Working Paper Series no. 249: Bayesian estimation of Cox models with non-nested random effects: an application to the ratification of ILO conventions by developing countries

Abstract

We use a multivariate hazard model to analyse the ratification behaviour of ILO conventions by developing countries. The model accounts for two random effects: one at the country level, the other at the convention level. After investigating identification, we use a semi-parametric Bayesian approach based on the partial likelihood. We find diverging results between Bayesian and frequentist estimates concerning the importance of the two unobserved heterogeneities.

Bernhard Boockmann, Dragana Djurdjevic, Guillaume Horny and François Laisney
August 2009

Classification JEL : C11, C14, C15, C41, D78, J80, O19.

Keywords : Gibbs sampling, partial likelihood, frailties, duration analysis.