High and persistent unemployment has been a principal feature of most major industrial economies in the 1980s and one of the phenomena associated with `Eurosclerosis'. Even recent reductions in unemployment have been accompanied by the emergence of inflationary pressure at unemployment rates far higher than in the 1960s and 1970s, implying an increase in the natural rate of unemployment. Some researchers have argued that the economies which have suffered most from persistently high unemployment have been those which are least flexible in matching their unemployed with available employment openings. Explanations for labour market mismatch have related to inadequate education and training or insufficient geographical and occupational labour mobility.

Related to mismatch are variations in unemployment rates across regions, occupations, industries and educational groups. Does the remarkable persistence of such differentials in many countries reflect substantial job mismatch, helping explain the growth and persistence of unemployment? Or is it simply due to long-standing differences in human capital requirements across occupations and in technology across regions, with no role in explaining the recent malaise in labour markets? An answer to these questions requires detailed examination of sectoral wage differentials, labour mobility, the relationship between unemployment and job vacancies, and the effectiveness of education, training and other labour market programmes. To produce robust conclusions, this investigation must cover a variety of economies with different experiences of unemployment in the 1980s.
This was the motivation for a conference on `Mismatch and Labour Mobility', held in Venice on 4/6 January 1990, sponsored jointly by CEPR, the Centre for Labour Economics at the London School of Economics, and the Centro Interuniversitario di Studi Teorici per la Politica Economica (STEP). The organizers were Fiorella Padoa Schioppa (Università di Roma `La Sapienza' and CEPR) and Gianni Toniolo (Università di Venezia and CEPR), with financial support from Directorate General V (Employment, Social Affairs and Education) of the Commission of the European Communities as well as the UK Department of Employment.

Unemployment Dispersion and Aggregate Unemployment
In their paper on `Mismatch: A Framework for Thought', Richard Jackman (LSE), Richard Layard (LSE and CEPR) and Savvas Savouri (LSE) reviewed evidence on unemployment differentials across regional, skill and age groups from a large number of industrial economies. In the United Kingdom and the United States, for example, the unskilled suffer unemployment rates four times larger than professional and managerial workers. The unemployment rates of manual workers are double those of non-manual workers in all the countries examined except Germany, where extensive training schemes may explain the lack of a skill-based disparity.
The authors argued that the variance of relative unemployment rates across sectors provided the best index of mismatch, capturing its effects on overall unemployment and wage pressure. They showed that, provided sectoral wages are determined by the sectoral labour market rather than by wage determination in a `leading sector', any increase in the dispersion of unemployment rates across sectors would, at a given rate of overall unemployment, be associated with greater wage pressure and a higher natural rate of unemployment.
Jackman, Layard and Savouri constructed time series data for this index for both occupational and regional disparities in unemployment. Though the level of mismatch is significant in all countries, only Sweden displayed any evidence of increased occupational mis- match since the late 1970s. Turning to regional disparities, while unemployment in Germany, Italy, Japan and the United Kingdom exhibited persistent regional differences, in Australia and the United States it did not. Again, in none of these countries has unemployment dispersion by region increased since the mid-1970s. This evidence warns against simply attributing the rise in unemployment persistence to increasing job mismatch. Several of the other papers presented at the conference used the index developed in this paper to construct new estimates of mismatch in various developed economies, with similar results.
Variations in unemployment rates across occupations, regions and demographic groups are closely related to variations in flows into unemployment, according to the authors' estimates. Variations in unemployment duration appear to play only a minor role. Jackman et al. also considered the role policy could play in alleviating mismatch. Where membership of a group is exogenous, as in age-groups, there is a case for subsidizing employment for some groups and taxing the employment of groups enjoying higher employment. Where workers choose their sectors, however, as with occupations and regions, interventions are only justified by specific externalities, such as to discourage congestion in low-unemployment regions and under-provision of skills training.
Sherwin Rosen (Chicago University) wondered why the authors' measure of mismatch had exhibited a downward trend in countries where unemployment levels had increased. Moreover, the link in their model between dispersion of unemployment rates and inflationary tendencies was based on a stable relationship between real wages and unemployment, which he believed was not supported by empirical evidence.
