Bayesian Estimation and Smoothing of the Baseline Hazard in Discrete Time Duration Models

This paper proposes a Bayesian approach for estimating and smoothing the baseline hazard in a discrete time hazard model. The hazard model is specified as a multiperiod probit model and estimated using a Gibbs sampler with data augmentation. The baseline hazard specification is smoothed using the smoothness priors introduced by Shiller (1973). The methods proposed in this paper are then used to study the effect of Canadian Unemployment Insurance eligibility rules on employment durations from New Brunswick, Canada. © 2000 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology