the labour force. On the other hand, in the twentieth century there seems to have been significant increases both in the length of jobs and in the prevalence of long-term jobs. These developments cannot be accounted for by structural or compositional changes in the workforce alone. They must have been the result of changes in employer-worker relations, an inference consistent with Jacoby’s emphasis on the rise of internal labour markets and the bureaucratizing of employment (1985). Alfred Marshall noted that the late nineteenth century labour market did not function like a spot market. He recognized the continuity of association between employee and employer in many jobs, which he attributed in effect to asset specificity (Matthews 1990). The relatively small proportion of Kansas building trade workers in 10+ year jobs would seem to support this view that firm-specific capital was the binding force in long-lasting employment relationships. While firm-specific capital may well have been important for the lifetime jobs achieved by a relatively small proportion of furniture industry workers, substantial accumulations of idiosyncratic capital there were probably not very widespread. Firms apparently did not try to protect employees with only moderately long tenures from lay-offs. Reasonably long-term jobs were primarily for middle-aged workers and most did not continue into old age. The value of relatively long-term workers to the (furniture manufacturing) firm seems to have been fairly limited. It would therefore be premature to attribute the fairly widespread incidence of reasonably long-term (as opposed to lifetime) jobs to the influence of firm-specific capital. Further research is necessary to determine the extent to which the lengthy jobs in the Gilded Age were in fact the product of such productivity— enhancing considerations. Okun (1981), for example, argued that long-term 201
JOHN A.JAMES
employment relationships could be in the interest of both employer and worker if substantial screening (for the firm) and search (for the worker) costs were incurred as a result of high turnover. But according to Jacoby, screening costs were minimal before the First World War (1985:17–18). If so, the attachment between most workers and their employer may have been the result of little more than ‘mutual familiarity’ or a reflection of the desire of middleaged workers to settle down. NOTES Financial support from the John Simon Guggenheim Foundation is gratefully acknowledged. I would also like to thank particularly Stanley Engerman, Mark Thomas, Robert Whaples, and my discussant, Michael Huberman, for their comments, and Phil Heap for excellent research assistance. As usual, however, all responsibility for what appears here must rest with the author. 1 2
3
4 5 6 7
8 9 10 11
The implications of these long-term employer-worker relationships are discussed in more detail in Hall (1980). It is difficult to determine the implications of Keyssar’s (1986) findings of substantial levels of unemployment among Massachusetts workers at the end of the nineteenth century for the nature of the labour market. While they are consistent with high turnover and short periods of employment, they could also be consistent with longer-term employment relationships punctuated by lay-offs. Using the surveys of the Metropolitan Life Insurance Co. (more comprehensive than those cited by Slichter or Brissenden and Frankel), I calculate the average monthly separation rate of production workers over the decade of the 1920s as 4.95 workers per 100 employees, a figure not out of the ball park from the 3.9 rate for the ‘average’ year of 1964 (US Bureau of the Census 1975:181–2). Strictly speaking these figures are not comparable because of differences in coverage, but they suggest nevertheless that the contrast between the earlier and the modern period may have been exaggerated. For more information on these surveys see Carter and Sutch (1989). More detailed information on the Michigan furniture workers survey may be found in Ransom and Sutch (1989). As noted earlier, I am constructing a steady state based on the average US population growth rate (reflected in the age distribution). Chi-squared goodness of fit tests could not reject the hypotheses that the age distribution of Michigan males in 1890 matched that of the US male population and that the Kansas male age distribution in 1900 was the same as that of the United States as a whole, so I used the US age distribution. To be definite about this we would need information on gross rather than net flows. It is possible that there may have been large inflows masked by large outflows. For purposes of comparison, similar figures calculated from the unadjusted data, that is uncorrected for industry growth, are given in the Appendix. Curiously, however, turnover among the younger groups, aged 15–29, appears to be higher in the modern labour market. This comparison of 1889 Michigan furniture workers with 1978 US male workers probably understates the true difference in the incidence of lifetime jobs. Job tenure among male workers in durable manufacturing in 1978 was longer than for all males. If it were possible to compare the job experience of 1889 furniture
202
JOB TENURE IN THE GILDED AGE
12 13
14 15
16
17 18
19
20
workers with 1978 furniture workers, we would find a more pronounced difference than that described in the text. Carter (1988:296) finds twenty year retention probabilities for non-union native-born men from the 1892 California survey to have been at least 24 per cent (which we have argued is an overestimate) and probabilities decline with age from the 15–19 cohort to the 40–49 cohort. Brissenden and Frankel (1922:72), among others, support the inference that unskilled labour was more mobile or ‘unstable’. Because of the relatively small number of observations, some of the cells in these more finely drawn tabulations are quite small. As a result there is more variability in the computed levels from age class to age class, although the patterns are still clear. Union status was not reported for the Michigan furniture workers, so I take them to have been non-union. While the existence of internal labour markets does not seem consistent with the evidence presented here for Michigan furniture workers, I do not claim they were unimportant everywhere. Gitelman (1974:75), for example, finds fairly substantial upgrading at the Waltham Watch Co. in the late nineteenth century. Such a comparison, of course, tells us at best only about the importance of firmspecific capital in one manufacturing industry, furniture, relative to the building trades. The importance of firm-specific capital varied substantially from industry to industry. Firm-specific capital may have been important in other industries of the late nineteenth century, but evidence is virtually non-existent. The Michigan survey distinguishes between days lost due to ‘plant shut down or closure’ and those due to ‘no work’, but since each worker reported only one reason for time lost there could have been a mingling of motives affecting the reported figures. Nevertheless, within each firm there is a dispersion of days lost reported by workers due to ‘no work’ and almost all firms had some workers who did not report any days lost due to ‘no work’. The observed distribution of days lost due to ‘no work’ across workers is not then simply the result of different rates of shutdown across firms. Rather, it seems reasonable to attribute it to selective lay-offs (among those continuously employed). Time lost reported due to factory being shut down or closed may be taken as representing general or unselective lay-offs. The lengths of time spent with present employer reported by workers represent interrupted spells of employment. Table 9.8 allows us to address an underlying, but as yet unstated, question of the comparability of our figures on the length of job tenure with modern data. The Bureau of Labor Statistics considers lay-offs only in excess of thirty days to represent a break in tenure (US Bureau of Labor Statistics 1980:48). The average number of days lost due to lack of work by Michigan furniture workers on the job at least a year generally ranged between ten and twenty days. We do not know how many spells of unemployment they suffered over the year, but even if it were only one, the experience of the average Michigan furniture worker would still certainly be in the ball park of being considered as having been steadily employed by the same employer under the current definition. We may therefore take the reported time with present employer to be comparable with contemporary data on job tenure. Ransom and Sutch (1986) are ambiguous on whether such changes were generally voluntary or not. While at one point they suggest that skill downgrading among older workers may have been the result of an involuntary loss of a more skilled job (1986:27), later they point to the demanding nature of industrial jobs and argue the plausibility of such changes having been voluntary (1986:30).
203
JOHN A.JAMES 21 22
James (1991) finds the probability of Massachusetts workers in the late nineteenth century leaving employment to have been substantially higher than today. See also Margo (1990b). Turnover may be calculated as the inverse of the average job length. These figures therefore imply annual rates of 5.5 separations per 100 employees for 1968 durable manufacuturing workers compared with 15.7 for the 1889 Michigan furniture workers. This is the appropriate comparison because the Akerlof and Main turnover estimate is inferred from those currently holding jobs (instead of observing actual separations) as is the figure for the Michigan workers.
204
10 THE GREAT WAR AND THE CANADIAN LABOUR MARKET Railway workers 1903–39 Mary MacKinnon
The First World War is in many ways a watershed in the history of labour markets. The war marked an end to mass emigration from Europe, the war years saw a sharp increase in trade union activity, and war-induced economic change required a rapid reallocation of workers across occupations and industries. Everyone agrees that labour markets functioned differently in the inter-war period than they had before the war, and researchers often ask whether and how labour market developments of the 1920s and 1930s raised unemployment rates during the depression (e.g. Eichengreen 1992:217). How wartime conditions and wartime developments influenced inter-war labour markets is unclear, because a lack of data makes it almost impossible to compare the pre-war, war, and interwar years. This chapter uses Canadian firm-level data to study changes in a very large firm’s workforce between the early 1900s and the late 1930s. I would like to know whether there were distinct eras in the types of workers and lengths of jobs, the ways the First World War marked a substantial change, and whether wartime developments were reversed soon after the war ended. Following Chapter 9’s focus on job duration in the late nineteenth century United States, I am particularly interested in changes in job durations and reasons for separations. The war may have marked a break in the composition of the labour force, so the second section here concentrates on workers’ ethnic and occupational characteristics. Canada experienced very high rates of immigration between 1900 and 1914, both from the UK and from continental Europe and the United States. Real GNP per capita rose by over 3 per cent per annum from 1900 to 1910, and both population growth and real income growth were greater in Canada than in the United States (Urquhart 1986:32). Settled territory expanded greatly: the population west of the Ontario-Manitoba border roughly trebled between 1900 and the outbreak of war (Mackintosh 1964:40–70; Safarian 1970:19–26). 205
MARY MACKINNON
In a rapidly growing economy, one would expect high job turnover, with many workers willing to quit in search of more favourable prospects. One can put a positive or negative interpretation on high quit rates as they may show that workers believed they could go from good to better jobs, or from bad to slightly less bad ones. Native-born English-speaking Canadians and British immigrants probably had the greatest access to skilled jobs (Reynolds 1935). French Canadians and European immigrants, especially those from southern and eastern Europe, were less likely to have, or be thought by employers to have, the necessary attributes to find highly paid employment. For unskilled work, however, many employers preferred non-Englishspeaking immigrants, who were believed to be willing to work hard and complain little (Avery 1979:25–32). During the First World War immigration virtually ceased, and about a third of the men of military age served in the Armed Forces (Kerr and Holdsworth 1990:Plate 26). Recruiting rates were much higher for Britishthan Canadian-born males. Francophone Canadians and immigrants from continental Europe were less likely to volunteer than native-born anglophones.1 In the later years of the war a severe labour shortage developed, and in response, one would expect to see both francophones and European immigrants moving into jobs previously dominated by anglophones. The immediate post-war years saw a short boom, a slump from 1920 to 1922, and then a return to growth. The initial reintegration of veterans into the civilian labour force went quite smoothly, although many veterans found themselves in great difficulty in 1920–2 (Morton and Wright 1987:115; Struthers 1983:25–7). Immigration increased substantially in the mid- and late 1920s. We have no idea whether the presumed labour market gains made by nonanglophones during the war were maintained in the 1920s. Overall, the depth and length of the Great Depression in Canada was broadly similar to the depression in the United States (Safarian 1970). In the 1930s, European immigrants, especially those who had arrived in the 1920s, suffered much higher unemployment rates than the British and Canadian born (Green and MacKinnon 1988). Woytinsky (1942) thought that there were three distinct phases in the employment relationship in the United States. Before and during the First World War, turnover rates were very high, and most job separations were voluntary.2 Therefore, employers could usually adjust the size of the workforce simply by changing hiring rates. He believed that turnover fell in the 1920s. While most separations were still quits, not lay-offs, there were far fewer ‘unstable’ workers filling a succession of short-term jobs. Quit rates rose only after hiring rates increased, which suggests that workers were becoming sensitive to cyclical changes in job openings. In the 1930s, workers were extremely reluctant to quit their jobs and firms had to use lay-offs as the main method of cutting the size of their workforce. 206
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
According to Woytinsky: This change in the distribution of separations by quits and lay-offs was of considerable importance…. Before the depression, workers available for employment were mainly persons who had quit of their own free will, and it is likely that among them were not only the unstable (and marginal) employees but also some of the most efficient workers. In the 1930’s, on the contrary, almost all workers whose employment was terminated were laid off, and very probably a selection was made by the management to assure that the most efficient workers were retained. This created a prejudice against the separated workers. (Woytinsky 1942:55) Woytinsky plays down the extent of change in the structure of labour markets just after the First World War—there were fewer footloose workers, but many were still willing to quit in search of better options. Relative to the pre-war years, fear of unemployment may have reduced mobility (1942:4). Jacoby contrasts the brisk labour market of 1916–20 with ‘labor market stagnation’ in manufacturing industry in the 1920s (1985:168). He sees the war years as an important transition in American industrial relations, with many large firms adopting policies designed to reduce turnover. In the 1920s, by contrast, firms had little incentive to introduce new personnel policies in part because the macroeconomic climate had changed and labour was in plentiful supply (1985:205). Woytinsky based his conclusions on evidence gathered in a variety of ways by different agencies at different times. Therefore, changes in the nature of the data, not the behaviour of workers and firms, may explain some of the contrasts he finds (1942:11–14, 30–2, 43–7). There is no Canadian evidence comparable with that Woytinsky used. The only way one can study hiring and separation in a Canadian context is to examine job durations and reasons for separation within a firm. To make a virtue out of a necessity, using a continuous time series for a firm ensures that observed changes are not due to differences in sampling. There are, however, two obvious problems with firm-level records. The experience of the workers at the firm in question may be atypical, and a firm’s characteristics change over time. By studying a firm, we learn about employment relationships over its life cycle, and we may confuse firmspecific with economy-wide developments. Very few firms created employee records in this period (and fewer have kept them), so we start with the knowledge that the firms we can study are unusual.3 Extrapolating from one firm to a large sector of the labour force is obviously risky, but when there is no aggregate information, such records provide a starting point. From 1910 to 1930, the Canadian Pacific Railway employed about 3 per cent of Canadian male non-agricultural wage earners.4 It holds summary employment records for virtually every worker present from 1903 on. Since 207
MARY MACKINNON
workers could move from place to place and be hired many times by the same firm, railways had to set up centralized records to keep track of disciplinary infractions—carelessness and disobedience could so easily spell disaster on a railway—as well as pension entitlements. The Canadian Pacific set up its system of staff record cards in 1903, when it instituted a pension plan (MacKinnon 1993b). I have constructed a sample of about 7500 workers employed by the CPR between 1903 and 1945.5 These workers began, or spent most of, their careers in the Mechanical Department, the department responsible for building, maintaining, and repairing locomotives, passenger coaches, and freight cars. While some men in the sample mainly worked on trains (as firemen, locomotive engineers, conductors, porters, and the like), most are effectively manufacturing workers—skilled men such as carpenters, boilermakers, and machinists, helpers to the tradesmen, and labourers and cleaners. The sample includes men from almost every CPR station in the country, but given the location of the railway repair shops, the largest concentrations of workers are in Montreal (Quebec) and Winnipeg (Manitoba). The CPR recorded the birth date and (from 1910) nationality of its employees, and listed each job held, with information on the occupation, location, wage rate, start and stop dates, and reason for leaving each job. Railways adopted bureaucratic employment policies very early. Given the large number of employees and their geographic dispersion, senior management had great difficulty controlling the hiring, promotion, and firing of workers. Workers, especially engineers and conductors, built up a great deal of firm-specific human capital, so that lengthy attachments were common, as James (Chapter 9) finds for Kansas railwaymen. Perhaps because they had many long-service employees, perhaps because they wanted more longservice employees, North American railways began to adopt senioritybased promotion and lay-off policies in the late nineteenth century, and these policies became more important over time. Unionization also began early, which again encouraged the use of seniority systems and at least rudimentary grievance procedures. Railways were the first companies in North America to set up pension plans covering manual workers (Craven and Traves 1980; Latimer 1932:21–34; Licht 1983; Mater 1940:397–404). For all these reasons, quit rates were probably lower, and job durations longer, for railwaymen than for virtually any other group of manual workers. In the steel industry (Heron 1988), where there was little (successful) unionization, and many unskilled jobs, qualitative and fragmentary quantitative evidence suggest much higher turnover rates than in the railways. Mechanical Department employees had less firm- and industry-specific human capital than did workers operating trains. Skilled workers were recruited from a variety of industries, and did not undergo on-the-job training in the way that prospective locomotive engineers and conductors did. At least in the early years of the century, seniority rules for tradesmen 208
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
were not as comprehensive as for workers in the running trades. Unskilled workers were never unionized. Therefore, job durations and reasons for separations in the CPR sample will be closer to what we would find for workers in heavy manufacturing than for a railway sample chosen from a wider variety of departments. EMPLOYMENT DURATIONS AT THE CPR, 1903–39 Figure 10.1 shows several measures of average employment durations for workers in the CPR sample. The top line (Total) shows the median length of time spent with the company by workers present on selected dates, and the second line (To date) months with the company to that date. This second measure is roughly what the Michigan and Kansas labour department surveys (Chapter 9) recorded; the first allows us to look into the future, as well as the past, in a way that contemporary surveys obviously could not. Both of these measures put heavy emphasis on the experience of longterm workers, who were most likely to be present on the sampled dates. The bottom line (Total new) on Figure 10.1 shows the median tenure of cohorts of new workers.6 Workers present in the years before the First World War spent an average of about ten years with the company, and at least a quarter stayed for about 209
MARY MACKINNON
thirty years. Given that CPR employment (At work) was rising rapidly in this period (Figure 10.2), it is not surprising that most workers were near the beginning of their tenure. By the beginning of the century, the company had built up a core of men who were already or would become long-service employees, but the majority of recruits worked for less than a year.7 The median total duration generally dropped somewhat when employment was expanding, and rose when the number of men at work stabilized or decreased. In 1915, total employment durations and employment durations to date increased dramatically. The obvious reason for this is that employment fell very sharply in 1914 and 1915. In 1914, the company hired only about a quarter as many new men (First hired) as in the boom years of 1912 and 1913. Even in 1915, when it had to replace men who had joined the Armed Forces, the number of new hires increased only modestly, although many former employees were rehired (Rehired) (Figure 10.2). Clearly, workers with greater seniority in 1914 were more likely to hold onto their jobs, but what is most interesting is that the men who were still at work in 1915 were far more likely to stay with the company for the rest of their working lives than was true for those present before the war. While in 1916, the recovery in employment cut job durations somewhat, 1915 is a clear break with the past. Thereafter, the typical worker spent at least 20–25 years with the company. There are further, but rather smaller, upward jumps in median time with the company in 1921–2 and 1930–2. By the 1930s, the median 210
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
worker spent over thirty years with the company, and the twenty-fifth percentile of the distribution was about twenty years.8 It is surprising that the break comes at the beginning of the war, not towards its end. One might expect that the upheavals of war would soon have affected many of the 1915 employees, but this was not the case. At the CPR, the shake-out of 1913–15 was extremely large9—indeed much greater than the first two years of the Great Depression. The men who left in 1914–15 were not, on the whole, going voluntarily (Table 10.1). As in the downturn of 1907–8, about half of all separations were lay-offs, and the men who were laid off were disproportionately fairly recent hires. 10 Only 6 per cent of the men leaving in 1914, and 11 per cent in 1915, went directly from the CPR into the Armed Forces. 11 The shake-out had little impact on the age, skill, or ethnic distribution of the workforce. During the war, railway wages lagged behind those of men in the same occupations in other industries. In the United States, high turnover of railway repair shop workers became a pressing concern (Railway Mechanical Engineer 1919:22–3; Hines 1928; Kerr 1968). At the CPR, median employment durations for new hires fell to as little as 5 months, from the pre-war norm of 7–8 months, but the overall workforce was apparently much less susceptible to the lure of high-paying job alternatives than were the recent recruits. It is possible that total time worked increased so much in 1915 because the men laid off in 1914–15 were poor matches. If so, the company rid itself quickly of men who would in any case have soon left. Given the very slight changes in workers’ characteristics between 1913 and 1915, this interpretation is implausible. It is more likely that the massive lay-offs frightened the remaining employees. Having survived the cuts (probably largely because they had been at work somewhat longer than most of the men who were let go), they realized (perhaps for the first time) the benefits of seniority. These men stayed with the CPR throughout the war, even though many could have moved to higher-paying employers. After the war, as we shall see, they had much more reason to stay put. It is clearer why job attachments should have increased around the end of the war, when railway workers in Canada and the United States gained big wage increases and reductions in hours of work (MacKinnon 1993a). Employment possibilities in other Canadian railways nearly disappeared at about the same time because virtually all competitors to the CPR were nationalized (the companies were bankrupt or operating at a substantial loss (Stevens 1973)). In consolidating these lines into one system, the new Canadian National Railways had surplus employees, and certainly no need to hire men away from the CPR. The immediate post-war boom was short lived, and there was another sharp depression in 1920–1. Despite all this, 211
MARY MACKINNON
212
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
average eventual job attachments rose only modestly between 1918 and 1922, and then stabilized during the 1920s at an average of almost thirty years. Technological improvements in the railway industry in the 1920s (notably the use of more powerful locomotives and all-steel passenger coaches and freight cars) reduced the amount of repair work, while motor vehicles were beginning to cut into the demand for rail services (Rountree 1936:137–60). The CPR sample (and total Canadian railway employment) fluctuate with business cycles over the decade, but have no general upward trend (Green 1986:800). In the early 1930s, the CPR cut its workforce drastically, and there was no recovery until after the Second World War broke out. Not surprisingly, average total time worked rose to almost thirty-five years for those lucky enough still to have a job. During the 1920s (except in the depression of 1921) the time the average worker had already spent with the company drifted up only gradually, but in the early 1930s, the combination of the virtual cessation of hiring, and the extensive use of lay-offs by inverse seniority, greatly increased the average time workers had been with the CPR (Figure 10.1). If the big jump in average total time worked had come around 1920, it would be impossible to speculate on the relative importance of companyand railway-specific, versus general macroeconomic, factors, because so much changed between 1918 and 1922. There were, however, no dramatic developments in the Canadian railways in 1914–15. The new transcontinentals were in precarious financial shape, but were still under construction and in business (Regehr 1976), and both the framework of industrial relations and the wage structure were static. Therefore, it seems reasonable to assume that the trade cycle caused the change, and to infer that what we see at the CPR was quite possibly common to many large firms. Canadian labour markets became much less fluid around the beginning, not after the end, of the First World War. During and just after the war, the ‘floating’ workforce was extremely mobile, but the job attachment of the ‘core’ workforce increased substantially. The post-war regime at the CPR differs from the earlier period, but Woytinsky’s observations about changes in the reasons for separations seem to have been drawn too starkly. As Table 10.1 and Figure 10.2 show, in all cyclical downturns, most separations were involuntary and the company cut back sharply on hiring, but lower quit rates did not persist. After the war layoffs accounted for a large fraction of departures every year, especially in 1921 and after 1928.12 Since the company’s growth had levelled off, hiring rates were always much lower after the war, but in 1921, and especially during the 1930s, the CPR virtually stopped hiring.13 Before the war, it had been possible to make big reductions in the size of the workforce without resorting to such draconian measures. The propensity to quit possibly fell more in the railway industry than elsewhere, because in the 1920s railways were relatively good employers, 213
MARY MACKINNON
and railway jobs were very scarce.14 In addition, the workforce was ageing. The average man in the sample was about 40 by the mid-1920s, and by the late 1930s, about 50, compared with about 30 in the decade before the war. This was well above the Canadian average for men in manufacturing and transportation in the inter-war years: 1921 Census data for these industries show the average man in his low 30s; by 1931, the average had crept up to about 35 (Dominion Bureau of Statistics 1929: Table 4; DBS 1936: Table 56).15 Since the company was no longer expanding and workers were reluctant to leave, the workforce aged, and given that the propensity to quit falls as the workforce ages, men in the sample were progressively less likely to resign.16 Aggregate US data suggest that in the 1920s about 70 per cent of separations were quits, but in the 1930s only about 20 per cent (Woytinsky 1942:36, 49). Counting resignations and leaves of absence as quits, my data show similar proportions for the 1930s, but quits were much less common in the 1920s. The firms providing Woytinsky’s data for the 1920s expanded very much more quickly than the US average, and this could account for part of the difference (1942:31). The only other Canadian evidence on reasons for departure comes from the aluminium works at Arvida, Quebec, which were only established in the mid-1920s and where business picked up sharply in the late 1930s. There, voluntary departures accounted for about half the total in the late 1920s and in some years of the late 1930s. While the CPR was not the typical Canadian employer in the 1920s and 1930s, neither was the Arvida plant, which was located in a new company town in a remote part of Quebec (Igartua 1987). It is thus certainly possible that for the large number of mature Canadian (and American) firms, a long-term change in the composition of departures occurred just after the end of the war. CHANGES IN THE WORKFORCE DURING THE FIRST WORLD WAR In Canada, as in most nations involved in the First World War, the later war years were characterized by extremely high demand for labour (Heron 1988:115; Thompson 1978:62). Wartime demand increased production in many branches of manufacturing and in agriculture (Mackintosh 1964:56–7; Safarian 1970:27–32). By 1916, the Armed Forces had drawn in very large numbers of men and the result was a serious labour shortage. Wartime conditions often require employers to hire workers from previously untapped labour sources (in terms of sex, race, age, background, and region). Sometimes, a temporary emergency sets up a long-term flow that permanently alters the characteristics of the labour force; in other cases, returning soldiers displace most of their temporary replacements. The changing composition of the workforce was probably partly responsible for the many complaints voiced in the later war years about decreased productivity (Railway 214
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
Mechanical Engineer 1919:22–3; Heron 1988:115). Workers failed to report for duty, and loafed on the job. There is no evidence that the CPR suffered particularly from these problems. Periods of peak labour demand can change the types of jobs many people hold; they can also change the structure of pay. Reder’s (1955) explanation of changes in occupational wage differentials relies heavily on US experience in the two world wars.17 According to Reder, in periods of very high labour demand, employers tend to substitute less skilled (or more generally, less desirable) workers for more highly skilled and desirable workers. At full employment, employers cannot increase the supply of labour for the least skilled jobs because they have reached the bottom of the ladder. Therefore, ‘rates on unskilled jobs are driven up relatively to those on semiskilled (which compete most keenly for labor with the unskilled jobs) and…this, in turn, results in a diminution in the spread between the rates on skilled and unskilled jobs’ (Reder 1955:838). Compression of wage differentials occurred at the end of the First World War in many countries, and Canada is no exception (Phelps Brown 1977:73–81). For the few Canadian occupations with wage data, and certainly in the CPR Mechanical Department, skill differentials between skilled and unskilled workers fell sharply (especially in the west) at the end of the war (Table 10.2). Reder argued that margins should decrease for semi-skilled relative to unskilled wages, but he presented evidence for skilled and unskilled wages (1955:841). At the CPR, helpers and labourers both improved their position relative to skilled workers, but helpers gained relative to labourers, especially in central Canada. If we accept Reder’s argument, the pattern of skill differentials does not suggest a critical shortage of unskilled, relative to semiskilled, employees. My data can be used to see how hiring and promotion standards were adjusted in a time of severe labour shortage. Women were one potential source of industrial workers during the war, and in Canada were used extensively in munitions production. Fragmentary quantitative information suggests that relatively few women found jobs in other heavy manufacturing firms (Ramkhalawansingh 1974:275–9). This was certainly the case at the CPR. Not quite 3 per cent of all workers in the Mechanical Department sample hired between 1915 and 1918 were women.18 Most worked as car cleaners, and a very few as labourers, seamstresses, and painters. Men and women in the same occupation were generally paid the same rate. Virtually all of the women were anglophones or francophones; there were no immigrants from eastern or southern Europe. A handful stayed with the company after the war, but, like male recruits at the same time, most worked for only a few months. Clearly, women were not an important source of replacement labour in the repair shops. There is no evidence that the CPR tried to hire men in small towns to work in the main railway workshops, or indeed that it took any active steps 215
MARY MACKINNON
to recruit workers.19 The other ways of dealing with the labour shortage were to use internal promotions to move existing workers from less skilled to more skilled jobs, and to lower hiring standards. I will first consider observable changes in the types of workers hired, and the jobs they were hired to do. Workers had to give information about their previous employment when they were hired, but this was not recorded on staff record cards.20 European immigrants, especially those from eastern and southern Europe, and perhaps French Canadians to some extent, would have been considered as lowerquality job applicants for the more skilled jobs. Therefore, an increase in the hiring of French Canadians and Europeans during the war may be a sign of relaxed hiring standards. However, even if ethnicity had been irrelevant for hiring decisions before the war, the proportion of French Canadians and Europeans hired would likely have risen somewhat in 1916–18. The Britishborn were more likely to volunteer for the Armed Forces, so that fewer of them could have been applying for work at the CPR. Unfortunately, one can only guess at changes in the total labour force, because we have only rough estimates of the ethnic composition of the Canadian Expeditionary Force (Brown and Loveridge 1982). The company could also substitute boys and older men for prime-aged workers. Boys lacked experience, judgement, and sometimes strength, while men over about 50 were liable to have health problems and be unable to do heavy work. That men in their 50s were looking for a new job may also have been a sign of poor previous employment records. Table 10.3 shows the characteristics of workers hired from 1903 to 1928. Not surprisingly, young and old workers were always most likely to be 216
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
hired for unskilled jobs, and least likely to be hired for skilled jobs.21 During the later years of the war, at all skill levels, the company hired substantially more boys and older men. This is consistent with Reder’s story about worker substitution, but the greatest increase in ‘undesirable’ workers was not among the unskilled, where he suggested the shortage should be most severe. In 1917–18, the company hired fewer skilled, and substantially more semiskilled, workers than it had before the war (part (B) of Table 10.3). It also hired slightly fewer unskilled workers. Given differential separation rates by skill group and possibilities for internal promotion, changes in the composition of the workforce show up more clearly in Table 10.4. The company maintained the proportion of skilled workers during the war, but allowed the proportion of unskilled workers to fall. From 1916 there was a dramatic decline in the hiring of British-born workers—the proportion was cut roughly in half (part (C) of Table 10.3). The Canadian born (both French- and English-speaking) approximately doubled their proportion among workers hired, while the European-born proportion was fairly stable.22 As soon as the war was over, the British were once again hired in large numbers. Overall, the Canadian-born percentage was maintained at roughly wartime levels, although anglophones appear to have done better than francophones.23 In the years just before the war, 40–50 per cent of semi-skilled and skilled hires, and about 30 per cent of unskilled hires, were British. In the later years of the war, the proportion dropped to about 20 per cent for all skill levels. After the war, unskilled hires were more likely to be British or other Canadian than had been the case before 1914. Skilled jobs were much more likely, and semi-skilled jobs somewhat more likely, to be given to French and other Canadians, so the British lost much of their dominance in filling vacancies at higher levels. The overall shift towards hiring unskilled workers masks these changes in the ethnic composition of each group hired. Unlike the contrast between the skill composition of new hires and that of the total workforce, the ethnic composition of the total workforce moved as one would expect from an examination of changes in hiring patterns. Far fewer anglophone workers were present during the war, but the proportion returned to its pre-war level by the early 1920s.24 Clearly, Europeans were squeezed out after the war. There was considerable opposition to hiring or retaining foreigners at a time when veterans were unemployed (Avery 1979:77). The CPR did not dismiss foreigners, but hired large numbers of veterans (who were rarely European immigrants). Another reason why the European share of the workforce fell in the 1920s was that immigration levels were much lower than in the years to 1914 (Urquhart and Buckley 1965: series A254, A337). Recent immigrants were likely to apply for work with the CPR. Most immigrants passed through Montreal or Winnipeg, where the CPR was an extremely important 217
MARY MACKINNON
218
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
employer, so reduced numbers of recent immigrants tended to cut the proportion of European applicants. Employers sometimes complained about immigrant workers’ short job durations, and it might be thought that the CPR moved away from hiring immigrants because it wanted a more stable workforce. Although recent scholarship emphasizes the prevalence of rapid turnover among European immigrants (Ramirez 1986), their employment durations were only a little shorter than average. Should one characterize Canadians as inferior workers, hired only because of wartime emergency? The British workman was highly regarded in Canada but it seems doubtful that the British born were considered greatly superior to native-born English-speaking Canadians, although British foremen might have preferred craftsmen with British training. While one might expect the British to have been preferred to French Canadians, many francophones held skilled jobs before the war. At this time, about half of all francophone CPR workers had skilled jobs, compared with about 40 per cent of anglophones, and about 20 per cent of Europeans. Roughly 20 per cent of all skilled jobs were held by francophones. Only jobs as foremen were and remained an anglophone preserve. At all times, about 90 per cent of foremen were anglophones. It is possible that the native born hired during the war were undesirable because they had little experience as industrial workers. The final way to fill vacancies was to rely on internal promotion. Table 10.5 shows the extent of occupational upgrading at the CPR. Columns (l)– (3) compare the occupational groups of workers present on two dates. For example, 48 per cent of workers in unskilled jobs in April 1905, who were still with the company in April 1907, were in semi-skilled jobs at the later date. Columns (4)–(6) show the proportion of workers in each skill group present at the first date who were also at the CPR two years later. For example, 61 per cent of workers in unskilled jobs in 1905 were also 219
MARY MACKINNON
employed in 1907. Not surprisingly, unskilled workers were always least likely to be with the company two years later. Those who did stay were so much more likely to be promoted to semi-skilled jobs that promotion probabilities (including the possibility of departure) were generally higher for unskilled than semi-skilled workers. Table 10.5 suggests that wartime experience was not unusual, relative to pre-war conditions. The proportions of workers reappearing in 1917 and 1918 were about the same as in pre-war expansions. Promotion possibilities for the workers who stayed were also about the same over 1915–17 and 1916–18 as in periods of growth before the war, and (for unskilled workers) were actually much lower over 1918–20. Chances of promotion were vastly greater before and during the war than after it, but much of the change occurred because workers were more likely to stay at the CPR, whether or not they were promoted. The extent of occupational upgrading was always sensitive to the business cycle. The effects of each depression show up clearly in Table 10.5. In 1913–15, 1920–2, and 1929–31, few of the workers who stayed were promoted, but only over 1913–15 was there a marked drop in the proportion of workers staying with the company. Can we put all this together, and is it consistent with Reder’s explanation for the compression of skill differentials? As already noted, the semi-skilled– unskilled differential did not fall. The skilled-unskilled differential declined mainly during 1919–21, after the period of labour shortage was over. From 1918 to 1922 wages were set by US awards, so we cannot expect shortterm changes in wage relativities, especially in this period of inflation and increased union strength, to reflect closely supply and demand conditions in Canada. 220
THE GREAT WAR AND THE CANADIAN LABOUR MARKET
Skilled workers may have been most frightened by the downturn of 1913– 15, and most aware of the benefits of staying with the CPR, which could explain the increase in their retention rates during the war. The higher retention rates meant that the company could maintain its skilled workforce fairly easily by using internal promotions (as it always had in expansions), and also by hiring more non-British and fewer prime-aged workers. Since some semi-skilled workers were moving up into skilled jobs, and more semi-skilled than skilled workers were leaving, a lot of semi-skilled workers had to be hired. Again, the company used internal promotion and hired a lower-quality group to maintain its semi-skilled workforce. To the extent that changing wage differentials tell us about relative shortages, it appears that semi-skilled workers were hardest to find and keep. At the bottom of the heap, it looks like the company decided to operate with fewer unskilled workers rather than scrape the bottom of the barrel. While the proportion of prime-aged adults and British workers hired fell, the company certainly did not have to rely entirely on what appear to be substandard workers. The shortage of British workers was mainly met by hiring more Canadians, not more Europeans. Fairly limited use was made of women for unskilled tasks. It may be that some semi-skilled workers in the later years of the war were actually performing tasks previously done by unskilled workers, and perhaps the company was able to alter its structure of production so that fewer unskilled workers were needed. CONCLUSIONS There were dramatic changes in employment relationships at the Canadian Pacific Railway around the time of the First World War, and it is likely these changes were common to other large firms. Obviously, firms experience high turnover when they grow quickly, regardless of overall macroeconomic conditions, but many of Canada’s other large employers also grew most rapidly from 1900 to 1914, which suggests the experience of the CPR was probably fairly typical. Hiring patterns changed considerably during the war, in the directions one would expect given the labour shortage. While skill margins declined, their pattern does not suggest that unskilled workers were relatively scarce. The decline in quits, and the end of the company’s expansion, meant that hiring rates were much lower in the 1920s than in the years up to 1920. The type of worker hired was also quite different than before the war, with a permanent increase in the share of the Canadian born.25 A decreased supply of recent British and European immigrants helps to explain this change. Contacts developed by Canadians during the war, which could later have been used to find jobs for friends and relatives, were probably also important. Finally, in the slacker labour market of the 1920s, there may have been greater reluctance to hire Europeans. 221
MARY MACKINNON
The depressions of 1913–15 and 1920–1 were extremely important. While neither was directly caused by the war, wartime conditions undoubtedly increased the extent of macroeconomic dislocation. During the earlier depression, job tenures of CPR employees rose dramatically and permanently, and from then on a large proportion of workers held lifetime jobs. Other industries and countries need to be studied to see how widespread was this break at the beginning of the war. The 1914–15 downturn may have been exceptionally severe in Canada, given the country’s dependence on capital imports and their sudden cessation at the outbreak of war. Increased job tenures help to explain the upsurge in union strength and activity around the end of the war. The recently enlarged core workforce, who had already spent several years with their present employer and likely intended to stay, would have felt they had a lot at stake in 1918–19, and therefore strong reason to press for better wages and working conditions. In the second depression, voluntary departures dropped dramatically and permanently. At the CPR, the 1920s were clearly a new era, and conditions looked a lot more like those of the 1930s than of the pre-war years. The CPR is not an example of gradual evolution of an employment regime. If the maturation of the firm and its workforce had been the main reason for the changes, most developments would have occurred over several years. Instead, when macroeconomic shocks affected the company, big adjustments took place within one or two years, and once made, these changes persisted. NOTES This research project is funded by the SSHRC and the McGill Faculty of Graduate Studies and Research. I thank Anna Alfano, Mark Antaki, Audrae Erickson, Catherine Jarmain, Valerie Jarvis, Alain Meere, Adrian Owens, Rebecca Toppa, and especially Louis Bouchard for their excellent research assistance. The staff of the CP Rail Archives and Pension Benefits departments, and of the Faculty of Arts Computer Laboratory, have been very helpful. Seminar participants at Northwestern, Illinois, Queen’s, Berkeley, Cornell, the Université de Montreal conference on incentives and contracts, and at the McGill conference have given many useful suggestions. 1 Conscription was in effect in 1917–18. 2 Both Keyssar (1986) and Jacoby (1985) stress the prevalence of dismissals and layoffs in the late nineteenth century. Jacoby views many quits as acts of desperation. 3 In Chapter 13, Whatley and Wright use a similar type of sample constructed from Ford Motor Co. records. Igartua (1987) has constructed a sample of worker records from the Arvida (Quebec) aluminium smelter. 4 Innis (1923) and Lamb (1977) are the standard histories of the CPR, Canada’s first transcontinental railway. The CPR was one of the two largest employers in Canada in this period. Nahuet (1984) discusses pre-war industrial relations and the extent of scientific management at the CPR. 5 All workers in the sample had left the company, or, if pensioned, had died by 1992.
