The worldwide rise in unemployment in the last decade has rekindled the interest of economists and historians in the Great Depression of the 1930s. The experience of the thirties is often seen as a watershed, a traumatic experience which influenced social and economic policy for more than a generation. The sudden and unexpected emergence of mass unemployment; its uneven incidence across industries, occupations and individuals; the rise of an underclass composed of the long-term unemployed; and the belated and partial responses of public authorities are the building blocks of the mythology of the 1930s. Yet all too often such caricatures are based on an inadequate understanding of exactly how, why and by whom the burden of unemployment was shared.

During the 1930s, just as today, unemployment rose worldwide, though the experience of different countries varied. Much of the renewed interest in the 1930s has focused on the functioning of labour markets, which may offer the key to understanding depression and recovery. This emphasis has raised new questions concerning the operation of these markets. Analysis of the UK and US labour markets at a macroeconomic level has proved particularly controversial and has already stimulated a large and growing literature. Yet many other equally important aspects of interwar labour markets remain unexplored. The labour market experiences of Continental Europe, Canada and Australia, for example, have been almost entirely neglected. In the absence of an international perspective, there is a temptation to describe each country's experience in terms of specific national circumstances; in consequence, the more fundamental characteristics of interwar unemployment can easily be overlooked.

These concerns provided the motivation for a recent conference on 'Interwar Unemployment in International Perspective', held at the Center for International Affairs, Harvard University, on 7-8 May. The conference heard eleven papers, two providing a comparative analysis and nine focusing on the unemployment experience of individual countries. The conference was attended by economists, historians and sociologists from nearly a dozen countries. The conference was organized by Research Fellows Barry Eichengreen (University of California, Berkeley) and Tim Hatton (University of Essex), with the assistance of CEPR and the Center for International Affairs. Financial support was provided by the Center for International Affairs, NATO and the Harvard/Mellon Economic History Center.

In their introductory paper, 'International Perspectives on Interwar Unemployment', Barry Eichengreen and Tim Hatton provided an overview and comparative perspective on some of the key issues of the conference. Above all, perceptions of the severity, incidence, and character of unemployment are influenced by the limited statistics available for the interwar period. These arise primarily from the operation of trade union and unemployment insurance funds and usually focus exclusively on the industrial sector. For most countries (the United States and Canada serving as notable exceptions), such data indicate much higher unemployment rates than population censuses and labour force surveys. Other evidence of this discrepancy can be found in regression estimates of 'Okun's Law' for the interwar period, relating changes in output to changes in unemployment: these regressions imply implausibly large elasticities of unemployment with respect to output. It would be dangerous to conclude, however, that differences across countries reflect only the peculiarities of the data gathered during the 1930s. Looking across countries, the slump and recovery in output within the manufacturing sector were accompanied by very different patterns of unemployment. In some countries, such as the United States and Italy, work-sharing was a common response to the slump, giving rise to a permanent decline in hours worked per employee; in other countries, such as Australia, workers tended to be laid off, with relatively little adjustment in the length of the working week for the employed. As a result, productivity per worker behaved very differently across countries, slumping badly where work-sharing was the rule.

Labour supply responses are more difficult to identify. The issue is important, in view of the controversial claims that the 1930s were characterized by a substantial rise in hidden unemployment. Current labour supply behaviour provides little guidance in interpreting the interwar period; responses then may have differed from those of the postwar years because of changes in alternative income sources (either from the state or from secondary employment), in demographic composition and in family structure. This presumption is consistent with a number of other distinctive characteristics of interwar unemployment. Evidence from the United States and United Kingdom suggests that turnover among the unemployed was higher than in the postwar period. At the same time, however, a growing proportion of the 1930s unemployed had been jobless for long periods and faced a very low probability of re-employment. Unemployment rates also show a distinctive age pattern in the 1930s: while recorded rates of youth unemployment were low in several countries, the age profile of adult unemployment was 'U'-shaped - falling with age until the 35- to 45-year age bracket, rising thereafter. Convincing explanations of this have not yet been discovered. Despite the declining income and morale of many of the unemployed in the 1930s, the effects of unemployment on individual health, family structure, social attitudes and institutions are by no means well understood.

The discussion which followed focused on definitions of unemployment and on comparability across countries. Mark Thomas (University of Virginia) pointed out that the statistics available from unemployment insurance systems are a political and administrative artefact and do not necessarily correspond with postwar definitions. Gianni Toniolo (University of Venice) and Robert Salais (Institut National de la Statistique et des Etudes Economiques (INSEE), Paris) argued that in many countries there was no clear definition of an 'unemployed' individual; imposing contemporary definitions may distort our understanding of interwar labour markets. It is still not clear how regional and industrial differences in unemployment incidence should be interpreted. Did 'structural mismatch' exacerbate interwar problems or merely reflect the concentration of the depression in certain sectors?

