Publications
Foreign Aid and the Performance of Bureaucrats(Job Market Paper)
Forthcoming, International Organization
AI Surveillance: A Threat to Bureaucratic Autonomy
Accepted, Journal of Democracy
Globalizing Fairness Attributes in Machine Learning: A Case Study on Health in Africa
(with Mercy Nyamewaa Asiedu, Awa Dieng, Iskandar Haykel, Negar Rostamzadeh, Stephen Pfohl, Abigail Oppong, Sanmi Koyejo, and Katherine Heller) .
EAAMO '24: Proceedings of the 4th ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization, Article No.: 10, Pages 1 - 24, October 2024
Under Review
“We sleep with the head inside, but the brain outside.”: Local Integration in the Shadow of Resettlement Hopes
with Manya Kagan and Noa Rubinstein
Revised and Resubmitted, Journal of Refugee Studies
How do refugee perceptions of resettlement policy influence local integration? In this paper, we argue that UNHCR's three durable solutions—integration, resettlement, and repatriation—are interdependent rather than separate pathways, as refugees' perceptions of one solution directly shape their engagement with others. We investigate our argument in Uganda, drawing on in-depth qualitative data from 20 key informant interviews and 5 focus group discussions with 34 participants conducted in Kampala and selected rural settlements. Our findings reveal that many refugees prioritize third-country resettlement, resulting in underinvestments in local integration pathways. This dynamic is fueled by a problematic information environment where unclear eligibility criteria, limited transparency, and refugee reliance on informal information networks lead to fraud and strategic vulnerability performances among refugees. Our research demonstrates how refugees proactively navigate policy spaces, indicating that effective support should enhance rather than diminish their agency.
The Effect of Government Intervention on the Operational Decisions of NGOs: Evidence from a Survey Experiment in Three Countries (Under review)
with Simon Hoellerbauer, Graeme Robertson, Jeremy Springman, and Erik Wibbels.
Repressive governments seek to influence the behavior of domestic NGOs with both carrots and sticks. How do these efforts shape NGO operations? We identify common actions to influence NGOs: repressive and accomodative interventions in NGO operations, positive and negative rhetoric, and cooptation. Using a survey experiment of 425 NGO directors in Cambodia, Uganda, and Serbia, we investigate how community-level variation in the prevalence of these actions shape NGO preferences over where, how, and with whom they work. As expected, government interventions shape where NGOs prefer to work, raising concerns about how NGO benefits are distributed. Additionally, cooptation isolates NGOs, making them less likely to involve the public in planning or partner with other NGOs. However, moderate repression increases NGO preferences for organizing public action, suggesting NGOs see public mobilization as an effective strategy to resist some forms of repression. Importantly, these results hold across NGOs operating in very different sectors and countries. Panel data from Cambodia documents this finding in self-reported real-world behavior. These findings shed light on how NGOs navigate democratic backsliding.
Merit or Oversight? Citizen Perceptions of Bureaucratic Management and Trust in the State (Under review)
with Diego Romero
How do citizens interpret public management practices, and how are those perceptions related to trust in government? We argue that citizens use recruitment and promotion practices as signals of competence, fairness, and accountability. Using a household survey and embedded conjoint experiment with 1,500 respondents across 30 Guatemalan municipalities, we find that citizens perceive municipal hiring as shaped by both merit and political ties. Consistent with our preregistered expectations, respondents who perceive recruitment as more merit-based evaluate government as more capable and less corrupt, report greater satisfaction with services, and express greater trust in the mayor. Contrary to our pre-registered expectations, respondents who perceive promotion as more insulated from mayoral discretion evaluate government less favorably on the same dimensions. These findings indicate that citizens value both professionalization and political control over bureaucratic careers. This study challenges purely technocratic approaches to state building, brings promotion, as distinct from selection, to the forefront of our understanding of bureaucratic management, and contributes to literature on representative bureaucracy.
Are Women Targeted More for Corruption? Experimental Evidence from Traffic Officers in Nigeria (Under review)
with Aliyu Ali Bawalle, Harunobu Saijo, Diego Romero, and Ghulam Dastgir Khan
Are women more exposed to extortion by street-level bureaucrats? Conventional resource-based models predict that officials exploit structurally vulnerable actors, which would make women frequent targets. Gendered norms, however, may instead restrain predatory behavior. We adjudicate these competing expectations with an original survey and an embedded randomized conjoint experiment among 1,078 traffic officers in Nigeria, eliciting the targeting decision directly rather than inferring it from citizen accounts. Officers evaluated pairs of driver profiles that varied randomly on gender and other attributes to estimate the causal effect of gender on perceived likelihood of being stopped and asked for a bribe. Contrary to our preregistered expectations, women are significantly less likely to be targeted. Mechanism tests and interviews from 80 officers and drivers indicate that protective gender norms, alongside officers' doubts about women's capacity to pay, drive this result. These findings run counter to vulnerability accounts of corruption and show that informal norms can constrain predatory behavior even where petty corruption is pervasive.
