Bayesian Analysis of Econometric Models for Count Data: A Survey

Abstract

The paper reviews recent advances in analyzing complex count data models using computer intensive Bayesian methods, such as Gibbs sampling or Markov Chain Monte Carlo. These methods provide a powerful estimation tool for models characterized by an underlying latent structure. Examples are count data with random effects, correlated count data, counterfactuals in treatment models, latent class models, endogenous switching models and models with underreporting, to name but a few. In all of these cases, the Bayesian approach can be implemented by a simple extension of the parameter space to include the latent effects, a case of data augmentation. As a by-product of the MCMC simulation, the posterior distribution of the latent effects is obtained, which may be very useful in some problems.