Publications

This paper investigates how having more White coworkers influences the subsequent retention and promotion of Black, Asian, and Hispanic women and men. Studying 9,037 new hires at a professional services firm, we first document large racial turnover and promotion gaps: even after controlling for observable characteristics, Black employees are 6.7 percentage points (32%) more likely to turn over within two years and 18.7 percentage points (26%) less likely to be promoted on time than their White counterparts. The largest turnover gap is between Black and White women, at 8.9 percentage points (51%). Drawing on conditional random assignment of new hires to initial project teams, we then show that a one standard deviation (14.0 percentage points) increase in the share of White coworkers is associated with a 10.6 percentage point increase in turnover for Black women. These effects are similar in magnitude to the overall turnover gap between White and Black women, and asymmetric: Black women are the only race-gender group whose turnover and promotion are negatively impacted by the racial composition of their coworkers. We explore potential pathways through which these peer effects may emerge: while the share of White coworkers does not affect formal task assignment, Black women who were initially assigned to Whiter teams subsequently report fewer billable hours and more training hours, and are more likely to be labeled as low performers in their first performance review. Our findings call for more research on how peer effects early in one’s career shape longer-term racial inequalities at work.

Many low-income households in the US miss out on social safety net benefits because of the information, compliance, and psychological costs associated with take-up of government assistance. Yet, the empirical evidence on the impact of learning and psychological costs on take-up, and how to reduce them, is mixed. Leaning on an administrative burden framework, this paper measures the role of reducing learning costs and stigma on demand for rental assistance in two field experiments (N = 117,073) conducted in two US cities. We find that providing information about emergency rental assistance increased program application requests by 52% compared to a no-communication control group. Moreover, subtle framing changes aimed at destigmatizing rental assistance increased engagement with the communication by 36% and increased application requests by about 18% relative to an information-only group, with potentially larger effects for renters of color. In two subsequent online experiments (N = 1,258), we document that the destigmatizing framing reduces internalized stigma, without affecting perceptions of the program itself.