Addresses the problem of modeling students' choices among institutions of higher education. This study offers a methodological approach which obviates certain difficulties encountered in previous studies, where the primary tool of analysis has been conditional logit. A parametric model for P(j;i), the probability that student i chooses institution j is developed: the parameters of P(i;j), the distribution of student characteristics at institution j, are estimated via ordinary linear regression; Bayes' Theorem is then used to invert this. The regression models describe student ability, income, and distance from home as functions of the characteristics of chosen institutions. The approach is demonstrated with data that have been used in previous studies. The results show this model to have substantially greater predictive power than existing conditional logit models, while also being easier to use, more flexible, and less expensive to apply.