Awarded a Dissertation Research Grant by the Russell Sage Foundation.
I conduct an online hiring experiment with a geographically representative U.S. sample, where White employers decide whether to hire White, Black, and Hispanic workers. In the experiment, guided by a Bayesian framework featuring asymmetric updating and multiple worker groups, I manipulate the worker groups available for hire. I employ the experiment first to identify and compare biases in beliefs about productivity and hiring against Black and Hispanic workers. I then investigate whether hiring experiences with Black (Hispanic) workers have positive or negative spillovers on beliefs about productivity and the hiring of Hispanic (Black) workers, and whether the answer differs by minority. Results from the experiment show that employers are biased against both Black and Hispanic workers, with the bias being stickier for Hispanic workers. I find positive hiring spillovers from Hispanic to Black workers only: while employers are more likely to hire Black workers after hiring experiences with Hispanic workers, the converse is not true. I provide suggestive evidence that the latter finding is driven by a more pronounced bias against Hispanic workers.
We examine the impact of an intervention targeting all economics majors at a large public university, designed to provide encouragement during challenging periods in the semester. We randomly selected students who would receive six weekly emails on behalf of the department of economics during the latter half of the Fall 2021 semester. Drawing insights from studies in social psychology, we designed the emails with the aim of enhancing students' sense of belonging and academic support, while providing information about resources available to students. In addition, for a subsample of the treated students, each email also featured a peer role models, i.e., a former student who had recently graduated from the same university. We chose the role models to be primarily women and Hispanic. A third treatment group also received an encouragement message from the featured role model. We find evidence of a positive impact of the encouragement emails on women's academic performance, notably driven by under-represented minority (URM) women. Our analysis of mechanisms points at improvements in URM women's growth mindset and emotional well-being. These results underscore the critical role of a supportive learning environment in promoting academic success, particularly among students from traditionally marginalized backgrounds.
Awarded a Graduate Student Small Research Grant from the Race & Ethnic Studies Institute at Texas A&M University.
I employ an online hiring experiment to study the short- and long-term effects of different hiring policies that could be used to incentivize the hiring of Hispanic workers. In particular, I contrast a hiring quota that imposes 50 percent of hired workers are Hispanic, an “extra-hire” policy that allows employers to hire an extra worker (of any race or ethnicity) if they meet a 50 percent quota of Hispanic workers, and an extreme 100 percent quota that forces employers to only hire Hispanic workers. While the 50 percent quota and extra-hire policies have similar effects on the hiring of Hispanic workers after they are lifted, lasting improvements in beliefs are only observed with the extra-hire policy. On the contrary, the 100 percent quota backfired and increased biases in hiring against Hispanic workers.
Informal labor workers face multiple sources of uncertainty (e.g., client affluence, stocking patterns, competition for the use of space) that ultimately lead to income fluctuations and, in some cases, vulnerability. We designed an exploitation-exploration dilemma task and conducted a lab-in-the-field experiment with street vendors in Bogotá, Colombia, to investigate behavior under uncertainty and increasing scarcity: in each turn, participants must choose between selling to a number of clients that will be halved for the following turn (i.e., to exploit the resource), and moving to a different sales location where the number of clients will be reset to an uncertain new level (i.e., to explore the resource). Although the decision setting is quite simple to grasp, the dynamic and uncertainty components favor the gap between the observed and optimal behavior. In fact, the design is very helpful to test backward-looking rules of optimal behavior, standard in Economics, and forward-looking rules of optimal behavior grounded in the marginal value theorem, popular in Biology for modelling patch-foraging decisions.
Many interventions aimed at diminishing bias against a minority group hinge on improving individuals’ understanding of the skills of minorities (in gender or abstract group settings). Specifically, a growing number of studies test whether providing employers (or CV evaluators) with information about the average productivity of different groups can reduce bias towards certain groups. In this paper, we ask whether part of the observed bias against Black and Hispanic workers is due to the incorrect belief that, while majority and minority groups may be equally productive on average, top workers (the stars) are more likely to be found in the majority (a.k.a. white) group. If this is the case, can correcting these beliefs reduce bias and discrimination against Black and Hispanic workers? Our data so far show that when only white workers of average productivity are available for hire, but minority workers across the productivity distribution can be hired, discrimination against minority workers vanishes. This suggests that employers want to find and hire stars, and they tend to believe that stars are more likely to be found among white workers. When white stars are unavailable on the market, employers turn to minority workers. Motivated by these findings, we test the impact of an information intervention aimed at de-biasing beliefs about the distribution of stars across groups. Our analysis suggests that while the intervention is successful at decreasing bias in beliefs, it proves unsuccessful in decreasing bias in hiring. Data collection completed
We examine how different forms of communication affect work dynamics between men and women. We employ a laboratory experiment to investigate which method of communication – face to face, audio call, video call or instant messaging – is most effective at reducing possible bias against women in the contribution of ideas to team tasks. In the experiment, men and women are paired and form a team. They engage in a task consisting in providing answers to Family Feud style questions. Team members first provide individual answers; they then meet and communicate with their partner in order to agree on an answer to submit as the team answer. Incentives are such that the two team members are equally rewarded for submitting the correct team answer, but the person whose individual answer is submitted as the team answer receives a bonus. In different treatments, we manipulate whether communication happens in person, via audio call, via video call or via instant messaging. We investigate whether the method of communication impacts 1) the extent to which men’s or women’s individual answers are submitted as the team answer; 2) team performance in the task; and 3) the occurrence and possible resolution of conflict, as perceived by each team member.
We exploit the random assignment of residential housing at Rice University to investigate how college friendships affect labor market outcomes and postgraduate study decisions of minority students. We use a unique dataset generated by the Rice Preference Study, which consists in surveys of (the same) students at Rice University from 2016 to 2020, combined with institutional administrative data on demographic characteristics, academic performance, and housing arrangements (all incoming freshmen at Rice University are randomly assigned to one of eleven residential colleges). Approximately two-thirds of the 2016 entering class were randomly selected to participate in a survey that elicited a battery of incentivized preference measures. Subjects had approximately three annual surveys over the subsequent four years, with some of them including network elicitation: students had to identify up to 10 friends from Rice undergraduates. Our data also include an exit survey administered by Rice to all graduating students, with information about their employment status: whether they are continuing education, working, or still looking for a job.