There has been a growing consensus that in most countries the rise in unemployment in the 1930s was due partly to the sluggish behaviour of nominal wages. The rapid price deflation of the early 1930s, coupled with nominal wage inertia, led to a sharp rise in the real wage which priced workers out of jobs. In their paper, 'The Macroeconomics of the Interwar Years: International Comparisons', Andrew Newell (University of Sussex) and Jim Symons (University College, London) argued that the behaviour of real wages and real interest rates was the proximate determinant of both the depression and the recovery. Newell and Symons estimated a set of equations designed to explain the time-series behaviour of employment, the real wage and price changes, using data for 14 countries. They found that the real wage and real interest rate had strong negative effects on employment. (Their estimates suggested that the long-run elasticity of employment was -0.7 with respect to the real wage and -1.3 with respect to the real interest rate). Surprisingly, in view of the attention paid to these effects in the UK and US in earlier literature, the results were weakest for these countries. The real wage equations in turn revealed that changes in the price level had powerful effects on the real wage, confirming a substantial degree of nominal wage inertia. Unemployment, in contrast, had only a weak effect on real wages. Price changes in Newell and Symons's model were determined by world prices (adjusted for the effective rate of protection) and world trade. For individual countries, Newell and Symons argued, the contraction in world wealth and world trade led to the fall in the price level and the increase in real wages and real interest rates which brought about the decline in employment. At the world level, the dominant causal factor in Newell and Symons's model was a contraction of aggregate demand, due to the fall in the value of assets and wealth, and disturbances in the monetary sector, which arose from banking failures in the United States.

In his comments, Robert Gordon (Northwestern University and CEPR) argued that Newell and Symons's model involved an asymmetry. The labour demand function was based on a model of optimizing behaviour while the labour supply (wage) equation was not. He suggested that the price terms in the labour demand model were acting as proxies for aggregate demand and that the model could be interpreted as a disequilibrium model driven by variations in output demand. Stephen Nickell (Institute of Economics and Statistics, Oxford, and CEPR) argued that the alternative disequilibrium model was largely tautological because output and employment were linked through the production function. Newell and Symons replied that their model was not an attempt to test competing theories; insofar as rival interpretations implied different sets of restrictions on the estimated coefficients of the model, theirs were satisfied by the data. Stefano Fenoaltea (Swarthmore University) and Brad DeLong (Harvard University) wondered whether the model provided a useful interpretation of the global causes of the depression. The financial collapse was important, but why were its repercussions so much more severe than those of other international crises?

Most of the debate on the causes and consequences of high unemployment in the 1930s has focused on the United States and United Kingdom and has used annual time-series data at an aggregate level: microeconomic evidence has been relatively neglected. Such aggregate analyses have, however, left the puzzles of interwar unemployment largely unsolved. The remaining papers presented at the conference compared results already obtained for the United Kingdom and the United States with those for other countries: Canada, Australia, Germany, Italy, France and Belgium. The papers emphasized in particular the impact of and responses to unemployment at the microeconomic level; previous analyses had tended to neglect responses at the level of the industry and region and at the level of the firm and household.

In his paper on the US labour market, entitled 'Interwar Unemployment in the United States: Evidence from the 1940 Census Sample', Robert Margo (Colgate University) provided the first analysis of unemployment based on the 'public use sample' of 100,000 individuals drawn from the 1940 census. Margo contrasted the personal, regional and industrial characteristics of three groups of household heads: the employed, the unemployed and persons on 'New Deal' relief projects. His data revealed that, compared with the employed, the unemployed tended to be older, foreign-born, single, urban, less geographically mobile and more likely to be living in the North-East or the West. Michael Darby has argued that US unemployment in the 1930s is exaggerated by labour force statistics which treat those working on relief projects as unemployed: relief work was equivalent to employment, Darby claims. Margo's analysis suggests this view is incorrect: it is hard to reconcile the Census evidence with the view that individuals were indifferent between relief work and regular employment. Those working on relief projects differed from the unemployed: they were more likely to be under 45, married and owner-occupiers but less likely to be either white or foreign- born. While Margo's analysis indicated that higher levels of schooling and geographical mobility lessened the risk of unemployment, he also found that they lowered the chances of working on relief projects when unemployed. Those on relief work tended to experience very long spells without regular work, indicating that for many, relief work was virtually a permanent occupation. Overall Margo's results suggest that relief work was not equivalent to regular employment and that the conditions and characteristics of relief workers were substantially different from the wholly unemployed.

