Semiparametric Bayesian Time-Space Analysis of Unemployment Duration

In this paper, we analyze unemployment duration in Germany with
official data from the German Federal Employment Office for the
years 1980-1995. Conventional hazard rate models for leaving
unemployment cannot cope with simultaneous and flexible fitting of
duration dependence, nonlinear covariate effects, trend and
seasonal calendar time components and a large number of regional
effects. We apply a semiparametric hierarchical Bayesian modelling
approach that is suitable for time-space analysis of unemployment
duration by simultaneously including and estimating effects of
several time scales, regional variation and further covariates.
Inference is fully Bayesian and uses recent Markov chain Monte
Carlo techniques.