222
THE GREAT WAR AND THE CANADIAN LABOUR MARKET 6
7
8
9
10
11 12
13
14
None of these is an estimate of continuous spells—while most workers have only one or two spells, a few have twenty or even forty. The nineteenth century American surveys, and modern surveys, generally look only at the current spell. I report median rather than mean rates because the presence of a few very longservice workers pulls the mean a long way above the median. The CPR began in 1881, and a few men in the sample had employment stretching back to the 1870s as they began with companies that were purchased by the CPR. The employment histories of workers present before 1903 were usually recorded only in summary form. It is not clear how this biases the results in Figure 10.1. While gaps in service were probably not noted for some workers, a good many start and stop dates are missing for the years before 1903, so that I cannot include all early spells in my estimates of total employment durations. Estimated job durations for the period after 1945 may be slightly inaccurate. Only a summary of jobs held after 1945 has been entered in the database. This means that a considerable number of leaves of absence, and a few lay-offs, have been omitted, which exaggerates job duration. However, the ends of a few records are missing, leading to underestimation in these cases. It is most unlikely that these omissions change median total time worked by more than a few months. The depression was probably more severe in western than central Canada (Thompson 1978:47–9), but conditions in Ontario were bad enough that the provincial government set up a Royal Commission to investigate the nature of unemployment (Struthers 1983:12–16). At the CPR, employment in Quebec fell much more than in Ontario or the west. Since we have no pre-war unemployment estimates, it is impossible to compare this depression with those of 1920–1 or 1929–39. The strike of August—October 1908 obscures the trend towards lay-offs in 1908, as many departures were caused by the strike. From January to July 1908, 56 per cent of departures were lay-offs, and 14 per cent were dismissals. Only eightytwo men were hired, and sixty-seven rehired in this seven-month period. Like many large firms, the CPR promised to re-employ those who joined up, and paid their wages to their families for at least six months (Railway Age Gazette 1917:1301–2). Very few workers were dismissed (for disciplinary infractions or incompetence) after 1914. This is consistent with increased effort by workers in 1914–15, tight labour markets which led foremen to overlook minor infractions from 1916 to 1919, and greater union involvement in disciplinary matters by 1918. In the 1920s, some workers listed as having been laid off had the notation ‘not to be rehired without further investigation’, or ‘unsatisfactory’ on their card. I suspect that in earlier years they would have been dismissed. The downturn of 1920 began in the summer. Only 30 per cent of departures from January to May 1920 were lay-offs. The company’s heavy reliance on rehiring former workers, especially in the 1930s, offers some support for Baily’s (1983:48) claim that during the depression, ‘layoffs followed by rehires (the contract theory or labor pool model) were very important relative to layoffs followed by search and labor mobility’. In the mid1930s, rehired workers averaged about ten years’ accumulated experience at the time of rehiring. Workers returning from a leave of absence are included as rehires, although their reappointment was virtually guaranteed. As noted in the introduction, railway jobs were always relatively ‘good’ jobs, so they may not have become relatively more desirable in the 1920s. They did, however, definitely become much scarcer.
223
MARY MACKINNON 15
16 17 18 19
20
21
22 23
24
25
Census data do not give us a very accurate guide to the average age of the labour force (for adults, ages are grouped into ten- or fifteen-year intervals). These estimates are for the gainfully occupied, not those employed on the census date. Unfortunately, there are no pre-war census data on workers’ ages. Voluntary departures fell in the 1920s, and especially in the 1930s, for workers of all ages and lengths of service. In Chapter 7, O’Rourke considers the Reder hypothesis in its geographic form as applied to Ireland and Britain. David (1987) incorporates the Reder hypothesis in his study of the Chicago labour market in the 1850s and 1860s. The CPR hired seamstresses and the very occasional female car cleaner before the First World War. In 1917, the company stated that mainly women ‘in special need’ had been hired, and that there had been no scarcity of women applying for jobs. Special facilities (including a dining hall and library) were provided for women in the Angus shops in Montreal, and a special uniform was designed for them (Railway Age Gazette 1917:1303–4). Very few application forms remain (these are for pensioners who have died since about 1970). Applicants had to give references, but my impression is that references were actually taken up only for apprentices, as these are the only letters of reference I have found. The company did not ask about formal education or technical qualifications. Most skilled workers were tradesmen, such as machinists, boilermakers, carpenters, and painters. Locomotive engineers and conductors are also classed as skilled men. Locomotive firemen, car repairers, machine operators, and mechanics’ helpers are the vast majority of semi-skilled workers. Most unskilled workers are labourers, cleaners, or watchmen. The definitions of skill levels are unavoidably somewhat arbitrary. I have assigned skill levels according to conditions prevailing at the beginning of the century. In this chapter, I concentrate on movements between categories. Some workers were promoted within a category, for example from freight carpenter to coach carpenter. Such promotions do not show up in Table 10.5. No Germans were hired, but some Austrians and others from the Austro-Hungarian empire did get jobs during the war. A lot of hiring was in the west in the 1920s, which reduced the French Canadian proportion. In the 1920s, the CPR did not always distinguish French Canadians from other Canadians, so some francophones are listed as other Canadians in this period. In the 1920s, some Canadian-born children of European immigrants were hired, and they are classified as other Canadians. I have to use language groups, rather than nationalities, when considering the ethnic composition of the workforce because nationality was not given for workers hired before 1910. I have inferred ethnicity from names. This is obviously a somewhat less accurate classification, but is the only possible method. Immigrants continued to make up a very large proportion of the Canadian labour force. From 1911 to 1931, about a third of male workers were born outside the country (Trade and Commerce 1915: xx; DBS 1929: lxvi; DBS 1936:310).
224
11 WAGE WOES IN WEIMAR? A partial investigation David Corbett
Between the end of the hyperinflation in 1923 and the onset of the Great Depression, the Weimar Republic experienced a brief period of relative prosperity. But while output grew at a reasonable if unsteady pace, the unemployment rate remained well above the level normal before the First World War. The situation began to worsen dramatically in 1929, as spreading joblessness seemed to bring political radicalism in its wake. By 1931 a quarter of the industrial labour force was out of work. Unemployment did not begin to fall until after the Nazi seizure of power, a coincidence that did much to win popular support for the Third Reich.1 Germany was of course not the only country to suffer severe unemployment in the inter-war years. The depression of the 1930s was a world-wide catastrophe, but Germany’s experience of high rates of joblessness even before the depression and the unusual speed of the Nazi recovery make her an important test case for competing macroeconomic theories. Germany’s inter-war unemployment fluctuations also have a special contemporary relevance. The labour market to which East Germans returned in 1989 has begun to look disturbingly like that of the Weimar Republic. The similarities extend beyond the common manifestation of high unemployment, for in both cases excessive real wages imposed by unions and the state stand accused of being a main causal culprit.2 During the 1920s and 1930s, movements in real wages figured prominently in explanations of Germany’s employment fluctuations. Businessmen, orthodox economists, and some Marxists argued that high and rising real wages were reducing investment, exports, and employment. Until well into the depression, Social Democrats, unionists, and a variety of academics countered with the claim that more generous pay for workers would stimulate both aggregate demand and output.3 Both sides in the dispute about wages shared a fundamental assumption: real wages were a causal force in their own right rather than an outcome of competitive pricing decisions. The Weimar Constitution of 1919 had included significant new protection of workers’ rights to collective bargaining, and a 225
DAVID CORBETT
series of laws and decrees in the early 1920s gave government arbitrators the ability to impose wage contracts on entire industries.4 Unions and the state—or so went the common claim—were setting real compensation levels without regard for labour supply and demand conditions.5 This chapter accepts the notion that German real wages were ‘politically determined’ during the inter-war years, and asks whether such wages could have had any of the serious consequences alleged by the contending sides of the Weimar wage debate. The first section briefly surveys recent discussions of the role of real wage movements in inter-war Germany. It introduces new evidence bolstering Knut Borchardt’s contention that rising real wages reduced investment between 1924 and 1932. The second and third sections turn to the real wage-employment nexus, and discuss several simple supply and demand models of the labour market which take the real wage as given. The rise in the real wage prior to 1931 is shown to have been a proximate cause of the rise in German unemployment and its subsequent fall a stimulus for the recovery.6 The countercyclical real wage behaviour underpinning this last conclusion is consistent with a broad range of economic theories, including both Keynes’ position as articulated in The General Theory (1936) and that of his ‘classical’ predecessors. But it undermines the relevance of the underconsumptionist theories and policy prescriptions favoured by the Social Democrats at the time. If real wages in Weimar were in fact largely ‘political’ and hence unresponsive to economic conditions, the evidence presented here suggests that unions and the state bear a significant share of the blame for the collapse of the German economy between 1928 and 1931. Conversely, the Nazi-enforced wage restraint after 1933 would appear to deserve a good deal of the credit for the rapidity of the German recovery.7 WAGES, PROFITS, AND INVESTMENT During the late 1920s, German businessmen raised a variety of objections to what they took to be excessively high real wages. Their main concern was the detrimental impact of high labour costs on investment.8 Economic historians writing in the immediate post-war period tended to ignore the issue, but Knut Borchardt revived it in a seminal article on the constraints facing German policy makers on the eve of the depression (Borchardt 1979, 1991). Borchardt’s argument was primarily focused on the impact of wage increases in the mid- to late 1920s on the distribution of income. He observed that in the Weimar Republic, labour was on average so expensive that incomes from entrepreneurial activity and property were relatively reduced…. In conjunction with the rise of other costs, the result was a 226
WAGE WOES IN WEIMAR?
squeeze of those kinds of income from which normally investments were financed. (Borchardt 1991:156) A low level of investment, in turn, helped slow German economic growth well before the country plunged into depression. Borchardt gave empirical support for his contentions by showing that between 1925 and 1931 the growth in output per worker had not kept pace with the growth in real hourly wages, and that as compared with its pre-First World War levels, investment was low in the late 1920s. Borchardt’s arguments prompted a number of responses. Most followed his lead in two ways: by focusing on the relationship between real wages and investment, and by emphasizing comparisons of wage and productivity growth. In his critique of Borchardt’s hypothesis, Carl-Ludwig Holtfrerich (1984) points out that it is inappropriate to compare output per worker with hourly wages, since the length of the average work week was not constant, but rather fell abruptly during the period in question. Albrecht Ritschl (1990) has since convincingly demonstrated that real wages did rise more rapidly than productivity in the mid- to late 1920s.9 Table 11.1 presents several new industrial unit labour cost series that bolster Ritschl’s findings. The series are based on annual rather than hourly magnitudes, and thus avoid possible distortions induced by errors in measuring hours. More importantly, they calculate labour costs with a variety of price deflators. Until now scholars have calculated German real wages and real industrial product using the consumer price index (CPI), but unit labour cost series should be based on the prices employers realize for their outputs. There are two available alternatives to the CPI: the wholesale price index (WPI) and an index of industrial finished goods prices (PIF), both of which result in a substantial upward revaluation of the estimated increase in unit labour costs between 1925 and 1931.10 Industrial unit labour costs thus appear to have been high in the late 1920s and early 1930s, but this in itself does not establish that a rising labour share—even if exogenously determined—reduced industry profits and consequently investment. Proponents of this chain of causality have generally been content to point out the low levels of the two last variables between 1924 and 1933 and have assumed that correlation implies causation. But did profits move inversely with labour costs within the period, and did profit rates and investment rise and fall together in the inter-war German economy? To explore these issues, consider some simple aggregative accounting relationships linking wages and profit rates. In accounting terms, the profit rate π/K is given by
227
DAVID CORBETT
Accordingly, the effect of a rise in unit labour costs on profits could be offset either by a decline in the materials share or by a fall in the capital-output ratio. There are strong negative correlations (from -0.57 to -0.87, with most close to -0.8) between the different measures of industrial unit labour cost defined in Table 11.1, and the rate of profit in industry between 1926 and 1938. This supports, although it does not prove, the contention that changes in labour costs affected profits.11 How were profits related to investment? A lack of relevant data again forces one to interpret the evidence of simple correlations. Using data from Maxine Sweezy and the Statistisches Jakrbuch, one can construct a crosssectional sixteen-industry sample of matched average profit and net investment rates over the late 1920s and early 1930s.12 Table 11.2 shows that the correlations between profit rates and investment are unifor mly positive. As 228
WAGE WOES IN WEIMAR?
with the correlations between labour costs and profits, this evidence is consistent with Borchardt’s hypothesis that exogenously determined wage levels had a deleterious impact on investment. To summarize, unit labour costs were high in Germany in the late 1920s and early 1930s. During the inter-war period as a whole, profits moved inversely with labour costs, and net investment rates by industry moved positively with profits. However, before concluding that high wages undermined the health of the Weimar economy, one must investigate the direct relationship between real wages and employment as well as the determinants of the wage level. The next section takes up the first of these issues. The second is not addressed here. REAL WAGES, EMPLOYMENT, AND LABOUR SUPPLY To those familiar with the literature on British real wages in the inter-war period, the focus of the recent German debate on unit labour costs and investment may be mildly surprising. Students of Britain’s inter-war unemployment problems tend to ignore the determinants of investment on the grounds that they are unquantifiable. Beenstock and Warburton (1986:165) are typical in holding that ‘In the absence of a complete econometric model for the U.K. in the inter-war period the indirect effects [of wages] on investment cannot be calculated.’ Instead of addressing the connection between wage levels and capital accumulation, the British literature has focused on the supposedly direct relationships between real wages, employment, and labour supply. Real wages and unemployment moved closely together during the Depression in Britain, and the German data display a similar coincidence. Could increases in the real wage have been a proximate cause of the sharp rise in German unemployment after 1928? And did the fall in real wages after 1932 play a key role in stimulating Germany’s economic recovery? Not surprisingly, many German businessmen answered the first question in the affirmative. The Verein deutscher Maschinenbau-Anstalten (the association of German machine building firms) attempted to prove the point 229
DAVID CORBETT
in a pamphlet published in 1931. The Verein presented a graph—reproduced here (Figure 11.1) with translated legends—showing wholesale prices of industrial finished goods, average contracted wage rates, wages divided by prices, and seasonally adjusted unemployment among union members (Verein…1931, no page). The latter two curves move in remarkably parallel fashion, evidence which the Verein took to support its claim that rising wages were causing unemployment. Of course such correlations do not constitute proof of causation. In order to construct a more convincing test of the possible effects of exogenous real wage movements, we must estimate structural models of both labour demand and labour supply. Unemployment, the prime variable of interest, emerges in such a model as the difference between the quantity of labour demanded and supplied at the prevailing wage. Readers should be forewarned that an econometric approach is seriously hampered in this context by the inadequacy of the data and especially by the very small number of available observations. However, explicit structural modelling offers a way of evaluating the simultaneous impacts of the different factors influencing the inter-war German labour market. Labour demand Assuming perfectly competitive product markets and that production can be described by a Cobb-Douglas function of labour, capital, and materials, profit maximizing labour demand takes the following form:
where L is employment or labour demand, K is the capital stock, W the wage, P the price of gross output, and Pm the price of materials inputs. This ‘basic neoclassical’ labour demand schedule assumes that output is endogenously determined and therefore does not belong in the labour demand function. If prices do not adjust to clear product markets, and if firms minimize the costs of producing a given level of output, labour demand becomes dependent on expected output, as shown in the ‘Keynesian’ labour demand function:
where Q* is expected output. Layard and Nickell (1985) show that if product market prices are flexible but competition imperfect, labour demand is described by a formula similar to equation (11.5). Tables 11.3–11.5 present econometric estimates of several variations of the neoclassical and ‘Keynesian’ labour demand equations, using annual data from inter-war Germany. Because real wages have been assumed to be 230
WAGE WOES IN WEIMAR?
exogenously determined, the estimates can be made using ordinary least squares. The presumed exogeneity of the real wage eliminates the most obvious simultaneity problems, while the presence of excess supply of labour at the prevailing wage makes it likely that observations of employment identify the labour demand curve. The most serious remaining econometric problem is that of serial correlation. Where static estimation results did not allow a clear acceptance of the null hypothesis of no positive first 231
DAVID CORBETT
order autocorrelation at the 5 per cent level, the same specifications were tried using the Cochrane—Orcutt correction or including a lagged dependent variable.13 Table 11.3 shows that the neoclassical specification provides strong evidence for a negative correlation between the real wage (adjusted for employers’ non-wage labour costs) and the quantity of labour demanded by industry as a whole. Significantly, the choice of wage deflator makes little difference under the assumption of perfect competition. One feature brought out by Table 11.3 is the deterioration in both the overall fit of the regressions and in the magnitude and significance of the real wage coefficients as the sample period is extended from 1925–36 to 1925–38; 1937 may have been an important turning point of the German industrial labour market. In that year unemployment reached its lowest level since 1925, but, in contrast to the earlier year, many of those who remained out of work were long-term unemployed. There were roughly five times as many registered vacancies outstanding in 1937 as in 1925, and contemporary observers noted that firms were increasingly resorting to spurious job reclassifications and hidden side-payments to evade wage regulations (see 232
WAGE WOES IN WEIMAR?
Mason 1977 and Hamburger 1940). All these signs suggest a switch to excess labour demand in industry in 1937 as German rearmament went into high gear. Such a switch would account for the decline in the magnitude and statistical significance of the coefficient measuring the effect of real wages on employment levels, and imply that the normal inter-war elasticity of labour demand is best estimated using the years 1925 to 1936. Table 11.4 presents the results from estimating the ‘Keynesian’ specification of labour demand for the industrial sector. The table suggests that aggregate demand had a powerful influence on employment. But real wages continue to be significantly negatively correlated with labour demand. Including output reduces the explanatory power of the capital stock—and to a lesser degree, that of relative materials prices—by more than it does that of real labour costs. Turning from industry to the total economy, we again find evidence that real wages (adjusted as always for employment taxes) were negatively correlated with the level of employment. Table 11.5 reports labour demand equations for both the neoclassical and the ‘Keynesian’ specifications of the total labour market. One final source of information on labour demand deserves a brief discussion. The regression results referred to so far are based on highly aggregated data. If overall employment was negatively correlated with the 233
DAVID CORBETT
real wage over time, one should also expect to find the same relationship within individual industries. Unfortunately the difficulty of finding appropriately matched employment, output, and price series limits the scope of industry-specific analysis. Table 11.6 presents the results of a variety of industry-specific estimations. The static results tend to confirm the negative relationship between real wages and employment, although the omission of input prices and capital stocks renders the specifications suspect and perhaps accounts for the evident serial correlation. When a time trend and lagged output are added to the explanatory variables, the Durbin-Watson statistics improve, while the coefficients on the real wage terms tend to decrease in size and in statistical significance. Taken as a whole, the results from regression analysis of inter-war German labour demand are similar to those found for Britain in the 1920s and 1930s by Beenstock and Warburton (1986) and Casson (1983). In both Germany and Britain the inter-war period was marked by strong countercyclical movements in the real wage. This result holds for Germany under both neoclassical and Keynesian specifications, and for all price deflators. If wages were determined by events originating outside the labour market, their movements played a role in pushing Germany into and out of the 234
WAGE WOES IN WEIMAR?
depression. But before we can evaluate the magnitude of such an effect, however, we need to examine the determinants of labour supply. Labour supply A complete evaluation of the relation between real wages and unemployment requires knowledge of the determinants of labour supply as well as of labour demand. Given the difficulty of defining or quantifying the industrial labour force, this section focuses on the total civilian labour force. Figure 11.2 shows that the number of persons working or seeking work rose and fell in step with the net returns to employment. The coincidence between the two series is roughly similar to the pattern for inter-war Britain. Econometric analysis should reveal whether this crude correlation persists when the effects of other factors influencing labour supply are taken into account. The total civilian labour force, L*, is simply the sum of employment 235
DAVID CORBETT
and unemployment, and may in general be thought to depend positively on the size of the population of working age and the real returns to employment net of taxes. If the persistence of unemployment causes discouraged workers to cease searching for work, the labour force will depend negatively on unemployment. On the other hand, rising rates of male unemployment may push the wives and children of the jobless into the labour market. From 1933 to 1938, the Nazis’ efforts to build up the Armed Forces and keep women at home may well have reduced the size of the civilian labour force. Summarizing in equation form, for inter-war Germany one might expect that
where the sign next to each independent variable shows the expected direction of its impact on L*.14 Table 11.7 reports the results from estimating a number of variants of equation (11.6) using annual data for inter-war Germany. In every regression the coefficient on the real earnings term is positive, and is statistically significant at the 5 per cent level. The size of the labour force thus responded positively to rates of compensation. The unemployment coefficients are uniformly negative, indicating that the ‘discouraged worker effect’ outweighed any tendency for male unemployment to encourage greater female participation. Positive estimated coefficients on the ratio of employment to working age population (TW/POP) in equations (11.1) and (11.3) are further evidence of the importance of discouragement. 15 236
WAGE WOES IN WEIMAR?
QUANTIFYING THE EFFECTS OF EXOGENOUS REAL WAGE MOVEMENTS While the regression results discussed in the previous two sections support the view that movements in the real wage alternately exacerbated and alleviated unemployment, they shed no direct light on the magnitude of such effects. Even if real wages were countercyclical, they may not have changed enough to contribute much to labour market fluctuations. If we assume that labour costs and incomes were exogenously determined by political factors, one can assess the consequences of the movement in wages using a simulation model and well-specified counterfactual real wage series. The simple model used here consists of a labour force equation, an employment equation, and an identity stipulating that unemployment is equal to the difference between the quantity of labour demanded and the quantity supplied at the given wage. If the overall German economy approximated perfectly competitive conditions, the appropriate model is16
If output was not endogenously determined, labour demand is more appropriately specified as 237
DAVID CORBETT
Figure 11.3 compares the actual total unemployment rate with the results of simulations of the neoclassical and Keynesian models using the historical paths of all the explanatory variables. Each of the simulated unemployment rates conforms well to the actual rate, although the simulated rate based on equation (11.7') is clearly superior in this regard. The simulation models can help answer two basic questions about real wage movements and the German depression. First, to what extent were increases in labour costs and the returns to employment prior to 1932 proximately responsible for the rise in unemployment? Second, how important was the modest decline of those two variables after 1932 to the recovery under the Nazis? To address these questions, both neoclassical and Keynesian variants of the simulation model were solved for the following counterfactual wage series. CF1: After 1928, real wages were maintained unchanged at their 1928 level. CF2: After 1928, real wages grew at their trend growth rate determined over the period 1925–38. CF3: After peaking in 1930–31, real wages were maintained unchanged at that level. CF4: After reaching their 1930–1 level, real wages continued to grow at the trend rate determined over the period 1925–1938. ‘Real wages’ is used here as shorthand for both hourly wage and non-wage 238
WAGE WOES IN WEIMAR?
labour costs to employers, which peaked in 1931, and net annual earnings of workers, which reached their high point in 1930. If the neoclassical specification of labour demand is correct, unemployment would have risen much less rapidly in the early years of the depression if both types of wage had been frozen at their 1928 levels. As reported in the column headed CF1 in Table 11.8, under this scenario unemployment would have been about the same in 1930 as in 1928. Tables 11.9 and 11.10 show the sources of this effect: compared with the simulated actual levels, freezing real wages and incomes would have saved 1,500,000 jobs and discouraged 750,000 people from joining the labour force by 1930. Restricting wage and income growth to their trend rates after 1928—as stipulated by CF2—would have reduced the unemployment rate in 1930 by more than a third. From 1931 on, the counterfactual unemployment rates found for CF1 and CF2 under the neoclassical labour demand specification are much closer to the simulated actual rate (the rate implied by CF2 surpasses the actual one in 1933). For CF2, this is partly due to the fact that the actual rates of growth of wages and incomes had fallen below trend by 1931, but another reason for the convergence is the decline in the capital stock that set in by 1931. If the real wage was truly exogenous, it may have caused a good deal of the shrinkage in the capital stock. Tables 11.8– 11.10 focus on the direct impact of wages on employment and the labour force, and hence underestimate their overall effects. While growth of wages above trend between 1928 and 1931 made a significant contribution to the rise of unemployment, it also appears that the decline in real wages (net incomes) after 1931 (1930) played a large role in the upturn. Assuming the neoclassical specification of labour demand is correct, the unemployment rate would have been almost five percentage points higher than it actually was in 1934 had real wages maintained their peak 239
DAVID CORBETT
values. If employers’ labour costs and workers’ net incomes had resumed growing at their trend rates after 1930–1, the unemployment rate would have been more than seven percentage points higher in 1934. Equation (11.7), the employment equation in the neoclassical specification, was chosen in large part because of its low, but reasonably well estimated, real wage coefficient. Its use understates the impact of real wage movements on employment. Since the real wage coefficient in (11.7') is slightly larger than that in (11.7), the Keynesian simulation results presented in Table 11.11 suggest even more scope for movements in the real wage to determine employment than do the neoclassical simulations. 240
WAGE WOES IN WEIMAR?
