Monday, 21 May, 2012

16:30 | Applied Micro Research Seminar

Dr. Martin Burda: “A Bayesian Semiparametric Duration and Competing Risk Model with Unobserved Heterogeneity”

Dr. Martin Burda

University of Toronto, Canada

Authors: Martin Burda, Matthew Harding, and Jerry Hausman

Abstract: This paper introduces a new estimation method for the censored mixed proportional hazard model and its generalization — a competing risk model — allowing for a nonparametric baseline hazard, time-varying regressors, and unobserved individual heterogeneity with a flexible semiparametric form. Specifically, unobserved heterogeneity is distributed as an infinite mixture with a Dirichlet process prior. The model nests the specification of Han and Hausman (1990) as a special parametric case. We derive a tractable model likelihood for mixing kernels in the class of Generalized Inverse Gaussian distributions, using a series representation of scaled Laplace transforms and kernel moments. We apply our approach to analyzing the determinants of unemployment duration among unemployment insurance recipients using data from the U.S. Department of Labor, based on nationally representative individual-level surveys.


Full Text IS AVAILABLE in the CERGE-EI Library and will be available (hard copies) in the seminar room