“Selective Stayers: Understanding the Persistence of Christians After ISIS.” Revise and resubmit, Journal of Peace Research.

Why do some members of persecuted groups remain in their homeland while others flee? Drawing on an original 2017 survey of n = 1,129 Christians internally displaced by ISIS, conducted just weeks after the liberation of Christian towns, I propose a theory of selective stayers in which two layers of filtering — an initial economic selection at the height of violence, followed by an attitudinal selection among those who remain displaced — shape who ultimately persists. The rich exit early, leaving behind a population that was already resource-constrained by 2017. Within this group, staying correlates with a distinctive attitudinal profile: warmer views toward Muslim civilians, sharper demands for accountability from perpetrators, greater optimism about future safety and stability, and a pragmatic preference for political integration. The selective stayers framework helps explain minority persistence in the wake of genocide and clarifies how the composition of those who remain shapes prospects for coexistence, reconstruction, and the survival of pluralism in Iraq and beyond.

“Bad Days, Not Bad Groups: Parasocial Contact and Prejudice in the English Premier League” (with Ala Alrababah, William Marble, and Alexandra Siegel). June 2026.

Positive intergroup contact is thought to reduce prejudice, but the theory rests on a paradox: positive outgroup exemplars are often dismissed as exceptional while negative exemplars are considered representative. We ask whether this asymmetry imposes a burden of representation on minorities. Drawing on millions of social media posts, newspaper articles, and Fantasy Premier League transfer data, we find no evidence that minority players are punished more harshly for poor performances. Survey experiments among fans and non-fans go further: negative coverage of a minority player reduces perceived typicality, and among those with deep parasocial investment, this individuation is accompanied by increased empathy and tolerance. We attribute this to parasocial buffering. Because fandoms are voluntary and organized around shared passion, they encourage observers to individuate outgroup members and respond to failure with empathy. By treating typicality as an outcome rather than a precondition, we find that negative experiences need not reinforce intergroup stereotypes. Parasocial exposure to celebrities may be uniquely well-suited for reducing prejudice in naturalistic, scalable, and ethical ways.

“Drama and Doctrine: Lessons from an Edutainment Experiment in Egypt” (with Ahmed Ezzeldin Mohamed and Don Green). March 2026.

Can narrative entertainment dislodge harmful practices --- and if so, how can such changes be measured unobtrusively? We report results from a randomized edutainment intervention in Egypt, in which participants were assigned to view a state-backed drama targeting female genital mutilation (FGM) and early marriage, family planning, or a placebo storyline. Outcomes were measured through surveys and interviewer-blinded conversations six months later, analyzed at scale using large language models. The FGM drama significantly increased opposition to the practice, driven by corrections to health and religious misperceptions --- corrections the state cannot reliably engineer due opposition from local clerics and doctors. Family planning effects were weak and inconsistent. These findings replicate under more challenging field conditions, among a population with deeper baseline resistance to targeted practices, boosting external validity. Interview-based measures validate and expand the constructs captured by surveys alone. Interviews reveal the spontaneous salience of FGM as a topic without showing elevated signs of respondent re-traumatization, establishing an ethical pathway for integrating open-ended measurement into experimental evaluations of harmful practices.

“Equal Contact, Unequal Lessons: Motivated Updating, Power, and Race in the U.S. Military” (with Eddie Yang). March 2026.

The contact hypothesis posits that cooperative, equal-status contact reduces prejudice. We argue that whether and how contact generalizes depends on where one stands in the racial hierarchy — a dynamic we call motivated updating. Leveraging data from newly-digitized surveys, intermarriage records, and testimonials from servicemembers in WWII and the Vietnam War, we develop and test this framework. For Black soldiers, equal contact improves both immediate and generalized intergroup attitudes, while unequal contact sparks backlash. For White soldiers, equality improves immediate outcomes but triggers resistance to advancing Black rights more broadly. The asymmetry follows from what equality means for each group: for Black soldiers, a permeable hierarchy is aspirational — evidence that racial barriers can yield; for White soldiers, it is a threat to the status they stand to lose. These results challenge the assumption that equal-status contact uniformly reduces prejudice, and underscore the complexities of designing institutions to foster social cohesion among minorities and majorities alike.

“Random Acts of Kindness? How Gender Shapes Everyday Helping Behaviors.” (with Saad Gulzar). March 2024.

