Estimating Demand Systems when Outcomes Are Correlated Count

We develop a Bayesian posterior simulator for fitting a high dimensional system of ordinal or count outcome equations, illustrating its use by modeling the multiple site recreation demands of individual agents to a set of twenty-nine Iowa lakes. The model flexibly adjusts to match observed frequencies in trip outcomes, permits a flexible correlation pattern among the visited sites, and the posterior simulator for fitting this model is relatively easy to implement. We also describe how the model can be used to conduct counterfactual experiments, including predicting behavioral changes and describing welfare implications resulting from shifts in demographic and site characteristics.