Two of the country studies presented at the conference examined whether rising unemployment during the 1980s could be partially explained by the dispersion of unemployment. In his paper on `Mismatch in Japan', Giorgio Brunello (Osaka University) found that the index of mismatch suggested by Jackman, Layard and Savouri displayed no upward trend from the mid-1970s to the 1980s across sexes, ages or regions. Brunello argued that previous indices had failed to capture more or less permanent regional differences in unemployment and in the unemployment/vacancy relationship, or Beveridge curve. He believed the former were caused by differences in amenities and the latter by differences in matching technologies. The suggestion that mismatch was zero if unemployment rates were the same across regions or if regional shares of vacancies equalled their shares of total unemployment failed to account for such permanent structural effects.
Brunello provided empirical evidence on several measures of mismatch in the Japanese labour market during 1970-87. Though the unemployment/vacancies curve shifted outwards over this period, there was no evidence that increases in mismatch played a role in these shifts. The estimates suggested that most regional differences in unemployment were equilibrium differences and that the disequilibrium phenomena which the Jackman/Layard/Savouri index sought to capture accounted for only 4% of the total. A cursory examination of regional matching technologies showed a wide dispersion: the proportion of engagements accounted for by public agencies, for example, varied across regions from 62% to 7%.
Sushil Wadhwani (LSE and CEPR) criticized the inclusion of structural unemployment differentials in the indices used by Brunello, because it was the persistence of such differentials that one wished to explain. He also commented on the finding that regional wages did not appear to depend on regional unemployment. If regional wages in Japan respond only to the national rate of unemployment, then on the definition proposed by Jackman, Layard and Savouri, there is no mismatch.
In his paper on `Match and Mismatch in the German Labor Market', Wolfgang Franz (Universität Konstanz) focused more on the Beveridge curve. Data on vacancies suggested some evidence for a sequence of outward shifts in this curve after 1982, which Franz argued reflected increases in unemployment caused by structural imperfections in the labour market. Using a disequilibrium macroeconomic model to test for the presence of rationing in the German labour market, he found evidence of rationing and of upward shifts in equilibrium unemployment throughout the 1980s. Conventional mismatch indices showed no trend in Germany in the 1980s and, as with most of the countries under study, inter-regional migration exhibited a trend decline, correlated with the growth in unemployment. Franz used regression techniques to investigate alternative explanations for the outward shifts in the Beveridge curve. The proportion of long-term unemployed exhibited a secular trend along with overall unemployment and explained quite well the outward shifts in the Beveridge curve. The argument that employer choosiness might be to blame was also supported by evidence that vacancy durations had exhibited a positive time trend independent of the overall increase in unemployment.

Regional Mismatch and Labour Migration
Issues of job mismatch are closely related to those of labour mobility. Several of the papers presented at the conference examined skills training and occupational mobility, while others looked in more detail at the relationship between regional variations in unemployment and wages and patterns of inter-regional migration. Samuel Bentolila and Juan Dolado (Banco de España) examined `Mismatch and Labour Mobility: The Case of Spain', where unemployment increased from 4% to 20% over 1977-86. This has been associated with increases in unemployment persistence and in the natural rate, as witnessed by recent inflationary pressure at 17% unemployment. Unemployment disparities in Spain are wide and have increased with the overall rate. But Bentolila and Dolado's calculation of Jackman, Layard and Savouri's index of mismatch across sex, age, educational, regional and skill groups suggested that measured in this way, mismatch in Spain fell consistently between 1977 and 1986.
Survey evidence on labour mobility in Spain suggests that the costs associated with geographical movement are high and that inter-regional migration flows have been declining since the mid-1960s. In attempting to explain migration behaviour, the authors estimated pooled equations for inter-regional migration and for the dynamic adjustment of regional wage differentials. They found that migration does respond to economic variables such as regional variations in unemployment and wages, but only slowly. Wage differentials are also slow to respond to an exogenous increase in unemployment in one region. Credit constraints or risk aversion during periods of high unemployment might explain slow inter-regional migration, Bentolila and Dolado suggested. Policies designed to move jobs to people may be more effective in the short to medium run than those intended to encourage people to move towards jobs.