CONCLUSIONS A variety of contemporary observers of the German inter-war economy agreed that real labour costs and incomes were determined by ‘polities’ rather than the ‘market’. The supposed insulation of wages from the forces of economic competition justified viewing them as a major cause of cyclical fluctuations. Opinions sharply differed, however, over the direction of the impact of real wage movements. Those on the far left joined with conservatives and most orthodox economists in holding that wage increases beyond certain limits squeezed profits, reduced investment, and restricted employment. Through much of the same period, Social Democrats and unionists stuck by their classic reformist belief that increasing wages was in everyone’s interest. The evidence presented here discredits the old Social Democratic perspective. What data there are on unit labour costs, profits, and investment suggest that rising unit labour costs coincided with both a profit squeeze and a slowdown in capital accumulation in the late Weimar Republic. More convincingly, econometric analysis reveals that German real wages moved countercyclically during the inter-war period, a result only loosely dependent on the specification of the regression equation. Simple simulation models show that increases in real labour costs and incomes between 1928 and 1930 were at least proximately responsible for a large share of the concurrent rise in unemployment. Similarly, it can be shown that the moderation of wage growth after 1930–1 was an important proximate cause of the Nazi recovery. While the econometric and simulation results challenge the validity of underconsumptionist theories of the German depression, they also have a 241
DAVID CORBETT
limited bearing on the dispute between Keynesian and ‘classical’ interpretations of the Great Depression. Like his classical predecessors and their recently reinvigorated descendants, Keynes would have been quite comfortable with the finding of a countercyclical real wage. But if real labour costs and incomes were in fact more a product of power politics than of ‘normal’ market forces, their movements become final causes rather than merely proximate ones. Excessive wage increases, not insufficient aggregate demand, would bear the lion’s share of the responsibility for the onset of the German depression. And dictated wage restraints, begun under Bruening and continued by Hitler, would have to be credited with a large part of the recovery. DATA APPENDIX CPI: Index of consumer prices, 1913–14=100 (Bry 1960:465). IH: Hours worked in industry (Loelhoeffel 1974:135, col. 4). Excludes construction. INDUSTRY-SPECIFIC OUTPUT PRICES: The price deflators used in Table 11.6 are wholesale prices from the Konjunktur-Statistisches Handbuch (1936:99–107), or from the statistical supplements to the IfK Wochenberichten. IPROD: Index of industrial production from the IfK Wochenbericht, vol. 8, no. 24, p. 96, and from the ‘C’ supplements to later numbers. IW: Industrial employment (Hoffmann 1965:205–6, cols 2 plus 3). K: The commercial capital stock (Hoffmann 1965:254, col. 2). This excludes residential, agricultural, and transportation infrastructure capital. KT: The total non-residential capital stock (Hoffmann 1965:254, col. 7 minus col. 3). M: Index of materials prices, 1913=100. From the Konjunktur-Statistisches Handbuch (IfK 1936:101), and supplements to the IfK Wochenbericht. NAZI: dummy variable set to 0 for 1925–32, and to 1 for 1933–8. NETW: Net average earnings of dependent employees. Gross average annual earnings are from Hoffmann (1965:495, col. 9). These were then adjusted for income taxes and contributions to social insurance programmes without taking account of the fact that some non-industrial workers were exempt from contributing. NNP1: Net social product at factor costs, in 1913 prices (Hoffmann 1965:455). NNPPD1: An NNP price deflator, 1913=100 (Hoffmann 1965:601, col. 15). Available for the period 1925–38. PIF: Index of the prices of industrial finished goods, 1913=100. From the KonjunkturStatistisches Handbuch (IfK 1936) and supplements to the IfK Wochenbericht. POP: Population 15–65. Based on the census of 1925 and 1933 as reported in the Statistisches Jahrbuch (1935). TH: Total hours worked (Loelhoeffel 1974:135). TW: Total civilian employment (Hoffmann 1965:205–6). TLF: Total civilian labour force. The sum of the total number of unemployed (see ‘U’) and civilian employment (see ‘TW’). U: The unemployment rate as a percentage of the total labour force. The number of unemployed given by the 1933 census was extrapolated using the movements in the number of registered job seekers (Statistisches Jahrbuch 1935:13; Reichsarbeitsblatt 1925–38). The total labour force used as the denominator is described above as ‘TLF’. W: The nominal hourly wage and non-wage labour cost for industry (Loelhoeffel 1974:135). WPI: An index of wholesale prices, 1913=100 (Bry 1960:327).
242
WAGE WOES IN WEIMAR? WT: The nominal hourly wage and non-wage labour cost for the total economy (Loelhoeffel 1974:150).
NOTES This chapter is based on Chapter 2 of my Ph.D. thesis (‘Unemployment in Interwar Germany 1924–1938’, Harvard University, 1991). I would like to thank Jeffrey Williamson, Peter Temin, Bradford DeLong, and the editors of this volume for helpful comments. 1 2
3
4 5 6 7 8 9 10 11
12
13
Harold James (1986) provides a good overview of economic and political developments in Germany between 1924 and 1936. Sinn and Sinn (1992:29–31) provide an interesting comparison of the dimensions of the current economic collapse in eastern Germany with that of the depression of the 1930s. But while they see unjustifiable increases in real wages in the east as a major source of present difficulties, they do not comment on this possible similarity with the Weimar period. Tarnow (1929) gives a straightforward exposition of the Social Democratic ‘purchasing power’ theory. The opposing view of business interests is clearly articulated by the Reichsverband der deutschen Industrie (1929) and the Verein deutscher Maschinenbau-Anstalten (1931). Borchardt (1991:182, 257) discusses the arguments of Eugen Varga, a prominent Marxist of the Weimar years who thought rapid wage increases were jeopardizing capitalist growth. See Hartwich (1967) for a description of labour market institutions in the Weimar years. For evidence that this view was widely shared, see James (1986:190–2, 209–13). The methods used and the results obtained are similar to those presented by Casson (1983) and Beenstock and Warburton (1986) for inter-war Britain. This finding is consistent with Peter Temin’s assertion that ‘The fascist organized destruction of unions led to low wages and the rapid recovery of employment’ (Temin 1989:114). See Reichsverband der deutschen Industrie (1929). See Chapter 12 for a discussion of how different measures of wages and prices in Britain affect estimates of real wages in the inter-war years. The wholesale price index is heavily influenced by agricultural prices, and hence is probably not a good choice for constructing an industrial unit labour cost series. Another way rising labour costs might have hurt the German economy was by reducing exports. The fragmentary data available, however, do not allow a straightforward corroboration of the possibility: over the period 1925–35, real goods exports were positively correlated with unit labour costs. Maxine Sweezy (1940:390–1) reports annual profit rates for twenty-five different industries and sectors between 1926 and 1936. The Statistisches Jahrbuch (1934:509) makes possible the calculation of net fixed investment rates for some forty industries during two periods: 1924–8, and 1924–31. From these two groupings, I found sixteen industries that seemed sufficiently similarly defined to justify matching profit and investment rates. Unfortunately the periodization of the data does not allow one either to match investment and profit figures for exactly the same time period or to match investment with lagged profit. Experiments with more complicated dynamic structures led to no clear pattern of results, although the negative sign on the real wage term was generally robust to changes in the specification.
243
DAVID CORBETT 14
15
16
Although it is customary to use the unemployment rate to capture the discouraged worker effect, the ratio of employment to the population of working age might be more appropriate. The unemployment rate is itself influenced by the discouraged worker effect, whereas the employment ratio is not. Table 11.7 reports results for both specifications. I am grateful to David Green for suggesting the use of the employment ratio. POP and NAZI variables work less well, with the former being well estimated but of the wrong sign. Estimates for NAZI were close to (and not significantly different from) zero, and are not included in Table 11.7. POP works much like a time trend, as is confirmed by regression results not shown in Table 11.7. POP’s puzzling negative sign may in large part reflect the increasing impact of the Nazi policy measures alluded to above. The estimated coefficients come from Tables 11.5 and 11.7, and variables are defined in the Appendix.
244
12 WAGE BEHAVIOUR IN INTER-WAR BRITAIN A sceptical enquiry Mark Thomas
Were real wages too high in inter-war Britain? Was real wage behaviour at the core of Britain’s fundamental economic and social problem of the 1920s and 1930s—massive unemployment? Or was the real wage an innocent bystander while the British economy was assaulted by external forces that laid claim to jobs and profits? These are questions that actuated political economists in the 1920s and 1930s, and which still generate considerable controversy among economists and economic historians. This chapter is not about to resolve these disputes (although the author’s sympathies will undoubtedly surface). Its purpose is much more modest. It starts from the proposition that these are essentially empirical questions, that accurate answers depend on accurate information, and that too little attention has been given by most macroeconomic modellers to the strengths and weaknesses of the data they employ. In an earlier paper (Thomas 1988), I attempted to go behind the standard unemployment series, to see what could be learned about the operation of the inter-war British labour market by looking at its component parts. In this chapter, I extend this sceptical enquiry to the pattern of wages in the same period. Were wages as rigid as contemporaries claimed? Or will some statistical archaeology reveal an entirely different picture of wage behaviour from the accepted norm? Are the data series used in aggregate wage equations and labour demand functions representative of what was actually happening in the inter-war marketplace? And should any single index be preferred to the others? These are important questions because the interwar period, the years of Keynes and Pigou, were formative years in the evolution of the belief in rigid wages; a belief that shaped post-war policies towards the labour market, as well as academic thinking about how labour markets work. 245
MARK THOMAS
THE RECEIVED WISDOM The standard interpretation of nominal wages in inter-war Britain is that they were, for all intents and purposes, stationary for most of the period. Figure 12.1 presents the standard (Ramsbottom) index of average nominal wages received by (employed) British workers.1 Four phases are apparent: a plateau in 1920–1, followed by a precipitous decline through 1923, a slight downward drift during the rest of the 1920s and into the early 1930s, and a final phase of gradually rising wages after 1934. The stability of nominal wages as depicted here is indeed remarkable—wage rates varied by no more than 10 per cent between 1923 and 1938 (the coefficient of variation of the series is 0.03).2 Most political economists and commentators of the period were, however, more concerned with the behaviour of real wages. Precisely how real wages should be measured was a matter both of debate and of divergent practice. Jacques Rueff, who made the most important empirical contribution to the inter-war literature, measured the real wage by dividing the nominal index by a wholesale price index. Rueff demonstrated a remarkable consilience (the word choice is Pigou’s) of real wages with unemployment (as captured in Figure 12.2). Stamp (1925) calculated the correlation between the two series at 0.95 (for quarterly observations, 1919–25), and estimated that a 10 per cent increase in the real wage was associated with a 4 per cent rise in unemployment. Rueff argued that the high British unemployment of the 1920s was directly caused by increased real wages. His analysis was introduced into the British academic literature by Pigou (1927), albeit with
246
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
reservations. Pigou characterized Rueffs use of wholesale prices to deflate the nominal wage (without explanation) as ‘an unsatisfactory proceeding’, and substituted a retail price deflator (an adjustment, however, that made little difference to the overall pattern).3 More significantly, Pigou rejected the relevance of Rueffs argument for the crucial period of rising unemployment, 1919–21. Pigou argued that the post-war slump was demand driven, causing employment and prices to fall precipitously. The rise in the real wage was a result, not a cause, of the slump, and was due to the tendency of wages to lag behind prices (1927:358–9).4 Pigou did agree, however, that real wages had not fallen far enough after 1922. He estimated that unemployment was about 5 per cent above equilibrium (the insured unemployment rate stood at 9.3 per cent in mid-1927), concluding that: If wage-earners insist on maintaining a real rate of wages above the economic level…and if no mitigating action is taken by the State, an abnormal volume of unemployment, with all the material and moral waste that this implies, is the inevitable concomitant. (Pigou 1927:365–6) This general presumption was shared by other commentators in the 1920s. Beveridge, for example, in his monumental study of Unemployment, updated Rueffs chart with the languid comment that ‘it hardly seems reasonable…to doubt the connection [between real wages and unemployment]…any residuum of post-war unemployment which cannot be explained otherwise must be explained in this way’ (1930:371). However, in 247
MARK THOMAS
the wake of a further French contribution to the debate in 1931, Stamp cautioned against too easy acceptance of simple correlations between wages and unemployment, identifying ‘the use of alternative wage indices, [price] indices of different kinds of commodities differently weighted, lagging the wage changes, a critical search for statistical “traps” when comparing time series and dividing the indices by each other’ (letter to The Times 18 June 1931, cited in Stamp 1931). The pungency of Stamp’s warning is made all the more clear by the patterns revealed in Figure 12.3, which depicts the main real wage series for inter-war Britain. A comparison of these series demonstrates that the choice of deflator can matter a great deal. In particular, employing the wholesale price index (WPI) presents an entirely different picture of real wage behaviour from the other series. The wholesale price deflated series posts a 56 per cent increase for 1924–32, compared with 14 per cent for wages deflated by the price of final domestic output (FOD). The respective coefficients of variation are 0.17 and 0.06. By contrast, the choice between a national account deflator and the retail price index (RPI) makes very little difference. The GDP (GDPD) and final output deflators provide essentially identical series for the 1920s, and only a minor divergence during the 1930s; a retail price deflator produces a real wage that closely tracks the national accounts series,5 albeit at a higher level of real wages throughout the period (implying a terms of trade shift between 1913 and 1920). Are there good reasons to prefer one series over the rest? There are indeed. Since we are primarily interested in the demand for labour in relation to wages, it is an index of the producers’ (or own-product) wage that we 248
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
require. The retail price index is, of course, the appropriate deflator for investigating labour supply questions and issues of household welfare. For the calculation of own-product wages, however, we need a measurement of the price of the average worker’s output. And, for both theoretical and empirical reasons, the GDP price deflator is to be preferred over the wholesale price index. The various wholesale price indices available for the inter-war period have been criticized severely and widely. Roy Allen, in his discussion of the standard Board of Trade series, focused especially on the inadequacy of an index of manufactured goods that incorporated ‘practically no [price] quotations for finished manufactures’ (1948:146); Stafford (1951) emphasized the duplication implicit in the calculation of composites of inputs and outputs; Devons (1956:173) objected to the weighting scheme that ‘gave some very queer results’, and to certain other strictures that ‘tend to discredit the indices altogether’. To this catalogue we might add the failure of the Board of Trade to measure prices consistently across products;6 its use of a geometric mean, rather than the theoretically preferred arithmetic mean, to combine subindices; and the difficulties of working with two different series, spliced in 1930 but not truly comparable. The other wholesale price indices available for the period (which include the Sauerbeck-Statist and the Economist) are even weaker, and serve as little more than curiosities, tracking commodity prices continuously back to the mid-nineteenth century. The prices that are covered by these two indices are almost entirely composed of imported foods and raw materials, and bear little or no relation to the price movement of British-made goods. That these are significant problems is best demonstrated by comparing the behaviour of the various wholesale price indices with the one reliable source of industrial price data for the 1920s, the Census of Production. The 1930 Census (Board of Trade 1935) estimated the value of total physical output in 1924 at both current and constant (1930) prices. The implicit price deflator for aggregate output declined by 15.6 per cent between 1924 and 1930, compared with recorded price falls of 28.1 per cent for the Board of Trade index, and 30.2 per cent for the Sauerbeck-Statist series. The gap is actually even larger, however, since the Census price index overstates the true decline in industrial prices, being calculated on a gross output, rather than a net output (value-added) basis. The gross output valuation incorporates declining materials prices, including imported raw materials. It is superior to any of the wholesale price indices, given that it includes finished as well as intermediate products, and correctly weights imported materials. It is, however, inferior to an accurate index of value-added prices, which if it were available would exclude imported materials while retaining the appropriate weighting scheme for domestically produced goods (materials, intermediate and finished).7 Value-added prices are the theoretically preferred deflator for wages (Hill 249
MARK THOMAS
1971). From the perspective of the firm, however, it is unit labour costs that matter. In textbook parlance, workers are hired up to the point at which wages equal the physical product of the marginal worker. Employment depends on the relationship between the wage and productivity. Figure 12.4 shows the pattern of unit labour costs in British industry. Two series are shown—one based on the behaviour of wage rates relative to real productivity, the other using the average wage bill (i.e. wages and salaries divided by employment) as numerator. The patterns exposed in Figures 12.1–12.4 show how it has been possible for so many different things to have been said about wages in the inter-war labour market. For those pre-Keynesians (such as Rueff, but also Clay) who made use of the wholesale price deflated series, the complaints about the high and rising level of real wages, relative to 1913 experience, appear well motivated, if not well founded. Pigou’s argument to the Macmillan Committee that real wages behaved asymmetrically to excess demand and supply of labour is captured by the various real wage series, all of which exhibit very little downward drift after 1923, and a consistent upward trend during the recovery of the 1930s. Keynes’ rejection of the pre-Keynesian framework of rigid real wages (even if based on an unfair characterization) rests firmly on a comparison of the retail price deflated real wage and the nominal wage: Yet it might be a provisional assumption of a rigidity of money-wages, rather than of real wages, that would bring our theory nearest to the facts. For example, money-wages in Great Britain during the turmoil 250
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
and uncertainty and wide price fluctuations of the decade 1924–1934 were stable within a range of 6 per cent., while real wages fluctuated by more than 20 per cent. (Keynes 1936:276) Recent debates are similarly data dependent. Thus, the renaissance of the pre-Keynesian perspective as represented by the work of Michael Beenstock (see Beenstock et al. 1984; Beenstock and Warburton 1986), in which movements in the terms of trade (creating a rising wedge between consumption and own-product real wages between 1929 and 1933, and a falling wedge thereafter) are considered pivotal to the behaviour of unemployment,8 was originally based on the contrasting patterns of the wholesale price and retail price deflated wage series. The critics of this variant of the ‘new classicism’ (e.g. Dimsdale 1984; Dimsdale et al. 1989) instead use national account deflated real wages. This gives much less force to the behaviour of the wedge and, pari passu, much less force to the Beenstock argument. Indeed, a comparison of Figures 12.3 and 12.4 removes all remaining support for Beenstock’s argument—that rising labour costs caused the collapse in manufacturing employment, and that the recovery was stimulated by the halting and reversal of rising wage costs. The trends in unit labour costs, appropriately deflated, are gradually downwards throughout the period (with a minor hiccup in the strike year of 1926, when productivity fell sharply, and again in 1931). None of the drama of Figure 12.3 survives. Unit labour costs bear instead a strong family resemblance to the tranquillity (rigidity?) of the nominal wage series. Much then rests on the choice of data. And, as the next section will make clear, this warning applies as much to the numerator as the deflator in the real wage series. THE NOMINAL WAGE SERIES The natural resistance of wages to falling prices has been seriously increased, with a prolongation of economic disturbances. But Mr. Keynes exaggerates the extent of this resistance. Apparently he bases his statement that labour costs have not declined during the past 3 years on Professor Bowley’s Index of Wages. It only shows how fallacious such indices are. (F.Leith Ross, Economic Advisor to the Treasury, 1928, quoted by Clarke 1990:180).
There are significant problems with the standard wage data in this period. The criticism is not directed to the wage bill series developed by Chapman (1953) and Feinstein (1972) for use in the national accounts. Rather, it is directed at the wage rate data commonly employed in tests of competing macroeconomic models for the inter-war years, and in particular in tests of 251
MARK THOMAS
the ‘rigidity’ of the British labour market. There are good reasons to believe that wages were more flexible than the standard wage rate series implies. There are three major series of wage data for inter-war Britain. The most venerable is that of Arthur Bowley, the pioneer of so much statistical research in Britain. The original Bowley series, produced for 1914–24, was calculated as a simple average of the wage rates of eleven groups of operatives. This index was replaced in 1929 by a weighted average of the rates of twenty groups of wages chosen from selected occupations and regions thought to be representative of the entire labour market (in most cases, the rates refer to time workers, although some piece workers were also included). The other two wage series were produced by the Statistical Department of the Ministry of Labour—the first derived as a weighted average of wages in thirty-two occupations, the second, much more thorough, series based on ‘over 900 separate wage records’ covering sixtyfive industries (Ramsbottom 1935:693). The Ramsbottom data (depicted in Figure 12.1) are clearly the most thorough of all wage series, and have become the most regularly cited.9 The greater commitment to detail involved in the Ramsbottom index makes very little difference to the general impression of wage stability created by the original Bowley series; all three series are basically substitutable.10 It is unsurprising that the series conform so closely, since they were based on a common source—‘information systematically collected from the employers, or their organizations, and from the trade unions concerned, in all cases in which changes in recognized rates of wages come to the notice of the Ministry of Labour’ (Ramsbottom 1935:641). Inspection of the summaries of the principal changes reported to the Ministry (as reproduced in the monthly Gazette) indicates a number of drawbacks to these series. To begin with, the records concentrate on changes in basic, minimum, or standard rates of pay.11 However, by the 1920s most workers earned more than the basic rate. There were substantial margins for skill or productivity, margins that were sensitive to the state of demand. As an instance, in the engineering industry in October 1936, the margin between the full-time weekly wage of fitters and their negotiated minimum wage was greater than 42 per cent in Coventry and Derby, 27 per cent in Birmingham, and over 15 per cent in London and the Bedford/Luton labour market— among the most prosperous of labour markets—and below 5 per cent in the still depressed towns of Blackburn, Bolton, and Aberdeen. There were also significant margins for labourers—over 12 per cent in the three Midlands markets mentioned (Hart and Mackay 1975:44–8; Labour Gazette 1937). As a preliminary test of the impact of economic conditions on the margins in engineering, I ran a cross-section regression of the size of the margin in local negotiating districts in 1936 against the logarithm of the relevant local unemployment index. It was possible to match up data on margins (above local minimum wages) and unemployment (LUI) for fifteen local labour 252
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
market areas. Separate regressions were run for fitters (MARGINF) and labourers (MARGINL), with the following results (t-statistics in parentheses):
The equations support the hypothesis that margins were sensitive to economic conditions. This was especially the case for skilled workers. Some 86 per cent of the total variance in full-time weekly earnings of fitters was due to variations in margins; only 5 per cent can be accounted for by regional differences in minimum wages.12 The sensitivity of the margins to fluctuations in unemployment is quite marked: a rise in the local unemployment index from 8 to 23 per cent (roughly the increase in national unemployment between 1929 and 1931) is associated with fitters’ wages being lower by 12.6 per cent, and labourers’ wages by 5.4 per cent. Drawing inferences from cross-section to time series is, of course, notoriously tricky. Most obviously, factors other than local unemployment (such as cost of living differentials) affected local margins. Nonetheless, a comparison of the actual movement in the margins during the early years of the depression for fifteen local labour markets supports the general tenor of the cross-section results. Male unemployment in engineering increased nationally from 15.0 to 24.1 per cent between June 1930 and June 1933, while the (unweighted) average local unemployment rate for all workers in the fifteen localities increased from 14.2 to 27.4 per cent. During the same period, the average margin above basic wage fell from 20.2 to 14.9 per cent for fitters, and from 10.4 to 5.5 per cent for labourers. These figures are consistent with a fall in average wages of a little over 4 per cent for both groups. The margin shrank in all labour markets bar one (fitters in prosperous Bedfordshire, which had the lowest local unemployment rate of the sample); in Liverpool and Coventry, two of the most hard-hit towns, margin reductions effectively lowered wages by 10.1 and 11.2 per cent respectively. Although there is a need for further work with the engineering data and also with evidence drawn from other industries, it seems highly probable that the Ramsbottom data, by excluding one of the main mechanisms by which the labour market responded to changing circumstances, systematically understate wage flexibility. The overall scale of this distortion should not be exaggerated, however. For the entire labour force, a reasonable guesstimate would be that wages fell by a further percentage point or two across the deepest part of the Great Slump. 253
MARK THOMAS
The source data are, however, likely to be further contaminated by the inadequate treatment of piece rate wages in the Ramsbottom series. In particular, a wage series that was based solely on changes in the base rate of piece work pay would tend to distort the true behaviour of payment systems. Higher earnings caused by productivity improvements (which ought to be measured as an increase in the effective full-time wage) will be ignored by the wage series. If productivity gains were sterilized (partly or wholly) by some alteration of piece rates, the wage index would fall, while the ‘true’ effective wage rate could have risen, fallen, or remained constant. The extent of this difficulty cannot be estimated, but it should not be lightly dismissed. J.H.Richardson (1935:676) suggested that piece rates were the more flexible of the payments systems (since piece workers earn more of their pay through bonus schemes than their time card colleagues), and were more commonly applied than their weight in the wage index allowed. Certainly, the fragmentary evidence on this question supports Richardson. Roger Penn’s investigation of engineering wage structures in Rochdale indicates that piece rate earnings fell much further than time rate pay during the early depression (1971:200); at the same time, piece rate work was becoming more prevalent among the skilled labour force. 13 In 1938, about 25 per cent of industrial workers were paid on piece rates (Ainsworth 1949:44), while in certain sectors the proportion was much higher (e.g. textiles, 45 per cent, clothing, 35 per cent, building materials, 35 per cent. 14) Piece wage workers were also among the highest paid (average weekly earnings were 18.2 (13.7) per cent higher than for male (female) time workers, despite significantly lower hours). The most serious problem of all with the standard wage series is, however, one of omission rather than commission. Table 12.1 reproduces the statistics on the total number of wage changes reported to the Ministry of Labour, and also the estimated weekly change in the amount of wages paid. These lie at the heart of the Ramsbottom index. When the Ministry of Labour made their report on the volume of wage changes in a given period, they were careful to point out that the data represented ‘in the main those [wage changes] arranged between organised groups of employers and workpeople, and that many changes among unorganised workers, especially those affecting only employees of single firms’ were excluded (Labour Gazette 1940:104, emphasis added).15 The majority of reported wage changes was the result of collective agreements, either through direct negotiation between unions and employers, or through automatic adjustments (normally COLAs and selling price scales) established in earlier agreements. Collective agreements were not, however, the only form of institutionalized wagesetting. Workers in a wide variety of unorganized or poorly organized industries were covered by statutory wage legislation, such as the Trade 254
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
Boards, the Agricultural Wages (Regulation) Acts, and the Road Haulage Wages Act.16 Table 12.2 provides estimates of the number of workers covered by the collective and institutional arrangements.17 The data follow a distinctly elongated U-shape (see Figure 12.5). The main explanation for this pattern lies in the behaviour of union membership between 1920 and 1938. As is well known, one of the major casualties of the depressed state of the interwar labour market was the trade union movement, which lost some 4 million members (net)—or rather more than 47 per cent of its peak membership— between 1920 and 1933. Bain and Price’s estimate of trade union density falls from 45 per cent in 1920 to 22.6 per cent in 1933, gradually picking up thereafter as recovery got under way. There is also a hint of this general pattern in the coverage of the Trade Boards, reflecting the unequal incidence of unemployment on the low skilled. Overall, fewer workers were party to 255
MARK THOMAS
the formal institutional arrangements of the labour market in 1930 than a decade previously. Consequently, a smaller proportion of workers were covered by the wage change data compiled by the Ministry of Labour.18 But this is not the entire story—as Figure 12.5 makes clear, the drop in union membership is insufficient to explain the entire decline in the number of wage changes recorded—especially in the late 1920s. Figure 12.5 plots the number of wage changes reported by the Ministry of Labour alongside the number of workers covered by formal wage-fixing machinery. A strong, positive, if non-linear relationship emerges (the simple correlation between the two curves over 1920–38 is 0.58). For 1915–40, the entire period for which these data are available, I tested the relationship more formally by regressing NUMBER, the number of workers affected by wage changes (as a proportion of the total workforce) against several variables, including the logarithm of the density of collective coverage (In C), unemployment, inflation, and dummy variables to capture the effects of war (WAR is 1 if 1915–18, 0 otherwise), and COLAs (1 in 1921–5, 0.5 in 1920 256
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
and 1926–9, 0.25 otherwise). A standard result was (t-statistics in parentheses): No additional variables were appropriately signed and statistically significant in other variants of this regression. Clearly the number of individual wage changes was very sensitive to the institutional regime of wage determination. Although this equation makes no pretence to forecasting abilities, it seems reasonable to infer that in the absence of a declining density of union membership, the Ministry of Labour’s database would have been richer, and the resulting wage series would have shown greater movement over time.19 The Ramsbottom wage series reveals the behaviour of wages only for the (declining) segment of the labour force stored in the Ministry of Labour’s card file. The index is therefore biased in two different directions. First, it makes the doubtful assumption that the wage changes faced by unionized and uncovered workers were the same; second, it fails to correct for the shrinking denominator in the estimation of the average wage change. It is no easy matter to correct the wage data for these deficiencies. The fundamental problem is that we have no information on the behaviour of wages in the uncovered sector. We also lack information on the behaviour of skill and productivity margins over the business cycle, and adequate data on 257
MARK THOMAS
certain key sectors of the economy (notably the service sector, including domestics and those employed in distribution). The rest of this section will focus on matters of coverage. The preliminary task in correcting for inconsistent coverage is to allocate the recorded wage changes to the appropriate subpopulation. Figure 12.6 shows a first approximation at a wage series for the covered workforce, using the data presented in Tables 12.1 and 12.2.20 Nothing major comes of this—the new series shows only a little more movement than the Ramsbottom index. Between 1929 and 1933, for example, the Ramsbottom data post a 4.5 per cent decline, compared with a 6.5 per cent drop in my figure. The revisions to the standard data make the British labour market seem a little more flexible—but certainly nothing on the level of wage cuts after 1930 in the United States (Lebergott 1990; Mitchell 1986).21 All we have done thus far is to examine the wage behaviour of what, a priori, we would expect to be the most insensitive part of the labour market. It has often been noted that union wages tend to be more sticky than nonunion wages across the business cycle (e.g. Lewis 1963). This was clearly the case in the United States during the Great Depression, when the union mark-up rose dramatically in the light of differential responses of the average wage in the covered and uncovered sectors to the slump in labour demand (Pencavel and Hartsog 1984). Similarly, in their work on the period between 1967 and 1977, Moore and Raisian (1983) found that non-union wages were clearly the more sensitive to unemployment. We cannot assume, as the Ramsbottom index implicitly does, that the covered wage series is perfectly collinear with pay patterns in the rest of the economy. 258
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
The next stage then is to estimate the behaviour of wages for those excluded by the data. Since we lack any direct evidence on this part of the population, an indirect approach is required. The method used is derived from the crosssection procedure for extracting the union mark-up from standard undifferentiated wage data. This approach, pioneered by Lewis (1963) and first confronted with British data by Pencavel (1974), starts from the observation that, although the wage differential is hidden from view, the known wage (Wi) is a geometric average of the union and non-union wages. Thus
where Ui is the proportion of the industry’s labour force unionized. By rearrangement of (12.1), we have
where . Since (12.2) is an identity, and is not observable, the standard procedure is to define an estimating equation for ln , which captures the observable worker and industry characteristics that determine non-union wages. In this way, the identity is convened to an estimating equation that (at least in intent) fulfils the demands of statistical inference. The size of the union mark-up (ri) can then be derived from simple manipulation of the coefficient on Ui. This method provides an estimating procedure for determining the changing relationship between union and non-union wages over time, and in particular allows us to see whether union wages were stickier over the cycle. This allows us to test whether the standard wage indices used by historians and macroeconomists of the inter-war period understate the true flexibility of the British labour market. I have applied a form of this model to the results of the Ministry of Labour’s wage inquiries in the inter-war period (Table 12.3). The Ministry undertook four inquiries as a counterpart to the Censuses of Production, sponsored by the Board of Trade. These took place in 1924, 1928, 1931, and 1935 (all in October). (The results were published in various issues of the Labour Gazette 1926–7, 1929, 1933, 1937.) In addition, the Ministry of Labour undertook a fully fledged Wage Census in 1938 (Ainsworth 1949). In each case, estimates of the average wage by industry are provided in a (reasonably) consistent form, along with evidence on hours, which enables the construction of cross-sections of hourly earnings figures for over a hundred industries. The model estimated is a variant of that put forward by Pencavel (1974). The vector of characteristics chosen to represent the wage-setting process in the non-unionized world were the proportion of female employees (F), the proportion of young adults (21–25) (YADULT) and juveniles (14–20)
259
MARK THOMAS
