Research

Why do individuals support redistribution in some settings but not others, even under persistently high inequality? This paper distinguishes between two foundations of redistributive preferences—beliefs about inequality and aversion to it—and shows that each operates differently at the municipal and the national level in the context of Brazil. I combine a survey informational experiment with a structural choice model to first identify individuals’ beliefs about inequality and to quantify their aversion to inequality. I use a Difference-in-Differences approach to identify belief effects and a GMM estimator derived from the choice model to recover inequality aversion parameters as a function of individual-level and municipal-level covariates. Downward belief updating reduces support for national but not local redistribution. Respondents exhibit stronger aversion to local inequality than to national inequality. Together, the results show that beliefs drive national preferences, whereas aversion anchors redistribution locally. These results challenge a unitary view of redistributive preferences and show how assessments of the problem’s magnitude and preferences over where to address it interact across levels of government.

Why do voters who want high incomes taxed so often vote against the parties that promise to tax them? I argue this results from two features of unequal federations: citizens hold different income positions locally and nationally, and parties set taxation and allocation as one package. A voter's position in the national income distribution shapes what she wants from taxation; the gap between that position and her position at the local level shapes where she wants the money spent. Because candidates bundle these dimensions together, voters whose two demands point to different candidates find no match, and part of the demand for redistribution never reaches the redistributive party. I test the model's mechanisms in Brazil, combining records of formal employment contracts from the Annual Report of Social Information (RAIS), an original survey, and election results from the Superior Electoral Court (TSE). Preferences for progressive taxation predict support for Lula in the 2022 presidential election, while preferences for local delivery predict voting against him, including among voters who want high earners taxed. In the 2020 mayoral elections, spatial gaps are less favorable to the left where municipalities can deliver. The paper shows when taxing and spending sit at different levels of government, spatial inequality splits redistributive majorities and shapes how much of their demand elections can represent.

Do hate crimes carry economic and political consequences beyond their immediate victims?  We argue religion-based hate crimes act as identity-specific signals that depress housing demand among the targeted group.  We test this "targeted withdrawal" mechanism using the surge in anti-Jewish and anti-Muslim hate crimes in New York City following October 7, 2023.  Using an instrumental variables strategy, we find that predicted anti-Jewish hate crimes reduced average unit sale prices by 5.46 percent and total sale prices by 9.74 percent.  Deed-level evidence shows these declines reflect withdrawal by prospective Jewish buyers in already visible, established neighborhoods. Yet, established residents in high-share-Jewish neighborhoods show no political disengagement: turnout and communal investment rose after October 7; anti-Muslim effects are statistically inconclusive.  In sum, hate crimes can deter homebuyers and undermine the residential concentration underlying minority political representation, even as they sustain rather than suppress political voice among residents already there. 

Do hate crimes increase political mobilization?  And, if so, who do they mobilize? We address these questions using 12.3 million linked voter-election records and New York City Police Department hate-crime complaint data across four mayoral contests (the 2021 and 2025 primaries and general elections). We find a robust, precisely estimated association between hate-crime exposure and turnout (β̂ = +0.038, p < .001). We then ask a sharper question that this linked administrative data uniquely allows: does this mobilization concentrate among the group a given hate crime targets, or is it diffuse across the electorate?  For six inferred voter groups (Jewish, Muslim, White, Black, Asian, and Hispanic), we distinguish each group’s response to hate crimes targeting its own group from its response to hate crimes targeting other groups. In a joint structural model of turnout and candidate choice, we find no jointly significant evidence that voters respond more strongly to hate crimes targeting their own group specifically (p = 0.239). Asian and Hispanic point estimates are the largest and only marginally suggestive (p ≈ .07–.10). We argue this pattern is consistent with a diffuse-threat mechanism, in which the political salience of hate crime — plausibly transmitted through media coverage rather than direct victimization — mobilizes the electorate broadly rather than along group lines, and we outline a media-salience research design for testing that mechanism directly in future work.

Why do left governments in oil-dependent states maintain fuel subsidies while cutting social programs for the poor? We develop a model of fiscal allocation under political constraints in which a resource-dependent government distributes oil rents between progressive transfers to its electoral base and a regressive fuel subsidy captured by a non-base group with credible destabilization capacity. The model generates three results: (i) when revenues are high, the government rationally sustains both transfers; (ii) when an exogenous resource shock tightens the budget, it cuts progressive transfers first; (iii) when revenues fall further, it converts the universal subsidy into a conditional, clientelist instrument. We derive the welfare cost of this appeasement equilibrium: the gap between realized welfare and the counterfactual in which the government could commit to the first-best allocation. We show that it falls disproportionately on the poor. We calibrate the model to Venezuela, exploiting the 2014 global oil price collapse as an exogenous shock, and validate its predictions against household survey data. 

This study explores the role of entertainment media in shaping public perceptions of inequality and social mobility, focusing on Brazil, one of the most unequal countries in the world. Despite extensive research on traditional determinants of redistributive preferences such as income, education, and political ideology, the influence of media and cultural representations has been largely understudied. We argue that popular media, particularly telenovelas, subtly influences viewers’ attitudes toward inequality by normalizing wealth disparity and promoting "rags-to-riches" narratives, which frame upward mobility as attainable while downplaying structural barriers. By examining how entertainment media affects perceptions of economic disparities, this study seeks to uncover the mechanisms by which media consumption shapes attitudes toward redistribution, potentially reinforcing or challenging public support for redistributive policies. Using Avenida Brasil, a widely popular Brazilian telenovela, as a case study, we conduct a survey experiment in which participants are exposed to contrasting media portrayals of social mobility—one highlighting manipulative paths to wealth and the other focusing on merit-based success. Through this approach, we aim to isolate the causal effects of media exposure on viewers' perceptions of economic realities and their preferences for redistribution. In doing so, we extend current research by examining the impact of entertainment media on political outcomes, with a particular focus on Latin America, a region characterized by high levels of income inequality. This research contributes to a deeper understanding of how media representations can shape societal attitudes toward inequality and social mobility in contexts marked by significant wealth disparities.

We have collected data spanning 13 years from the Court of Justice of São Paulo, where we observe promotion outcomes, whether promotions were based on seniority or merit, the number of cases closed by judges, the number of hearings held, and the hierarchy of judges each quarter. Our findings indicate that the higher the ranking of judges in cumulative closed cases, the higher their probability of promotion tends to be. Moreover, the additional number of cases a judge closes decreases as their seniority ranking increases. These findings point out that the rate at which judges close cases depends on their anticipated probability of promotion. We will structurally estimate the case-closing decisions made by judges at various stages of their tenure to empirically test whether these patterns arise from judges inflating their perceived merit by prematurely closing cases (without acquiring the socially optimal level of information) or whether merit-based promotions effectively increase the productivity of public servants.