Do Optional Information Policies Increase Equity? Evidence From Two Large-Scale Pass-Fail Grading Experiments

Abstract
Many universities created temporary optional information policies – e.g., by allowing students to replace grades with ``credits" and by making test scores optional – during the early part of the COVID pandemic. As universities consider whether to make some of these policies permanent, it is essential to understand their equity implications. To study the equity implications of information-optional policies, we leverage data from two large-scale natural experiments at two highly selective universities that introduced optional "credit" (pass/fail) grading that students could choose after already learning their grades. We find that female students are significantly more likely to reveal negative performance information (i.e., grades below their GPA) when given access to the optional grading policy, which reduces their GPA relative to men. We also document similar gaps in take-up for traditionally disadvantaged groups including underrepresented minorities as well as first generation and low-income students. That is, optional information policies can create unexpected disparities because some groups are more transparent than others about their performance. We argue that these patterns are consistent with women (and disadvantaged groups) anticipating discrimination in the absence of performance information and choosing to reveal more in an effort to mitigate discrimination. We conduct a student survey for this purpose, and find evidence consistent with this channel: students anticipate that female students, particularly in STEM, Business, and Economics, will face labor market discrimination which makes them less likely to mask. The survey-based measures of anticipated discrimination are systematically correlated with females’ decision to reveal grades in STEM fields. Finally, using a field experiment involving real employers, we show that employers in STEM indeed discriminate against women in the absence of information, and that performance information reduces this discrimination. Joint work with Christine Exley, Raymond Fisman Judd Kessler, Louis-Pierre Lepage, Xiaomeng Li, Corinne Low, Xiaoyue Shan, and Mattie Toma.