Research

Abstract. Using a vignette-based survey experiment on Amazon’s Mechanical Turk, we measure how people’s assessments of the fairness of race-based hiring decisions vary with the characteristics of the discriminator, the race of the discriminatee, and the reasons why a discriminatory action was taken. We find that conservative respondents are much more accepting of discriminatory actions than others, and that only moderates’ and liberals fairness ratings vary with the discriminatee’s race. In contrast, respondents of all political leanings react in very similar ways to the reasons why a discriminatory action was taken: Statistical discrimination is fairer when more accurate information is used, and indulging one’s own racial animus is less fair than accommodating one’s customers’ animus. Overall, we argue that a two-group framework, in which one group (mostly self-described conservatives) values employers’ decision rights and the remaining respondents value utilitarian concerns, explains our main findings well. In this framework, both groups also value applying a consistent set of fairness rules in a race-blind manner. 

Abstract. The population of unmarried heterosexual cohabiting women has nearly tripled in the US over the past two decades. While previous studies have tended to ignore these women, or treat them as single/married, this paper examines the labor supply responses of cohabiting women, single women, and married women from 1996 to 2016 using March-CPS. A comparison of the three groups finds that cohabiting women have the lowest labor force participation elasticity with respect to after-tax wages. That cohabiting women would work more hours if their partners earned more annually and married women would not, points to another behavioral difference between the two groups. Results from ATUS-CPS 2003–2017 further imply that cohabiting women share some of the same characteristics of single and married women. We conclude that unmarried heterosexual cohabiting women should be classified as a separate female group.

Abstract. This paper studies gender disparities among engineers in the technology sector, using granular data from a Chinese tech company that tracks precise work hours. I first document a raw gender gap in hours worked, with female engineers working 10 fewer hours per month than their male counterparts. This gap narrows slightly to 9 hours after adjusting for demographic and job-related controls but doubles for married women. To analyze how hours worked influence salary progression—a proxy for career progression, as performance rewards and promotions typically lead to higher base salaries—I construct salary progression spells and apply discrete hazard models. Results reveal persistent gender gaps across various measures of salary increases (‘raises’), even after controlling for hours and other observables. In particular, women have a 12 percent lower likelihood of receiving a raise. While hours worked positively affect the likelihood, women’s return to hours is three times that of men’s. Decomposition results further indicate that gender differences in hours account for 15% of the adjusted gender gap in raise likelihood that cannot be explained by other observables. These findings suggest that hours worked serve as a signal of commitment or productivity.

Abstract. This paper studies the effects of time worked—daily hours and days worked—on workers’ bonuses and promotions in an environment where productivity data is not collected, no formal link exists between time worked and pay, and all absent days are approved uses of leave to which workers are entitled. Using digitally collected sign-in/sign-out data, we find that managers implicitly penalize leave-taking when awarding both bonuses and promotions: One additional absent day per month reduces the probability of receiving a bonus by 1.16 percent, the size of positive bonuses by 0.87 percent, and the probability of being promoted to a higher position by 0.79 percent. Mean hours on days worked do not affect promotions, but have a small positive effect on bonus rates among workers whose absences are above average. Finally, we find that workers’ absence rates are highest in the first half of each promotion cycle, suggesting a strategic response to the managers’ implicit reward policies.

(Previously as "Financial Concerns, Cognitive Abilities, and Economic Decisions")

Abstract. This paper investigates the psychological effects of alleviating financial concerns on economic decisions and labor productivity among the very poor in Nairobi, Kenya. We use a lab-in-the-field experiment, where we reduce immediate financial concerns, without altering wealth or consumption, by varying the timing of a future, unconditional, and high-stakes cash transfer. Participants were randomly assigned to receive an immediate or delayed transfer. We collect our measurements after they know their transfer timing but before they are paid. Our findings show that individuals with higher financial concerns violated rationality axioms less frequently when choosing food bundles and health insurance policies, but only after being allowed to revise their decisions. In labor tasks, the group with high financial concerns outperforms in tasks that require purely mechanical effort. However, there are no significant differences in performance in cognitively demanding tasks. We interpret these results with a model of labor productivity that incorporates effort and mental bandwidth as complementary inputs. These results suggest that financial concerns impact economic decisions and labor outcomes by shifting the allocation of cognitive resources and effort, with implications for policies to alleviate poverty.