(JUV), and the proportion of unskilled workers in the labour force (UNSK). Each of these variables was derived from the 1931 Population Census (Registrar-General 1934), and applied to each of the cross-section equations. I also incorporated the proportion of the industry labour force that lived in the (relatively prosperous and expensive) Southeast of England (SOUTHEAST), and the industry unemployment rate (UNEMP) to capture any differential cross-section effects of economic conditions. This vector was chosen as a parsimonious representation of the entire range of influences, a range that had to be severely curtailed given the small sample size available.22 The degree of disaggregation was determined by the availability of reliable data on union density at the industrial level. There are two series of union membership classified by industry for inter-war Britain, Burkitt (1974) and Bain and Price (1980). Burkitt’s series has the attraction of being very disaggregated, especially within certain key sectors (such as metals). Such disaggregation has its costs, however, and it is clear from a close inspection of Burkitt’s methods that there are a few too many (dangerously) heroic adjustments in his data to guarantee the accuracy that this model demands.23 Bain and Price’s data set is better motivated and more sensitively derived. It does provide estimates of union density (TU) for thirty-five sectors, but, given the wage data available, we have to rely on twenty-three observations. Table 12.3 sets out the OLS regression results. The results are very satisfactory. All of the variables were signed as expected, and almost all 260
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
were significant. Only the measure of the ‘unskilled’ nature of the industry systematically failed.24 Most valuably from our perspective, the estimate of the union mark-up was positive and significant. Moreover, the coefficient estimate on TU only moved within a narrow range from year to year. If we believe these results, then unions raised wages by somewhere around 25 per cent in inter-war Britain. Moreover, as Table 12.3 makes clear, the level of the mark-up was somewhat sensitive to the onset of the slump. There is virtually no change between 1924 and 1928, but the coefficient on TU in 1931 (while it would not pass a difference of means test) is markedly higher. The mark-up appears to decline as recovery gets under way—it is back to 1920s levels by 1935. These results may seem a thin reed upon which to place any weight of interpretation. There are numerous problems with drawing inferences from the cross-section procedure laid out above. There was, it should be said, no problem of parameter instability—changes in model specification and sample size made very little difference to the sizes or significance of any of the variables (and least of all to the mark-up estimate). But the difficulties of assuming exogeneity of unionism to changes in the mark-up are evident, as too is the lack of success of using 2SLS and other non-linear methods to get around this problem (Hirsch and Addison 1986:122–3). The results of Table 12.3 need to be treated with considerable caution. That said, they are the only estimates we have, and the only ones that we are liable to have for the union wage gap over time.25 It is worth, moreover, reiterating that the motivation for this study is frankly experimental—its purpose is to ask how data determined our priors are about inter-war labour markets, and whether there is good reason for doubting the implications of our current ‘best-guess’ data set. I shall, therefore, take the estimates of Table 12.3 as reliable, and apply them to the business at hand—which is making the transition from the wage series for the covered workforce (Figure 12.6) to an economy-wide series. This is achieved by manipulating equation (12.1) to produce
where Ni is the proportion of the labour force that is not unionized. Since we have only four data points, it is necessary to apply some form of interpolation between years. The value of ri was closely tracked by the square root of the unemployment rate, so this was used as a general interpolating device between observations. The annual series was then applied to the (revised) Ramsbottom series to produce our final wage index, reproduced as the ‘final series’ in Figure 12.6. This new series, as we would expect, shows rather more variation than the standard nominal wage index. The drop in nominal wages between 1929 and 1932 is not, however, especially large—rather less than 10 per cent. This is almost twice as great as a decline as that posted by the 261
MARK THOMAS
Ramsbottom index, yet it is less than the wage cuts experienced by Germany, Australia, the United States, Canada, Japan, and four other countries reporting to the International Labour Office in Geneva (Eichengreen and Hatton 1988:21). By international standards, at least, Britain could be construed to have rigid wages, even on the terms of the new index. As I have already noted, the intention of this section is not to forge a replacement wage series for Ramsbottom’s. The picture drawn in Figure 12.6 makes no pretence to be definitive, and is meant to be no more than suggestive. So what does it suggest? Oddly enough, it suggests that wages were actually even less flexible in inter-war Britain than we have previously thought! For, although the revised index shows rather more of a decline in wage rates during the depression, it also indicates a much smaller increase in wages up to 1936. The responsiveness of wages to excess supply (as reflected in a rising unemployment level) is, therefore, marginally increased, while their responsiveness to excess demand (as reflected in a falling unemployment rate) is lowered.26 Since it is the recovery period that dominates the overall picture of upwardly flexible wages, any reduction in the rate of wage growth after 1932 will dilute the appearance of wage flexibility. The revised pattern of real wage behaviour indicates a small fall in wages between 1929 and 1933 (9.7 per cent, while unemployment more than doubled), and a slow recovery in wages between 1933 and 1937 (8 per cent, while unemployment was halved). The source of the flatter response of wages during recovery is twofold: the denominator of the Ramsbottom index increases after 1933 as union membership reversed its downward trend, and the union/non-union mark-up shrank after 1931. Is it possible that this conclusion is misplaced, the result perhaps of placing too much emphasis on a single, aggregate figure? Beveridge stated that ‘it comes out clearly…that different industries have fared [in respect of real wages] very differently’. He noted the curious implication of Ramsbottom’s data that, ‘the industries which have maintained wages are also those with the least unemployment, and…the industries in which wages have fallen most are also those of most unemployment’ (1935:674).27 Is wage flexibility buried beneath a smooth surface? A DISAGGREGATED APPROACH In one of the few analyses of the dispersion of wages across industries for this period, Georges Rottier concluded that ‘the stability of the general level of wages…concealed numerous changes in the relative levels of wage rates and earnings in different industries’ (1957:239). Rottier worked from the results of the Ministry of Labour wage censuses referred to above. Consequently his analysis refers to the behaviour of earnings rather than wage rates. He found that average earnings for all workers fell slightly between 1924 and 1931—by less than 0.8 per cent for men, by about 3.2 per 262
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
cent for women—and rose sharply between 1931 and 1935–17 per cent for men, 15 per cent for women. But more significantly, his data show considerable variation in earnings movements across industries, especially between 1924 and 1931. The median dispersion (the ratio of the interquartile range to the median) of average industrial earnings for male (female) workers rose from 0.11 (0.07) in 1924 to 0.25 (0.12) in 1931.28 In his analysis of the sources of the increased dispersion, Rottier distinguished three groups of industries: the staple sectors (coal, woollen and cotton textiles, shipbuilding, general engineering), the ‘new industries’ (chemicals, autos, non-ferrous metals, electrical engineering, building), and all others. Table 12.4 sets out the median earnings and the interquartile range (IQR) for each industrial category. The increased dispersion between 1924 and 1931 was caused primarily by an increase in intergroup variability, originating in the sharp decline in staple earnings. There was some added dispersion created by earnings behaviour within the categories (although earnings in the residual category converged in the 1920s), but this was a secondary factor. During the period between 1931 and 1935, in contrast, there was a marked increase in intragroup earnings dispersion, on top of a continued differential between wage movements in the staple and non-staple sectors. Apparently, recovery intensified the divergence in earnings behaviour from industry to industry. The use of earnings figures is very helpful, in the light of my criticisms of the Ramsbottom index. The Ministry of Labour’s surveys incorporated nonunion as well as union members; the earnings figures reported include bonuses and margins, the impact of piece rate changes, and so on. The 263
MARK THOMAS
clearest problem involved in working from earnings data is that they are current weighted rather than base weighted; they are therefore contaminated by compositional effects, such as differential lay-offs across skill, gender, and age categories. A fall in average earnings may be caused not by downward flexibility in pay, but by a reduction in the proportion of higher-paid workers in the labour force. This problem is best corrected by working with unit labour costs, to incorporate the productivity effects of such composition effects.29 A further problem with Rottier’s calculations involves the use of actual average weekly earnings data. These clearly exaggerate the volatility of earnings across the business cycle. The practice of short time during downturns (which tended to be concentrated in the staple sectors), whereby workers were employed for less than a full week, both pulls down average actual earnings and increases reported dispersion, since short time was worked with different intensity in different industries. In October 1931, for example, 21 per cent of woollen workers and 30 per cent of carpet workers were on short time, compared with less than 10 per cent in staple sectors outside of textiles (including pig-iron and shipbuilding). I have recalculated Rottier’s figures, on the basis of average full-time earnings, and estimated coefficients of variation, rather than the uninformative median dispersion. The pattern that emerges from these calculations (Table 12.4, ‘Rottier revised’) suggests neither a high degree of earnings variability across industries, nor much sensitivity of earnings across the cycle. The coefficients of variation fall within a narrow range, while the standard deviation actually falls slightly between 1928 and 1931. The image that Rottier’s calculations create, of intense turbulence beneath a smooth surface, does not survive close scrutiny. But is it still possible that Beveridge was correct? Will a shift of focus from earnings to wage rates, and from levels at one point in time to changes over time, create a more compelling picture of hidden turbulence? There was certainly more volatility in wage rates at the sectoral level than the stationary aggregate index implies. The obvious anomalies involve cotton, coal, and woollens, whose average wages fell by 14, 15, and 16 per cent respectively between 1928 and 1932 (1926–32 for coal), while the aggregate index fell by less than 5 per cent. But these are hardly representative sectors, as Table 12.5 makes clear. This table presents summary statistics on the variability of wage rate changes across the sixty-five industries covered by the Ramsbottom study. The three time periods (1924– 8, 1928–32, 1932–7) were chosen to reflect the ‘normal’ 1920s, the depression, and the recovery. The greatest degree of variability is to be found between 1924 and 1928, as measured both by the coefficient of variation of wage changes, and by the number of sectors experiencing offsetting changes. In contrast, both the depression and recovery were marked by remarkable uniformity of wage movements, both in scale and direction. 264
WAGE BEHAVIOUR IN INTER-WAR BRITAIN
Certainly, different sectors of the economy show different profiles of wage behaviour. But we should be careful not to exaggerate the significance of these figures. The extent of wage movements in the various industries was, all in all, quite small.30 The aggregate index is not severely distorted by composition effects. It is certainly not true, as Rottier would have us believe, that the stability of the general indices of wage rates resulted from movements in opposite directions, wages decreasing in the depressed staple industries, while they were increasing in the new industries located in the southern part of the country and in the building industry. (Rottier 1957:239) Between 1928 and 1932, no sectors experienced wage rate increases, while all sectors (except aerated waters) shared in the recovery in wages between 1932 and 1937. The real reason that the aggregate index shows such tranquillity is that most sectors of the economy experienced minimal wage fluctuations: during 1928–32 wages fell by more than 1 per cent a year in only twenty-six of the sixty-five sectors, and by more than 2 per cent in only seven. Even during recovery (1932–7), wages rose at an annual rate above 2 per cent in only seventeen industries, and by less than 1 per cent a year in twenty-one. Real GDP per head rose during the same period at 3.4 per cent a year, and output prices (final output deflator) at 1.1 per cent.31 The evident dispersion in wage rate behaviour across industries during the 1920s and 1930s should not be confused with wage flexibility.
265
MARK THOMAS
CONCLUSIONS The general impression of inflexible nominal wages in inter-war Britain stands up to close scrutiny, even after recognizing the empirical deficiencies of the standard series. An investigation of wages at the disaggregated level, moreover, indicates inflexibility of wage structure. But what does this imply about the relationship between the labour market and macroeconomic performance? Was unemployment created by greedy unions who raised the real wage above feasible levels, pricing British producers out of markets and British workers out of jobs? Or was the initial problem created by exogenous forces, such as the Great War and the Great Slump, pushing against a rigid wage structure, formed by institutional sclerosis? Although the pattern of wages revealed in this chapter provides little support for a supply-based explanation of Britain’s economic problem between the wars, it does not follow automatically that demand factors were to blame. The relationship of wages and employment depends not only on the institutional structures of the labour market, but on the product market as well (Nickell 1988; Thomas 1992). If prices are more (less) flexible than wages in response to changes in the real economy, then the real wage will rise (fall). It is not enough, then, to focus on the behaviour of labour market variables if we wish to understand the macroeconomics of unemployment. Price setting is equally deserving of attention. And, as the contemporary studies of Hall and Hitch (1938), Saxton (1942), and others indicate, it should not be presumed that prices were any more flexible than wages. However, as Austin Robinson reminds us (1984), the findings of the Oxford School on price setting were by no means universally accepted, and certainly not in Cambridge, the home of Marshall, Pigou, and Joan Robinson. Unfortunately, as I have suggested above, reliable and relevant price data are extremely scarce for the inter-war period. But, unless and until such data become available, we shall be unable to offer definitive answers to the questions that introduced this chapter. NOTES I am grateful to Mary MacKinnon for helpful comments on the conference version of this chapter. 1 The series is calculated as ‘the average movement in the level of full-time weekly rates of wages’ (Feinstein 1972: T141, n. 1). It excludes both changes in hours and in the composition of the labour force. 2 There is slightly more variation in the pattern of average weekly earnings, reflecting in part the (limited) practice of short-time working during the depression, and of overtime during the later stages of recovery, as well as the changing composition of the labour force, and the lower elasticity of demand for skilled than for unskilled workers. 3 Pigou seems unclear on the distinction between own-product real wages, as faced by the employer, and consumption real wages, as faced by the worker.
266
WAGE BEHAVIOUR IN INTER-WAR BRITAIN 4
5 6 7
8
9
10 11
12
13
Benham noted that the high correlation recorded between 1919 and 1925 reflected ‘big corresponding movements in the earlier years, due to monetary policy, which swamp smaller dissimilar movements [after 1922]’ (1931:1315). The correlation between Rueff’s wage and unemployment data between 1923 and 1925 was 0.73 (quarterly basis), and between 1925 and 1929, -0.36 (quarterly data from Beveridge (1930:460)). Especially the final output deflator. Some were measured at factory gate prices, others at delivered, others still at retail values. The importance of excluding imported goods reflects both the degree of openness of the UK economy—roughly 30 per cent of industrial raw materials in 1924 were imported (Flux 1929)—and the strong shift in the terms of trade between imports and British manufactures. Imports became 25 per cent cheaper between 1924 and 1933 (Dimsdale 1984:98, Table 3). The mechanism for 1929–33 is that secondary workers, attracted by higher consumption real wages, enter the labour market faster than employers, distressed by rising own-product real wages, are prepared to hire them—leading to an increase in (voluntary) unemployment; the process is reversed after 1933 when the wedge falls, reducing the incentive for secondary workers to attach themselves to the labour force. Ian Gazeley at the University of Sussex is currently revising the Ramsbottom series. Gazeley is working to produce more detailed quarterly series, disaggregated by industry, occupation, region, and (where possible) age and gender. The preliminary results of this exercise (Gazeley and Rice 1992) do not indicate any very radical changes from the received doctrine (the absolute mean discrepancies between the Gazeley and Ramsbottom series are 4.3 per cent [shipbuilding], 1.4 per cent [wool and worsted], 1.0 per cent [fine chemicals], and 2.2 per cent [vehicles]). Moreover, any such changes may not represent improvements to the Ramsbottom series. Gazeley is using the same basic data source (the changes in the rates of wages reported to the Ministry of Labour), but while Ramsbottom utilized all the reported wage changes, Gazeley is restricted to those that were published in the Ministry’s Gazette. The largest discrepancy between the two Ministry of Labour series during the period from 1924 to 1934 is less than 0.6 per cent; between Ramsbottom’s and Bowley’s series, it is 1.8 per cent. Of the thirty-nine principal changes reported during March 1933, for example, eighteen were changes in minimum rates, one (in coal) was a basic rate change, and four represented changes in standard rates (Labour Gazette 1933:148–9). The point is stated explicitly in the heading to the Department of Employment’s historical series (1971:53), namely ‘changes in basic rates of wages’. Var[earnings]=var[base wage]+var[margin]+covar[base wage,margin]. Over 85 per cent of the earnings variance for fitters can be traced to margin adjustments. As might be expected, margins accounted for less of the variance in weekly wages for labourers; nevertheless, the proportion was close to 50 per cent. Note also that skilled workers, of whom fitters were the most prominent, accounted for 52.8 per cent of the engineering labour force in 1936, and unskilled workers only 13.2 per cent (Engineering Employers’ Federation archive, Modern Records Centre, University of Warwick, MSS 237/13/3/44). The proportion of fitters on piece rate rose from 27 per cent in 1925 to 44 per cent in 1930; the proportion of pattern makers rose from 56 per cent to 87.5 per cent (Penn l971:108, Table 7.6).
267
MARK THOMAS 14 15
16
17
18 19
20 21 22 23 24 25 26 27
Some 41 per cent of coal miners were on piece rate systems in the late 1920s (ILO 1931:18). Note, also, that the figures ‘are exclusive of changes affecting agricultural labourers, Government employees, domestic servants, shop assistants and clerks’ (Labour Gazette 1933:120 n. 1). For information on the wage (and salary) behaviour of these groups for the inter-war period, see Chapman (1953) and Thompson (1937) on agricultural workers; Marley and Campion (1940) on shop assistants and clerks; and Chapman (1953) on domestic servants. Sells argued that about 11 million workers were ‘covered by some definite form of wage-fixing machinery’ (1939:49). Elsewhere (Thomas 1992) I have suggested that this estimate is far too high. A better approximation for 1937 (even including workers covered by such informal arrangements as the Fair Wages Clause) would be somewhat below 9 million, or about 44 per cent of the total workforce. These figures are predicated on the assumption that the relationship between union membership and bargaining coverage was constant, such that the decline in union activity was accompanied by a decline in the intensity of collective bargaining overall. The strong correlation between the level of unionization and the intensity of wage negotiations would indicate that this is reasonable (especially as we would anticipate a cet. par. decrease in wage changes as unions subsided in visibility). Table 12.2 shows that an increasing number of agricultural workers were brought under statutory wage control by the Agricultural Wages (Regulation) Acts of 1924 and 1937, but the wage change data specifically exclude agricultural workers. It could be argued that the number of wage changes (as a measure of union success in pay negotiations) determined the level of unionization, rather than vice versa. Such an interpretation receives no support from the econometric work on inter-war union membership, which places weight on unemployment and prices, rather than wages (Bain and Elsheikh 1974). In conjunction with data on the average wage in manufacturing, I also used data derived from the earnings inquiry of 1924 on average wages in manufacturing and the extractive industries (Labour Gazette 1928). These figures still understate downward flexibility, given the sensitivity of margins. These are likely to present second-order problems, however, in the overall assessment of wage behaviour. The cross-product of women and unionization was never significant, suggesting that union wage effects were not gender specific. I did run some equations using Burkitt’s data set, but could make no sense of any of the results. Not surprisingly, given the almost non-existent skill classification of the Census of 1931 and the very rough nature of the variable. UNSK is not included in the regression estimates of Table 12.3. The microeconomic approach is possible only for an extremely limited number of data sets, namely the BLS survey of British households in 1890–1 (Haines 1979) and the New London Survey of 1928 (Eichengreen 1986). This result seems entirely plausible given the small amount of excess demand in the economy up to 1936. Similarly, Clay (1929:112), in a rebuttal to Pigou (1927), argued that, ‘so far as “high wages” are the explanation of the present unemployment, it is the low wage-rates of the depressed industries that are “too high”, not the high wages of the prosperous industries’.
268
WAGE BEHAVIOUR IN INTER-WAR BRITAIN 28
29
30 31
The figures for 1935 and 1938 were 0.24 (0.15) and 0.25 (0.10) respectively. (Calculated from Rottier (1957:241, Table 3).) Rottier’s median for 1935 appears to be too high, biasing the dispersion measurement downwards. A quick calculation suggests a more accurate figure of 0.26. As an example of the impact of productivity on measured real wages, I calculated crude unit labour costs for nineteen industrial sectors, 1924–8. Average nominal earnings for these sectors remained stationary; productivity-corrected earnings fell by 10.4 per cent; unit labour costs (using gross output prices as the deflator, other than for raw material intensive sectors, when a crude value-added deflator was applied) declined by 2.7 per cent. (Calculated from Labour Gazette (1928); London and Cambridge Economic Service (1930:187–9); Board of Trade (1935:142– 6).) These calculations cannot make allowance for the inaccuracies in the Ramsbottom series identified in the second section. Unfortunately, in the absence of reliable sectoral value-added price indices, it is not possible to determine the cyclical behaviour of real wages at the disaggregated level. Given, however, the very strong correlation between movements in gross output and raw material prices, it seems unlikely that moving from nominal to real sectoral wages would severely distort the pattern revealed here, other than perhaps to bring import-dependent sectors closer to the economy-wide mean.
269
13 RACE, HUMAN CAPITAL, AND LABOUR MARKETS IN AMERICAN HISTORY Warren Whatley and Gavin Wright
What happens to economic relationships among people when they think in racial terms? Do the competitive pressures of markets tend to erode the economic effects of racialistic thinking, or does racialistic thinking evolve into a parameter of market structure? What role do markets play in the ‘social construction’ of race? These questions are both economic and historical in character, and we will argue that they can only be answered in historical context. Economists tend to start from the premise that racialistic thinking leads to racial discrimination, which in turn leads to racial segregation as a way of avoiding the costs of mixing antagonistic groups of workers. The resulting segregation can be innocuous in its economic effects (as in separate firms for black and white workers) or it can be harmful (as in inferior jobs for black workers). In the simplest models, non-discriminating firms will tend to drive out discriminators (whether segregated or not), if the market structure is competitive. In more complex models, discrimination may persist or even worsen, because of non-convexities in production relationships, communication problems, or racial stereotypes, combined with feedback processes affecting the experience and incentives of different racial groups.1 The question is thus empirical, and as intricate as the theoretical possibilities have become, the bottomline empirical question has remained the same: how much of observed racial inequality is attributable to discrimination, defined as unequal rewards to otherwise identical workers? The operative phrase here is ‘otherwise identical’. The ideal test case would be two workers—one white and one black—each of the same age, with the same skills, the same marital status, the same labour market experience, the same gender, the same height and weight, and so on. In addition, they work at the same job, doing the same work, under the same pay scheme, at the same firm/farm, at the same time. If the otherwise identical 270
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
black worker earned the same as the otherwise identical white worker, then race is an inconsequential descriptive characteristic. The employer ‘discriminated’ on the basis of experience, skill, education, or any other characteristic related to expected productivity, but not race. If we could find such ‘otherwise identical’ black-white twin-pairs, then finding that they earned identical wages would have implications that go beyond the workers’ immediate set of contractual relations to imply that other employers paid them equal wages as well. Why else would the observed employer pay them equally, except that he has been forced to do so by other employers who have bid up the wages for whites and blacks to the point that their opportunity wages are equal? Otherwise, one of the workers could have been hired for less than the other. If so, even a well-intentioned employer would be sorely tempted to pay black workers less, if no one else wants to hire them. If it takes substantially less to keep blacks on the job (i.e. if discrimination prevails elsewhere), then even an employer with nondiscriminatory intentions would end up treating blacks differently. So a finding of equal pay for identical work, for identical black and white workers in identical situations, not only implies an absence of racial discrimination in the situation under study (by definition), but an absence of discrimination by other firms drawing workers from the same labour market. A finding of racial discrimination in wages, therefore, has implications for our conception of the world beyond the particular workplace at hand. It leads us to ask questions about how labour markets develop and operate: how information about worker performance generated in one corner of the world is transmitted to another corner of the world; how firms identify worker characteristics and learn about their productivity; what sort of proxies for ‘ability’ emerge, and what their implications are for the incentives facing workers. These topics are intrinsically historical: they depend on the particular institutional structures of the labour market, and the particular state of knowledge or beliefs about the races on which these structures operate. Once we leave the comfortable world of wage equality and nondiscrimination, it is far less certain what information about one set of employers tells us about the behaviour of others. Unfortunately, the empirical programme still begins with conventional measures of wage discrimination, and that programme has proved difficult because information on black and white twin-pairs is scarce. The problem is not merely a lack of data, but the more substantive historical fact of racial segregation. Historically, blacks and whites have seldom been in the same place at the same time, and have seldom done the same work. Some of the reasons for this separation may be historically ‘accidental’, but others clearly derive from the very labour market processes we are trying to understand. To approach the subject by ‘controlling’ for such factors as region, urbanrural location, or industry, is to pull out of the analysis many of the major components of life experience that have come to define the meaning of race 271
WARREN WHATLEY & GAVIN WRIGHT
in American history. In one form or another, segregation has also prevailed in residence, occupations, and schooling. We believe that the impact of race cannot be adequately understood as a residual, and that an empirical strategy focused on a narrow concept of discrimination between otherwise identical workers should be the beginning rather than the end of the analysis. Indeed, the most thorough recent study of education and jobs in the south, Robert Margo’s Race and Schooling in the South 1880–1950, finds that an index of employment segregation increased between 1900 and 1950 even after controlling for racial differences in schooling (Margo 1990a: ch. 6). Schooling is at best an imperfect measure of ‘human capital’, but segregated schools were undoubtedly a key component of racially segregated life in the south, and Margo confirms that these schools were separate and unequal in the provision of educational services. But Margo also reports that ‘race per se’ had a powerful effect on the distribution of employment, and that ‘pure racial segregation’ continued to increase until at least 1950 (1990a: 97, 102). He writes: ‘Race, not schooling, was the principal factor limiting the participation of black labour in certain industries and occupations’ (104). But how are we to interpret the increasing race effect? Statistically, it appears to mean that over time employers tended to give race increasing weight in hiring decisions and job assignments, so that there was an economywide tendency towards separate black and white jobs. But does this trend mean that white workers were able to enforce a racial division of labour that preserved the better jobs for themselves? Or does it mean that the experiences of blacks and whites in the south became increasingly different over the Jim Crow era, not just in schooling, but also in work experience, family status, jobs applied for, and the like? Did race become an increasingly accurate proxy for a well-defined set of expectations that differed markedly by race? Did the old phrase ‘know your place’ increasingly capture the economic as well as the political essence of the Jim Crow south? ‘Controlling for education’ just does not tell us enough. Even if Margo had found that years of schooling completely accounted for racial wage differences, this would still not imply the irrelevance of race in the labour market, because of the possibility of feedback effects between job opportunities and educational achievement, along racial lines. In short, we need to know the historical process in order to understand the nature of the observed outcome. Segregation and discrimination in the south have come in for detailed study by labour economists recently, because of the evident importance of the civil rights revolution for black economic progress in the 1950s and 1960s.2 In contrast, the experience of black workers in the labour markets of northern cities has been neglected, a status we hope to improve on through our ongoing study of major employers of black labour. Although one might have suspected that race was a less salient factor in labour markets outside of the Jim Crow south, our findings suggest that this presumption is far from true. In the sections that follow, we attempt to follow through on the 272
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
arguments sketched here: that segregation gives rise to discrimination as well as the other way around, so that market forces can actually tend to widen rather than diminish racial differences. This at least is our reading of race and the American labour market over the past century. We begin with the south, move on to the north, and draw some inferences for the literature on post-Second World War economic developments. THE SOUTH: FROM HORIZONTAL TO VERTICAL SEGREGATION Racial segregation was the norm in the industries of the south. In his study of Virginia firms in 1900 and 1909, Robert Higgs found that ‘occupational workforce segregation was overwhelmingly the rule’. In 1900, 76 per cent of the cases employed only one race or the other at a given job, and the figure had increased to 81 per cent in 1909 (Higgs 1977:241). There are even more dramatic illustrations at the state level: in Florida, Georgia, Louisiana, Mississippi, North Carolina, Tennessee, and Virginia, the Census of 1930 recorded over 4000 male telegraph operators, not one of them black; the southern states reported nearly 15,000 locomotive engineers, but less than fifty were black (0.3 per cent); only four of 5600 loom-fixers were black (0.07 per cent). These extreme cases may involve access to special skills, but the evidence is nearly as strong in unskilled and semi-skilled categories: more than 90 per cent of the operatives were white in such industries as furniture, textiles (including cotton, silk, knitting, and rayon), shoes, printing, and car and railroad shops. At the same time, more than 75 per cent of the labourers were black in tobacco manufacture, iron and steel, fertilizers and turpentine (US Bureau of the Census 1933: Table 11 for each state). These patterns of segregation, which persisted for decades, cannot readily be attributed to legal barriers, unions, or an absence of labour markets. In looking for an economic logic behind these patterns, it is notable that racial separation was far more prevalent and clearly delineated by industry than by location. The region of all-white cotton mills had broad geographic overlap with tobacco manufacturing, which was predominantly black, and these racial features persisted wherever the industry spread. Though the white-majority state of North Carolina had white furniture factories, its fertilizer producers were overwhelmingly black. Employment data from Virginia show striking continuities in the racial compositions of the workforce by industry: although whites were about two-thirds of the wage-earning labour force, they comprised more than 85 per cent in fifteen industries, including cotton mills, knitting mills, silk mills, woollen mills, furniture, leather, metal signs, boots and shoes, garments, printing, and tin boxes. Blacks were just under one-third of the wage earners for the state, but seven industries were more than 70 per cent black, and sixteen were black majority. It is sometimes suggested that industrial segregation reflected the uneven geographic 273
WARREN WHATLEY & GAVIN WRIGHT
distribution of the races within states (Higgs 1977:241). But one could observe industry-based segregation even within particular cities and towns. White cotton mills and black tobacco factories co-existed in places like Durham, North Carolina, and Danville, Virginia, and in Birmingham, Alabama, where two-thirds of iron and steel workers were black, and the Avondale cotton mill was 98.1 per cent white (Worthman 1969:392). In one sense, explaining segregation has never posed a serious problem for economic theory. If there are no differences in productivity, but whites demand a premium for working in close association with blacks, segregation will dominate mixed alternatives (Arrow 1972, 1973; Stiglitz 1973). Other costs of mixing, such as work disruption or problems of communication between races, will lead to the same result (Lang 1986). What is less often appreciated is that segregation can actually raise average productivity, if race is a good predictor of traits that are correlated with productivity, and if individual productivities are not readily observed. Suppose, for example, that the only relevant distinction is between experienced and ‘green’ workers; employers cannot distinguish them by sight, and cannot believe the workers’ own claims. In this case, the average productivity of a randomly chosen workforce would be
where U is the productivity level of an inexperienced worker, S is the productivity of an experienced worker, and p is the percentage of experienced workers in the labour pool. If many whites have previous experience, but few blacks, then the productivity advantage of an all-white labour force is
where FB is the percentage of blacks in the pool, and pw and pB are the percentages of blacks and whites who are experienced, respectively. It is immediately evident that segregation raises average productivity for the firm, the more so as experience matters for productivity (S-U), as whites have more initial experience (pw-pB), and as the relative size of the black population increases. This simple model is surprisingly powerful. It explains segregation independently of worker prejudice; it explains the absence of all-black plants in industries where blacks had little experience (since such plants would be no cheaper, but less productive than randomly selected plants); and it explains the self-perpetuating character of industry segregation, since the initial difference in experience would be rapidly magnified by the effects of the segregation policy. These specifications seem to fit the cases of the cotton textile and tobacco industries, which grew up in the post-Civil War south, but had clear roots in the ante-bellum period. At the time of the post-war 274
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
revival of the cotton textile industry in the 1870s, virtually all of the initial pool of experienced labour was white. Slaves had been used in the cotton mills during the 1840s, but by the time of the Civil War the textile labour force had become almost entirely white.3 In contrast, the tobacco industry in Richmond and nearby towns continued to use slave labour. After the war, many former slaves continued as free labourers at the very tasks they had done before, and tobacco manufacturing remained predominantly black (Taylor 1926:117–18). As the industries developed, the emergence of a regional pool of experienced labour became an increasingly important factor (Carlson 1981; Hall et al. 1986:260–3). Employers came to believe that the traits required for the two industries corresponded to ‘natural’ characteristics of people— defined in racial terms. The belief by economists that such inaccurate beliefs cannot persist in the face of experimentation and competition seems naive (e.g. Lang 1986:385). A series of experiments with all-black mills around the turn of the century were watched with great interest and genuine curiosity by industry people. Since all of these attempts were failures, however, their net effect was to reinforce the beliefs that blacks were unsuited to mill work. Exactly why these efforts failed is open to debate, but it is surely relevant that their labour was completely inexperienced, yet they were competing with mills that by then had second- and even third-generation ‘cotton mill people’. Of course, if blacks could have been gradually worked into ongoing mills, the results would surely have been different. But this path was ruled out by vehement objections from whites. (Only at a later point was racial integration actually prohibited by law.) When labour shortages began to appear after the turn of the century, the response of employers was not to press for racial integration, but to widen their recruiting networks to bring in whites from more distant parts of the south. In stressing the early historical origins and perpetuation of segregation lines in employment, we do not mean to imply that these origins were entirely arbitrary and accidental, nor that the positions of the two races were symmetrical. Still less do we mean to deny the reality of racial hostility, nor the desire among whites to maintain racial supremacy, in all of its many manifestations. The rise of the cotton textile industry, for example, featured bursts of civic enthusiasm and community involvement, and was portrayed by boosters as a means of advancing the status of poor whites (as well as a way to make money). Across the entire region, the prospects for blacks may well have been better in agriculture than in even those industries with a strong historical black presence. Our purpose is not to highlight economic motivations as opposed to racial ones, but to ask how market forces and racial segregation interacted historically, from these beginnings. The issue that economists have wrestled with is not segregation per se, but wage discrimination: did segregation serve to support an ‘unjustified’ wage differential, or was segregation merely the market’s device for avoiding the 275