Martin Baily (The Brookings Institution) suggested that the regression analysis of the incidence of unemployment was largely a descriptive exercise and was not capable of discriminating between alternative hypotheses about the macroeconomic causes of persistent unemployment. A number of participants questioned the interpretation of Margo's results. Even if those employed for long periods on relief projects had characteristics different from the wholly unemployed, their incomes were only about half of those in regular employment. Since relief workers also had to work, their incentive to seek real employment would not necessarily differ from that of the totally unemployed. John Wallis (Maryland University) emphasized the variety of work relief schemes and argued that the characteristics of the workers employed and the instability of employment reflect in part the nature of the projects and the process of selecting workers for employment on them.

The next session featured two papers on the character and consequences of unemployment in interwar Britain. Mark Thomas, in his paper 'Labour Market Structure and the Nature of Unemployment in Interwar Britain', examined measures of labour turnover and unemployment duration using data from the records of the labour exchanges. Thomas's analysis emphasized the flows of individuals into and out of unemployment, in contrast to previous studies focusing on 'stock' analysis of the employed and unemployed. Strikingly, his calculations revealed that inflows to unemployment in the early 1930s were two or three times higher than the levels observed in the 1980s. In consequence the average (completed) duration of unemployment in the 1930s was lower than might have been expected from recent experience: about 10 weeks for all unemployed but 20 weeks for the wholly unemployed (which excludes the temporarily stopped and casual workers). But for the wholly unemployed the length of unemployment spells (weighted by work experience) rose from 15 weeks in 1929 to nearly 50 weeks in 1936. How can these observations be reconciled? The unemployed, at least in Britain, appear to have been divided according to the duration of their unemployment into two groups with very different characteristics. The first group experienced rapid movements into and out of employment with repeated brief spells of unemployment. The system of unemployment benefits, which imposed no waiting period on most claimants experiencing recurrent spells of unemployment, may have contributed to these frequent short spells of unemployment. For the long-term unemployed, however, the benefit system was less important: for all individuals the probability of re-employment declined steadily with the length of their unemployment.

In his comments, Charles Feinstein (Nuffield College, Oxford) emphasized that particularly in the period before 1914, many labour markets had been characterized by rotating underemployment: as many as two million workers were out of work for 20-25% of the time. In the interwar years, intermittent unemployment or underemployment was still present but the collapse of certain staple industries was an additional factor, and long-term unemployment was closely associated with these industries. Several participants pointed also to the different labour market conditions faced by different demographic groups, particularly juveniles. Barry Eichengreen noted the lower cyclical sensitivity of juvenile unemployment and the lower participation rates among this group, which may partly reflect measurement problems.
Though it is difficult to argue that the long-term unemployed were induced by the insurance system to stay on the register for such long periods, the average income loss from unemployment was limited to about 25%. Does this imply that unemployment was not a major cause of poverty and related social ills? In his paper, 'Unemployment, Insurance and Health in Interwar Britain', Bernard Harris (Birkbeck College, London) reviewed contemporary studies of the poor, arguing that these tended to underestimate poverty. Harris based much of his empirical analysis on data from the reports of school inspectors. He noted that although general mortality declined after 1911 the decline was smallest in areas of high unemployment. Maternal mortality and that among men over 55 appeared to be the most sensitive to unemployment. It is difficult, however, to obtain reliable indicators of general health, the variable presumably linking unemployment and mortality. Variations in nutrition have been found to influence the early growth and therefore the height of children, which may therefore provide a useful indicator of general health. Harris's time-series analysis for certain areas in the 1930s revealed that unemployment levels had a negative effect on stature, although this effect was weak.

Harris had alluded to the issue of 'benefit-induced' unemployment in his paper, but several participants thought that one could not really address this issue by looking at health or mortality. Nicholas Crafts (University of Leeds and CEPR) commented that although poverty lines are always somewhat artificial, nonetheless the interwar period as a whole saw a general rise in children's heights. He noted that elderly workers without children suffered the most from long-term unemployment: their experience would not be expected to influence heights. If there were any effects of unemployment on heights it would occur only with a lag and might not be picked up in analyses of annual time series.

