Race, gender and beauty in instructional ratings: a quantile regression approach

The objective of this paper is to apply quantile regression (QR) to analyze the effect of ascriptive characteristics such as beauty, gender and race on teaching evaluations. QR offers significant methodological advantages for studying issues that involve unequal outcomes at various points of a skewed distribution. While QR has been applied to other forms of discrimination (age, caste, obesity, etc.), it has not been specifically applied to discrimination in teaching evaluations. Using an unusual dataset originally compiled by Hamermesh and Parker (H–P 2005) on beauty, I re-analyze the dataset using QR. While the original paper by H–P focused only on the effect of beauty on teaching evaluations, I find evidence of other biases. The striking result here is not the impact of beauty on evaluations. Beauty, by itself, does not exert much of an effect. Gender, however, has a more pronounced effect and when combined with race tends to result in significant negative effects at various points of the quantile distribution. When contrasting these findings of bias with more recent ones, not much seems to have changed over the last two decades.