WARREN WHATLEY & GAVIN WRIGHT
costs of mixing the races, and gaining the benefits of specialization? If blacks were not gaining experience in textiles, they were instead working in tobacco, mining, steel, or agriculture, and if the labour market cleared, in equilibrium wages should be equal for equivalent labour. There is considerable evidence that equilibration of this sort did prevail in the south before the First World War. Figure 13.1 displays frequency distributions for black and white male wage earners in 1908 Virginia, estimated from that state’s annual labour reports. The white distribution is bimodal, the black unimodal; but the peak for the black distribution is at virtually the same wage level as the left-hand peak for the white distribution. If we interpret the first peak as the ‘unskilled’ market and the second peak as the ‘skilled’ market, then we may say that despite the prevalence of racial segregation, unskilled wages were effectively not differentiated by race. Apparent racial wage differences were, in reality, skill differentials, or so it would be easy to believe in the world portrayed by Figure 13.1. The finding of ‘no wage discrimination’, however, seems less pertinent than the observation that virtually no black workers had access to the high paying jobs. In the Virginia sample, the skilled jobs were primarily in the iron industry and the building trades. The job hierarchies in these industries were highly structured, and subject to tight control by craft unions. In these respects they resembled the job structures of the railroads, examined by William Sundstrom (1990). Sundstrom finds that a high degree of wage discrimination prevailed in railroad work, generated by an extreme racial exclusiveness in promotion to the top 276
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
jobs, as well as by unequal pay within job categories. Whites sometimes argued that blacks were not qualified for these jobs, but Sundstrom points out (and cites evidence to this effect) that poor black work performance may have been attributable to the absence of a promotion incentive. If one could trace the course of skills and wages over time in such a structure, one might well find an increasingly close match between the racial distribution of skills and the racial distribution of wages. But it is evident that the skill distribution was endogenous to the structure of segregation. Hierarchies such as these long pre-dated the major effects of inequality in schooling, as dated by Smith (1984). The majority of southern white wage earners were in the unskilled category, however, and in the ununionized, labour-intensive industries. When we isolate these cases by removing iron and the building trades from the sample, the wage distributions do show a strong tendency towards equilibration (Figure 13.2). Even here, whites had an edge on the right-hand side, presumably reflecting opportunities for advancement within these industries. But the ‘base wage level’ appears to have been similar for blacks and whites. One should not assume that the equilibrating pressures were coming mainly from these industries themselves. Wages in southern industry during this period were closely tied to the wages of farm labour. The existence of a large agricultural sector was an important background condition facilitating equilibration across segregated industries. Despite the many forms of racial inequity in southern agriculture at this time, wage-paying farm labour jobs had virtually no entry requirements, and rarely paid black and white labourers differently (Higgs 1978). This labour market structure had undergone an essential qualitative change by the 1920s. Figure 13.3 displays the black and white wage distributions for 1926, for the identical Virginia industries covered in Figure 13.2. It is evident that a ‘racial wage gap’ had come to prevail, in contrast to the ‘skill gap’ of the pre-war era. We believe this change is of fundamental importance, not because of the direct implications for the racial distribution of income, but because it represented the crystallization of a racial stereotype that made the low-wage status of blacks self-justifying. When we try to identify the reasons for this divergence, many proximate factors can be listed: the stagnant world market for cotton during the 1920s had a disproportionate effect on blacks, as did the depressed conditions in lumber and saw milling; real wages in cotton textiles, on the other hand, rose to a relatively high plateau during the wartime boom, and proved resistant to downward adjustment for a number of reasons (Wright 1986:147–55). Invoking these immediate circumstances, however, begs the question of why segregation could serve to support a racial wage differential in the 1920s, when it had not done so earlier. Among the many possible answers, the one most persuasive to economists is education, specifically the wide racial gap 277
WARREN WHATLEY & GAVIN WRIGHT
278
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
in years and quality of schooling for the cohorts entering the labour force after the First World War (Smith 1984; Margo 1990a: ch. 2). The key role of public education is to alter the standards of ‘ability’ that firms could expect from a randomly selected new worker with no previous work experience. The basic issue is not whether education is ‘required’ for work in (say) the textiles industry; job specifications can be adjusted to accommodate the background of the ‘average’ worker in the relevant labour pool. But if employers can safely assume that the ‘ability’ of the average white is higher than that of the average black, and this difference is not merely a proxy for experience, then they can set wages so as to equate the price of a unit of effective labour for each group. In such a setting, segregation will be productive (as well as conducive to good relations with white workers), by reducing heterogeneity within workforce components, even if the degree of dispersion within each of the two groups is identical. We have no desire to minimize the significance of educational discrimination in the history of black labour in the south. What we do want to emphasize is that educational policies were largely endogenous to the previously established racial division of labour by industry and occupation. The largest employers of black labour, planters and farmers, were the strongest opponents of expanded black schooling opportunities, on the grounds that ‘educated negroes, in nearly all cases, become valueless as farm laborers’ (quoted in Anderson 1988:96). When northern reformers began to pour money into southern schools after 1900, the policy implications dictated by the labour market were self-evident: in the words of the philanthropist William H.Baldwin, ‘it is a crime for any teacher, white or black, to educate the negro for positions which are not open to him’ (Anderson 1988:84, emphasis added). By contrast, cities with a substantial black industrial labour force, like Durham, became centres of black education as well. In Birmingham, schools for both blacks and whites were better funded than in rural areas. The largest employer, Tennessee Coal and Iron (US Steel after 1907), played an active role in promoting, funding, and administering black schools (Harris 1985). Perhaps the most dramatic illustration of endogenous schooling is provided by Price Fishback’s recent study of educational discrimination in West Virginia (Fishback 1989). That state’s economy was dominated by coal mining, in which the bulk of the workforce were labourers, including many blacks recruited from plantation areas. Fishback finds little evidence of wage discrimination against blacks. Even though schools were segregated, black schools in West Virginia came closer to parity with white schools than in any other southern state. Indeed, the quality of black schools was used in recruiting efforts by the mining companies. Over most of the south, however, the emerging logic of racial dualism was the more powerful underlying tendency. Table 13.1 displays average wage rates for entry-level workers in twenty industries as of July 1937, from a rare 279
WARREN WHATLEY & GAVIN WRIGHT
Bureau of Labor survey that requested race-specific information. Nearly 60 per cent of the southern firms reported hiring only whites or only blacks; those hiring only blacks paid starting wages fully one-third lower than those hiring only whites. Of the firms hiring both blacks and whites, nearly 30 per cent paid blacks a lower starting wage, a figure that must be regarded as a lower-bound estimate. Equally striking in Table 13.1 is the absence of an explicit racial wage gap in the northern states, despite black educational levels far below national standards at that time. If one were to look only at this table, it would be tempting to attribute the regional contrast to ‘southern racism’. But as shown above, there was no comparable entry-level gap thirty years before, when ‘southern racism’ was surely at its historic peak. The ‘separate wage rates for Negroes’ that observers took to be ‘a fixed tradition’ (Johnson 1943:90) had in reality developed and become institutionalized in the twentieth century. It was not so much a direct reflection of southern race prejudice, as a market outcome that crystallized what had come to be the accepted expectations about the quality of black workers and the types of jobs to which they would be assigned. When the Rosenwald Fund sought to provide financial support for black high schools in the south during the 1920s and 1930s, the fund sponsored a series of surveys of ‘Negro jobs’ in various cities, so as to adapt the curricula to the jobs actually available. In place after place the response indicated that there were no black jobs for which a high-school education would be useful. Thus, black schools typically 280
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
did not offer training in such subjects as stenography, accounting, bookkeeping, printing, or typing. The fund’s curriculum expert acknowledged: ‘If commercial courses were offered in the negro school there would no doubt be tremendous pressure to get into them and the only result would be keen disappointment for nearly everyone.’ Thus the black high school in the textile centre of Greenville, South Carolina, excluded textiles entirely. The Rosenwald report on Greenville recommended a course in ‘washing and greasing’ for black students, plus training in ‘public service occupations such as those of janitor, waiter, bellman, cook, bootblack, etc.’. The report noted with regret: ‘A course in shoe shining would prove unpopular among certain classes of Negroes in any southern community, because they do not realize its economic worth, and the splendid possibilities for developing a real vocation’ (Anderson 1988:223, 226). FROM SOUTH TO NORTH: SEGREGATION TO EXCLUSION Between 1890 and the First World War, American industry surged into a position of global pre-eminence, but black labour played a relatively small role in this history. Prior to the First World War, black representation was negligible in the major industries of the northeast and midwest. In 1910, blacks comprised 1 per cent or less of the industrial workers in New York, Cleveland, and Chicago; and less than 4 per cent in Pittsburgh and Philadelphia. Most of this relatively small number were concentrated in such low-paying occupations as domestic and personal service, or in eating/ drinking establishments or hotels. Why did so few blacks migrate to the better-paying jobs in the north? Wright (1986) lays out an interpretation based on the conception of labour markets as networks of communication and mutual assistance, especially among unskilled, poor, and poorly educated people. These networks took on a regional character for southern whites as well as blacks during this era, for historical reasons. This view is confirmed by the high levels of labour mobility within the south, and by the evidence summarized above that unskilled black and white labourers operated within a market that was economically integrated though racially segregated. Margo (1990a) has questioned this interpretation, citing evidence that higher levels of schooling greatly increased the probability of black out-migration from the region, and drawing the inference that ignorance and illiteracy were the key barriers to inter-regional mobility. Even Margo agrees, however, that this effect operated mainly for southern blacks rather than southern whites, and that ‘the bulk of the movement and its particular timing’ are not explained by changing levels of black education. Only with the ending of mass European immigration and the beginnings of active southern recruitment during the First World War, did blacks migrate north in substantial numbers. 281
WARREN WHATLEY & GAVIN WRIGHT
When they arrived, they found that race mattered in the north, just as in the south. True, explicit racial wage differentials were rare in the north. But those African-Americans who did reside in northern cities were excluded outright from a wide range of occupations and firms, as thoroughly and completely as in the former confederacy. In his detailed study of the history of labour recruitment practices in Philadelphia, Walter Licht (1992) reports that blacks were systematically excluded from industrial jobs in that city. With rare exceptions, the major Philadelphia manufacturing companies did not hire a single black worker until the late 1930s (1992:45, 141). The lower rate of labour force participation for black youths was directly caused by this fact; in the words of W.E.B.DuBois, ‘there is really very little that Negro children may do’ (quoted in Licht: 43). Even in a city like Detroit, where foreign-born workers were coming from many exotic parts of the world to work on the assembly line, resistance to hiring blacks in the factories was so great that it took more than six months of combined effort by the Society of Prevention of Cruelty to Children, the Poor Commission, and the YMCA to secure a manufacturing job for a 15-year old black boy of unusual mechanical ability who wanted to secure a trade to support his mother and six younger brothers and sisters. (Katzman 1973:105) It is a great mistake to believe that the problem of black economic progress was all downhill, once the migratory floodgates opened during the First World War. In fact, the influx of the black population was much faster than the growth of industrial jobs for which they were eligible. Two recent studies of racial gaps in unemployment confirm that even before the Great Depression, unemployment rates in northern cities were 50 to 100 per cent higher for blacks than for whites. Vedder and Gallaway (1992) find that between 1890 and 1930, the national non-white unemployment rate was not substantially higher than the rate for whites. But this national average was dominated by the south; in the northeastern and midwestern regions, the ratios of non-white to white unemployment rates ranged from 1.27 to 2.12 in 1930 (1992:699). Sundstrom (1992a) confirms this finding, and shows that after controlling for occupational composition, the blackwhite gap was much greater in northern cities than in the south. Furthermore, the relative employment status of blacks worsened between 1930–1 and 1936–7. In the south it may have been ‘know your place’, but in the north, it was not clear that the economy really had a place for black workers. There were a limited number of cases in which black workers did establish a northern presence before the Second World War, and these serve to underscore the importance of historical continuity and self-perpetuating racial perceptions in labour markets. 282
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
One of the few avenues for black entry into northern industrial employment was strikebreaking. Whatley (forthcoming) identifies 111 examples of black strikebreaking between 1880 and 1929, predominantly in the north, and involving in total many tens of thousands of workers. Although these numbers are not large relative to the entire labour force, black strikebreakers were highly visible in almost all of the major industrial labour confrontations, including nine of the twentyfour most violent clashes. These events were significant beyond their numbers, for at least two reasons: they sometimes opened up job opportunities previously closed to blacks, and thereby initiated northern networks which became significant in facilitating migration at a later point; for white labourers, they confirmed their perception of blacks as threatening outsiders, unresponsive to the ethic of working-class solidarity. But since in most cases African-Americans had been previously excluded from the unions that were leading the strikes, they had little reason to reject the roles to which they had been assigned. For our present purposes, it is worth noting that these strikebreaking episodes reinforced another kind of continuity—in work experience and the perception of the appropriate uses of black labour. When employers recruited blacks to break a strike, they did not typically call on inexperienced farm hands, but on workers who had relevant job experience in the industry at hand. In fact, most of the cases clustered in a small number of industries that had southern branches employing large numbers of black workers: iron, steel, coal, railroads, and longshore work. (Meat packing, a relatively new industry in the north, was the only significant exception.) As early as 1878 African-American strikebreakers were being recruited directly from southern mills, playing a visible role in the 1892 Homestead strike. Spero and Harris (1931) assert that nearly all of the black strikebreakers used in the steel industry of Pennsylvania had prior training in the south (250–1). Thus, to a considerable degree the early movement was a northward extension of the same types of ‘Negro jobs’ that were open to blacks in the south, an expansion of earnings and perhaps horizons for the workers themselves, but not a broadening of perceptions with respect to the types of labour of which they were capable. This analysis implies that there were unexploited possibilities for the productive use of blacks by northern employers, and the evidence suggests that this was so. Whatley (1990) finds that for firms in Cincinnati who hired African-Americans for the first time during the First World War, the learning experience had a lasting effect on employment policies. This pattern stood in contrast to the experience in Atlanta, where firms who used blacks during the war were more likely afterwards to restore the pre-war lines of segregation. This finding suggests a somewhat greater northern potential for positive reinforcement, for those firms willing to buck the prevailing racial norms. And indeed, this is what we find for the most notorious of these, the Ford Motor Company. 283
WARREN WHATLEY & GAVIN WRIGHT
FORD: THE EXCEPTION THAT PROVED THE RULE Between 1920 and 1940, Ford actively recruited black workers, employing more than 14,000 by the time of the Second World War, accounting for roughly 40 per cent of all black male employees in Detroit at that time. All but a few of these worked at the giant River Rouge plant, which opened in the early 1920s. These actions were not the result of passive race-neutrality in hiring, but represented conscious decisions by management, dating from an agreement between Henry Ford and a group of black ministers during the depression of 1920–1. Before the Second World War, no other major auto company followed Ford’s lead on this score. Reflecting another dimension of Ford’s uniqueness, the company kept detailed employee records, including race, which have allowed us to examine wages, work histories, job tenure, and personal background variables for black employees in comparison with whites.4 Although Ford was by no means a representative company, and jobs in the auto industry were by no means typical for blacks, the information contained in these records allows us to draw inferences about the character and quality of the available black labour supply, and thus about the opportunities available to blacks in northern cities. Let us begin with a synopsis of what we have learned about Ford’s own policies. Consistent with Table 13.1, we find no indication of a systematic differential in entry-level wages by race. Although black auto workers’ average annual earnings were nearly twice the average for all black manufacturing workers, Ford evidently felt constrained to pay new black employees at roughly the same scale as everyone else. Nonetheless, we do find evidence of discrimination. When entry wages are regressed in conventional fashion against worker characteristics (age at first hire, marital status, gender, education, and race) and evaluated at the means, we do find an unexplained wage differential by race of just under 5 per cent. That is, a beginning black worker of ‘average’ characteristics received about 5 per cent less than a white twin with identical characteristics. A wage differential of less than 5 per cent may seem small, but this aggregate figure obscures a striking qualitative difference in the treatment of blacks and whites: attributes that warranted higher pay for whites (age, marriage, education) made no difference at all for blacks. Figures 13.4 and 13.5 display wage profiles for black and white male non-apprentice workers, married and unmarried, for various ages and education levels. These results indicate that even at Ford, a company that challenged prevailing racial employment practices, black workers were stereotyped by race. The form of discrimination corresponds to that predicted by statistical discrimination models, such as that presented by Lundberg and Startz (1983), but as we argue below, the reasons may have been somewhat different. What does Ford’s experience with black workers tell us about the larger racial contours of labour markets in northern cities? Clearly Ford was able 284
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
285
WARREN WHATLEY & GAVIN WRIGHT
to make successful use of a portion of the population that others shunned, suggesting that the demand side made a difference. But was it perhaps the case that Ford’s differential treatment of blacks actually represented a form of ‘statistical discrimination’, attributable to characteristics of the black population as a whole, if not to those of individuals? To answer these questions, we need to know something about performance on the job, and we need to know what sorts of samples Ford was drawing from the underlying populations. These issues are addressed in a recent paper by Maloney and Whatley (1992). The results are striking. Maloney and Whatley do not have direct evidence on job performance, but they do have evidence on job tenure. They find that contrary to the stereotype of unreliability, black workers at Ford consistently stayed longer on the job than whites. Not only was their tenure longer, but marital status at the time of hire was a better predictor of job tenure for blacks than it was for whites. Estimation of a proportional hazard model of job tenure for blacks and whites indicates that a 20 year old black worker who was married at the time of hire was more than twice as likely to stay after having been on the job two years than an otherwise identical white worker. Marital status among young married white workers had virtually no effect on expected tenure. Not only did blacks stay longer at Ford, but they stayed while working at the least desirable, highest-effort jobs, from which whites departed at the earliest opportunity. Thus, from Ford’s standpoint there was an element of ‘statistical discrimination’, but not of the kind normally modelled: Ford was gaining a mature, stable black workforce, and using these workers to fill jobs for which whites would have required a substantial premium (quite apart from their taste for racial proximity). Clearly Ford was able to take advantage of the limitations on black opportunities elsewhere. This interpretation is confirmed by pooled time series, cross-section regressions of monthly wages as a function of work experience. When we hold Ford tenure constant along with other worker characteristics, the agewage profile becomes an estimate of the return to labour market experience outside of Ford. We find that the two profiles had strikingly different shapes for the two races: holding Ford tenure constant, black work experience elsewhere in the economy had almost no positive value, in contrast to whites, who commanded higher wages as they gained general labour market experience. Outside of Ford, African-Americans did not have access to career lines or job ladders on which wages rose with experience; hence there was no need for Ford to pay a higher wage to attract more mature black workers, and no reason for Ford to give such experience a positive weight for internal purposes. In these circumstances, mature black workers who left Ford suffered a severe cut in wages. Mature white workers, on the other hand, were in a much stronger position. Because their outside opportunities improved with age, a separation from Ford was less likely to be costly. 286
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
THE RIVER ROUGE FOUNDRY AS A BLACK DEPARTMENT Ford’s early recruitment of black auto workers had a lasting influence. Job opportunities were opened up in many different departments, and as late as the 1960s, the company’s black representation in skilled and white-collar positions was significantly higher than that of other firms in the industry (Northrup 1968). Nonetheless, it is useful at this point to return to our original argument that even for a race-neutral, profitseeking employer, discrimination elsewhere in the economy creates an incentive for racial differentiation at the firm in question. Such a mechanism is illustrated by the way in which the River Rouge foundry became a black department. One may regard assignment to the foundry as another form of implicit discrimination, a non-pecuniary burden not reflected in the measured wage. The foundry was regarded as the most stressful, hot, dirty, and dangerous department at the plant, and as we have seen, the assignment of blacks to such jobs continued a tradition dating back to the iron mills of Virginia and Alabama. In his dissertation on black labour in the auto industry, Lloyd Bailer concluded that ‘one can be quite certain that an automobile plant employing sizable proportions of Negro Laborers includes a foundry department’ (Bailer 1943:60). At Ford, blacks were well represented at the foundry from the start, but one must remember that blacks comprised only a little over 10 per cent of total company employment. Between 1920 and 1932, whites accounted for between 20 and 50 per cent of foundry employment at Rouge. When the plant reopened from the shutdown of 1932–3, however, the foundry was for all practical purposes a black department. As shown in Figure 13.6, more than 80 per cent of the man-days worked in the foundry after that date were by blacks. It is certainly possible that the policy had some basis in objections from whites to working in close association with blacks. But this motivation does not seem to have been the proximate cause of the new policy. Figure 13.6 does not suggest that the black percentage gradually increased until a racial ‘tipping point’ into segregation had been reached. Instead, the initiative was with the company. Why would Ford have chosen to self-impose such a constraint on its own job assignments? One reason is suggested by Figure 13.7, which shows that the segregation policy was associated with a shift of the entire foundry from high-wage to low-wage status. By making the foundry black, Ford not only found it possible to fill its entire employment needs at the foundry without paying a wage premium but the company was actually able to pay wages that were below the levels of other departments.
287
WARREN WHATLEY & GAVIN WRIGHT
288
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
This interpretation is confirmed by a more detailed statistical decomposition of wages by race and department, using the monthly wage regressions described in the previous section. Controlling for individual characteristics, the wage differential between white and black foundry workers in 1920–7 was only -0.7 per cent. Over 1928–32 it was 11.3 per cent, and 10.2 per cent in 1933–41. Here is the unmistakable proximate cause of the move to segregation: Ford learned during 1928–32 that blacks could be placed in the foundry without paying a foundry premium. By filling the department almost exclusively with blacks, Ford was able to eliminate the premium for the foundry as a whole. Was segregation at the foundry an adverse development for blacks as a group? Yes and no. It appears to be a clear example of a company that gains in labour costs, by employing qualified minority workers who are excluded elsewhere. Black workers gained too, because segregation provided a vehicle for employing them in much larger numbers, at auto industry wages far higher than they would otherwise have received. These gains were surely real, yet we cannot overlook the crucial fact that the incentive to arbitrage the labour market was also an incentive to treat blacks differently. Segregation at the foundry not only perpetuated separation, but an age-old conception of the nature and value of the work for which blacks were suited. As such, it made the prophecy self-fulfilling, in that the economic conditions of the victims of racial thinking became the rationale for thinking in racial terms. CONCLUSIONS The literature on post-Second World War developments goes into a wide range of topics that we do not address here. Nonetheless we believe that the research presented here sheds light on some of the issues that have been debated. The most basic of these concerns the heavy weight assigned to the contribution of black schooling to black economic progress, particularly by Smith (1984) and Smith and Welch (1986, 1989). They have emphasized both the quantity and the quality of black schooling, as reflected in such measures as the pupil-teacher ratio, teachers’ salaries, and term length, and as manifested in pronounced Vintage effects’ in earnings regressions. It is undeniably true that there have been significant improvements in relative black schooling since 1940, but Card and Krueger (1992) report that improvements in the quality of black schooling can account at most for about 20 per cent of the narrowing of the gap between black and white earnings between 1960 and 1980. As Donohue and Heckman (1991) point out, the timing of relative black progress does not correspond to the picture of gradual, steady improvement in the quality of black labour. The most striking changes occurred in the south during the 1960s, and were occasioned by the integration of the textiles industry and other long-segregated white 289
WARREN WHATLEY & GAVIN WRIGHT
bastions. Because southern blacks did not passively accept the low-wage, uneducated job status assigned to them, but fought through political and legal channels for better black schooling, the white-only policy seemed economically irrational as well as morally unjustifiable by the 1960s. Heckman and Payner (1989) credit the role of federal pressures in pushing integration in the south, but they also note that these pressures were effective well beyond their objective strength, because employers had an underlying interest in compliance during a time of labour scarcity. The fact that much of the gain in relative black incomes has come through a rise in the return to black education strongly suggests that routine exclusion of blacks from skilled, white-collar, and professional jobs in both regions must have been an important part of the story. In the south, the overwhelming share of the ‘good jobs’ opened up by the post-war industrialization drive were reserved for whites. In the north, racial exclusion policies were nearly as common and thoroughgoing as they were in the south, and were so completely accepted as normal that they rarely came up for discussion. For example, Claudia Goldin reports that the major employers of clerical workers in all parts of the country routinely excluded black women as part of their standard policy, responding to surveys with statements such as ‘no colored hired in office, no discrimination as to race’ (Goldin 1990:147). Myrdal and his associates found, for the same era: ‘The North is almost as strict as the south in excluding Negroes from middle class jobs in the white-dominated economy’ (Myrdal 1944:306). They reported that whereas the black wage earner had a fighting chance to enter skilled blue-collar trades, and was always struggling to do so, the exclusion from white-collar jobs was much more complete and ‘settled’ (Rose 1948:107). Is it reasonable to suppose that the feedback effects of this long history of exclusion on black educational aspirations and achievements could be swept away overnight? Indeed, the entire history of racial employment policies and the struggle to integrate the workplace has been neglected by economists and economic historians, in favour of supply side studies. As late as the 1950s, the exclusion of blacks from white-collar jobs was so routine, in both north and south, that discussions of integration had an abstract, hypothetical quality similar to the debates about blacks in the cotton mills around the turn of the century. In 1956, for example, Eli Ginzberg wrote It is difficult to assess how near we in the United States are to the time when important companies will be willing to recruit young Negro college graduates as executive trainees in the same way that they have long been recruiting white graduates. He noted that some large companies had by then accepted qualified Negroes for ‘scientific and technical’ positions, and suggested that ‘if the trend continues, as it most likely will, a few of these staff people will undoubtedly 290
RACE, HUMAN CAPITAL, AND LABOUR MARKETS
begin to move over into “line” management just as white engineers and scientists frequently do’ (Ginzberg 1956:39–40). In a fascinating survey of business integration cases from the 1950s and 1960s, Steven Gelber (1974) cites example after example where private sector proposals to employ blacks in sales, clerical, or technical jobs were subject to active resistance and debate. The discussions seemed to take for granted that the elevation of blacks to these positions was a fit subject for public decision. The meaning of race was not just socially constructed in some metaphorical sense; it was the direct object of collective public choice. Even if one believes that in some aggregate, long-run sense, the advancement of blacks into these positions depended on prior educational attainments for blacks as a group, it would still be the case that the actual realization of that progress required the overthrow of long-standing and deeply entrenched attitudes and policies. We argue that such attitudes and policies had long been embedded as parameters in many other parts of the labour market, and hence, that there is no clean distinction between the role of prejudice and the role of market forces. The study of this history, however untidy or unconventional, should be high on the research agenda of our field. NOTES We thank the NSF (grants SES-8300696 and SES-8910514) and the University of Michigan (grants from the Office of the Vice President for Research and the Rackham School of Graduate Studies) for financial support. 1 2 3 4
For examples, see Becker (1957), Arrow (1972, 1973), Aigner and Cain (1977), Lang (1986), Milgrom and Oster (1987). See Butler et al. (1989), Heckman and Payner (1989), Donohue and Heckman (1991), Card and Krueger (1992). In addition to Margo (1990a), see Margo and Finegan (1993). An analysis of this development is presented in Wright (1986: ch. 5). We have constructed a sample of 4144 workers in the Detroit area plants of the Ford Motor Company in 1918–47. The records include information on worker characteristics at the time of hire, as well as complete information on workers’ job and wage experiences at Ford. The data and codebooks are stored at the Intercollegiate Consortium for Political and Social Research in Ann Arbor, Michigan.
291
BIBLIOGRAPHY
ABBREVIATIONS A.E.R. Ec.H.R E.E.H J.Ec.H. J.R.S.S.
American Economic Review Economic History Review second series Explorations in Economic History Journal of Economic History Journal of the Royal Statistical Society
OFFICIAL PAPERS International organizations International Labour Office (annual) Bulletin of Labour Statistics, Geneva: ILO. International Labour Office (annual) International Labour Review, Geneva: ILO. International Labour Office (annual) International Labour Review, Statistical Supplement, Geneva: ILO. International Labour Office (1931) Principles and Methods of Wage Determination in the Coal-Mining Industry, Studies and Reports (series D) No. 20, London: P.S. King. International Labour Office (annual) Year-book of Labour Statistics, Geneva: ILO. Organization for Economic Cooperation and Development (1987) National Accounts, vol. 1, Main Aggregates 1970–1985, supplement Purchasing Power Parities, Paris: OECD.
Australia Australia, Bureau of Statistics (1977), Official Yearbook of Australia, no. 61 (1975 and 1976), Canberra: Commonwealth Government Printer. Knibbs, G.H. (1911) Inquiry into the Cost of Living in Australia, 1910–11, Melbourne: Commonwealth Bureau of Census and Statistics. NSW, Government Statistician (1887–1914) Wealth and Progress of New South Wales, Sydney. ——(1919–20) Statistical Register, Sydney. ——(1905–6, 1911, 1913) The Official Year Book of New South Wales, Sydney. New South Wales Legislative Assembly (1891–2) ‘Report on the Census and Industrial Returns Act of 1891’, Votes and Proceedings, vol.7. Queensland, Government Statistician (1879–1914) Statistics of the Colony [of the State], Brisbane.
292
BIBLIOGRAPHY Tasmania, Government Statistician (1879–1914) Statistics of the Colony [of the State] of Tasmania , Hobart.
Canada Coats, R.H.(1915) The Rise in Prices and the Cost of Living in Canada, 1900–1914, Board of Inquiry into Cost of Living, vol.11, Ottawa: King’s Printer. Department of Employment and Immigration (1981) Unemployment Insurance in the 1980s: Report of the Task Force on Unemployment Insurance, Ottawa: Supply and Services. Department of Trade and Commerce, Census Statistics Office (1915) Fifth Census of Canada 1911, vol. 6, Occupations of the People, Ottawa: King’s Printer. Dominion Bureau of Statistics (1926) Canada Year Book, Ottawa: King’s Printer. ——(1929) Sixth Census of Canada 1921, vol. 4, Occupations, Ottawa: King’s Printer. ——(1936) Seventh Census of Canada 1931, vol. 7, Occupations and Industries, Ottawa: King’s Printer.
France Chambre de commerce de Paris (1851) Statistique de l’industrie à Paris résultant de l’enquête faite par la Chambre de commerce pour les années 1847–8, Paris. Ministère de l’agriculture (1897) Statistique agricole de la France. Résultats généraux de l’enquête décennale de 1892, Paris. Ministère du Commerce, de l’industrie, des Postes et des Télégraphes, Office du Travail (1893–7) Salaires et durée du travail dans l’industrie française (4 vols and a volume of figures), Paris: Imprimerie Nationale. Statistique générale de la France (1840–1) Statistique agricole, 1840–41, Paris. ——(1847–52) Enquête industrielle, 1840–45, Paris. ——(1858–60) Statistique agricole décennale de 1852, 2 vols, Paris. ——(1869) Résultats statistiques du dénombrement de 1866, Nancy: Berger-Levrault. ——(1870) Agriculture. Résultats généraux de l’enquête décennale de 1862, Strasbourg. ——(1873) Industrie. Résultats généraux de l’enquête effectuée dans les années 1861– 65, Nancy.
Germany Reichsarbeitsministerium (1924–38) Reichsarbeitsblatt, Amtliche und Nichtamtliche Teile, Berlin. Statistiches Reichsamt (1925–39) Statistiches Jahrbuch fuer des Deutsche Reich, Berlin: Verlag von Reimar Hobbing.
