We develop new methods for conducting a finite sample, likelihood-base
d analysis of the multinomial probit model. Using a variant of the Gib
bs sampler, an algorithm is developed to draw from the exact posterior
of the multinomial probit model with correlated errors. This approach
avoids direct evaluation of the likelihood and, thus, avoids the prob
lems associated with calculating choice probabilities which affect bot
h the standard likelihood and method of simulated moments approaches.
Both simulated and actual consumer panel data are used to fit six-dime
nsional choice models. We also develop methods for analyzing random co
efficient and multiperiod probit models.