A random-effects logit model of work-welfare transitions

Abstract

This paper models the transitions between states defined in terms of work and welfare status as a discrete-time competing-risk model with unobserved heterogeneity. A random-effects logit model is estimated with a panel drawn from Wisconsin administrative records. Unobservables are assumed to follow a discrete distribution. The most striking result is that welfare recipients were substantially less likely to start working while remaining on welfare afterthe 1981 changes in program rules. © 1990.