The economic policies undertaken by fascist regimes in the 1930s are often said to have promoted rapid economic recovery and to have virtually eliminated unemployment. Dan Silverman (Pennsylvania State University), in his paper 'German Unemployment in the 1930s', examined the policies for economic recovery under the Nazi regime. Could work-creation programmes explain the dramatic turnaround in the German economy, which saw unemployment fall by 60% in the first 18 months of the regime? Silverman's examination of the timing of these schemes suggested that they could have had only marginal effects on the jobless total until the rearmament drive began in 1935. Silverman also explored the suggestion that changes in the procedures used for recording unemployment could have exaggerated the 1932-3 fall in the numbers recorded as unemployed. Although deliberate statistical manipulation is impossible to establish, the government department concerned with these statistics was thrown into chaos during the process of Nazification. Coupled with changes in eligibility for and coverage of unemployment insurance, this obscures the true picture of the German labour market in the critical early years of recovery and has led many observers in the past to exaggerate the impact on unemployment of German labour market programmes, according to Silverman.

Harold James (Princeton University) pointed out that even though there is some doubt about the reliability of the employment statistics, there is little question that Germany experienced a dramatic economic recovery after 1932. Two major factors contributed to this. First, nominal wages were held down by the Nazis following the destruction of trade unions; as a result employment grew rapidly in low-productivity sectors. Second, trade and currency controls instituted in 1931 and intensified in 1934 and 1936 led to increasing isolation from the world economy and growing employment in import-substituting sectors.

Inadequate data make the unemployment picture for Italy even more obscure. In their paper, 'Italian Unemployment in the 1930s', Francesco Piva (University of Rome) and Gianni Toniolo emphasized the 'dual' nature of the Italian interwar economy. Until 1929 there were high rates of migration from the agricultural sector, where underemployment was common, to the industrial cities and abroad. Although industrial employment declined by only 6% between 1929 and 1932, disguised unemployment undoubtedly increased. There is evidence of job-sharing in the sharp decline in hours worked per employee and the high rates of employment turnover. Although the level of unemployment is unclear, the authors' estimates suggest a dramatic recovery between 1933 and 1937. As in the German case, however, Mussolini's public works policy had only marginal effects on industrial unemployment in 1931-3. Piva and Toniolo argued that in Italy there is evidence of a strong inverse correlation between the real wage and total hours worked, at both the aggregate and industry levels. But whether policies aimed at reducing average hours and holding down the nominal wage contributed to the Italian economic recovery remains unclear.

Stefano Fenoaltea commented that, in his view, the Italian labour market in the interwar period was not unusual. It shared many of the features of other European countries and of countries such as Britain at a much earlier stage in their development. In such cases the phenomena of 'rotating underemployment' and a high degree of seasonality in employment were common, but the characterization of the rural sector as holding huge labour reserves was misleading. Charles Maier (Harvard University) disagreed with this interpretation, arguing that there was less seasonal variation in agriculture in the Italian south than in the United States.

France and Belgium are two of the countries whose economic recoveries in the 1930s were hindered by their adherence to the gold standard. In both countries, however, unemployment rates remained surprisingly low. In his paper, Robert Salais asked 'Why Was French Unemployment So Low During the 1930s?'. One reason is that the French labour force declined dramatically between the 1931 and 1936 censuses. It is sometimes argued that the large exodus of foreign workers was the major reason for this decline, but there were other important factors, according to Salais. Recorded unemployment in France was also moderated by the return of urban workers to rural areas and by the prevalence of self-employment. Salais found little relation between employment loss and unemployment across départements of France, although recorded unemployment was strongly correlated with urbanization and with the mix and organization of industry. Home workers and those employed in small businesses who experienced a loss of work were less likely to appear as 'unemployed' due to the nature of their employment contract; those in large firms were more likely to be permanently laid-off and hence to be recorded as unemployed. Institutions devoted to the relief of unemployment and the concepts used to measure it were less developed in France than elsewhere; this may have led to serious undercounting of the jobless. Salais's analysis of the records of over 6000 individuals extended relief in Paris showed that those who did not head households often experienced termination of benefits after only brief periods of unemployment.

A number of participants questioned whether the low aggregate unemployment rate could be satisfactorily explained in terms of the peculiarities of different types of employment. Robert Gregory (Australian National University) pointed out that demographic variations and changing participation rates made more important contributions to changes in French unemployment than in most other countries. Charles Maier remarked that unemployment was unusually low even in the factory sector, where similarities with the experience in other countries might have been expected. He suggested that the existence of a large traditional sector disguised much of the unemployment and that the late onset of the depression delayed its recognition as an issue of public concern.

