The analysis of individual job search of workers and hiring and separation behaviour of firms has provided fruitful insights into the functioning and dynamics of labour markets. A workshop held on 'Entry and Exit of Firms and Workers and the Interaction between Labour and Product Markets' at the Humboldt University in Berlin on 9-10 June considered both theoretical as well as empirical aspects of the flow approach to labour market analysis. The workshop was organised by Michael Burda (Humboldt University in Berlin and CEPR) as part of the network on 'Product Market Integration, Labour Market Imperfections and European Competitiveness', funded by the European Commission's Human Capital and Mobility programme. The papers presented during the workshop focused on the interaction of job and worker reallocation, cyclical patterns of job and worker flows, the influence of labour market policies and regulations as well as explanations for the divergent unemployment performance of European and North-American labour markets.
Simon Burgess (University of Bristol and CEPR) stressed the difference between gross worker reallocation, defined as the sum of accessions and separations of workers across firms over some period of time, and job reallocation, defined as the sum of job creation and destruction undertaken by firms. In a paper 'Job Flows, Worker Flows and Churning' joint with Julia Lane (The American University) and David Stevens (University of Baltimore), he identified the difference between these two figures, which the authors call churning, as an important component of gross worker movements. According to the authors, churning accounts for about 2/3 of all worker flows, which, in the end have no net impact on employment levels. In addition, the authors present evidence that churning is pervasive, persistent over time, and particular to industries and even firms. Michael Burda welcomed the paper's contribution as correcting an important deficit in our understanding of gross worker flows. Both firm-initiated severance as well as worker-initiated quits are responsible for churning flows, although the authors' finding that churning flows are procyclical suggests that the latter dominates; the former are also important however, and probably occur predominantly in recessions when the associated opportunity costs of "cleaning house" are low.
In 'Gross Worker Flows: How Does the Spanish Evidence Fit the Stylised Facts' Pablo Antolin (Instituto de Analisis Economico, CSIC) examines empirically the cyclicality of employment and unemployment flows for the Spanish labour market. During the period of 1977 to 1994, flows into unemployment seem to be counter-cyclical whereas flows out of unemployment seem to be acyclical. The flows into employment show a pro-cyclical pattern, while flows out of employment are counter-cyclical. These findings are in contrast to evidence recently found for other countries that employment flows are pro-cyclical and unemployment flows counter-cyclical. Antolin points out that these results may be explained by the strongly pro-cyclical labour force growth and participation rates, and by the existence of dual labour market where the frontier is drawn between workers with permanent labour contracts and those hired on a temporary basis. A higher degree of on-the-job search in the secondary market, consisting primarily of workers with fixed-term contracts, dampens the effect of the cycle on unemployment-to-employment flows. Finally, both the magnitude and the variability of flows have increased during the last years. Karl Pichelmann (Institute for Advanced Studies, Vienna) wondered whether the author's interpretation was consistent with the high cyclicality of unemployment stocks in Spain compared with other OECD countries.
Jonathan Leonard (University of California, Berkeley) emphasized the complexity of labour market dynamics when the heterogeneity of labour demand on the firm level is allowed for. In recent studies which consider turnover of jobs and turnover of workers simultaneously, it is demonstrated that shrinking firms may be hiring personnel while expanding firms fire workers. This evidence challenges the results from standard macroeconomic representative firm models. In a study 'A Difference in Degree: Unemployment Despite Turnover in the Belgium Labor Market' co-authored by Marc van Audenrode (University of Quebec) the author investigates the links between the reallocation of jobs and the flows of individual workers in Belgium in the mid-1980s. Leonard reports a 20 percent annual rate of new hires for the Belgian labour market. While the degree of job reallocation in the Belgian economy is substantial, the figures are lower than those for the US. This finding holds as well for the ratio of employee turnover to job turnover being consistent with a greater investment in pre-employment screening in Belgium. The main explanation of persistent high unemployment in Belgium is, according to the author, that hiring and job creation rates have not increased sufficiently to work off the stock of unemployed created by excessive wage growth in the 1970s. Eric Smith (University of Essex) emphasized the surprising nature of the results but thought that variation in the underlying source of heterogeneity could be investigated in more detail. He remarked that existing models of matching -; in which firms consist of a single worker -; are insufficiently rich for understanding why firms hire and fire so much in supposedly sclerotic labor markets .
Monika Merz (Rice University) presented a neo-classical growth model in which hiring and firing decisions of firms are endogenised by assuming that each firm's production technology contains an idiosyncratic as well as an aggregate component of a technology shock. 'Heterogeneous Job-Matches and Cyclical Behavior of Labor Turnover'. Firms adjust their labour demand heterogeneously according to aggregate and idiosyncratic productivity shocks, by hiring, laying off or recalling workers, which, as simulations demonstrate, results in a counter-cyclical pattern of flows in and out of unemployment that is consistent with some recent empirical findings. Assuming that newly created matches take one period of time to become productive, the model is able to replicate the empirical evidence that flows into unemployment lead flows out of unemployment over the business cycle. Jonathan Leonard noted that countries where layoffs did not exist exhibit similar patterns of countercyclical worker flows. The audience exhibited some skepticism about the model's accounting for certain transitions between layoff (attached to a firm) and unemployed (unattached) and nonconvexities which might arise through mobility costs.
