Factors Affecting the Student Evaluation of Teaching Scores: Evidence from Panel Data Estimation

We use a large panel data, covering 6 semesters, 496 undergraduate courses related to 101 instructors and 89 disciplines. This allows treating adequately unobserved heterogeneity. We use a random-effects model estimated with feasible generalized least squares to find the factors that affect the student evaluation of teaching (SET) scores, including time-invariant instructors’ characteristics. Our empirical findings are: (i) controlling for the instructor’s status as full-time or part-time professor, the quality of his research affects positively the SET score; (ii) participating in training programs, designed to improve the quality of teaching, did not increase the SET scores; (iii) instructors can ‘buy’ a better evaluation by inflating students’ grade; (ii) the class size affects negatively the SET score; (iv) instructors with more experience are better evaluated, but these gains reduce over time. Finally, there are no significant changes in the rankings overall when we adjust the SET score to eliminate either the possible manipulation by the instructor or the effects of variables beyond his control. Despite some dramatic changes in some instructors’ positions, they are not statistically significant.