Ireland Central Statistics Office (1949) Census of Population of Ireland 1946, Dublin: Stationery Office. ——(1952–4) Census of Population of Ireland 1951, Dublin: Stationery Office. Commission on emigration and other population problems 1948–54 (1955) Reports, Dublin: The Stationery Office. Department of Industry and Commerce (1927–80) Statistical Abstract of Ireland, Dublin: The Stationery Office. (From 1967, issued by the Central Statistics Office. Also issued by the Irish Free State.)
293
BIBLIOGRAPHY Irish Free State, Department of Industry and Commerce (1928–34) Census of Population 1926, Dublin: Stationery Office. ——(1938–41) Census of Population 1936, Dublin: Stationery Office.
United Kingdom Central Statistical Office (1920–81) Statistical Abstract of the United Kingdom, London: HMSO. Department of Employment (1971) Historical Labour Statistics, 1886–1968, London: HMSO. England and Wales, Registrar-General (1934) Census of England and Wales, 1931, Occupation Tables, London: HMSO. ——(1957) Census of England and Wales, 1951, Occupation Tables, London: HMSO. Ministry of Labour (1921–75) Gazette, London: HMSO. ——(1924–38) Report, London: HMSO. ——(1926) Eighteenth Abstract of Labour Statistics, British Parliamentary Papers, vol. 29. Standing Committee on Trusts (1921) Report on the Soap Industry, British Parliamentary Papers, vol.16. UK, Board of Trade (1893–4) Report on the Wages of the Manual Labour Classes in the United Kingdom, with tables of the average rates of wages and hours of labour of persons employed in several of the principal trades in 1886 and 1891, British Parliamentary Papers, vol. 83, Part II. ——(1900) Report by Mr. Wilson Fox on the Wages and Earnings of Agricultural Labourers in the United Kingdom, British Parliamentary Papers, vol. 82. ——(1903) Report on Wholesale and Retail Prices in the United Kingdom 1902, British Parliamentary Papers, vol.68. ——(1905a) Second Report by Mr. Wilson Fox on the Wages, Earnings, and Conditions of Employment of Agricultural Labourers in the United Kingdom, British Parliamentary Papers, vol. 97. ——(1905b) Second Series of Memoranda, Statistical Tables, and Charts, British Parliamentary Papers, vol.84. ——(1907, 1908c, 1910, 1912–13) Abstract of Labour Statistics, British Parliamentary Papers, vols 80, 98, 107, 110. ——(1908a) Rates of Wages and Hours of Labour in Various Industries in the United Kingdom. Unpublished. ——(1908b) Enquiry into Working Class Rents, Housing and Retail Prices, British Parliamentary Papers, vol. 107. ——(1909, 1910, 1912–13, 1914) Reports on Standard Time Rates of Wages in the United Kingdom, British Parliamentary Papers, vols 80, 84, 92, 80. ——(1911) Report of an Enquiry…into Working Class Rents, Housing and Retail Prices…in the Principal Industrial Towns of the United States of America, British Parliamentary Papers, vol.88. ——(1913) Report of an Enquiry by the Board of Trade into Working-Class Rents and Retail Prices, 1912, British Parliamentary Papers, vol.66. ——(1925) Royal Commission on Food Prices, vol. III, Appendices, London: HMSO. ——(1935) 1930 Census of Production, London: HMSO.
United States Aldrich, Nelson W. (1892) Report on Retail Prices and Wages, US Congress, 52nd Congress, 1st Session, Senate Report 986.
294
BIBLIOGRAPHY Aldrich, Nelson W.(1893) Wholesale Prices, Wages, and Transport, US Congress, 52nd Congress, 2nd Session, Senate Report 1394. Bureau of the Census (1913a) Thirteenth Census of the United States: 1910, vol. VIII, Manufactures, 1909, General Report and Analysis, Washington: GPO. ——(1913b) Thirteenth Census of the United States: 1910, vol. X, Manufactures, 1909, Reports for Principal Industries, Washington: GPO. ——(1933) Fifteenth census of the United States: 1930, vol. IV, Occupations by States, Washington: GPO. ——(1975) Historical Statistics of the United States, Colonial Times to 1970, Bicentennial Edition, Washington: GPO. Bureau of Labor Statistics (1895–1925) Bulletin, Washington: BLS. ——(1980) Job Tenure Declines as Work Force Changes, Special Labor Force Report No. 235, Washington: GPO. Bureau of Labor and Industrial Statistics for the State of Virginia (1909) Twelfth Annual Report, Richmond: Davis Bottom, Superintendent of Public Printing. Bureau of Labor and Industrial Statistics for the State of Virginia (1927) TwentyNinth Annual Report, Richmond: Davis Bottom, Superintendent of Public Printing. Census Office (1895) Eleventh Census of the United States: 1890, vol. VI, Part 3, Report on Manufacturing Industries in the United States, Washington: US Census Office. ——(1902a) Twelfth Census of the United States: 1900, vol. 8, Manufactures, Part II, States and Territories, Washington: US Census Office. ——(1902b) Twelfth Census of the United States: 1900, vol. 10, Manufactures, Part IV, Special Reports on Selected Industries, Washington: US Census Office. Department of Commerce and Labor (1904) Eighteenth Annual Report of the Commissioner of Labor, 1903: Cost of Living and Retail Prices of Food, Washington: GPO. ——(1905) Nineteenth Annual Report of the Commissioner of Labor, 1904: Wages and Hours of Labor , Washington: GPO. Department of the Interior, Bureau of Labor (1888) Third Annual Report of the Commissioner of Labor, 1887: Strikes and Lockouts, Washington: GPO. Kansas Bureau of Labor and Industry (1900, 1901) Annual Report, Topeka: W.Y. Morgan. Michigan Bureau of Labor and Industrial Statistics (1891) Annual Report, Lansing: W.S.George and Co. Senate (1911) ‘Investigation relative to wages and prices of commodities’, Senate Document 847, 61st Congress, 3rd Session, vol. 3.
BOOKS AND ARTICLES Abel, W. (1935) Agrarkrisen und Agrarkonjunktur in Mitteleuropa vom 13. bis zum 19. Jahrhundert, Hamburg and Berlin. ——(1974) Massenarmut und Hungerkrisen im vorindustriellen Europa, Hamburg and Berlin. Aden, O. (1964) Entwicklung und Wechsellagen ausgewählter Gewerbe in Ostfriesland von der mitte des 18. bis zum Ausgang des 19. Jahrhundert, Aurich. Aigner, D.J. and Cain, G. (1977) ‘Statistical theories of discrimination in the labor market’, Industrial and Labor Relations Review 30, April: 175–87. Ainsworth, R.B. (1949) ‘Earnings and working hours of manual wage-earners in the United Kingdom in October 1938’, J.R.S.S. 112, 1:35–65. Akerlof, G.A. and Main, B.G.M. (1981) ‘An experience-weighted measure of employment and unemployment durations’, A.E.R. 71, 5:1003–11. Akerlof, G.A. and Yellen, J. (1988) ‘Fairness and unemployment’, A.E.R. 78, 2: 44–9.
295
BIBLIOGRAPHY ——(1990) ‘The fair wage-effort hypothesis and unemployment’, Quarterly Journal of Economics 105, May: 257–83. Allen, R.C. (1986) ‘The B.C. economy: past, present, future’, in R.C.Allen and G. Rosenbluth (eds) Restraining the Economy: Social Credit Economic Policies for B.C. in the Eighties, Vancouver: New Star Books. Allen, R.C. (1992) Enclosure and the Yeoman: The Agricultural Development of the South Midlands 1450–1850, Oxford: Clarendon. Allen, R.G.D. (1948) ‘Wholesale prices, 1938–1948’, Economic Journal 59, June: 137– 53. Altman, M. (1987) ‘A revision of Canadian economic growth: 1870–1910 (a challenge to the gradualist interpretation)’, Canadian Journal of Economics 20, 1:86–113. Anderson, J.D. (1988) The Education of Blacks in the South, 1860–1935, Chapel Hill: University of North Carolina Press. Ardouin-Dumazet, V.E. (1912) Les petites industries rurales, Paris: V.Lecoffre. Armstrong, W.A. (1989) ‘Rural population growth, systems of employment, and incomes’, in J.Thirsk (ed.) The Agrarian History of England and Wales, vol. 6, 1750–1850, Cambridge: Cambridge University Press. Arrow, K. (1972) ‘Some mathematical models of racial discrimination in the labor market’, in A.H.Pascal (ed.) Racial Discrimination in Economic Life, Lexington, MA: Lexington Books. ——(1973) ‘The theory of discrimination’, in O.Ashenfelter and A.Rees (eds) Discrimination in Labor Markets, Princeton: Princeton University Press. Ashby, J. and King, B. (1893) ‘Statistics of some Midland villages’, Economic Journal 3, March, June: 1–22; 193–204. Atack, J. and Bateman, F. (1989) ‘How long did people work in 1880?’, paper presented at the annual meeting of the Economic History Association, Washington, DC. ——(1992) ‘How long was the workday in 1880?’, J.Ec.H. 52, 1:129–60. Avery, D. (1979) ’Dangerous Foreigners’: European Immigrant Workers and Labour Radicalism in Canada, 1896–1932, Toronto: McClelland and Stewart. Baars, C. (1973) De geschiedenis van de landbouw in de Beijerlanden, Wageningen. Bailer, L. (1943) ‘Negro labor in the automobile industry’, unpublished Ph.D. thesis, University of Michigan. Baily, M.N. (1983) ‘The labor market in the 1930s’, in J.Tobin (ed.) Macroeconomics, Prices, and Quantities, Washington, DC: The Brookings Institution. Bain, G. and Elsheikh, F. (1974) Union Growth and the Business Cycle, Oxford: Basil Blackwell. Bain, G.S. and Price, R. (1980) Profiles of Union Growth, Oxford: Basil Blackwell. Baines, D.E. (1985) Migration in a Mature Economy, Cambridge: Cambridge University Press. Bairoch, P. (1976) ‘Europe’s Gross National Product: 1800–1975’, Journal of European Economic History 5, 2:273–340. Balke, N.S. and Gordon, R.J. (1989) ‘The estimation of prewar Gross National Product: methodology and new evidence’, Journal of Political Economy 97, 1:38–92. Banerjee, A.V. and Newman, A.F. (1993) ‘Occupational choice and the process of development’, Journal of Political Economy 101, 3:274–98. Béaur, G. (1984) Le marché fonder à la veille de la Revolution. Les mouvements de propriété beaucerons dans les regions de Maintenon et de Janville de 1761 à 1790, Paris: Editions de l’Ecole des Hautes Etudes en Sciences Sociales. Becker, G. (1957) The Economics of Discrimination, Chicago: University of Chicago Press.
296
BIBLIOGRAPHY Beenstock, M. and Warburton, P. (1986) ‘Wages and unemployment in interwar Britain’, E.E.H. 23, 2:153–72. Beenstock, M., Capie, F., and Griffiths, G. (1984) ‘Economic recovery in the United Kingdom in the 1930s’, Bank of England Panel Paper 23, April: 57–85. Benham, F.C. (1931) ‘Memorandum on wages, prices and unemployment’, The Economist, June 20:1315–16. Benjamin, D. and Kochin, L. (1979) ‘Searching for an explanation of unemployment in interwar Britain’, Journal of Political Economy 87, 3:441–70. Berman, H.J. (1983) Law and Revolution: The Formation of the Western Legal Tradition, Cambridge, MA: Harvard University Press. Bernanke, B.S. and Powell, J.L. (1986) ‘The cyclical behavior of industrial labor markets: a comparison of the prewar and postwar eras’, in R.J.Gordon (ed.) The American Business Cycle: Continuity and Change, Chicago: Chicago University. Bertram, G.W. (1965) ‘Economic growth in Canadian industry, 1870–1915: the staple model and the take-off hypothesis’, Canadian Journal of Economics and Political Science 29, 2:159–84. Bertram, G.W. and Percy, M.B. (1979) ‘Real wage trends in Canada, 1900–26: some provisional estimates’, Canadian Journal of Economics 12, 2:299–312. Beveridge, W.H. (1930) Unemployment: a problem of industry, London: Longmans, Green. ——(1935) ‘Comment’, J.R.S.S. 98, 4:674–5. Biddick, K. (1985) ‘Medieval English peasants and market involvement’, J.Ec.H. 45, 4:823–31. Bienefeld, M.A. (1972) Working Hours in British Industry: An Economic History, London: Weidenfeld and Nicolson. Blinder, A.S. (1988) ‘The challenge of high unemployment’, A.E.R. 78, 2:1–15. Bompard, J.P., Magnac, T., and Postel-Vinay, G. (1990a) ‘Migrations saisonnières de main d’oeuvre. Le cas de la France en 1860’, Annales d’économic et de statistique 19, juillet/sept.: 97–128. ——(1990b) ‘Emploi, mobilite et chômage en France au xix siècle’, Annales, ESC 45: 55–76. Borchardt, K. (1979) ‘Zwangslagen und Handlungsspielraeume in der großen Wirtschaftskrise der fruehen dreißiger Jahre: Zur Revision des Ueberlieferten Geschichtbildes’, Jahrbuch der Bayrische Akademie der Wissenschaften, 87–132. ——(1991) Perspectives on Modern German Economic History and Policy, Cambridge: Cambridge University Press. Bottin, J. (1983) Seigneurs et paysans dans l’ouest du pays de Caux 1540–1650, Paris: Le Sycomore. Bowley, A.L. (1898) ‘The statistics of wages in the United Kingdom during the last hundred years. Part I. Agricultural wages’, J.R.S.S. 61, December: 702–22. ——(1900a) Wages in the United Kingdom in the Nineteenth Century, Cambridge: Cambridge University Press. ——(1900b) ‘The statistics of wages in the United Kingdom during the last hundred years. Part VI. Wages in the building trades—English towns’, J.R.S.S. 63, June: 297– 315. ——(1901) ‘The statistics of wages in the United Kingdom during the last hundred years. Part VIII. Wages in the building trades, London’, J.R.S.S. 64, March: 102–11. ——(1914) ‘Rural population in England and Wales: a study of the changes of density, occupations, and ages’, J.R.S.S. 77, May: 597–645. ——(1937) Wages and Income in the United Kingdom since 1860, Cambridge: Cambridge University Press.
297
BIBLIOGRAPHY Boxer, C.R. (1965) The Dutch Seaborne Empire, 1600–1800, New York: Knopf. Boyer, G.R. (1986) ‘The determinants of labor migration in southern England: 1851– 1901’, unpublished paper, Cornell University. Boyer, G.R. and Hatton, T.J. (1994 forthcoming) ‘Did Joseph Arch raise agricultural wages? Rural trade unions and the labour market in late-nineteenth century England’, Ec.H.R. Braudel, F. and Spooner, F. (1967) ‘Prices in Europe from 1450 to 1750’, in E.E.Rich and C.H.Wilson (eds) The Cambridge Economic History of Europe, vol. 4, Cambridge: Cambridge University Press. Brissenden, P. and Frankel, E. (1922) Labor Turnover in Industry, New York: Macmillan. Broadberry, S.N. (1983) ‘Unemployment in interwar Britain: a disequilibrium approach’, Oxford Economic Papers 35 (Supplement): 463–85. ——(1986) The British Economy between the Wars: A Macroeconomic Survey, Oxford: Basil Blackwell. Brody, D. (1960) Steelworkers in America: The Non-Union Era, Cambridge MA: Harvard University Press. Brown, J.C. (1990) ‘The condition of England and the standard of living: cotton textiles in the northwest, 1806–1850’, J.Ec.H. 40, 3:591–614. Brown, M. and Phillips, P. (1985) ‘The evolution of labor market structure: the California canning industry’, Industrial and Labor Relations Review 38, October: 392–404. Brown, R.C. and Loveridge, D. (1982) ‘Unrequited faith: recruiting the CEF 1914– 1919’, Revue Internationale d’histoire militaire 54:53–79. Bruijn, J.R. and Lucassen, J. (1980) Op de schepen der Oost-Indische Compagnie, Groningen. Bry, G. (1960) Wages in Germany 1871–1945, Princeton: Princeton University Press. Buckley, K. and Wheelwright, T. (1988) No Paradise for Workers: Capitalism and the Common People in Australia, 1788–1914, Melbourne: Oxford University Press. Burkitt, B. (1974) ‘The calculation of the degree of unionization’, British Journal of Industrial Relations 12:449–58. Butler, R., Heckman, J., and Payner, B. (1989) ‘The impact of the economy and the state on the economic status of blacks’, in D.W.Galenson (ed.) Markets in History, Cambridge: Cambridge University Press. Butlin, N.G. (1962) Australian Domestic Product, Investment and Foreign Borrowing, 1861–1938/39, Cambridge: Cambridge University Press. Byington, M.F. (1910) Homestead. The households of a Mill Town, New York: Columbia University Press. Bythell, D. (1969) The Handloom Weavers. A Study in the English Cotton Industry during the Industrial Revolution, Cambridge: Cambridge University Press. Cahill, M.C. (1932) Shorter Hours: A Study of the Movement Since the Civil War, Columbia University Studies in History, Economics and Public Law, no. 380, New York: Columbia University Press. Caird, J. (1852, 1967 edn) English Agriculture in 1850–51, Reprinted 1967. London: Frank Cass. Cairncross, A.K. (1949) ‘Internal migration in Victorian England’, Manchester School of Economic and Social Studies 17, 1:67–87. Campbell, B.M.S. (1988) ‘Towards an agricultural geography of Medieval England’, Agricultural History Review 36:87–98. ——(1992) ‘The impact of concentrated urban demand upon agricultural production in late thirteenth century England: some preliminary thoughts’, paper presented to the annual conference of the Economic History Society, Leicester, April, 1992.
298
BIBLIOGRAPHY Card, D. and Krueger, A.B. (1992) ‘School quality and black-white relative earnings: a direct assessment’, Quarterly Journal of Economics 107, February: 151–200. Carlson, L. (1981) ‘Labor supply, the acquisition of skills and the location of southern textile mills’, J.Ec.H. 41, 1:65–71. Carmichael, H.L. (1989) ‘Self-enforcing contracts, shirking and life cycle incentives’, The Journal of Economic Perspectives 3, 4:65–83. Caron, F. (1973) Histoire de l’exploitation d’un grand réseau: la compagnie du Nord, Paris: Mouton. Carter, S.B. (1988) ‘The changing importance of lifetime jobs, 1892–1978’, Industrial Relations 27, 3:287–300. Carter, S.B. and Savoca, E. (1990) ‘Labor mobility and lengthy jobs in nineteenthcentury America’, J.Ec.H. 50, 1:1–16. ——(1991) ‘Gender differences in learning and earning in nineteenth-century America: the role of expected job and career attachment’, E.E.H. 28, 3:323–43. Carter, S.B. and Sutch, R. (1989) ‘Micro-level data sets suitable for investigation of macroeconomic issues extracted from reports of the State Bureaus of Labor Statistics, circa 1890’, University of California, Berkeley, Department of Economics Working Paper No. 89–114. ——(1991) ‘Sticky wages, short weeks, and ‘fairness’: the response of Connecticut manufacturing firms to the depression of 1893–94’, Historical Labor Statistics Project Working Paper 2, University of California. Casson, M. (1983) Economics of Unemployment. An Historical Perspective, Oxford: M.Robertson. Chambers, E.J. and Gordon, D.F. (1966) ‘Primary products and economic growth: an empirical measurement’, Journal of Political Economy 74, 4:315–32. Chapman, A. (1953) Wages and Salaries in the United Kingdom, 1920–1938, Cambridge: Cambridge University Press. Chatelain, A. (1976) Les migrants temporaires en France de 1800 à 1914, Villeneuved’Ascq: PUL. Clapham, J.H. (1938) An Economic History of Modern Britain, vol. 3, Cambridge: Cambridge University Press. Clark, K.B. and Summers, L.H. (1979) ‘Labor market dynamics and unemployment: a reconsideration’, Brookings Papers on Economic Activity, 1:13–60. Clarke, P. (1990) ‘The Treasury’s analytical model of the British economy between the wars’, in M.O.Furner and B.Supple (eds) The State and Economic Knowledge: The American and British Experiences, Cambridge: Cambridge University Press. Clay, H. (1929) The Problem of Industrial Relations, London: Macmillan. Clifford, F. (1875) ‘The labour bill in farming’, Journal of the Royal Agricultural Society 11:67–127. Coebergh van den Brink, S.M. (1988) Meer dan zes eeuwen Leids Gymnasium, Leiden. Coelho, P.R. and Shepherd, J. (1979) ‘The impact of regional differences in prices and wages on economic growth: the United States in 1890’, J.Ec.H. 39, 1:69–85. Coleman, D.C. (1956) ‘Labour in the English economy of the seventeenth century’, Ec.H.R. 8, 3:280–95. ——(1987) History and the Economic Past: An Account of the Rise and Decline of Economic History in Britain, Oxford: Clarendon Press. Collins, E.J.T. (1976) ‘Migrant labour and British agriculture in the nineteenth century’, Ec.H.R. 29, 1:38–59. Commons, J.R. (1907) Races and Immigrants in America, New York: Macmillan. ——(1910) A Documentary History of American Industrial Society, vol. 3, Cleveland, OH: A.H.Clark.
299
BIBLIOGRAPHY Cowling, K. (1966), ‘Determinants of wage inflation in Ireland’, Economic Research Institute Paper Series no. 31, Dublin. Crafts, N.F.R. (1982) ‘Regional price variations in England in 1843: an aspect of the standard of living debate’, E.E.H. 19, 1:51–70. ——(1983) ‘Gross National Product in Europe, 1879–1910:some new estimates’, E.E.H. 20, 4:387–401. Craven, P. and Traves, T. (1980) ‘Dimensions of paternalism: discipline and culture in Canadian railway operations in the 1850s’ , in C.Heron and R.Storey (eds) On the Job: Confronting the Labour Process in Canada, Montreal: McGill-Queen’s University Press. Crebouw, Y. (1986) ‘Salaires et salaries agricoles en France des débuts de la Revolution aux approches du XX éme siècle’, unpublished doctoral thesis, Université de Paris I. Cross, G. (1988) ‘Worktime in international discontinuity, 1886–1940’, in G.Cross (ed.) Worktime and Industrialization: An International History, Philadelphia: Temple University Press. Cullen, L. (1987) An economic history of Ireland since 1660, 2nd edn, London, Batsford. David, P.A. (1987) ‘Industrial labor market adjustments in a region of recent settlement, Chicago 1848–1868’, in P.Kilby (ed.) Quantity and Quiddity: Essays in U.S. Economic History, Middletown, CT: Wesleyan University Press. David, P.A. and Solar, P. (1977) ‘A bicentenary contribution to the history of the cost of living in America’, Research in Economic History 2:1–80. Deane, P. and Cole, W.A. (1962) British Economic Growth 1688–1959, Cambridge: Cambridge University Press. Deceulaer, H. (1992) ‘Arbeidsregulering en loonvorming in de Antwerpse haven, 1585–1796’, Tijdschrift voor sociale geschiedenis 18:22–45. Delatouche, R. (1977) ‘Regards sur l’agriculture aux temps Carolingiens’, Journal des Savants, 73–100. Devons, E. (1956) British Economic Statistics, Cambridge: Cambridge University Press. De Vries, J. (1981) Barges and Capitalism: Passenger Transportation in the Dutch Economy (1632–1839), Utrecht. ——(1984a) ‘The decline and rise of the Dutch economy, 1675–1900’, in G.Saxonhouse and G.Wright (eds) Technique, Spirit and Form in the Making of the Modern Economies: Essays in Honor of William N.Parker (Research in Economic History, Supplement 3). ——(1984b) European Urbanization, 1500–1800, Cambridge, MA: Harvard University Press. ——(1993) ‘The labour market’, Economic and Social History in the Netherlands 4, 55–78. Dickens, W.T. and Lang, K. (1988) ‘The reemergence of segmented labour market theory’, A.E.R. 78, 2:129–34. Dicks-Mireaux, L.A. and Shepherd, J.R. (1962) ‘The wages structure and some implications for incomes policy’, National Institute Economic Review 22:38–48. Dimsdale, N. (1984) ‘Employment and real wages in the interwar period’, National Institute Economic Review 110:94–103. Dimsdale, N., Nickell, S., and Horsewood, N. (1989) ‘Real wages and unemployment in Britain during the 1930s’, Economic Journal 99, June: 271–92. Donohue, J.H. III and Heckman, J. (1991) ‘Continuous versus episodic change: the impact of civil rights policy on the economic status of blacks’, Journal of Economic Literature 29, 4:1603–43. Douglas, P.H. (1921) American Apprenticeship and Industrial Education, New York.
300
BIBLIOGRAPHY ——(1930) Real Wages in the United States: 1890–1926, Boston: Houghton Mifflin. ——(1957) The Theory of Wages, New York: Kelly & Millman. Douglas, P.H. and Director, A. (1931) The Problem of Unemployment, New York: Macmillan. Duby, G. (1962, 1972 edn) L’économie rurale et la vie des campagnes dans l’occident medieval, Paris: Flammarion. Dunlevy, J.A. and Saba, R.P. (1992) ‘The role of nationality-specific characteristics on the settlement patterns of nineteenth-century immigrants’, E.E.H. 29, 2: 228–49. Dyer, C. (1988) ‘Changes in diet in the late Middle Ages: the case of harvest workers’, Agricultural History Review 36:21–37. Eichengreen, B. (1986) ‘Unemployment in interwar Britain: new evidence from London’, Journal of Interdisciplinary History 17, Autumn: 335–58. ——(1992) ‘The origins and nature of the Great Slump revisited’, Ec.H.R. 45, 2: 213– 39. Eichengreen, B. and Hatton, T.J. (1988) ‘Introduction’, in B.Eichengreen and T.J. Hatton (eds) Interwar Unemployment in Historical Perspective, Dordrecht: Kluwer. Elbaum, B. (1989) ‘Why apprenticeship persisted in Britain but not in the United States’, J.Ec.H. 39, 2:337–49. Elsas, M.J. (1936–49) Umriss enier Geschichte derPreise und Löhne in Deutschland, 2 vols, Leiden. Emery, J.C.H. (1993) ‘The decline of fraternal methods of social insurance: a study of the Independent Order of Odd Fellows sickness insurance in British Columbia, 1874–1951’ , unpublished Ph.D. thesis, University of British Columbia. Engerman, S. (1992) ‘Coerced and free labor: property rights and the development of the labor force’, E.E.H. 29, 1:1–29. Engerman, S. and Goldin, C. (1991) ‘Seasonality in Nineteenth-Century Labor Markets’, NBER Working Paper Series on Historical Factors in Long Run Growth #20 (January, 1991). Engle, R.F. and Granger, C.W.J. (1987) ‘Cointegration and error correction: representation, estimation and testing’, Econometrica 55, 2:251–76. Engrand, C. (1979) ‘Concurrences et complementarités des villes et des campagnes: Les manufactures picardes de 1780 à 1815’, Revue du Nord 61, 240:61–82. Erickson, C. (1957) American Industry and the European Immigrant, 1860–1885, Cambridge, MA: Harvard University Press. Feinstein, C.H. (1972) National Income, Expenditure and Output of the United Kingdom 1855–1965, Cambridge: Cambridge University Press. ——(1988) ‘The rise and fall of the Williamson curve , J.Ec.H. 43, 3:699–729. ——(1990) ‘What really happened to real wages?: trends in wages, prices, and productivity in the United Kingdom, 1880–1913’, Ec.H.R. 43, 3:329–55. ——(1991) ‘A new look at the cost of living, 1870–1914’, in J.Foreman-Peck (ed.) New Perspectives on the Victorian Economy, Cambridge: Cambridge University Press. Feldman, H. (1925) The Regularization of Employment, New York: Harper and Brothers. Firestone, O.J. (1958) Canada’s Economic Development, 1867–1953, London: Bowes & Bowes. Fishback, P. (1989) ‘Can competition among employers reduce governmental discrimination? Coal companies and segregated schools in West Virginia in the early 1900s’, Journal of Law and Economics 32, October: 311–28. Fisher, L.H. (1953) The Harvest Labor Market in California, Cambridge, MA: Harvard University Press. Fitzgerald, S. (1987) Rising Damp: Life and Work in Sydney, 1870–90, Melbourne: Oxford University Press.
301
BIBLIOGRAPHY Flux, A.W. (1929) ‘The national income’, J.R.S.S. 92, 4:587–93. Fox, A.W. (1903) ‘Agricultural wages in England and Wales during the last fifty years’, J.R.S.S. 66, June: 273–348. Friedlander, D. and Roshier, R.J. (1966) ‘A study of internal migration in England and Wales, Part 1’, Population Studies 19, 3:239–79. Friedman, G. (1992) ‘Dividing labor: urban politics and big-city construction in latenineteenth century America’, in C.Goldin and H.Rockoff (eds) Strategic Factors in Nineteenth Century American History: A Volume to Honor Robert W.Fogel, Chicago: University of Chicago Press. Gallman, R.E. (1966) ‘Gross national product in the United States, 1834–1909’, in Output, Employment and Productivity in the United States after 1800, NBER Studies in Income and Wealth, vol. 30, New York: Columbia University Press. Ganshoff, F.-L. (1975) Le polyptyque de l’Abbaye de Saint-Bertin, Paris. Gazeley, I. (1989) ‘The cost of living for urban workers in late Victorian and Edwardian Britain’, Ec.H.R. 42, 2:207–21. Gazeley, I. and Rice, P. (1992) ‘Wages in the interwar period and the determination of employment’, in N.Crafts and S.Broadberry (eds) Britain in the World Economy, 1870–1939: Essays in honour of Alec Ford, Cambridge: Cambridge University Press. Geary, P. and McCarthy, C. (1976) ‘Wage and price determination in a labour-exporting country: the case of Ireland’, European Economic Review 8, 3:219–33. Geary, P. and Ó Gráda, C. (1989) ‘Post-war migration between Ireland and the United Kingdom: models and estimates’, in I.Gordon and A.P.Thirlwall (eds) European factor mobility: trends and consequences, New York: St Martin’s Press. Geary, R. (1935–6) ‘The future population of Saorstat Eireann and some observations on population statistics’, Journal of the Statistical and Social Inquiry Society of Ireland 15:15–35. Geary, R. and Hughes, J.C. (1970) ‘Certain aspects of non-agricultural unemployment in Ireland’, Economic and Social Research Institute, Paper Series no. 52, Dublin. Gelber, S. (1974) Black Men and Businessmen, Port Washington, NY: Kennikat Press. Gerber, J. (1991) ‘Public school education in the plantation states’, E.E.H. 28, 3: 309– 22. Ginzberg, E. (1956) The Negro Potential, New York: Columbia University Press. Gitelman, H.M. (1974) Workingmen of Waltham, Baltimore: Johns Hopkins University Press. Goldin, C. (1990) Understanding the Gender Gap, New York: Oxford University Press. Goldstone, J. (1984) ‘Urbanization and inflation: lessons from the English price revolution of the sixteenth and seventeenth centuries’, American Journal of Sociology 89, 5:1122–60. Gordon, R.J. (1982) ‘Why U.S. wage and employment behaviour differs from that in Britain and Japan’, Economic Journal 92, 1:13–44. Gourvish, T.R. (1979) The standard of living, 1890–1914’, in A.O’Day (ed.) The Edwardian Age: Conflict and Stability, 1900–1914, London: Macmillan. Grantham, G. (1989) ‘Jean Meuvret and the subsistence problem in early modern France’, J.Ec.H. 49, 1:184–200. ——(1993) ‘Divisions of labour: agricultural productivity and occupational specialization in pre-industrial France’, Ec.H.R. 46, 3:478–502. Green, A.G. (1976) Immigration and the Postwar Canadian Economy, Toronto: Macmillan. ——(1986) ‘Growth and productivity change in the Canadian railway sector, 1871– 1926’, in S.L.Engerman and R.E.Gallman (eds) Long-Term factors in American Economic Growth, Chicago: University of Chicago Press.
