Publications

Abstract: To investigate the heterogeneous impacts of test-optional policies, I construct a network of private colleges based on their self-reported peer institutions and detect communities within it using a network science algorithm. I then estimate treatment effects by comparing adopters and non-adopters within the same community. I show how restricting to within-community comparisons drastically attenuates the severe pre-trends that arise when all untreated units serve as controls, suggesting that, within community, a parallel counterfactual trends assumption is satisfied. Overall, adoption substantially increases reported SAT scores, making adopting schools appear more selective, but decreases the share of bachelor’s degree completions in STEM, suggesting that marginally enrolled students are much less likely to complete a STEM degree than those they have displaced. Effects vary by community; at the most selective liberal arts colleges, policy adoption causes an 8 (26) percent decline in the share of all (under-represented minority) bachelor’s completions with a STEM major.