There is a large literature evaluating job-training programmes. In this paper, we evaluate three such job-training programmes that are used in Slovakia. Individuals participating in a job-training programme during their unemployment spell are usually not a random subsample of the population. To determine the treatment effect of a job-training programme on unemployment duration, one has to correct for this selection effect. Therefore, we use a multivariate mixed proportional hazard-type model to describe the hazard of getting a job simultaneously with the hazard of entering a programme. We allow for a heterogeneous treatment effect and for the treatment effect to vary over time. Furthermore, we add the monthly unemployment rate as a time-varying explanatory variable. The estimation results show that two of the three programmes shorten the unemployment duration quite substantially.

duration analysis, program evaluation, selection bias, time-varying variables
dx.doi.org/10.1111/j.1467-9574.2006.00311.x, hdl.handle.net/1765/19275
Econometric Institute Reprint Series
Statistica Neerlandica
Erasmus School of Economics

van Dijk, A. (2006). Treatment effect of job-training programmes on unemployment duration in Slovakia. Statistica Neerlandica, 60(1), 57–72. doi:10.1111/j.1467-9574.2006.00311.x