Sample selection with binary engonenoux variable: a Bayesian analysis of participation to timber auctions.

Summary We propose a Bayesian MCMC (Metropolis-Gibbs Monte Carlo Markov Chain) algorithm to estimate the parameters of a selection model in which the selection equation has an endogenous binary explanatory variable, using a model of three simultaneous equations. We apply our methodology to participation in timber auctions in which some lots are not bid (these lots are censored), others receive one bid (no competition), and others receive 2 or more bids. We find that the MCMC algorithm gives stable results for several model specifications, while the Heckman selection procedure does not give reliable results for the coefficient associated with the endogenous binary variable, nor for the correlation coefficient.