Everyday acts of cooperation between strangers from different social groups form the crux of social cohesion, a key driver of economic, polit- ical, and social development. This study examines how gender shapes these behaviors. Randomizing the gender, class, and ethnic identities of research assistants approaching over 25,000 pedestrians worldwide, we measure whether strangers provide directions, assist with dropped groceries, or lend their cell phone. Across all countries and experi- ments, gender consistently emerges as a stronger determinant of helping behavior than class or ethnicity: women are more likely to be helped — but less likely to help a stranger in need — compared to men. These gender gaps grow with the riskiness of the interaction, but can- not be explained by safety concerns alone. We also uncover nuanced patterns of in-group cooperation, with some groups, such as ethnic minorities, exhibiting solidarity, while others, like lower-income individ- uals, reinforcing existing hierarchies. Our results suggest that random acts of kindness are not random; they are gendered, first and foremost.

“Inequality without Solidarity: Can Correcting Misperceptions Erode Sectarianism in Lebanon?” (with Lydia Assouad, Augustin Bergeron, and Giulia Buccione). January 2026.

In divided societies, high income inequality rarely produces support for redistributive and cross-group policies. We test whether misperceptions about out-groups’ economic conditions help explain this lack of solidarity. In a representative survey of 3,300 adults in Greater Beirut, respondents were randomly assigned to a seven-minute video demonstrating that income inequality between Lebanon’s main religious groups are small, while overall inequality is high. The information produces a strong first stage: perceived inequality between sects falls by about 0.4 standard deviations and perceived similarity rises by about 0.2 standard deviations. These belief updates increase support for pro-poor redistribution, including higher top marginal tax rates and expanded social protection, and raise the salience of class identity and anger toward the rich. Neither costly civic nor political behaviors move in the short run, however. In a candidate-choice conjoint experiment, shared religion remains the dominant predictor of vote choice, and the treatment failed to encourage attendance at local anti-corruption workshops or applications to a civic internship program. Light-touch interventions correcting misperceptions can move redistributive preferences and identity salience, but are insufficient to erode deeper attachments to sectarian politics.

“Do refugee co-sponsorship programs improve economic integration? Causal evidence from the U.S.” (with Jeremy Ferwerda, Jens Hainmueller, Duncan Lawrence, and Jeremy Weinstein). January 2025. [Pending approval].

The global rise in forcibly displaced people has led many receiving states to adopt community-based sponsorship models, in which local volunteers partner with professional agencies to support refugee integration. While community-based sponsorship holds the potential to facilitate refugee integration, concerns remain about outsourcing responsibilities from professional caseworkers to volunteers. Leveraging U.S. administrative data, we compare outcomes between refugees assigned to cosponsorship, a U.S. form of community-based sponsorship, and those who were not, employing a selection-on-observables design using agency assignment criteria. Our analysis reveals that cosponsorship improves refugee employment, reduces out-migration from assigned locations, and increases educational enrollment for both parents and children. We also find, however, that cosponsored refugees are less likely to repay their initial travel loan, potentially harming their credit scores. Viewed together, the results suggest that community-based sponsorship programs can be an effective complement to government resettlement but would benefit from enhanced training and oversight to mitigate unintended consequences.

“Compliance without Enforcement? Experimental Evidence from a Co-Production Program in Lebanon” (with Kristen Kao and Trevor Incerti). December 2024.

How can we encourage citizens to take costly action for the environment? With waste management systems collapsing in Lebanon, individual action is all the more urgent. We leverage a field experiment in the Lebanese city of Bickfaya to answer this question. Partnering with the local municipality and a grassroots NGO, we evaluate a program that tracks citizens’ waste, inspects their waste bags, and sends them personalized feedback on how to improve their sorting. Two months after the intervention, we find that randomly inviting citizens to join the program improves their sorting quality by an average of 1.5 out of 5 stars (~160% relative to the control group mean). We also find effects on environmentally-conscious behaviors outside of waste sorting – four months after the intervention, treated households are double as likely to sign up for a raffle where prizes are explicitly “green” (4% vs. 8%). These effects disappear at the 12 month mark, however. A survey of the city’s residents explores mechanisms driving the initial effects — and why they did not last. The results suggest that monitoring, combined with knowledge on how to comply, can boost environmentally-conscious behaviors – even against the backdrop of economic and political crisis — but that financial incentives are likely needed to maintain these effects.

“Promoting Reconciliation in Conflict-Affected Communities in Mali” (with Chad Hazlett and Daniel Posner). In the field.

The ongoing conflict in Mali has deepened inter-group tensions in many communities. In partnership with USAID, we are designing, implementing, and evaluating the impact of an intervention designed to promote reconciliation in conflict-affected communities in the Mopti and Ségou regions of the country. The intervention and accompanying evaluation will allow us to test the relative and joint impacts of cash transfers and more traditional conflict regulating programming on inter-group tensions. 

“Can Vouching for Stigmatized Individuals Overcome the Trust Deficit in Post-ISIS Iraq? Evidence from a Survey Experiment in Mosul.” (with Vera Mironova).