Working Papers
Managing Elite Defection: Coalition Diversity and Punishment Severity in Authoritarian Regimes
with Diego Romero
Why do authoritarian regimes vary in their severity of punishment toward elite defectors? We argue that the diversity of elite coalitions determines punishment severity. Minority-led regimes, facing heightened threat perceptions, employ harsh punishments to deter elite defection. In contrast, regimes with broad coalitions demonstrate greater flexibility and leniency, including reintegrating defectors. We test this theory with novel data from two electoral autocracies with divergent punishment strategies. We utilize the Paths to Power dataset to track ministerial appointments, supplemented by manually coded ethnicity data and automated web scraping with natural language processing of media sources to categorize defection outcomes. This research advances understanding of authoritarian stability by identifying the structure of the coalition as pivotal to elite punishment strategies and coalition management. I presented the pre-analysis stage of this work at the WhoGov Workshop at University of Oslo.
Traditional Gender Norms, Clientelism, and Corruption: Survey of African Entrepreneurs
with Diego Romero, Harunobu Saijo, and Ghulam Dastgir Khan
This two-part study investigates how traditional gender norms—captured by the Male Dominance Index (MDI)—shape African entrepreneurs’ involvement in (1) corruption and (2) clientelism. First, in the corruption component, we ask whether patriarchal norms exacerbate gender disparities in bribery-related behaviors and attitudes. Specifically, we consider whether women in more male-dominant societies (a) report higher social pressure to offer bribes, (b) show lower willingness to engage in corrupt deals, (c) suffer higher levels of predatory extortion, (d) perceive reduced efficacy from corrupt exchanges, and (e) encounter harsher punishment for corrupt acts. Second, in the clientelism component, we probe whether patriarchal settings discourage female entrepreneurs from participating in electoral exchanges, such as vote-buying and turnout-buying, or from leveraging personal political connections. Here, we test whether women are targeted less by these inducements, have weaker ties to local powerbrokers, and face stronger social sanctions if they engage in clientelistic behavior. To test our hypotheses, we conduct a large-scale survey of 2,000–4,000 entrepreneurs (data collection is ongoing) across more than 15 African countries, matching each respondent’s self-identified ethnic affiliation to MDI scores. This allows us to pinpoint how male-dominant norms at the group level intersect with individual gendered experiences. Our measurement strategy includes direct questions about corruption and clientelism, as well as conjoint experiments.
The Violence of the Durable Solutions: How Refugee Protection Regimes Reproduce Harm
with Manya Kagan
Refugee governance rests on three durable solutions: resettlement, local integration, and repatriation. Although they are presented as pathways out of danger, for many refugees, they deliver neither permanence nor safety. Drawing on 23 key informant interviews, 20 in-depth interviews with refugees, and five focus group discussions with 34 refugees in Kampala and two Ugandan settlements (a total of 77 participants), this paper argues that each solution generates its own form of violence after flight. Resettlement makes suffering institutionally valuable, rewarding displayed vulnerability, while its opacity breeds mistrust and psychological harm. Integration exposes refugees to organized attacks, economic exploitation, and stereotyping, all softened by a hospitality narrative that demands gratitude and mutes complaint. Repatriation looms as the threat of return to unresolved conflict. Refugees respond with strategies shaped by these harms. They perform vulnerability for resettlement officers while attempting to conceal refugee status from neighbors, withdraw from local investment, and orient their futures toward departures that are highly improbable. This study shows that even a celebrated model host reproduces violence through the very architecture meant to protect refugees.
Ongoing Projects
AI-Assisted Detection of Tax Evasion in Uganda
with Carlos Schmidt-Padilla and Diego Romero
Across the developing world, cuts to foreign aid and mounting debt burdens are forcing governments to raise more revenue domestically. Better detection of tax fraud is one of the most direct ways to do this, yet the states under the most fiscal pressure are often the ones with the least capacity to audit effectively. Machine learning offers a novel tool for exactly these settings, giving revenue authorities a way to detect evasion at scale. This project takes Uganda as its case. Working in partnership with the Ministry of Finance, Planning and Economic Development and the Uganda Revenue Authority, we link disparate tax records to build profiles of taxpayers. We will conduct a randomized controlled trial to test whether algorithmic risk scores improve auditors' detection compared with current practice. The project is designed to not only develop theoretical insights into how technology boosts state capacity, but also help governments meet their own revenue targets.
