This paper examines empirically the determinants of student evaluation of teaching (SET). Empirical models were specified and estimated using the SET data collected in Hong Kong over six academic years. A key finding is that three different origins of students had a differentiated impact on teaching evaluation. In particular, students from mainland China appreciated and rated teaching favorably, and hence the more mainland talents in the class, the higher the class-average SET scores. However, local Hong Kong students valued teaching and learning effectiveness unfavorably. Exchange students from abroad also dropped the class-average SET scores, as well as class-average student performance. The results suggest that raw SET scores should be used with care if classes are unbalanced with a large group of atypical students who work less but blame instructors for everything.