Working Papers

  • Development via Administrative Redistricting: Evidence from Brazil. (Job Market Paper) [PDF] (with Christiane Szerman)

    Recipient of the Susan Schmidt Bies Prize for Doctoral Student Research on Economics and Public Policy, 2018

    Coverage: BBC Brasil

    We exploit a large redistricting episode in Brazil to examine if, and how, administrative unit splits impact local development. Using a rich panel of administrative and spatial data, we first document that requests to split are more likely to be initiated by poor and rural districts. Employing a difference-in-differences strategy with areas whose requests to split were never approved serving as a control group, we find that splitting leads to an expansion of the public sector, some improvements in public service delivery and children's education attainment, but no impacts on the private sector. Meanwhile, outcomes are unaffected in parent municipalities. Results are consistent with adaptations of policy to local preferences. Our results inform the equity-efficiency trade-off embedded in decentralization reforms worldwide.


  • Selecting Top Bureaucrats: Admission Exams and Performance in Brazil. April 2020 [PDF] [SSRN] (with Laura Schiavon and Thiago Scot) Submitted

    In the absence of strong incentive schemes, public service delivery depends crucially on bureaucrat selection. Despite being widely adopted by governments to screen candidates, it is unclear whether civil service examinations can predict performance on the job. This paper investigates this question focusing on a highly prestigious and influential set of bureaucrats in Brazil: state judges. We first explore rich data on judges’ monthly output and cross-court movement to separately identify what share of observed performance is explained by judges and courts. We estimate that judges account for at least 23% of the observed variation in the number of cases disposed. Using a novel data set on examinations, we then show that, within cohorts of candidates taking the same exam, those with higher grades perform better than their lower-ranked peers. Our results suggest that competitive examinations can be an effective way to screen candidates, even among highly qualified contenders.


  • Changing Racial Identity in the United States, 1880-1940. June 2020 [PDF] [NBER Working Paper w26465] (with Emily Nix and Nancy Qian) Submitted

    Coverage: The Weeds (40:45)

    This paper documents that a large number of African American men experienced a change in racial identity to white during 1880 to 1940, while analogous changes were negligible for other races. We provide descriptive evidence that is consistent with the conventional wisdom that “passing” for white was a response to severe discrimination, and came at great personal cost. The findings suggest that contrary to traditional economic thinking, racial identity is neither entirely exogenous nor fixed over the lifetime, and responds to incentives.


  • The Political Effects of Policy: Evidence from the Brazilian Amazon. New version coming soon! (with Arthur Bragança)

    Public policy often generates effects beyond its initial scope by changing the behavior of politicians and voters in response to it. This may be beneficial in a context of decentralized public good provision where individual and social incentives misalign. In this paper we study how a centralized environmental policy, which synced real-time satellite deforestation data with enforcement on the ground, further reduced forest loss by causing pro-agriculture politicians to cut back on effort. We assemble a comprehensive data set covering six electoral terms and show that, before the reform, municipalities governed by pro-agriculture mayors earmarked more resources to agriculture and had higher deforestation rates. However, after the reform, these differences disappear. We provide suggestive evidence that the policy affected political entry by making running for office less attractive for pro-agriculture candidates. Our findings indicate that the decline of the political influence of interest groups representing agriculture was an important mechanism through which the increases in the enforcement of the environmental legislation reduced deforestation.


  • Value of a Statistical Life Under Large Mortality Risk Change: Theory and an Application to COVID-19. August 2020 [PDF] [SSRN] (with Diego S. Cardoso) Submitted

    Coverage: CNN Brasil, Instituto Mercado Popular

    Benefit analyses of mortality reduction policies typically use multiples of the value of a statistical life (VSL). This approach approximates risk premia for small changes in mortality, but inaccurately characterizes premia for large risk changes because it implies increasing marginal utility and a risk-loving attitude. We propose a method to calculate the benefits of large mortality reductions adjusting for risk aversion. We apply this method to calculate the benefits of social distancing and other mitigation strategies to combat COVID-19 in the US and 42 other countries. Our findings show that the typical approach underestimates the benefits of social distancing in the US by a factor of 4.1 and in other countries by a factor of 2.2 on average.


Work in Progress

  • The Impact of 3G Mobile Internet on Educational Outcomes in Brazil. (with Pedro Bessone and Lisa Ho)

    Does the availability of mobile broadband internet affect children's test scores? We compare Portuguese and math scores before and after the staggered entry of 3G into Brazil's 5,570 municipalities using an event study design. We find that there is no effect of mobile internet on Portuguese or math scores, and can reject effect sizes of 0.02 standard deviations for 5th grade students, and 0.01 standard deviations for 9th grade students. Taken together, our results indicate that simply offering high-speed mobile internet is not enough to improve educational outcomes.


  • Choosing Institutions Locally: Determinants of Legislative Size in Brazil. April 2015 (available upon request)

    How are institutions determined? This paper studies how legislators locally choose an important dimension of local electoral systems, namely legislative size. To achieve this end, I construct a novel data set comprised of seat proposals and individual legislators’ votes to increase or not legislative size during the pre-2012 election period for a sample of municipalities in Brazil. I then outline and estimate a structural discrete choice model of legislative vote, in which legislators play a strategic game and also decide whether to run for reelection or not. I find that legislators weigh on average approximately 34% reelection payoffs and 66% social welfare when choosing seats. With these results, I run some counterfactual analyses varying the population caps’ function that federal government chooses.


  • Genocide and the Demand for Formal Institutions: Evidence from the Legacy of the Khmer Rouge (with Dennis Egger and Joris Mueller)

    The Cambodian genocide under the Khmer Rouge regime is remembered as one of the worst human tragedies of the 20th century. This paper examines the social and institutional legacies of this violent episode. We document that spatially more intense violence committed by the Khmer Rouge, proxied by exogenous adverse rainfall shocks during 1975–1977, is associated with a higher fraction of land covered by government-backed land titles in Cambodia today. We provide micro-level empirical evidence that social capital may play a role in explaining this result: Communities that experienced more violence had more of their social capital destroyed, increasing the demand for formal titles. We conclude that social capital may serve as a substitute for formal institutions in a context of weak state capacity. We complement our analysis by discussing several alternative mechanisms.