Projects

Making credentials visible is a common, low-cost remedy for evaluation bias, resting on the premise that a verifiable signal moves every evaluator toward merit. I test this premise in a randomized field experiment with 970 first-year MBA students at a leading Indian business school, who were randomly assigned, within 20 class sections over two academic years, to see either their current instructor’s name or a standardized profile showing the instructor’s photograph, doctoral institution and professional record before completing a teaching evaluation. Disclosure raised evaluations by 0.14 standard deviations on average. The gain was not shared uniformly. Male students rewarded male professors’ credentials by 0.28 standard deviations, while female students’ evaluations of the same professors did not rise. Measures of information receipt show why this is not a failure of attention - female students identified the credential at least as often as male students, yet disclosure raised male students’ rating of male professors’ achievements by 0.27 points and did not raise female students’ rating, a gap of 0.39 points (p = 0.02). The divergence arises when a signal is valued, not when it is seen, and it is not explained by evaluators’ stated beliefs about gendered access to academic careers. Information interventions, widely used in development policy and typically judged by whether recipients know more, therefore cannot be evaluated by receipt alone; what matters is whether the information is credited.

Discretionary professional interactions, requests for mentorship, guidance, or support with no formal obligation to respond, shape advancement, yet whether they follow the equal-treatment norms enforced for formal decisions remains largely untested, since discrimination research is built almost entirely around settings where a decision-maker holds formal authority over a material outcome. We study this through a preregistered field experiment in which 7,281 Indian university faculty received outreach emails from real male student interns disclosing their own genuine caste identity, crossed with task complexity and financial incentives. Faculty engage similarly with advantaged and disadvantaged-caste students at the checkpoints an organizational audit would capture, clicking and completion. An exploratory attention measure, email reopens, tells a more tentative story - disadvantaged-caste emails are reopened somewhat fewer times, a gap concentrated under a financial honorarium, counter to the standard prediction that incentives narrow such disparities. A further, descriptive divergence separates behavior from belief among survey completers - STEM faculty click most often, yet report the least inclusive attitudes and lowest willingness to sustain engagement. Behavior, attention, and belief thus move independently under the same identity treatment, helping explain why the discrimination literature finds mixed results, and showing that response or completion rates, an audit’s most available metrics, need not reveal where equal treatment breaks down.

When individual contributions are difficult to observe, evaluators may rely on demographic cues in personnel decisions. This paper studies how male leaders, operating in a male-typed domain where they remain the modal decision-maker, respond to information and incentives when repeatedly evaluating team members whose perceived gender is randomly assigned through digital avatars in a preregistered incentivized experiment. Under performance ambiguity, predictors assigned female avatars are 6.9-8.5 percentage points (33-40.5%) more likely to be replaced than otherwise identical predictors assigned male avatars, with the disparity concentrated among mid-performing teams. When leaders receive partial individual-level performance information, the estimated female-avatar penalty is smaller and no longer statistically detectable, although female-perceived predictors still require 16.8-21% higher observable performance to achieve comparable retention outcomes. A similar pattern emerges when replacement decisions carry financial consequences for leaders, despite no change in the information environment. The results suggest that information and incentives can shape evaluative decisions under uncertainty and influence the extent to which perceived gender affects retention decisions.

Firms designing incentive programs face important questions about their effectiveness - When do financial rewards work, when does social information work, and does combining them do more than either alone to influence consumer behavior? We address these using a randomized field experiment at a U.S. university recreation center, assigning participants of a group exercise program to one of four conditions - financial incentives (FI), rank information (RI), both (combined), or neither (control). Our results show that neither FI, RI, nor their combination significantly increases workout frequency or variety over the four-week intervention. However, FI and the combined treatment each improve goal completion (i.e., meeting weekly workout targets), relative to control, reflecting a shift in the temporal distribution of effort rather than an increase in total activity. Only the combined treatment shows persistence of effects on goal completion in the two weeks after the incentive period ends. Our mechanism analyses show stronger effects among participants who engage early on with the incentive program and explain the added value of rank information when paired with financial incentives for high-ranking participants and for openers of rank emails. The treatment effects are heterogeneous; the combined treatment works best for participants with salient intrinsic goals such as weight loss, whereas FI alone is effective for those without such goals. Both work better for consumers with past experience in the workout program. Our findings suggest that the effectiveness of incentive programs depends on consumer motivation and program engagement with implications for how firms should target and structure wellness offerings.

Gender disparities in tenure and promotion persist across STEM fields, including economics, partly due to statistical discrimination when review committees cannot clearly assess individual contributions to collaborative work. In Study 1, using publication data from 104 RePEc-ranked economics journals (October 2020–March 2021), we find that 74% of female economists (vs. 19% of male) coauthor in male-dominated teams, settings where attribution ambiguity may exacerbate evaluative bias. In Study 2, a randomized survey experiment (N = 233) with female economists shows that exposure to empirical evidence on attribution bias significantly shifts coauthorship preferences toward female-dominated teams, especially among early-career researchers. A follow-up survey (N = 54) indicates that the intervention provided new information, retained for months, that influenced how participants think about coauthorship strategically. These findings suggest that integrating attribution-bias awareness into mentoring and professional development, particularly for early-career scholars, can promote strategic coauthorship choices, enhance individual visibility, and help address the leaky pipeline reinforced by evaluative bias, much like the advice for Ph.D. candidates to produce solo-authored job market papers to signal individual ability.

While it is well established that a prison incarceration often causes a life-long stigma, there is only a limited understanding of how the prison inmates anticipate such mistreatment. In Study 1, we investigate the beliefs of Czech inmates (N=297) about trust and altruism by the general public to them while varying if their prison identity is revealed, using an incentivized lab-in-the-field experiment. We find strong evidence of optimism bias in anticipated altruism. Furthermore, we examine the causal impact of a self-affirmation intervention but find no significant effects on any of the outcomes. In Study 2 (N=489), we conceptually replicate the optimism bias in altruism and identify the likely underlying mechanism to be the perception that the public sympathizes with their experiences of prison hardships and the ambiguity about the crime they committed. The overoptimism dissipates in a more dynamic setting when directly questioned. Our findings underscore the importance of addressing optimism bias in stigma expectations to improve the effectiveness of reintegration programs for inmates.