Association for Public Policy Analysis & Management

Over the past 40 years, institutions of higher education in the U.S. have experienced a dramatic shift in the racial composition of students enrolled in both undergraduate and graduate programs. In this paper, we identify the extent to which the academic outcomes of undergraduates are affected by the race of their graduate student teaching assistants (TAs). Our primary analyses utilize over 600,000 student-class observations from a public university in California, coupled with TA assignment data from the university's Department of Economics.

We consider several empirical strategies to causally identify the effects of TA-student racial interactions and to overcome concerns of potential selection bias. Our primary analyses focus on models with class fixed effects, where we estimate differences in outcome variables between students across different races when assigned to the same TAs within the same class. Furthermore, we find no evidence of endogenous sorting into classes by student race when predicting the race of the class' TAs with a full set of controls. The lack of evidence of endogenous sorting is unsurprising for several reasons. First, the primary course registration period for undergraduates occurs before TA assignments are generated by the department. Secondly, once generated, the department only privately reveals the TA assignments to the corresponding professors and TAs.

Our results show that students perform better in classes taken with TAs who are of a similar race. We predict a 7.7% of a standard deviation increase in course grade for students who are assigned to TAs of similar race, relative to being assigned to TAs of dissimilar race. Race interactions have no impact on course withdrawal rates or likelihood of enrolling in subsequent related courses. Results also show that students are more likely to attend their TAs' optional discussion sections and office hours when the TA is of a similar race, providing direct evidence of students responding to similarly-raced TAs. We also see that racial interaction effects are especially prominent in classes where TAs had been given a copy of the exam prior to the exam date. We interpret this result as evidence of “teaching to the exam", where TAs, perhaps unconsciously, divulge information that is pertinent to the exams.

Lastly, racial interaction effects are strongest in classes which had no multiple choice questions on the exams. For these classes, we predict a 0.217 standard deviation increase in course grade for students who are assigned to TAs of similar race. This result could stem from several possible explanations. First, critical thinking is typically a key component to success on essay-based questions, and critical thinking skills may be fostered in settings where students discuss and ask questions about the course material, such as in TA discussion sections and office hours. Another explanation suggests that TAs are responding to students of similar race through grading. Classes with no multiple choice exams are classes where TAs exercise subjective judgment when grading, and students of specific races may be more likely to answer non-multiple choice questions in a manner which TAs of similar race favor.