Reducing and Widening Disparities with Blind Evaluations: Evidence from a Field Experiment

Abstract

When true quality is subjective or costly to discern, individuals may use group identities to discriminate. Blind evaluations are often proposed as a potential remedy; while non-blind scores are contaminated by reviewer biases, blind review limits the use of group identity to assess quality. I assess the efficacy of blinding evaluations using a natural field experiment at an academic conference. I randomly assign each submitted paper to both blind and non-blind reviewers. This design allows me to identify exactly which authors would be crowded out of the conference due to evaluator bias. I find that when author identities are known, reviewers score papers by male and prominent authors more favorably than papers by female and less-prominent authors, especially students. Blinding lowers scores for male and prominent authors, revealing reviewer favoritism towards these groups. While blinding particularly benefits students, effects by principal investigator gender are heterogeneous, suggesting that females face different levels of bias along the quality distribution. Blind scores perform at least as well as non-blind scores in predicting a paper's citations and publication status two years later.

Keywords: Discrimination, Academia, Blinding, Field Experiment

JEL Classification: C93, I24, J71