Unemployment Duration and Competing Risks: Evidence from Sweden

Abstract

The determinants of unemployment duration are studied in the presence of multiple exits out of unemployment. The results indicate that aggregation may be a severe problem in empirical analysis of a labor market such as the one in Sweden. Estimating a joint model for all transitions out of unemployment leads to biased estimates, since different transitions are governed by different mechanisms. The results also indicate positive duration dependence in both transitions to regular employment and transitions out of the labor force, but no duration dependence in transitions to public labor market programs. I. Introduction During the last decade we have witnessed a rapidly expanding literature on the determinants of unemployment duration. The main motivation for this development is that a large share of the increasing unemployment rates in most Western economies since the mid-1970s may be attributed to increased unemployment duration; see e.g. the survey by Lofgren and Engstrom (1988). Most empirical studies formulate an estimating equation for the determinants of the outflow from unemployment in terms of the hazard function, which describes the conditional probability of leaving unemployment. These studies are typically based on an explicit or implicit assumption that the only relevant transition out of unemployment is reemployment. In this paper it is argued that such an assumption, which amounts to assuming that all transitions from unemployment are governed by the same mechanism, may lead to state aggregation bias. We have no a priori reason to believe that e.g. increased education has the same effect on transitions from unemployment to employment, as on transitions from