Duration Analysis on Multiple Welfare Spells : A Repeated Hazard Model with Heterogeneity

This paper presents one econometric solution to the problems of estimating repeated hazard models with heterogeneity for on- and off-welfare spells, using data from The Washington State Family Income Study, a five-year longitudinal study. We developed a closed form probability that allows for unobserved heterogeneity in two related spells. Estimation results show that the hazard rates of both on-welfare and off-welfare duration increase at the beginning of the spells and then decrease over time, suggesting that efforts to help people get off and stay off welfare will be most effective early in the spells. We also found striking evidence of heterogeneity; that is, those who are likely to remain on welfare for lengthy periods are also likely to return quickly if they do leave welfare