Does unemployment cause future unemployment ? The empirical debate goes from (unobservable) heterogeneity to the so-called duration dependence. We survey and illustrate the empirical literature and build a simple model allowing to distinguish the four main arguments previously advanced by the theoretical literature : the ex ante heterogeneity, the unemployment duration viewed as a signal, the discouraged unemployed, and the human capital losses during unemployment spells. We also investigate policy implications. Classification JEL : J21, J64

This paper considers a novel theoretical argument allowing for the presence of duration dependence in individual exit rates of unemployment. If unemployed are ex ante heterogenous and if their skills are imperfectly observable, firing of the least able is an endogenous phenomenon that unemployed have an incentive to hide. Then, the unemployment duration conveys a (negative) signal about the quality of the worker, which firms use to evaluate his employability. A decrease in firing costs can make the economy topple from an equilibrium with unemployment, but without state dependence, over an equilibrium with discrimination against the long-term unemployed. Ce travail propose un mécanisme nouveau permettant d’expliquer la présence de dépendance temporelle dans les risques individuels de sortie du chômage. Lorsque les chômeurs sont ex ante hétérogènes et que leur compétence est imparfaitement observable, le licenciement des moins doués est un phénomène endogène qu’ils s’efforcent de masquer. La durée de chômage véhicule alors une information sur la qualité du travailleur considéré, dont les firmes se servent pour jauger de son employabilité. Une baisse des coûts de licenciement peut faire basculer l’économie d’un équilibre avec chômage, mais sans exclusion, à une économie caractérisée par l’exclusion des chômeurs de longue durée.

Our aim is to explain why the pattern of relative unemployment rates by education groups was non monotonic in most of the OECD countries. In a two-sector matching model, a simple unexpected productivity shock biased against unskilled labor can replicate the observed dynamics. Demographic effects of skill-biased shocks can be related to inequality in the distribution of wealth.

We consider the search behaviour of an infinite lifetime worker whose expected matching gains are assumed to be decreasing over the spell of unemployment. The search effort is linked positively to future earnings and negatively to the worker's reservation wage. We show that this may give rise to non monotonous hazard rates.