Nicola Rossi (Università di Venezia) noted that the authors had used data on population migration rather than worker migration. Their migration equation should therefore explain household decisions, not just workers' decisions, by including demographic variables in the model. Stephen Nickell (Nuffield College, Oxford, and CEPR) suggested that information on job quit rates could also be used to measure labour mobility.
The paper by Orazio Attanasio (Stanford University and CEPR) and Fiorella Padoa Schioppa on `Regional Inequalities, Migrations and Mismatch in Italy, 1960-1986' used a comprehensive data set which provided important new evidence on mismatch. Aggregating the data into six economically and sociologically homogeneous regions, Attanasio and Padoa Schioppa generated series for the variance of relative unemployment rates, as suggested by Jackman, Layard and Savouri. This index showed no trend over the 1970s and 1980s. After a long period up to the early 1970s, characterized by strong migration flows in the directions suggested by inter-regional unemployment differentials, inter- and intra-regional migration both fell throughout the 1980s. Patterns of migration exhibited a high level of persistence. The authors attributed the decline in migration and the breaking of its relationship with regional unemployment variations partly to increasing overall unemployment and partly to substantial reductions in regional real wage differentials. This was the result of progressive taxation, increased cost-of-living differentials, the growth of national wage bargaining, and rising government transfers to the South.
Giuseppe Bertola (Princeton University and CEPR) argued that the initial aggregation of regions should have been derived from more formal econometric analysis, since there seemed an element of circular reasoning in grouping according to economic characteristics and examining for dispersion along economic lines. He also observed that no theory existed that could explain the persistence observed in the migration equations.
In their paper on `Housing and Regional Migration to and from the South-East', John Muellbauer (Nuffield College, Oxford, and CEPR) and Anthony Murphy (Nuffield College) analysed the effects of house prices and labour demand on migration into and out of the most prosperous and congested region of the United Kingdom in the period 1971-87. They constructed a model of inter-regional migration in which they took considerable care in the choice of variables capturing relative labour demand. The authors found that regional differentials in the ratio of house prices to earnings were the single most important determinant of regional migration flows. Regional labour demand growth and unemployment rates were also important determinants of migration. Regional variations in the house prices to earnings ratio discourage migration into regions with higher employment, Muellbauer and Murphy argued, and so worsen job matching.

Skills Mismatch and Training Programmes
Charles Bean (LSE and CEPR) and Christopher Pissarides (LSE) presented their paper on `Structural Unemployment and Skill Shortages in Britain: A (Mis)Matching Approach'. Although conventional indices suggested that mismatch had not risen in the UK labour market, there were several explanations why shocks hitting one sector could increase overall unemployment without affecting relative unemployment rates. This could happen, for example, if workers migrated out of the affected sector or if offsetting movements in relative wages left relative unemployment rates unchanged. UK data suggest that the duration of vacancies rises with the level of skill, so the authors outlined a simple general equilibrium model of unemployment which distinguished two sorts of labour. Unskilled workers were governed by a conventional union bargaining model and skilled workers by a search/matching model. The model showed how technology shocks that affected the two sorts of workers unevenly could affect equilibrium unemployment, even though technology shocks that affected both sorts equally did not.
In their empirical work, Bean and Pissarides had to use data on manual and non-manual workers, due to the lack of data distinguishing skill levels. They estimated two different regression equations for manual and for non-manual wages, using data on wages, productivity, unemployment, benefits and skilled labour shortages in 15 industrial sectors over 1971-88. The results showed that non-manual wages were strongly influenced by economy-wide non-manual wages and by industry-level skill shortages, and only weakly influenced by productivity. Manual wages, however, were strongly influenced by overall unemployment and by the level of non-manual wages in the same industry. This `comparability' effect suggested that technology shifts favouring greater use of non-manual labour could well produce significant rises in manual unemployment.
