Unemployment duration among immigrants and natives: unobserved heterogeneity in a multi-spell duration model

This paper studies whether the unemployment dynamics of immigrants differ from those of natives, paying special attention to the impact of accounting for unobserved heterogeneity among individuals. Using a large administrative data set for Spain, we estimate multiple-spell discrete duration models which disentangle unobserved heterogeneity from duration dependence. Specifically, we estimate random effects models assuming that the distribution of the effects is discrete with finite support, and fixed effects models in which the distribution of the unobserved effects is left unrestricted. Our results show the importance of accounting for unobserved heterogeneity and that mistaken policy implications can be derived due to improper treatment of unmeasured variables. We find that lack of control for unobserved heterogeneity leads to the conclusion that immigrant males have a higher probability of leaving unemployment than natives and that the negative effect of unemployment benefits for immigrants lasts longer than for natives. Nonetheless, the estimates which do control for unobserved heterogeneity show the opposite results.