This paper develops an econometric framework to estimate student-professor match effects, and uses the framework to evaluate how students' choice over instructors contributes to learning. I extend the literature on teacher value-added by showing how to use sequences of subject-related courses to non-parametrically identify instructor-specific learning production functions when instructors differ in both their grading policies and teaching abilities. The framework accommodates endogenous course selection, course dropout, and discrete scoring, all ubiquitous in higher education.

Using post-secondary academic records, I estimate the model and document the existence of substantial student-professor match effects. However, when allowed to choose, students do not always select the instructor from whom they will learn the most; they place as much weight on expected grades. Relative to the current assignment rule, assigning students to the predicted learning-optimal instructor leads to increases in academic achievement and reductions in the course dropout rate.

We obtain a modified Ramsey formula, which clearly shows that the mere existence of evaders dampens the use of the price as a means to finance the firm’s deficit. The regulated price is always below the monopoly price and, under sufficient conditions, also below marginal cost. Then, we generalize the model to incorporate moral hazard.

Finally, we undertake an empirical application of our results, which shows quantitatively that the downward tendency of regulated prices in a context of high evasion is significant.