In this paper we suggest using predictive accuracy for comparing and testing different binary choice models.
