Works in Progress and Working Papers
Compensation Structure and Firm Wage Premia
Some firms pay more than others for identical occupations or individuals. These ``firm pay premia'' have been offered as an explanation for key structural features of the labor market including persistent unemployment, wage rigidity, and rising wage inequality. However, the contractual channels through which firm-specific pay premia are constructed have received little empirical attention. Using a unique dataset of compensation records at over 1,600 US firms, I estimate that increases in performance pay relative to competitors account for approximately 9% of total firm pay premia, with the remainder attributable to base pay. To understand why firms treat these two components as complements, I embed an optimal contract model into a matching framework in which workers of heterogeneous ability are paired with firms that differ in productivity (e.g., due to differences in technology adoption, capital investment, or market power). The model illustrates that the sorting of high-ability workers into high-paying firms can explain why such firms simultaneously raise both base and performance pay relative to their competitors. The complementarity of base and performance pay premia have implications for the structure of the labor market as it relates to efficiency wages, earnings inequality, and technological change.
Bargaining on Behalf of Others: Incentives, Beliefs, and Gender Gaps
With Jeanna Kenney
Published in the Journal of Economic Behavior & Organization (2025)
Oftentimes people delegate negotiation to others (i.e., “agents”), whether formally or informally. This paper explores the impact of agents on gender differences in negotiation and how this varies with common incentive structures. Using a bargaining experiment with over 2,400 subjects, we find that, absent agents, males make more aggressive demands than females. Introducing agents who negotiate on behalf of the players entirely closes this gap. Although agent incentives affect overall aggressiveness, they do not induce gender gaps. Belief elicitations suggest that this is because agents underestimate reservation prices for both males and females and incorrectly believe that they have the same threshold for rewarding aggressive behavior. While males and females have similar expected outcomes, agents close a risk exposure gap by making proposals across genders that are equally likely to be accepted.
Elite Capture of Clean Water in Bangladesh
With Ahmed Mushfiq Mobarak and Alexander Van Geen
(Draft Available on Request)
Fifty-seven million residents of Bangladesh consume well water with arsenic concentrations exceeding WHO safety standards - the largest documented poisoning of a population in history. This widespread contamination has been shown to harm long-term health, cognition, and earnings. In response, over 200,000 low-arsenic deep wells have been installed to address this public-health challenge. However, the spatial distribution of these deep wells does not maximize their accessibility to populations currently drinking contaminated water. We investigate the extent to which elite capture (i.e. preferential distribution of public goods to elites) explains the existing inefficiency in deep-well placement. Using a triple differences strategy, we show that deep wells are built closer to local politicians' households when their political party is in power nationally. We then design a model where a social planner who may treat elites preferentially decides to place deep wells. Using the model to estimate counter-factuals, we find that elite capture accounts for about a fifth (18%) of current inefficiency in deep well placement.
Resume Screening and Labor-Market Signaling Models
With Judd Kessler, Corinne Low, and Xiaoyue Shan
(Draft Available on Request)
Job candidates typically choose how much information they reveal to employers on resumes. Under plausible assumptions, traditional signaling models predict that employers will infer that candidates who omit information in a given category (e.g., GPA) are of the lowest type for that category. We collect 2,840 employer ratings of resumes from 71 employers in a two-year field experiment and find that employers do not always treat omitted information as proof that a candidate is a lowest-type. This behavior by employers allows low-types to meaningfully benefit by omitting information which runs contrary to the prediction of standard signaling models. For example, we estimate that an applicant with a B-Average GPA benefits more from omitting their GPA than they would from reporting a prestigious internship at a company like Google or Morgan Stanley. Overall, our results suggest that existing signaling models may not adequately describe settings where senders can strategically choose to omit information.
Regressive Electricity Subsidies
With Eric Hsu, Ahmed Mushfiq Mobarak, and Abu Parves Shonchoy
Using administrative data on outages in Bangladesh we investigate how rationing may contribute to the regressive nature of subsidies.
(Please reach out if you would like to talk more about this project!)