Ugo Trivellato (Università di Padova) argued that the authors' model had not explained why relative unemployment rates had appeared to be so stable over time. Dennis Snower (Birkbeck College, London, and CEPR) argued that the model would be better suited to analysing the service and manufacturing sectors: the model suggested that a sector undergoing productivity growth should experience a decline in unemployment, but UK manufacturing has exhibited no such effect.
Richard Freeman (Harvard University) analysed `Labour Market Tightness and the Declining Economic Position of Young Less Educated Male Workers in the United States'. Since 1979, the relative earnings of less educated male `high school dropouts' had declined, whereas previously their relative earnings had been increasing. Freeman used cross-section data from the Current Population Survey to investigate whether this change was caused by structural phenomena such as skills mismatch or associated with swings in the business cycle. He found that declining real and relative earnings for this group over 1973-88 were accompanied by deteriorating employment prospects relative to more skilled workers, suggesting that the effect of falling real wages was dominated by an inward shift in labour demand for young unskilled males. The data also revealed that the earnings and employment of this group were noticeably higher in localities with lower unemployment, suggesting that macroeconomic factors were important in the overall decline of their economic position.
Michael Burda (Institut Européen d'Administration des Affaires and CEPR) argued that Freeman had overlooked the discrete nature of unemployment. Probit analysis should have been used instead of regressions based on continuous variables. Guy Laroque (Institut National de la Statistique et des Etudes Economiques) thought a skill premium had emerged in the 1980s, but Katharine Abraham (University of Maryland) questioned the extent to which improvements in technology had shifted demand away from the unskilled. Anecdotal evidence suggested technology had been aimed at less skilled, less scarce workers.
The importance of skills training in the occupational dimension of mismatch was taken up by David Soskice (University College, Oxford). The starting point of his paper on `Skills Mismatch, Training Systems and Equilibrium Unemployment: A Comparative Institutional Analysis' was to examine how closely national statistics on skills mismatch were related to consensus views on the effectiveness of training schemes. Countries often thought to have effective schemes included Germany, Japan and Sweden, while those in the United Kingdom and United States are generally seen as less effective. Aggregate statistics appeared ambivalent on this question. Soskice argued, however, that conventional measures of effectiveness overlooked differences in the demand for labour. In Germany, Japan and Sweden, where companies tend to aim at innovation-intensive international markets and to demand a highly skilled workforce, there is a firm-led drive for trained skilled workers. UK and US firms, however, aim more to produce standardized goods and services and appear less concerned with the skill levels of school-leavers.
Leonardo Felli (MIT) argued that the initial assumption that Germany, Japan and Sweden were diverse producers and the United Kingdom and United States were standard producers led inevitably to the conclusion that their education and training systems were efficient. He also noted that Soskice's model was observationally equivalent to a screening/incomplete information framework and that this similarity should be examined further.
Per-Anders Edin and Bertil Holmlund (Uppsala University) examined `Unemployment, Vacancies and Labor Market Programs: Swedish Evidence'. They focused on whether Sweden's widely acclaimed training programmes and `relief jobs' had helped to facilitate transitions from unemployment to regular employment. The counter-cyclical nature of relief jobs in Sweden, which last six months, is such that a 10,000 rise in unemployment is associated with a 5,000 rise in relief jobs. Edin and Holmlund found that those in relief jobs engaged in considerably less intense job search than those in open unemployment and that relief jobs appeared to contribute little to the flow of aggregate hirings. It is possible, however, that this is because individuals with unfavourable re-employment prospects are disproportionally selected for relief jobs. Labour market training programmes, in contrast, which accounted for 15% of the outflow from unemployment in 1988 and 1989, had accelerated the rate of inflows to employment relative to the outflow from open unemployment. Estimates for 1970-88 revealed no outward shift in the unemployment/vacancy curve and therefore no tendency for a reduction in the efficiency of matching.

'Mismatch and Labour Mobility' edited by Fiorella Padoa Schioppa.
Available from CEPR, 90-98 Goswell Road, London, EC1V 7RR
ISBN (hardback) 0 521 40243 4