Sample Selection with Binary Endogenous Variable : A Bayesian Analysis of Participation to Timber Auctions

We propose a Bayesian Metropolis-Gibbs Monte Carlo Markov Chain (MCMC) algorithm to estimate parameters of a sample selection model in which the selected equation include a binary endogenous explanatory variable, using a three simultaneous equation model. We apply our methodology to participation in timber auctions for which some lots receive no bid (these lots are censored), one bid (no competition) and two or more bids. We find that the MCMC algorithm provides stable results across different model specifications, whereas the Heckman sample selection procedure results in unreliable inference on the coefficient associated with the binary endogenous variable as well as the correlation coefficient