In 'Separation Cycles in a Two Sided Search Equilibrium' Melvyn Coles (University of Essex) emphasised the importance of partnership separation in matching models. Most studies have ignored this subject or modelled separations according to some exogeneous process. The purpose of the paper joint with Ken Burdett (University of Essex) is to integrate a theory of endogeneous separations into the matching framework. In a stylised model of the marriage market the authors demonstrated how equilibrium separation cycles can exist even when all agents are rational, forward-looking, expected utility maximizers who know all characteristics of their partner at the time of the marriage and in absence of exogeneous shocks. However, the authors assume that the value of the outside option of matching partners, i.e. the pay-off to separating, varies over time. It follows that agents form a match while the value of the outside option is low and separate when the value of the outside option is sufficiently high. Hence, sorting effects together with heterogeneous agents can yield periodic separation cycles. In addition, it is shown how multiple equilibria with different numbers of singles/unemployed exist without assuming increasing returns in the matching technology and that the realised steady state equilibrium depends on initial agents' beliefs concerning the selectivity of other market participants. Pietro Garibaldi (London School of Economics) wondered, as many others at the workshop, whether such a model could be distinguished from alternative models with stochastic stocks. In addition, the exclusion of bribes or side payments may limit the model's ability to mimic real labour markets.
In 'Job Dynamics, Correlated Shocks and Wage Profiles' Antonio Cabrales (Universitat Pompeu Fabra) together with his co-author Hugo Hopenhayn (Universitat Pompeu Fabra and University of Rochester) are concerned with the impact of labour market policies on unemployment and labour turnover. Reviewing the literature, the authors conclude that the analysis of the impact of labour market policies - more precisely wage taxes and layoff costs - is highly model-dependent and that results are sensitive to different assumptions made concerning wage formation. According to their model, the effect of a change in payroll taxes depends on the existence of tax exempt-parts income, such as unemployment benefits, layoff compensation and the utility of leisure while unemployed. If exempted income is non-negative, the authors show that an increase of payroll taxes increases unemployment and labour turnover. An increase in layoff compensation is unambiguously associated with an increase in unemployment and a reduction in turnover. Guiseppe Bertola (Università du Torino and CEPR) noted that while the authors' model has the pleasing empirical implication that wages rise with tenure, alternative interpretations of upward-sloping wage profiles unrelated to productivity but might affect the authors' findings significantly.
The presentation 'On the Job Search and Unemployment Duration' by Tito Boeri (European University Institute and OECD) focused on the links between job-to-job shifts and unemployment duration. Empirical evidence suggests that job creation and job destruction rates as well as labour turnover do not differ systematically between North-American and European labour markets. This finding is surprising, since the presence of restrictive employment security schemes should imply a smoother pace of job reallocation in Europe. However, European countries show much poorer employment performance and higher and persistent unemployment rates compared to other OECD countries. According to the author, employment security regulations may lead to a shifting of labor turnover in favor of job-to-job transitions in European labour markets. Endogeneous on-the-job search increases the impact of duration on outflows from unemployment at an aggregate level. Rudolf Winter-Ebmer (University of Linz and CEPR) remarked that Boeri's conclusions were not based on direct data on job-to-job transitions or on-the-job search. He also wondered why the mere presence of on-the-job searchers should positively affect job finding probabilities for the unemployed.
Hartmut Lehmann (Ifo Institut, München and London School of Economics) reported the results of a firm-level study of gross jobs and worker flows in Polish manufacturing between 1988 and 1991, and compared those to findings in Western economies. The analysis of job flows, in particular its variation on a firm-level compared to sectoral and aggregate variations, and between private and state-owned firms, is crucial to understand the restructuring and emerging growth processes in Eastern European transition economies. One of the main findings of the study 'Employment Growth, Job Creation and Job Destruction in Polish Industry: 1988-91' joint with Jozeph Konings (Katholieke Universiteit Leuven) and Mark Schaffer (London School of Economics) is the considerable degree of heterogeneity in the Polish industrial sector with simultaneous job creation and destruction occuring in the state and private sectors. In addition, in 1990 gross job destruction and reallocation rates increased sharply in the state-owned sector, while a disproportionately large fraction of job creation took place in the private sector. Small firms are reported to be more dynamic in terms of net employment growth. After controlling for size effects, privately owned firms are on average characterised by a higher net employment growth rate. Bruno Contini (Università du Torino) warned that administrative data sources may pose problems much more significant than in established economies. Spurious "demographic events" can also lead to excessively measured job reallocation on transition economies.
In 'Gross Job Reallocation and Labour Market Policy' Pietro Garibaldi examines international data on job creation and destruction for a sample of ten OECD countries. The purpose of the study co-authored by Jozeph Konings and Christopher Pissarides (London School of Economics and CEPR) is to explore the relation between gross job reallocation, unemployment and its duration, and active and passive labour market policies. Data on job creation and destruction show large variation across countries. The relation between job reallocation and unemployment is found to be loose whereas the authors find a strong positive connection between long-term unemployment and job reallocation. Employment protection schemes and long durations of unemployment benefits are associated with lower rates of job creation and destruction. In contrast, the level of income support for unemployed exerts a weakly positive influence on job reallocation. Finally, acknowledging the data restrictions, the authors find only weak linkages between active labour market policies on job turnover. Jan van Ours (Tinbergen Institute and University of Amsterdam) criticised the emphasis on bivariate relationships and wondered whether one can learn much from them. Especially the endogeneity of policy may induce some of the observed correlations.