Martine Goossens, Stefaan Peeters and Guido Peppermans (Université Catholique de Louvain) considered the Belgian experience in their paper, 'Interwar Unemployment in Belgium'. They provided the first time-series estimates of the Belgian labour force for the interwar period, using both census data and the records of the voluntary unemployment insurance system. This insurance system had been extended in the 1920s as a result of trade union pressure and government intervention; by the early 1930s, it covered nearly a million workers. As in France the census indicates a sharp drop in the Belgian labour force from 1930 to 1936. The authors' new estimates of the total number unemployed suggest that this contraction of the labour force made a major contribution to the fall in unemployment between 1933 and 1938.

In his comments Bradford Lee (Harvard University) suggested that, since the Belgian unemployment insurance system was linked with trade union membership, one might have expected lay-offs to be more prevalent than reductions in hours. Peeters indicated that there was insufficient data available either on lay-offs and hours or on the nature of employment contracts to address this question. Others questioned the authors' assumption, in constructing their estimates, that unemployment rates among insured and uninsured workers were the same.

Canada's unemployment in the 1930s was nearly as severe as, and largely originated in, conditions in the United States. In their analysis of 'Unemployment and Relief in Canada During the 1930s', Alan Green (Queen's University, Kingston) and Mary MacKinnon (Australian National University) made use for the first time of census figures and regional data on unemployment relief. They found that, despite the regional specialization of the Canadian economy and its reliance on a narrow range of staple exports, the regional diversity in unemployment rates was surprisingly small. Differences in unemployment rates were more marked across occupations and between the sexes, with females and those in non- manual occupations experiencing the lowest rates. In 1930-1, 44% of wage-earners lost hours of employment and on average their loss represented six months work. Despite this, Canadian work relief projects started in the early 1930s did not develop into a full-scale programme like that in the United States. Compared to the US 'New Deal', Canadian relief provision in cities was meagre, but in rural areas assistance to farmers was more generous. Although agricultural areas suffered a disproportionate income loss, the proportion of the Canadian labour force employed in agriculture rose between 1931 and 1941.

In his comments, John Wallis suggested that the regional dispersion of unemployment rates was more significant than the authors had maintained. He also raised the question whether female unemployment was low relative to that of males because of the industries in which they were employed, or whether these industries were less affected by the depression because they employed females. Gianni Toniolo and Charles Feinstein noted that lower female unemployment was found in most countries and suggested that this reflected strong 'discouraged-worker' effects in the early 1930s.

The distinctive feature of the Australian experience was the rhetoric of 'equal sacrifice', i.e. that the burden of unemployment should be shared equally. Robert Gregory, in his paper 'The Australian Labour Market During the 1930s', written jointly with V Ho, L McDermott and J Hagan, demonstrated that in practice sacrifice was very unequal. 'Job-sharing', as reflected in the decline in labour productivity and the fall in hours worked per employee, was not as common as in the United States. Had the US output-employment relationship also held in Australia, Gregory estimated, unemployment would have risen to only 12% in 1932 rather than the 19% observed. Moreover, once out of employment, Australians had very little chance of regaining jobs. By 1933 there were more workers who had been unemployed for two or three than for one or two years, reflecting the almost permanent nature of the job losses that had occurred in 1930-1. Gregory concluded that labour market theories emphasizing rapid employment turnover and search behaviour are particularly inappropriate to Australia. A distinctive feature of the Australian institutional framework was the centralized wage- setting process whereby wage rates were linked to the cost-of- living index in the previous quarter. Gregory pointed out that despite this unique feature, the course of real wages over the depression was remarkably similar to that in the United States, where no such institution existed. In neither country did real wages alone seem to account for much of the variation in employment.
Stephen Nickell remarked that unless perfect competition prevails in the product market, the product wage might either rise or fall as a result of a negative demand shock. Since the real wage is not determined solely in the labour market it should not be seen as measuring labour market flexibility. He also pointed out that labour hoarding is a form of inflexibility and should not necessarily be regarded as desirable. Ben Bernanke (Princeton University) suggested that labour hoarding in the United States might have reflected the climate of industrial relations, in which employers desired to retain a stock of skilled workers attached to the firm. John Wallis stressed that the rules governing work relief schemes under the New Deal tended to lower actual, or at least reported, hours.

These papers together provide a new and much richer picture of interwar unemployment, which emphasizes the diversity of experiences across countries, regions and individuals. Revised versions of the papers presented at the conference will be published for CEPR by Martinus Nijhoff in early 1988, edited by Barry Eichengreen and Tim Hatton.