302
BIBLIOGRAPHY Green, A.G. and Green, D.A. (1993) ‘Balanced growth and the geographical distribution of European immigrant arrivals to Canada, 1900–1912’, E.E.H. 30, 1: 31–59. Green, A.G. and MacKinnon, M. (1988) ‘Unemployment and relief in Canada’, in B. Eichengreen and T.J.Hatton (eds) Interwar Unemployment in International Perspective, Dordrecht: Kluwer. Green, D.A. and Riddell, W.C. (1993) ‘Qualifying for Unemployment Insurance: an empirical analysis for Canada’, unpublished paper, University of British Columbia. Habakkuk, H.J. (1962) American and British Technology in the Nineteenth Century, Cambridge: Cambridge University Press. Haig, B. (1989) ‘International comparisons of Australian GDP in the 19th Century’, Review of Income and Wealth 35, 2:151–62. Haines, M.R. (1979) ‘Industrial work and the family life cycle 1889–1890’, Research in Economic History 4:289–356. ——(1989) ‘A state and local consumer price index for the United States in 1890’, Historical Methods 22, 3:97–105. Hall, J.D., Korstad, R., and Leloudis, J. (1986) ‘Cotton mill people: work, community and protest in the textile south, 1880–1940’, American Historical Review 91, April: 245–86. Hall, R.E. (1972) ‘Turnover in the labor force’, Brookings Papers on Economic Activity 3:709–56. ——(1975) ‘The rigidity of wages and the persistence of unemployment’, Brookings Papers on Economic Activity 2:301–35. ——(1980) ‘Employment fluctuations and wage rigidity’, Brookings Papers on Economic Activity 1:91–123. ——(1982) ‘The importance of lifetime jobs in the US economy’, A.E.R. 72, 4: 716–24. Hall, R.L. and Hitch, C.J. (1938) ‘Price theory and business behaviour’, Oxford Economic Papers 2, 1:12–45. Hamburger, L. (1940) How Nazi Germany has Mobilized and Controlled Labour, Washington, DC: Brookings Institution. Hanes, C. (1990) ‘Explaining a decrease in cyclical wage flexibility in the late nineteenth century’, paper presented at the Thirteenth Annual Cliometrics Conference, Urbana, Illinois. Hannon, J.U. (1984) ‘Poverty in the antebellum northeast: the view from New York state’s poor relief rolls’, J.Ec.H. 44, 4:1007–32. Harley, C.K. (1974) ‘Skilled labour and the choice of technique in Edwardian industry’, E.E.H. 11, 4:391–414. Harris, C.V. (1985) ‘Stability and change in discrimination against black public schools: Birmingham, Alabama, 1871–1931’, Journal of Southern History 51: 375–416. Hart, R.A. and MacKay, D.I. (1975) ‘Engineering earnings in Britain, 1914–68’, J.R.S.S. 138, 1:32–50. Hartwich, H. (1967) Arbeitsmarkt, Verbaende, und Staat. 1918–1933, Berlin: de Gruyter. Hatton, T.J. and Williamson, J.G. (1991) ‘Unemployment, employment contracts, and compensating wage differentials: Michigan in the 1890s’, J.Ec.H. 51, 3: 605–32. ——(1992) ‘What explains wage gaps between farm and city? Exploring the Todaro model with American evidence 1890–1941’, Economic Development and Cultural Change 40, 2:267–94. Hatzfeld, H. (1971) Du pauperisme a la securite sociale, Paris: A.Colin. Hébert-Barrat, S. (1980) Evolution des traditions agricole en Champagne Pouilleuse 1750–1914, Chalons-sur-Marne.
303
BIBLIOGRAPHY Heckman, J. and Payner, B. (1989) ‘Determining the impact of federal antidiscrimination policy on the economic status of blacks: a study of South Carolina’, A.E.R. 79, 1:138–77. Heffer, J. (1986) Le port de New York et le commerce extérieur américain, 1860–1900, Paris: Publications de la Sorbonne. Heron, C. (1988) Working in Steel: The Early Years in Canada, 1883–1935, Toronto: McClelland and Stewart. Higgs, R. (1977) Competition and Coercion: Blacks in the American Economy 1865– 1914, New York: Cambridge University Press. ——(1978) ‘Racial wage differentials in agriculture: evidence from North Carolina in 1887’, Agricultural History 52:308–11. Hill, T.P. (1971) Measuring Real Product, Paris: OECD. Hines, W.D. (1928) War History of American Railroads, New Haven: Yale University Press. Hirsch, B.T. and Addison, J.T. (1986) Economic Analysis of Unions: New Approaches and Evidence, Boston: Allen and Unwin. Hobsbawm, E.J. (1964) Labouring Men: Studies in the History of Labour, London: Weidenfeld and Nicolson. Hodne, F. and Gjolberg, O. (1981) ‘Market integration during the period of industrialisation in Norway’, in P.Bairoch and M.Levy-Leboyer (eds) Disparities in Economic Development since the Industrial Revolution, London: Macmillan. Hoffmann, W. (1965) Das Wachstum der Deutschen Wirtschaft seit der Mitte des 19. Jahrhunderts, Berlin: Springer Verlag. Holtfrerich, C. (1983) ‘Arbeitslosigkeit, Sozialabbau, Demokratieverlust. Ergebnis zu hoher Lohn in der Weimarer Republik?’ Gewerkschaftliches Monatshefte 34: 714– 22. ——(1984) ‘Zu hohe Lohne in der Weimarer Republik? Bemerkungen zur BorchardtThese’, Geschichte und Gesellschaft 10:122–41. Hoover, E.D. (1960) ‘Retail prices after 1850’, in Trends in the American Economy in the Nineteenth Century, NBER Studies in Income and Wealth, vol.24, Princeton: Princeton University Press. Huberman, M. (1991) ‘How did labor markets work in Lancashire? More evidence on prices and quantities in cotton spinning 1822–1852’, E.E.H. 28 June: 87–120. Hunnicutt, B.K. (1988) Work Without End: Abandoning Shorter Hours for the Right to Work, Philadelphia: Temple University Press. Hunt, E.H. (1973) Regional Wage Variations in Britain 1850–1914, Oxford: Oxford University Press. ——(1981) British Labour History 1815–1914, London: Weidenfeld and Nicolson. ——(1986) ‘Industrialisation and regional inequality: wages in Britain, 1760–1914’, J.Ec.H. 46, 4:935–55. Igartua, J.E. (1987) ‘La mobilité professionnelle des travailleurs d’aluminium a Arvida, 1925–1940’, Labour/Le Travail 20, Fall: 33–60. Innis, H.A. (1923) History of the Canadian Pacific Railway, Toronto: McClelland and Stewart. Institut fuer Konjunkturforschung (1928–38) Wochenbericht, Berlin: IfK. ——(1936) Konjunktur-Statistisches Handbuch 1936, Berlin: IfK. Jackson, R.V. (1987) ‘The structure of pay in nineteenth-century Britain’, Ec.H.R. 40, 4:561–70. Jacoby, S.M. (1983) ‘Industrial Labor Mobility in Historical Perspective’, Industrial Relations 72, 2:261–82.
304
BIBLIOGRAPHY ——(1984) ‘The origins of internal labor markets in American manufacturing firms’, in P.Osterman (ed.) Internal Labor Markets, Cambridge, MA: MIT Press. ——(1985) Employing Bureaucracy: Managers, Unions, and the Transformation of Work in American Industry, 1900–1945, New York: Columbia University Press. Jacoby, S.M. and Sharma, S. (1992) ‘Employment duration and industrial labor mobility in the United States, 1880–1980’, J.Ec.H. 52, 1:161–79. James, H.D. (1986) The German Slump: Politics and Economics 1924–1936, Oxford: Clarendon Press. James, J.A. (1989) ‘The stability of the 19th-century Phillips Curve relationship’, E.E.H.26, 2:117–34. ——(1991) ‘Reconstructing the pattern of unemployment in postbellum Massachusetts’, unpublished paper, University of Virginia. ——(1992) ‘Cyclical wage flexibility in early industrial labor markets: the United States, 1840–1891’, unpublished paper, University of Virginia. James, J.A. and Skinner, J.S. (1985) ‘The resolution of the labor scarcity paradox’, J.Ec.H. 45:513–40. Johnson, C.S. (1943) Patterns of Negro Segregation, New York: Harper. Jones, E.B. (1963) ‘New estimates of hours of work per week and hourly earnings, 1900–1957’, Review of Economics and Statistics 45, November: 374–85. Kahneman, D., Knetsch, J.L., and Thaler, R. (1986) ‘Fairness as a constraint on profit seeking: entitlements in the market’, A.E.R. 76, 4:728–41. Kaim-Caudle, P.R. (1964) ‘Social security in Ireland and Western Europe’, Economic Research Institute Paper Series no. 20, Dublin. Kaplan, S.L. (1989) ‘La lutte pour le contrôle du marché du travail à Paris au xviiie siècle’, Revue d’histoire moderne et contemporaine 36, juillet-sept.: 362–412. Katzman, D. (1973) Before the Ghetto: Black Detroit in the Nineteenth Century, Urbana: University of Illinois Press. Keenan, J.G. (1978) ‘Unemployment, emigration and the labour force’, in B. Dowling and J.Durkan (eds) Irish economic policy: a review of major issues, Dublin: Economic and Social Research Institute. Kendrick, J.W. (1961) Productivity Trends in the United States, NBER General Series no. 71, Princeton: Princeton University Press. Kerr, D. and Holdsworth, D.W. (eds) (1990) Historical Atlas of Canada, vol. III, Toronto: University of Toronto Press. Kerr, K.A. (1968) American Railroad Politics 1914–1920: Rates, Wages, and Efficiency, Pittsburgh: University of Pittsburgh Press. Kesteloo, H.M. (1881) De stadsrekeningen van Middelburg, 10 vols, Middelburg. Keynes, J.M. (1936) The General Theory of Employment, Interest and Money, London: Macmillan. ——(1939) ‘Relative movements of real wages and output’, Economic Journal 49, March: 34–51. Keyssar, A. (1986) Out of Work: The First Century of Unemployment in Massachusetts, Cambridge: Cambridge University Press. Kinoshita, T. (1987) ‘Working hours and hedonic wages in the market equilibrium’, Journal of Political Economy 95, 6:1262–77. Klein, P.W. (1977) ‘De zeventiende eeuw’, in J.H.van Stuivenberg (ed.) De economische gescbiedenis van Nederland, Groningen. Kneisner, T.J. (1976) ‘The full-time workweek in the United States, 1900–1970’, Industrial and Labor Relations Review 30, October: 3–15. Kooijmans, L. (1985) Onder regenten. De elite in een Hollandse stad, Hoorn 1700– 1780, Hollandse Historische Reeks 4.
305
BIBLIOGRAPHY Kravis, I., Heston, A., and Summers, R. (1982) World Product and Income: International Comparisons of Real Gross Product, Baltimore: Johns Hopkins University Press. Krugman, P. (1987) ‘Increasing returns and the theory of international trade’, in T.F. Bewley (ed.) Advances in Economic Theory: Fifth World Congress, Cambridge: Cambridge University Press. ——(1991) Geography and Trade, Cambridge, MA: MIT Press. Kussmaul, A. (1981) ‘The ambiguous mobility of farm servants’, Ec.H.R. 34, 2: 222–35. Lamb, W.K. (1977) History of the Canadian Pacific Railway, New York: Macmillan. Lang, K. (1986) ‘A language theory of discrimination’, Quarterly Journal of Economics 101, May: 363–82. Latimer, M.W. (1932) Industrial Pension Systems in the United States and Canada, New York: Industrial Relations Counselors. Lawrence, F.W. (1899) Local Variations in Wages, London: Longmans, Green. Layard, R. and Nickell, S. (1985) ‘The causes of British unemployment’, National Institute Economic Review 111:62–85. Lazear, E. (1979) ‘Why is there mandatory retirement ?’, Journal of Political Economy 87, 6:1261–84. Lazonick, W. (1979) ‘Industrial relations and technical change: the case of the selfacting mule’, Cambridge Journal of Economics 3:231–62. ——(1981) ‘Production relations, labor productivity, and choice of technique. British and U.S. cotton spinning’, J.Ec.H. 41, 3:491–516. Lebergott, S. (1964) Manpower in economic growth, New York: McGraw-Hill. ——(1990) ‘“Wage rigidity” in the depression: concept or phase?’, unpublished paper, Wesleyan University. Lee, J. and Fahey, C. (1986) ‘A boom for whom? Some developments in the Australian labour market, 1870–91’, Labour History 50:1–27. Lequin, Y. (1977) Les ouvriers de la region Lyonnaise (1848–1914), Lyon: Presse Universitaire de Lyon. Le Roy Ladurie, E. (1966) Les Paysans de Languedoc, Paris: SEVPEN. ——(1974) ‘L’histoire immobile’, Annales Economies, Societés, Civilisation 29, 3: 673– 82. ——(1977) ‘Motionless history’, Social Science History 1:115–36 (translation of (1974)). Lewis, H.G. (1963) Unionism and Relative Wages in the United States, Chicago: University of Chicago Press. Lewis, W.A. (1978) Growth and Fluctuations, 1870–1913, London: George, Allen & Unwin. Licht, W. (1983) Working for the Railroad: The Organization of Work in the Nineteenth Century, Princeton: Princeton University Press. ——(1992) Getting Work: Philadelphia, 1840–1950, Cambridge MA: Harvard University Press. Lindert, P. (forthcoming) ‘Unequal living standards’, in R.Floud and D.McCloskey (eds) The Economic History of Britain since 1700, 2nd edn, vol. 1, Cambridge: Cambridge University Press. Loelhoeffel, M. (1974) ‘Zeitreihen fuer den Arbeitsmarkt. Lohnsatz, Beschaeftigungsfaelle, Arbeitskosten und Arbeitsstunden’, Ifo Studien 20:33–150. London and Cambridge Economic Service (1930) Bulletin, vol. 8. Long, C.D. (1960) Wages and Earnings in the United States: 1860–1890, NBER, General Series No.67, Princeton: Princeton University Press. Lucassen, J. (1984) Naar de kusten van de Noordzee: Trekarbeid in Europees perspektief, 1600–1800, Gouda.
306
BIBLIOGRAPHY Lundberg, S.J. and Startz, R. (1983) ‘Private discrimination and social intervention in competitive labor markets’, A.E.R. 73, 3:340–7. Lyons, J.S. (1989) ‘Family response to economic decline: handloom weavers in early nineteenth-century Lancashire’, Research in Economic History 12:45–91. McCaffray, S.P. (1987) ‘Origins of labor policy in the Russian coal and steel industry, 1874–1900’, J.Ec.H. 47, 4:951–65. McCloskey, D.N. (1993) ‘The extent of the market and the faculty of speech’, paper delivered to the European Historical Economics Workshop on Market Integration from the Renaissance to the Present, Lerici, Italy, April 1993. MacKinnon, M. (1993a) ‘New evidence on Canadian wage rates, 1900–1930’, McGill University, Department of Economics Working Paper 5/93. ——(1993b) ‘Providing for faithful servants: pensions at the Canadian Pacific Railway’, unpublished paper, McGill University. Mackintosh, W.A. (1964) The Economic Background of Dominion-Provincial Relations , Toronto, McClelland and Stewart. (First published 1939) McLean, I.W. and Pincus, J.J. (1983) ‘Did Australian living standards stagnate between 1890 and 1940?’, J.Ec.H.43, 1:193–202. Maddison, A. (1982) Phases of Capitalist Development, Oxford: Oxford University Press. Maddock, R. and McLean, I. (1987) The Australian Economy in the Long Run, Cambridge: Cambridge University Press. Main, B.G.M. (1981) ‘The length of employment and unemployment in Great Britain’, Scottish Journal of Political Economy 28, 2:146–64. Maloney, T.N. and Whatley, W. (1992) ‘Making the effort: the racial contours of Detroit’s labor markets, 1918–1948’, paper presented at the meetings of the Cliometric Society, Miami, Ohio. Marchand, O. and Thélot, C. (1991) Deux siècles de travail en France, Paris: INSEE. Margo, R. (1990a) Race and Schooling in the South 1880–1950, Chicago: University of Chicago Press. ——(1990b) ‘Unemployment in 1910’, unpublished paper, Vanderbilt University. Margo, R. and Finegan, T.A. (1993) ‘The decline in black teenage labor force participation in the south: the role of schooling’, A.E.R. 83, 1:234–47. Marley, J. and Campion, H. (1940) ‘Changes in salaries in Great Britain, 1924–1939’, J.R.S.S. 103, 4:524–33. Marx, K. (1887 edn, 1961) Capital, vol. 1, Moscow: Foreign Languages Publishing House. Mason, T. (1977) Sozialpolitik im dritten Reich: Arbeiterklasse und Volksgemeinschaft, Opladen. Mater, D.H. (1940) ‘The development and operation of the railroad seniority system’, Journal of Business 13, October: 387–419. Matthews, R.C.O. (1990) ‘Marshall and the labour market’, in J.K.Whitaker (ed.) Centenary Essays on Alfred Marshall, Cambridge: Cambridge University Press. Matthews, R.C.O., Feinstein, C.H., and Odling-Smee, J.C.(1982) British Economic Growth, 1856–1973, Stanford: Stanford University Press. Meenan, J. (1970) The Irish economy since 1922, Liverpool: Liverpool University Press. Meuvret, J. (1987) Le problème des subsistances à l’époque Louis XIV. La production des céréales et la société rurale, Paris: Mouton. Milgrom, P. and Oster, S. (1987) ‘Job discrimination, market forces, and the invisibility hypothesis’, Quarterly Journal of Economics 102, 3:453–76. Milward, R. (1981) ‘The emergence of wage labor in early modern England’, E.E.H. 18, 1:21–39.
307
BIBLIOGRAPHY Miskimin, H.A. (1977) The Economy of Later Renaissance Europe 1460–1600, Cambridge: Cambridge University Press. Mitchell, B.R. (1990) British Historical Statistics, Cambridge: Cambridge University Press. Mitchell, B.R. and Deane, P. (1962) Abstract of British Historical Statistics, Cambridge: Cambridge University Press. Mitchell, B.R. and Jones, H.G. (1971) Second abstract of British historical statistics, Cambridge: Cambridge University Press. Mitchell, D.J.B. (1986) ‘Explanations of wage inflexibility: institutions and incentives’, in W.Beckerman (ed.) Wage Rigidity and Unemployment, Baltimore: Johns Hopkins University Press. Mokyr, J. (1976) Industrialization in the Low Countries, 1795–1850, New Haven and London: Yale University Press. Mooney, P. (1978) ‘Incomes policy’, in B.Dowling and J.Durkan (eds) Irish Economic Policy: A Review of Major Issues, Dublin: Economic and Social Research Institute. Moore, W. and Raisian, J. (1983) ‘The level and growth of union/nonunion relative wage effects, 1967–1977’, Journal of Labor Research 4, Winter: 65–79. Moriceau, J.M. (1985) ‘“Les Baccanals” ou grèves des moissonneurs du pays de France’, in J.Nicolas (ed.) Mouvements populaires et conscience sociale, Paris: Maloine. Morton, D. and Wright, G. (1987) Winning the Second Battle: Canadian Veterans and the Return to Civilian Life 1915–1930, Toronto: University of Toronto Press. Myrdal, G. (1944) An American Dilemma, New York: Harper and Row. Nahuet, R. (1984) ‘Une experience canadienne de Taylorisme: le cas des usines Angus du Canadien Pacifique’, unpublished MA thesis, Université du Québec à Montreal. Nevin, E. (1963) ‘Wages in Ireland, 1946–62’, Economic Research Institute, Paper Series no. 12, Dublin. Newell, A. and Symons, J.S.V. (1988) ‘The macroeconomics of the interwar years: international comparisons’, in B.Eichengreen and T.J.Hatton (eds) Interwar Unemployment in Historical Perspective, Dordrecht: Kluwer. Nickell, S. (1979) ‘Unemployment and the structure of labour costs’, in K.Brunner and A.Meltzer (eds) Policies for Employment, Prices and Exchange Rates, Amsterdam: North-Holland. ——(1988) ‘Wages and economic activity’, in W.Eltis and P.Sinclair (eds) Keynes and Economic Policy: The Relevance of the General Theory after Fifty Years, London: Macmillan. Noordegraaf, L. (1980) Daglonen in Alkmaar, 1500–1850, Amsterdam. ——(1985) Hollands welvaren? Levensstandaard in Holland, 1450–1650, Bergen. Noordegraaf, L. and Schoenmakers, J.T. (1984) Daglonen in Holland 1450–1600, Amsterdam. North, D.C. (1981) Structure and Change in Economic History, New York: Norton. Northrup, H. (1968) The Negro in the Automobile Industry, The Racial Policies of American Industry, Report Number 1. University of Pennsylvania, The Wharton School. Nusteling, H.P.H. (1985) Welvaart en werkgelegenheid in Amsterdam, 1540–1860, Amsterdam. O’Brien, P.K. and Keyder, C. (1978) Economic Growth in Britain and France, 1780– 1914, London: George Allen and Unwin. Offer, A. (1989) The First World War: An Agrarian Interpretation, Oxford: Clarendon Press.
308
BIBLIOGRAPHY Ó Gréda, C. (1981) ‘Agricultural decline 1860–1914’, in R.C.Floud and D.N. McCloskey (eds) The Economic History of Britain since 1700, vol. 2, Cambridge: Cambridge University Press. ——(1988a) ‘“For Irishmen to forget?” Recent research on the great Irish famine’, Centre for Economic Policy Research Working Paper no. WP88/7. ——(1988b) Ireland Before and After the Famine: Explorations in Economic History, 1800–1925, Manchester: Manchester University Press. Okun, A.M. (1981) Prices and Quantities, Washington, DC: The Brookings Institution. O’Mahony, D. (1964) ‘Industrial relations in Ireland: the background’, Economic Research Institute Paper Series no. 19, Dublin. ——(1965) ‘Economic aspects of industrial relations’, Economic Research Institute Paper Series no. 24, Dublin. ——(1967) The Irish Economy: An Introductory Description , 2nd edn, Cork: Cork University Press. O’Rourke, K. (1989) ‘Agricultural change and rural depopulation: Ireland 1845–1876’, unpublished Ph.D. thesis, Harvard University. ——(1991) ‘Burn everything British but their coal: the Anglo-Irish economic war of the 1930s’, J.Ec.H. 51, 2:357–66. Owen, J.D. (1970) The Price of Leisure: An Economic Analysis of the Demand for Leisure Time, Montreal: McGill-Queen’s University Press. Paping, R.F.J. (1991) ‘Timmeren en metselen aan de kerk van Framsum: werken en verdienen, 1773–1891’, Jaarboek voor de geschiedenis van bedrijf en techniek 8: 38–62. Parker, W.N. (1972) ‘Agriculture’, in L.E.Davis, R.A.Easterlin, and W.N.Parker (eds) American Economic Growth: An Economist’s History of the United States, New York: Harper and Row. ——(1984) Europe, America, and the Wider World: Essays on the Economic Theory of Western Capitalism, vol. 1, Europe and the World Economy, Cambridge: Cambridge University Press. ——(1986) ‘An historical introduction’, in W.N.Parker (ed.) Economic History and the Modern Economist, Oxford: Basil Blackwell. ——(1987) ‘New England’s early industrialization: a sketch’, in P.Kilby (ed.) Quantity and Quiddity: Essays in U.S. Economic History in Honor of Stanley Lebergott, Middletown, CT: Wesleyan University Press. Patten, J. (1988) ‘Patterns of migration and movement of labour to three pre-industrial East Anglian towns’, in P.Clark and D.Souden (eds) Migration and Society in Early Modern England, Totowa, NJ: Barnes & Noble. Pedersen, J. (1926) ‘Irland som Danmarks konkurrent paa det Engelske marked’, Nationalokonomiske Tidsskrift 64:423–38. Pencavel, J. (1974) ‘Relative wages and trade unions in the United Kingdom’, Economica 41, May: 194–210. Pencavel, J. and Hartsog, C. (1984) ‘A reconsideration of the effects of unionism on relative wages and employment in the United States, 1920–1980’, Journal of Labor Economics 2, April: 193–222. Penn, R. (1971) Skilled Workers in the Class Structure, Cambridge: Cambridge University Press. Penn, S. (1987) ‘Female wage-earners in late fourteenth-century England’, Agricultural History Review 35:1–14. Perlman, J. and Frazier, E.K. (1937) ‘Entrance rates of common laborers in 20 Industries, July 1937’, Monthly Labor Review 45, December: 1491–1510.
309
BIBLIOGRAPHY Phelps Brown, E.H. (1977) The Inequality of Pay, Berkeley: University of California Press. Phelps Brown, E.H. and Browne, M.H. (1968) A Century of Pay, London: Macmillan. Phelps Brown, E.H. and Hopkins, S.V. (1955) ‘Seven centuries of building wages’, Economica 22, 87:195–206. ——(1956) ‘Seven centuries of the price of consumables, compared with builders’ wage-rates’, Economica 23, 92:296–314. ——(1957) ‘Wage-rates and prices: evidence for pressure in the sixteenth century’, Economica 24, 96:289–306. ——(1959) ‘Builders’ wage-rates, prices and population, some further evidence’, Economica 26, 101:18–38. ——(1981) A Perspective of Wages and Prices, London: Methuen. Phillips, W.H. (1985) ‘Southern textile mill villages on the eve of World War II: the Courtenay mill of South Carolina’, J.Ec.H. 45, 2:269–75. ——(1986) ‘The labor market of southern textile mill villages: some micro evidence’, E.E.H. 23, 2:103–23. Pigou, A.C. (1927) ‘Wage policy and unemployment’, Economic Journal 37, September: 355–68. Pinchemel, P. (1957) Structures sociales et depopulation rurale dans les campagnes Picardes de 1836 à 1936, Paris. Piore, M. (1979) Birds of Passage. Migrant Labor and Industrial Societies, Cambridge: Cambridge University Press. Piva, M.J. (1979) The Condition of the Working Class in Toronto, 1900–1921, Ottawa: University of Ottawa Press. Postel-Vinay, G. (1988) ‘Les rapports entre industrie et agriculture en Picardie au xix siècle’, in G.Carrier and R.Hubscher (eds) Entre faucilles et marteaux, Paris: PUL, Ed. de la Maison des Sciences de l’Homme. ——(1991) ‘L’agriculture dans l’économie franchise: crises et reinsertion’, in M. LevyLeboyer (ed.) Entre l’état et le marché. L’économic fraçaise des années 1880 à nos jours, Paris: Gallimard. Posthumus, N.W. (1908–39) Geschiedenis van de Leidsche lakenindustrie (3 vols), The Hague: M.Nijhoff. ——(1944–64) Inquiry in the History of Prices in Holland (2 vols), Leiden: E.J.Brill. Prest, A.R. (1954) Consumers’ Expenditure in the United Kingdom, 1900–1919, Cambridge: Cambridge University Press. Pringsheim, O. (1890) Beitrage zur wirtschaftlichen Entwickelungsgeschichte der Vereinigten Niederlande im 17. und 18. Jahrhundert, Leipzig. Purdy, F. (1861) ‘On the earnings of agricultural labourers in England and Wales, 1860’, J.R.S.S. 24, September: 328–73. Raff, D.M.G. (1988) ‘Wage determination theory and the five-dollar day at Ford’, J.Ec.H. 48, 2:387–99. Railway Age Gazette (1917) ‘The railroads of Canada and the war’, Railway Age Gazette 62, 25:1299–1306. Railway Mechanical Engineer (1919) ‘Shop labor situation in 1918’, Railway Mechanical Engineer 93, 1:22–4. Ramirez, B. (1986) ‘Brief Encounters: Italian Immigrant Workers and the CPR 1900– 1930’, Labour/Le Travail 17:9–27. Ramkhalawansingh, C. (1974) ‘Women during the Great War’, in J.Acton, P. Goldsmith, and B.Shepard (eds) Women at Work: Ontario, 1850–1930, Toronto: Canadian Women’s Educational Press.
310
BIBLIOGRAPHY Ramsbottom, E.G. (1935) ‘The course of wage rates in the United Kingdom, 1921– 1934’, J.R.S.S. 98, 4:639–73. ——(1938) ‘Wage rates in the United Kingdom, 1934–1937’, J.R.S.S. 101, 1:202–4. Ransom, R.L. and Sutch, R. (1986) ‘The labor of older Americans: retirement of men on and off the job, 1870–1937’, J.Ec.H. 46, 1:1–30. ——(1989) Codebook and User’s Manual: A Survey of 5,419 Workers in Michigan, 1889; Reported in the Seventh Annual Report of the Michigan Bureau of Labor and Industrial Statistics, Berkeley: Institute of Business and Economic Research. Ravenstein, E.G. (1885) ‘The laws of migration’, J.R.S.S. 48, June: 167–227. Reder, M.W. (1955) ‘The theory of occupational wage differentials’, A.E.R. 45, 5: 833– 52. Redford, A. (1926) Labour Migration in England, Manchester: Manchester University Press. Rees, A. (1961) Real Wages in Manufacturing:1890–1914, NBER General Series No.70, Princeton: Princeton University Press. Regehr, T.D. (1976) The Canadian Northern Railway: Pioneer Road of the Northern Prairies, Toronto: Macmillan. Reichsverband der Deutschen Industrie (1929) ‘Aufstieg oder Niedergang? Deutsche Wirtschafts und Finanzreform 1929’, Veroeffentlichungen des RDI, no. 49, Berlin: RDI. Reynolds, L.G. (1935) The British Immigrant: His Social and Economic Adjustment in Canada, Toronto: Oxford University Press. Richardson, J.H. (1935) ‘Comment’, J.R.S.S. 98, 4:676. Ritschl, A. (1990) ‘Zu hohe Lohne in der Weimarer Republik? Eine Auseinandersetzung mit Holtfrerichs Berechnungen zur Lohnposition der Arbeiterschaft 1925–1932’, Geschichte und Gesellschaft 16:375–402. Roberts, C.A. (1990) ‘Apparent real wage rate differentials and convergence, agricultural laborers: 1870–1940’, paper presented at the Economic History Association annual meeting, Montreal. Robertson, P.L. and Alston, L.J. (1992) ‘Technological choice and the organization of work in capitalist firms’, Ec.H.R. 45, 2:330–49. Robinson, A. (1984) ‘Comment on Peter Wiles’ paper’, in P.Wiles and G.Routh (eds) Economics in Disarray, Oxford: Basil Blackwell. Rodgers, D.T. (1978) The Work Ethic in Industrial America, 1850–1920, Chicago: University of Chicago Press. Roediger, D.R. and Foner, P.S. (1989) Our Own Time: A History of American Labor and the Working Day, Contributions in Labor Studies, no. 23, New York: Greenwood Press. Romer, C.D. (1989) ‘The prewar business cycle reconsidered: new estimates of Gross National Product, 1869–1908’, Journal of Political Economy 97:1, 1–37. Rorabaugh, W.J. (1986) The Craft Apprentice. From Franklin to the Machine Age in America, New York: Oxford University Press. Rose, A. (1948) The Negro in America, New York: Harper and Row. Rosenband, L.H. (1985) ‘Productivity and labor discipline in the Montgolfier Paper Mill, 1780–1850’, J.Ec.H. 45, 2:435–43. Rosenberg, N. (1972) Technology and American Economic Growth, White Plains, NY : M.E. Sharpe. ——(1976) Perspectives on Technology, Cambridge: Cambridge University Press. Rosenbloom, J.L. (1990) ‘One market or many? Labor market integration in the late nineteenth-century United States’, J.Ec.H. 50, 1:85–107.
