

https://www.cambridge.org/core/journals/ps-political-science-and-politics/article/exploring-bias-in-student-evaluations-gender-race-and-ethnicity/91670F6003965C5646680D314CF02FA4
There are A LOT of studies that document biases in teaching evals, esp with respect to gender. There is far less research on POC in part because the small number of faculty of color. (It's hard to study small populations.) You can read prev lit here: http://www.rebeccakreitzer.com/bias/
Some of the strongest evidence of bias comes from experiments where online students receive identical content but instructor gender varies. Prev studies with this research design are unequivocal in finding gender-based bias, but there hasn't been a similar study that varies race.
This study is great. It uses 14 (!!) sections of Intro to American & Intro to TX Gov. An instructor that varied in terms of race and gender gave a welcome video - but a single course admin ran the courses day-to-day.
In short - the research design maintained nearly identical courses with the exception of the visible instructor's identity.
At the end of the course, students were asked both open ended questions and a series of quantitative indicators. Even controlling for students' final grades, the race and gender of the instructor were significant predictors of course evaluations. Open-ended comments also varied.
TLDR: This is a well designed experimental study. It finds that women and people of color receive lower teaching evaluations. This matters because teaching evaluations are routinely used in CRITICAL decisions regarding hiring, promotion and tenure.