Internalized Colonial Hegemony: Conceptualization and Measurement in Postcolonial Societies
with Rikio Inouye and Gavin Medina
Colonialism was not merely an economic or political phenomenon but also a profound psychological and cultural imposition, embedding lasting hierarchies within formerly colonized societies. This project develops the concept of "internalized colonial hegemony," a term highlighting how the colonizer’s worldview continues to dominate political, cultural, and economic spheres in postcolonial contexts. We develop an original survey to measure this novel concept and examine its correlation with domestic and foreign policy support. Early results from a pre-test in South Africa show that internalized colonial hegemony is driven primarily by perceptions of cultural hegemony of the former colonizer. Higher scores are associated with preference for Britain and the wider West as diplomatic partners, policy advisors, and mediators, greater value placed on British recognition, more favorable ratings of British people than of fellow citizens, and weaker support for reparations, removing colonial symbols, and autonomy from Britain. Previous research has largely been limited to single case studies; however, our comparative approach will enable us to develop a more comprehensive framework for colonial psychological legacies that can be applied cross-nationally..
Donor Identity and Citizen Evaluation of Aid Projects
with Nicholas Bau
Chinese-financed development projects are often evaluated more negatively than Western ones, even where Chinese finance delivers stronger development outcomes. Existing work does not disentangle whether this “reputation deficit” reflects attitudes toward China as a funder or responses to racial identity. To determine what drives this reputation deficit, we rely on a forced conjoint experiment varying attributes of projects and their managers. We will field the survey experiment in three countries that rely heavily on Chinese financing, with a pre-test planned for September 2026. This research will contribute to growing literature showing the importance of race in shaping international relations.
Changing Minds About Harassment: Messenger and Message Effects on Victim Blaming in Bangladesh
with Harunobu Saijo, Ghulam Dastgir Khan, and Md Abdul Bari
Can messaging reduce victim blaming in settings where conservative gender norms make attitude change difficult? Persuasion research predicts that locally embedded authority figures are more likely to effectively deliver messaging against harassment. Appeals rooted in conservative traditions however, may activate the very norms that tie a woman's conduct to the blame she receives. We will adjudicate these competing expectations with a randomized messaging experiment and conjoint field experiment among 4,000 respondents in Bangladesh, a Muslim-majority country where public harassment of women is pervasive. Respondents will view an anti-harassment message that randomly varies content and source of delivery. Respondents will then evaluate profiles of women with randomized attributes.. The design will not only identify which victim attributes drive blame, but also whether the message or the messenger does more to reduce it. Data collection will begin in Fall 2026.
Who Applies for Government Jobs? How Civil Service Features Shape Self-Selection in Timor Leste
with Harunobu Saijo, Diego Romero, Fernando Martin, and Ghulam Dastgir Khan
Bureaucratic selection is shaped not only by who is hired, but also by who applies.. Existing studies suggest a common ordering in which politicized systems repel desirable applicants while professionalized ones attract them. However, recent studies show that high performing public servants are integral to patronage systems. We argue that various applicant types are drawn to different features of the public service. We will test our argument with a lab in the field experiment among university students in Timor-Leste, a country undergoing civil service reforms. Students will complete behavioral tasks then rank hypothetical civil service jobs against an outside option.. Data collection will begin in Fall 2026.
Resting Projects
Preferences for Corruption and Public Employment
with Diego Romero, Isabaeva Zharkyn, Harunobu Saijo, Ghulam Dastgir Khan, and Goto Daisaku
Interest in the public administrations of low-income countries has grown in recent years. While numerous studies have focused on how clientelism and opportunities for corruption and rent-seeking shape demand for public sector employees, fewer studies have examined how bureaucrats factor corruption into their decisions when choosing jobs. We study how bureaucrats make decisions about jobs and effort depending on the availability of corruption opportunities. Since corruption provides additional income to otherwise low-paid employees, we expect bureaucrats to prefer jobs with corruption opportunities. However, we expect this effect to be pronounced for bureaucrats with higher levels of education and muted for bureaucrats with higher prosocial motivation. To test our argument, we rely on a survey of approximately 1,000 bureaucrats in Kyrgyzstan, which included a conjoint experiment designed to elicit preferences over job offers and their willingness to apply effort given the presence (or absence) of opportunities for corruption.