311
BIBLIOGRAPHY ——(1991) ‘Occupational differences in labor market integration: the US in 1890’, J.Ec.H. 51, 2:427–39. Rothenberg, W.B. (1988) ‘The emergence of farm labor markets and the transformation of the rural economy: Massachusetts, 1750–1855’, J.Ec.H. 48, 3:537–66. Rottier, G. (1957) ‘The evolution of wage differentials: a study of British data’, in J.T.Dunlop (ed.) The Theory of Wage Determination, London: Macmillan. Rountree, G.M. (1936) The Railway Worker: A Study of the Employment and Unemployment Problems of the Canadian Railways, Toronto: Oxford University Press. Rowe, J.W.F. (1928) Wages in Theory and Practice, London: Routledge and Kegan Paul. Royer, J.F. (1975) L’exode agricole va-t-il bientôt tarir? Paris: INSEE. Russell, J.C. (1972) Medieval Regions and their Cities, Newton Abbot: David & Charles. Sachs, J. (1980) ‘The changing cyclical behaviour of wages and prices, 1890–1976’, A.E.R. 70:78–90. Safarian, A.E. (1970) The Canadian Economy in the Great Depression, Toronto: McClelland and Stewart. (First published 1959.) Salais, R., Baverez, N., and Reynaud, B. (1986) L’invention du chômage, Paris: PUF. Saxonhouse, G. and Wright, G. (1984) ‘Two forms of cheap labour in textile history’, in G.Saxonhouse and G.Wright (eds) Technique, Spirit, and Form in the Making of the Modern Economies: Essays in Honor of William N.Parker (Research in Economic History, Supplement 3). Saxton, C.C. (1942) The Economics of Price Determination, Oxford: Oxford University Press. Schwab, S. (1986) ‘Is statistical discrimination efficient?’, A.E.R. 76, 1:228–34. Sells, D.M. (1939) British Wages Boards: A Study in Industrial Democracy, Washington, DC: Brookings Institution. Shergold, P.R. (l982)Working-Class Life: The ‘American Standard’ in Comparative Perspective, 1899–1913 , Pittsburgh: University of Pittsburgh Press. Shiells, M.E. (1990) ‘Collective choice of working conditions: hours in British and US iron and steel, 1890–1923’, J.Ec.H. 50, 2:379–92. Shiells, M. and Wright, G. (1983) ‘Night work as a labor market phenomenon: southern textiles in the interwar period’, E.E.H. 20, 4:331–50. Shulman, S. and Darity, W.P.Jr (1989) The Question of Discrimination, Middletown, CT: Wesleyan Press. Sicsic, P. (1992) ‘City-farm wage gaps in late nineteenth-century France’, J.Ec.H. 52, 3:575–95. ——(1993) ‘Wage dispersion in France, 1850s–1930s’, paper presented at the conference on Real Wages, Migration, and Labour Market Integration in the XlXth and XXth Centuries, Bellagio, Italy. Sinn, G. and Sinn, H.-W. (1992) Jumpstart. The Economic Unification of Germany, Cambridge, MA: MIT Press. Slicher van Bath, B.H. (1963) The Agrarian History of Western Europe 500–1850, London: E.Arnold. Slichter, S.H. (1919) The Turnover of Factory Labor, New York: D.Appleton. Smith, A. (1904 edn, reprinted 1976) Wealth of Nations, Chicago: University of Chicago Press (Cannon edition). Smith, J.P. (1984) ‘Race and human capital’, A.E.R. 74, 4:685–98. Smith, J.P. and Welch, F. (1986) Closing the Gap: Forty Years of Progress for Blacks, Santa Monica: Rand Corporation.
312
BIBLIOGRAPHY ——(1989) ‘Black economic progress after Myrdal’, Journal of Economic Literature 27, 2:519–64. Söderberg, J. (1985) ‘Regional economic disparity and dynamics, 1840–1914: a comparison between France, Great Britain, Prussia, and Sweden’, Journal of European Economic History 14, 2:273–96. ——(1987) ‘Real wage trends in urban Europe, 1730–1850: Stockholm in comparative perspective’, Social History 12, 2:155–76. Sokoloff, K.L. (1984) ‘Was the transition from the artisanal shop to the nonmechanized factory associated with gains in efficiency? Evidence from the U.S. manufacturing censuses of 1820 and 1850’, E.E.H. 21, 4:351–82. Solow, R.M. (1990) The Labor Market as a Social Institution, Oxford: Basil Blackwell. Southall, H.R. (1988) ‘The origins of the depressed areas: unemployment, growth, and regional economic structure in Britain before 1914’, Ec.H.R. 41, 2:236–58. ——(1991) ‘The tramping artisan revisited: labour mobility and economic distress in early Victorian England’, Ec.H.R. 44, 2:272–96. Spero, S.D. and Harris, A.L. (1931) The Black Worker: The Negro and the Labor Movement, Port Washington, NY: Kennikat. Stacey, R.E. (1986) ‘Agricultural investment and the management of royal demesne manors 1236–1240’, J.Ec.H. 46, 4:919–34. Stafford, J. (1951) ‘Indices of wholesale prices’, J.R.S.S. 114, 4:447–67. Stamp, J. (1925) The Financial Times, March 15: v. ——(1931) The Economist, June 20:1314. Statist (1926) ‘Unemployment and real wages’, The Statist, 9 January: 50–1. Stevens, G.R. (1973) History of the Canadian National Railways, New York: Macmillan. Stiglitz, J.E. (1973) ‘Approaches to the economics of discrimination’, A.E.R. 63, 2: 287– 95. Struthers, J. (1983) No Fault of Their Own: Unemployment and the Canadian Welfare State, 1914–1941, Toronto: University of Toronto Press. Sturt, G. (1984) Change in the Village, London: Caliban. (First published 1912.) Sundstrom, W.A. (1988) ‘Internal labor markets before World War I: on-the-job training and employee promotion’, E.E.H. 25, 4:424–45. ——(1990) ‘Half a career: discrimination and railroad internal labor markets’, Industrial Relations 29, Fall: 423–40. ——(1992a) ‘Last hired, first fired? Unemployment and urban black workers during the great depression’, J.Ec.H. 52, 2:415–29. ——(1992b) ‘Rigid wages or small equilibrium adjustments? Evidence from the contraction of 1893’, E.E.H. 29, 4:430–55. Sundstrom, W.A. and Rosenbloom, J.L. (1993) ‘Occupational differences in the dispersion of wages and working hours: labor market integration in the United States, 1890–1903’, E.E.H. 30, 4:379–408. Sweezy, M. (1940) ‘German corporate profits: 1926–1938’, Quarterly Journal of Economics 54, May: 384–98. Tarnow, F. (1929) Warum Arm Sein?, Berlin. Taylor, A.A. (1926) The Negro in the Reconstruction of Virginia, Washington, DC: Association for the study of Negro life and history. Temin, P. (1989) Lessons from the Great Depression, Cambridge, MA: MIT Press. Terrier, D. and Toutain, P. (1979) ‘Pression démographique et marché du travail a Comines au XVIIIème siècle’, Revue du Nord 61, 240:19–26. Thernstrom, S. (1973) The Other Bostonians: Poverty and Progress in the American Metropolis, 1880–1970, Cambridge, MA: Harvard University Press.
313
BIBLIOGRAPHY Thirsk, J. (ed.) (1989) The Agrarian History of England and Wales, vol. 6, Cambridge: Cambridge University Press. Thomas, B. (1954) Migration and Economic Growth, Cambridge: Cambridge University Press. ——(1969) ‘The migration of labour into the Glamorganshire Coalfield, 1861–1911’, in W.E.Minchinton (ed.) Industrial South Wales 1750–1914, New York: Augustus Kelley. ——(1972) Migration and Urban Development, London: Methuen. Thomas, M. (1988) ‘Labour market structure and the nature of unemployment in interwar Britain’, in B.Eichengreen and T.J.Hatton (eds) Interwar Unemployment in Historical Perspective, Dordrecht: Kluwer. ——(1992) ‘Institutional rigidity in the British labour market, 1870–1939: a comparative perspective’, in N.Crafts and S.Broadberry (eds) Britain in the World Economy, 1870–1939: Essays in honour of Alec Ford, Cambridge: Cambridge University Press. Thompson, J.H. (1978) The Harvests of War: The Prairie West, 1914–1918, Toronto: McClelland and Stewart. Thompson, R.J. (1937) ‘The agricultural labour bill in England and Wales’, J.R.S.S. 100, 4:607–35. Todaro, M.P. (1969) ‘A model of labor migration and urban unemployment in less developed countries’, A.E.R. 59, 1:138–48. Toutain, J.-C. (1963) La population de la France de 1700 à 1959, Paris: Cahiers de l’Institut de Science Economique Appliquée . Trempé, R. (1971) Les mineurs de Carmaux, 1848–1914, Paris: Editions ouvrières. Ulrich, L.T. (1991) A Midwife’s Tale. The Life of Martha Ballard, based on her diary, 1785–1812, New York: Vintage. Urquhart, M.C.(1986) ‘New estimates of Gross National Product, Canada, 1870–1926: some implications for Canadian development’, in S.L.Engerman and R.E.Gallman (eds) Long-Term Factors in American Economic Growth, NBER Studies in Income and Wealth vol. 51, Chicago: University of Chicago Press. ——(1993) Gross National Product, Canada 1870–1926: The Derivation of the Estimates, Kingston and Montreal: McGill—Queen’s Press. Urquhart, M.C. and Buckley, K.A.H. (1965) Historical Statistics of Canada, Toronto: Macmillan. Van Dam, P. (1992) ‘Gravers, ofzetters en berriedragers. Werkgelegenheid aan de Spaarndammerdijk omstreeks 1510’, Tijdschrift voor sociale geschiedenis 18:447– 78. Van de Spiegel, L. (1772) Over de armoede en bedelaary met betrekking tot de provintie van Zeeland en byzonder tot de stad Goes en Eiland van Zuid-Beveland, The Hague. Van der Vlis, D. (1981) ‘Daglonen in en rond Kampen van 1526 tot 1811’, Overijsselse historische bijdragen 96:77–99. Van der Wee, H. (1977) ‘Prices and wages as development variables: a comparison between England and the Southern Netherlands, 1400–1700’, Acta Historiae neerlandica 10:58–78. Van der Woude, A.M. (1980) ‘Demografische ontwikkeling van de Noordelijke Nederlanden, 1500–1800’, in Algemene geschiedenis der Nederlanden, vol. 5, Haarlem. Van Deursen, A.Th. (1978) Het kopergeld van de Gouden Eeuw, vol. 1, Het dagelijks brood, Amsterdam. Van Dillen, J.G. (1970) Van rijkdom en regenten, The Hague.
314
BIBLIOGRAPHY Van Houtte, J.A. (1977) An Economic History of the Low Countries, 800–1800, New York: St Martin’s. Van Leeuwen, M.H.D. and Smits, F. (1987) ‘Bedeling en arbeidsmarkt in Amsterdam in de eerste helft van de negentiende eeuw’, Tijdschrift voor sociale geschiedenis 13:432–57. Van Ravesteyn, W. (1906) Onderzoekingen over de economische en sociale ontwikkeling van Amsterdam gedurende de 16de en het eerste helft der 17de eeuw, Amsterdam. Van Zanden, J.L. (1983) ‘Lonen en arbeidsmarkt in Amsterdam, 1800–1865’, Tijdschrift voor sociale geschiedenis, 1–27. Vaughan, W.E. and Fitzpatrick, A.J. (1978) Irish Historical Statistics: Population 1821– 1971, Dublin: Royal Irish Academy. Vedder, R.K. and Gallaway, L. (1992) ‘Racial differences in unemployment in the United States, 1890–1990’, J.Ec.H. 52, 3:696–702. Verein deutscher Maschinenbau-Anstalten (1931) Lohnpolitische Kurvenbilder zur Krisenlage, Berlin: VDMA. Verlinden, C. et al. (1959–73) Dokumenten voor de geschiedenis van prijzen en lonen in Vlaanderen en Brabant (XVe-XVIII eeuw), 4 vols, Brugge. Villermé, L.R. (1971 edn) Tableau de l’état physique et moral des ouvriers employés dans les manufactures de coton, de laine et de soie, Paris: UGE. (First published 1840.) Walsh, B. (1974) ‘Expectations, information and human migration: specifying an econometric model of Irish migration to Britain’, Journal of Regional Science 14, 1:107–20. Ware, C.F. (1931) The Early New England Cotton Manufacture: A Study in Industrial Beginnings, Boston: Houghton Mifflin. Weir, D.R. (1991) ‘Labor market performance and demographic change in nineteenth century France’, unpublished paper, Yale University. Whaples, R. (1990a) ‘Winning the eight-hour day, 1909–1919’, J.Ec.H. 50, 2:393–406 . ——(1990b) ‘The shortening of the American work week: an economic and historical analysis of its context, causes, and consequences’, unpublished Ph.D. thesis, University of Pennsylvania. ——(1990c) ‘No reason to stay? The determinants of employment tenure in the late nineteenth century’, unpublished paper, University of Wisconsin-Milwaukee. Whatley, W. (1983) ‘“Labor for the picking”: the New Deal in the South’, J.Ec.H. 43, 4:913–26. ——(1990) ‘Getting a foot in the door: ‘learning,’ state dependence and the racial integration of firms’, J.Ec.H. 50, 1:43–66. ——(1993) ‘African-American strikebreaking from the Civil War to the New Deal’, Social Science History 17, Winter: 525–558. Whatley, W., Wright, G., and Foote, C.L. (1990) ‘Getting started in the auto industry: black workers at the Ford Motor Company, 1918–1947’, paper presented at ASSA meetings, Washington, DC. Willemsen, R. (1988) Enkhuizen tijdens de Republiek, Hilversum. Williamson, J.G. (1985) Did British Capitalism Breed Inequality?, Boston: Allen and Unwin. ——(1988) ‘Migrant selectivity, urbanization, and industrial revolutions’, Population and Development Review 14, 2:287–314. ——(1992) ‘The evolution of global labor markets in the first and second world since 1830: background evidence and hypotheses’, NBER/DAE Working Paper No. 36. Williamson, J.G. and Lindert, P.H.(1980) American Inequality: A Macroeconomic History, New York: Academic Press.
315
BIBLIOGRAPHY Wood, G.H. (1909) ‘Real wages and the standard of comfort since 1850’, J.R.S.S. 72, March: 91–103. Worthman, P. (1969) ‘Black workers and labor unions in Birmingham, Alabama, 1897– 1904’, Labor History 10, Summer: 374–407. Woytinsky, W.S. (1942) Three Aspects of Labor Dynamics, Washington, DC: SSRC. ——(1953) ‘Seasonal and cyclical variations in the labor force’, in W.S.Woytinsky (ed.) Employment and Wages in the United States, New York: Twentieth Century fund. Wright, G. (1981) ‘Cheap labor and southern textiles, 1880–1930’, Quarterly Journal of Economics 96, November: 605–30. ——(1986) Old South New South: Revolutions in the Southern Economy Since the Civil War, New York: Basic Books. ——(1987) ‘Labor history and labor economies’, in A.Field (ed.) The Future of Economic History, Boston: Kluwer-Nijhoff. ——(1988) ‘American agriculture and the labour market: what happened to proletarianization?’, Agricultural History 62:182–209. Wylie, P.J. (1989) ‘Technological adaptation in Canadian manufacturing, 1900–1929’, J.Ec.H. 49, 3:569–91.
316
INDEX
age: Irish migrants 154; and job tenure 187, 190–202, 214; of retirement 36 age-wage curve: France 80 Agricultural Wages (Regulation) acts 254–5 agriculture: French seasonal employment 29–30, 64–82; land holding and labour markets 10–15; and racial discrimination 277, 279; regional wage variation 85–8, 92–3, 94–6, 97, 99, 100–2; and rural industry 19–20; seasonal migrations 16–18 Allen, R.C. 30–1, 32, 107–37 Allen, R.G.D. 249 Armed Forces: Canada 206; inter-war Germany 236 Arvida aluminium works 214 Ashby, Joseph 85 assembly line production: USA 8–10 Australia: consumer price index 132–3; gross national product 117; international labour market integration 30–1, 107–37; real wages 121–2 2 automobiles see car manufacturing
bonuses 35–6; French industry 73–4; pre-industrial Netherlands 49 Borchardt, Knut 226–7 Bowley, A.L. 87, 88; consumer price indices 109, 111, 113, 114; wage data 252 Boxer, Charles 43 Boyer, G.R. 30–1, 84–104 bricklayers: hours of work 164, 167, 168, 178, 180, 181–2;wages 87–8, 118–25, 129, 136–7, 143, 144 Britain: consumer price index 133–4; gross national product 117; inter-war real wages 34, 229, 234, 245–66; Irish migration to 30–1, 34, 139–59; real wages 40–2, 50–3, 55, 119–21; regional labour market integration 30–1, 84–104; skill premia 124–5 British emigrants: Canadian railway workers 217, 219 Burkitt, B. 260
Bailer, Lloyd 287 Bain, G.S. 260 Baldwin, William H. 279 Beenstock, Michael 251 Belgium: construction wages 50–1 Bertram, G.W. 113, 114 Beveridge, W.H. 247, 262, 264 black workers: racial discrimination 1–2, 31, 270–91 Board of Trade: wholesale price index 249
Cahill, M.C. 162–3, 175, 176 Caird, James 85 Canada 37; consumer price index 134; gross national product 117–18; international labour market integration 30–1, 107–37; real wages 32, 122–3; seasonal labour market 33; technological change 6–7; unemployment insurance 33, 34 Canadian Pacific Railway: job tenure 205–22; pension schemes 35–6 capital intensity: French industry 73–4, 75 car manufacturing: racial discrimination 284–9 Card, D. 289
317
INDEX carpenters: hours of work, US 167, 168, 174, 177, 178, 180; regional wage variation, Britain 86–8, 93–4, 96–9, 102 Carter, S.B. 187–8, 190–1, 196 Census of Production 249 Clapham, J.H. 85–6 clothing: consumer price indices 111; employment in industry 69 coefficient of variation: measure of labour market integration 89–90, 91– 4, 100–1 communication: networks 6, 72, 94, 101, 221, 281 congestion: labour market 19–20 construction labour 195; hours of work 162, 164, 166–8, 175–6, 178, 180; Irish migration 151–3; pre-industrial Netherlands 44–50; strikes 151; unemployment, Britain and Ireland 147–50; wages 40–2, 50–1, 118–25; wages, Britain and Ireland 142–4, 146, 147, 157; see also bricklayers; carpenters, hod carriers, labourers consumer price index 127; Englishspeaking world 108–1, 132–6; interwar Germany 227–8; see also retail price index and cost of living consumption, social norms of: USA 9 contemporary issues: contribution of economic historians 37 coordination: and hours of work 174, 181–2 cost of living: Britain and Ireland 144–7, 158–9; English towns 92; Netherlands 50–2; and real wages 40–1; USA 177; see also consumer price index cotton textile industry, USA 7; racial discrimination 273–5, 278; labour recruitment 22–3 craftsmanship: France 81; lack of in USA 7; wages 53, 55; see also skill premia cross-occupational effects: and hours of work 174, 181–2
depressions: Canada 206, 220, 221–2; Germany 225, 241–2; Britain 264; US 258 Devons, E. 249 discrimination, racial: evolution of 32, 37; USA 1–2, 31, 34, 270–91 Donohue, J.H.III 31, 289 Douglas, P.H. 7, 15, 18, 25, 113, 114, 162, 163, 165 duration, employment see tenure, job Dutch East India Company see VOC earnings: Britain and Ireland 155–6; dispersion of British industries 262– 4; pre-industrial Netherlands 47–50; manufacturing workers, Englishspeaking countries 120; see also wages economic history: and historical labour markets 3–4 education: and racial discrimination 272, 277, 279, 280–1, 281; and regional labour markets 5–6 efficiency wage 60 emigration: see migration engineering workers: Irish migration 151–3; strikes 151; unemployment, Britain and Ireland 147–50; wages 252–3, 254; wages, Britain and Ireland 142–4, 146, 147, 157; skill differentials, Canada 216; see also foundry and machine shop workers England see Britain Enkhuizen, Netherlands 57 error correction model 91 ethnicity: Canadian railway workers 216–19 Europe: labour markets (800–1800) 10– 15; seasonally 16–17; wages 50–1; see also specific countries eventual job tenure 190 experience, work: black workers 286–7
days worked: pre-industrial Netherlands 49–50; France 75–6; see also hours of work De Vries, J. 25, 30–1, 33, 39–63 deflators: real wages series 246–51
factories see manufacturing family holdings and labour supply 13 Feinstein, C.H. 113, 114, 119, 121, 125, 266 firm size: and job tenure 194 firm-specific human capital: and job tenure 197–200, 201, 208 Fishback, Price 279
318
INDEX fisheries: seasonal labour demand 56, 57 fitters: wages 143, 145, 252–3 Fitzgerald, S. 122 food 109–10, 112–13, 145; shortages 20 Ford Motor Company: assembly line production 8; racial discrimination 31, 284–9 foundry and machine shop workers: hours of work 164, 166–8, 178, 179, 180, 181; wages 169, 170, 171; see also engineering workers France: agricultural labourers’ wages 100; geography of seasonal migration 71–2; industrial employment 75–6; traditional labour markets 29–30; unemployment 66, 80–1 French Canadians: railway workers 216– 19 Friedman, G. 175, 181 furniture industry workers: job tenure 188–9, 190–202 Gallaway, L. 282 Gazeley, I. 113 Geary, R. 148–9 Gelber, Steven 290 Germany: historical economist school 3–4; inter-war wages 34, 225–43; inter-war labour demand 230–5; migrant labour from 58, 60, 224; preindustrial wages 50–1 Ginzberg, Eli 290 Goldin, Claudia 290 Grantham, G. 1–24, 28, 29, 36–7 Green, A.G. 37 Green, D.A. 27–38 gross national product, real, per capita: English-speaking countries 115–18, 128 Hall, R.E. 63, 190, 192 hand compositors: wages 143 Harris, A.L. 283 harvests 16; France 64–82; USA 17–18; shortfalls 20 harvest wages, France, 67–8 Hatton, T.J. 30–1, 84–104 Heckman, J. 31, 289–90 herring fishery 56, 57 Higgs, Robert 273
hod carriers: hours of work 164, 167, 168, 178, 180, 181 Holland see Netherlands Holtfrerich, Carl-Ludwig 227 home employment: French 76; and rural industry 19–21 Homestead strike (1892) 283 Hopkins, Sheila: time series of wages 40–2 hours of work: decline in USA 32–3, 161–82; pre-industrial Netherlands 49–50 Hughes, J.G. 148–9 Hunt, E.H. 85–8, 91–2 immigration: see migration implicit contracts 36 industrial finished goods price index (PIF): inter-war Germany 227–8 industry: dispersion of earnings in Britain 262–6; early urban 21–4; French seasonal employment 29, 64– 82; American and racial discrimination 273–4, 282; rural 19– 20; and seasonally 16–17, 18–19; see also specific trade e.g. textile information: on labour market 5–6 innovation clusters 6 instability, labour (1900s) 186–8, 206 institutions: evolution of 27–33; necessity for long view 33–4; wagesetting 56–7, 82, 220, 225–6, 254–62 integration, labour market: model 103– 4; see also international; regional internal labour markets: France 68, 80; and job tenure 196–7 internal promotion: Canadian railway workers 219–20 international labour market integration: English-speaking world 30–1, 107– 37; Ireland and Britain 30–1, 34, 139–59 investment: and German real wages 226–9 Ireland: famine 141–2; migration and wages 30–1, 34, 139–59 Jackson, R.V. 125 Jacoby, S.M. 186, 195, 202, 207 James, J.A. 35, 185–202 Kansas: job tenure 189–201
319
INDEX Keynes, J.M. 38, 40, 226, 250–1; ‘Keynesian’ labour demand 230, 233, 234, 241–2 King, Bolton 85 Klein, Peter 43 Knibbs, G.H. 108 Krueger, A.B. 289 Kuznets, Simon 37
103–4; rural-urban 21; temporary 14, 16, 18 mobility, labour (1900s) 186–8; see also migration; tenure, job monasteries: labour 11
labourers: pre-industrial Netherlands 45–7, 48–9; wages 118–25, 130, 136– 7, 252–3; see also agriculture, construction labour land holding: Europe (800–1800) 10–15 Lawrence, F.W. 86 Lazear, E. 35, 36 Lee, J. 122 legislation: hours of work 162–3, 174–5; and labour markets 5; statutory wage 254–62; Weimar Republic, wages 225–6 Licht, Walter 208, 282 machine tools: USA 7 MacKinnon, M. 35–6, 205–22 ‘makeshifts’ 13 Maloney, T.N. 286 management, black 290 manufacturing: American history 7–10; hours of work 162, 163; wages 118– 25, 131, 137; wholesale price index 249; see also specific trade e.g., furniture margins (wage): inter-war Britain 252–3 Margo, Robert 272, 281 Marshall, A. 201 Marx, Karl 1, 10, 23, 43 mass production, USA 8–10 mechanization: USA agriculture 17; USA labour supply 7–9 Michigan: furniture industry workers 188–9, 190–202 migration: black labour to northern USA 281–3; to Canada 37, 205–6, 216–19; French seasonal 71–4, 75, 77–8; and international labour market integration 107–8; Irish 30–1, 34, 139–59; and job tenure 187, 190–1; to pre-industrial Netherlands 58, 60; and regional labour market integration 29–31, 32, 90–1, 102,
national price indices 113 Netherlands: pre-industrial labour markets 30–1, 33, 42–63 nominal wages: Britain 40, 41–2, 246, 251–62; Ireland and Britain 142–4; Netherlands 44–5, 50, 61 Noordegraaf, Leo 43 occupational trends, USA: hours and wages 165–71 O’Rourke, K. 30–1, 34, 139–59 pay ratio: Britain and USA 124–5; see also skill premia Payner, B. 289–90 peasants 11–15 Pedersen, J. 149 Penn, Roger 254 pension schemes 35–6 Percy, M.B. 113–14 Phelps Brown, E.H.: time series of wages 40–2; 215 piece work 19, 254 Pigou, A.C. 246–7, 250 Piva, M.J. 123 politics: German real wages 225–6, 241; and hours of work 174–5, 181; USA manufacturing 7 population change: and real wages 40– 1, 61–2 Postel-Vinay, G. 33, 34, 64–82 pre-industrial labour markets 10–14, 19– 21, 39–42; Netherlands 30–1, 33, 42– 63 Prest, A.R. 108 Price, R. 260 price setting: inter-war Britain 266 Pringsheim, Otto 42–3 printing and publishing workers, Irish and British: migration patterns 151– 3; unemployment 147–50; wages 142–4, 146, 147, 157 productivity: and racial discrimination 274–5 profits: and German real wages 226–9
320
INDEX Prussia: agricultural labourers’ wages 100, 101 purchasing power parities (PPP) 115– 16, 158–9 Purdy, Frederick 85 race: discrimination in USA 1–2, 31, 270–91; see also ethnicity railways: employment policies 208; job tenure 195, 207–22; racial discrimination 276–7 Ramsbottom wage series index 246; drawbacks 252–62 real wages: British time series 40–2; Canada 32; English-speaking world 30–1, 118–25; inter-war Britain 34, 229, 234, 245–66; inter-war Germany 34, 225–43; pre-industrial Netherlands 50–4, 55 Reder, M.W.: skill differential 154, 215 Rees, A. 113, 114 regional labour market integration 5–6; England and Wales 30–1, 84–104; France 29–30, 78–82; USA 32–3, 165– 71 rents: 111, 112, 145–6 retail price index, inter-war Britain 247, 249, 251 retirement 35–6 Richardson, J.H. 254 Ritschl, Albrecht 227 Road Haulage Wages Act 254–5 Rosenbloom, J.L. 32–3, 161–82 Rosenwald Fund 280–1 Rottier, George 262–5 Rowe, J.W.F. 86, 88 Rueff, Jacques 246–7, 250 rural industry 19–21, 72–4, 79–80 rural-urban migration 21, 102, 152–3 salaried occupations: Black American 290–1; pre-industrial Netherlands 47, 48 Savoca, E. 187–8 sawyers 45–6 schools see education scope: of labour markets 2–3 screening costs 202 seasonal employment 15–19, 33; France 29–30, 64–82; Ireland 149; preindustrial Netherlands 48–9, 55–6, 57–8, 60; industrial USA 18
segmentation: pre-industrial Netherlands 58–61 segregation, racial: Ford Motor Company 287–9; USA 31, 270, 271–2, 273–83, 289–91 seigniorial demesnes: labour 11–12 serfdom 11–15 service workers: job tenure 189–201 sharing, work 33 Sharma, S. 186 Shergold, P.R. 108–11 Shiells, M.E. 5, 163 shortage, labour: World War I 214–21; and hours of work 163, 173 shutdowns, industrial: French 66–71 skill premia: Britain and Ireland 144–6, 154; Britain and USA 124–5; Canada 215–16; pre-industrial Britain 40; preindustrial Netherlands 45–7 skills: Canadian railway workers 216– 21; and job tenure 193, 194, 195, 197, 198; racial differentials 276–7; and USA manufacturing 7–10 slave labour 1–2, 275 small arms industry: USA 7 Smith, J.P. 289 space: role in defining labour markets 4–10 spatial integration: France 78–82 Spero, S.D. 283 Stafford, J. 249 Stamp, J. 246, 248 statistical discrimination 286 steel industry: Canada 208; Germany 235; USA 9 stickiness: and hours of work 164, 174; of nominal wages 40, 44–5; of trade union wages 258 street and sewer workers: hours of work 164, 166–8, 177, 178, 180; wages 169, 170, 171 strikebreaking: black 282–3 strikes: Britain and Ireland 151; hours of work 173 Sundstrom, W.A. 32–3, 161–82, 276–7, 282 supply, labour: and agrarian structure 21–3; and hours of work 163, 173; inter-war Germany 235–7; preindustrial Netherlands 54–6 Sweden: agricultural labourers’ wages 100
321
INDEX technology: development of 34; USA 7–9 temporary employment see seasonal employment tenure, job: black workers at Ford 286; Canadian railways 209–14; French 80–2; USA 35, 185–202, 206–7 textile industry: France 69, 73; labour recruitment 22–3; racial discrimination 273–5; USA 7, 278 tobacco manufacturing 278; racial discrimination 273–5 Todaro model 147–50 towns: growth of 11 Trade Boards 254–5 trade unions: hours of work 162–3, 174–6; inter-war Germany 225–6; Irish migration 150–1; and job duration 195; and non-union wage changes 254–62; strikebreaking 282–3 traditional labour markets: France 29– 30, 64–82 turnover, labour 22–3, 186–8; see also tenure, job
real wages 123–5; technological change 6 Van Deursen, A.T. 43 Van Dillen, J.G. 43 Van Houtte, J.A. 43 Van Ravesteyn, W. 43 Vedder, R.K. 282 Verein deutscher MaschinenbauAnstalten 229–30 Verenigde Oostindische Compagnie (VOC) 58, 59, 60
unemployment: Britain and Ireland 147–50; changing concept of 80; industrial, at harvest time 66; interwar Britain 246–8; inter-war Germany 229–41; and Irish migration 140; and margins 252–3; preindustrial Netherlands 46, 61; and race 282; and real wages 225–6, 229– 41, 246–8 unemployment insurance: Canada 33, 34 unit labour costs, inter-war: Britain 250; Germany 227–8 unit root tests 90 USA: agricultural labourers’ wages 100, 101; agricultural mechanization 17; consumer price index 134–6; decline in hours of work 32–3, 161–82; employment relationship 206–7; gross national product 117; history of manufacturing 7–10; international labour market integration 30–1, 107– 37; job tenure 35, 185–202; racial discrimination 1–2, 31, 34, 270–91;
wage drift: pre-industrial Netherlands 47–50 wage rate models: labour market integration 89–91 wages: pre-industrial Europe 13, 21–2, 50–1; pre-industrial Netherlands 33, 44–50; and racial discrimination 31, 270–91; relationship with hours of work 164, 174, 176–80; unionized and uncovered workers 254–62; see also international labour market integration; nominal wages; real wages; regional labour market integration Wales: regional labour market integration (1850–1913) 30–1, 84–104 water power 34 Weimar Republic: wages 34, 225–43 Welch, F. 289 Whaples, R. 163, 173, 198 Whatley, W. 6, 31, 32, 37, 270–91 wheat production: Canada 118; France 70; USA 17 Wheelwright, T. 122 wholesale price index: inter-war Britain 246–7, 248, 249, 251; inter-war Germany 227–8 Williamson, J.G. 24, 65, 102, 107, 124–5, 147, 158 women: Canadian railways 215; early industrial workers 22–3; job tenure 196 working class: rise of urban 23–4 World War I: changes in Canadian labour market 214–22 Woytinsky, W.S. 206–7 Wright, G. 4, 18, 23, 29, 31, 32, 37, 270–91
322