Reemployment Probability and Multiple Unemployment Spells: A Partial-Likelihood Approach.

An extended version of the mixed-proportional-hazards model is used to study the determinants of the conditional probability of reemployment of long-term unemployed young persons sampled in the first stage of the Australian National Longitudinal Survey. The survey data facilitate an econometric comparison between the effect of different factors on the conditional reemployment probability of those in their first unemployment spell with those who have experienced multiple spells. The article uses Cox's partial-likelihood approach. The empirical results support the hypothesis of lagged-duration dependence in that the length of previous job experience or the duration of previous unemployment is found to be an important determinant of reemployment probability. The results strongly suggest that fitting a common duration model to data from different spells involves a major misspecification.