Working Papers

Revise and Resubmit at AEJ: Economic Policy, with Gaurav Bagwe and Allison Stashko
Online Appendix

We study how distance to one’s polling place affects the likelihood of voting, either in-person or by mail. We use a border discontinuity design with data from over 15 million voters in Pennsylvania and Georgia. The average effect of distance to the polling place on turnout is small, partly because voters substitute for mail-in voting. A one-mile increase in distance to the polling place decreases the likelihood of voting in a general election by up to 0.99 percentage points. The effect is larger in areas with a heavy reliance on public transportation and in low-income areas. Using these estimates, we identify the turnout-maximizing locations of polling places and compute gains to turnout.

Conferences:  American Political Science Association Annual Meeting, Sep 2019


The paper provides the first assessment of: (i) America’s progress in lifting the lower bound—the floor—of the distribution of real income; (ii) whether the country’s largest antipoverty program, SNAP (“food stamps”), helped do so. An operational method of estimating the floor is implemented on micro survey data spanning 30 years, with various robustness and significance tests. SNAP partially compensated the poorest, and helped stabilize the floor. Nonetheless, the floor has been sinking over the last 30 years. The efficiency of SNAP in lifting the floor has declined over time. Full coverage of the poorest would lift the floor appreciable. 

Featured in: USDA ERS, Vox,, WSJ Blog: Economics, QuartzWorld Economic Forum

It is theoretically ambiguous whether people in richer countries have a higher floor to their living standards. Nor is it clear whether social protection spending reaches the poorest and thus lifts the floor. Across countries, the paper finds that higher mean incomes come with a higher floor. The bulk of this is direct rather than via public spending on social protection. Social insurance (mainly public pensions) does the “heavy lifting” of the floor. Social assistance (mainly targeted cash-transfers) lifts the floor by only 1.5 cents per day on average, which is less than 10% of mean spending on social assistance.  

Work in Progress

Where is the party? Campaign Visits in US Presidential Elections

Presidential candidates face crucial and difficult decisions when deciding which locations to include on their campaign trails. Stromberg (2008) Showed how the electoral college incentives play an important role in determining which states to visit during the campaign. However, it remains an open question how politicians strategically choose to spend time within those states in order to increase their election prospects. To answer this gap, I develop a theoretical model that describes the types of allocations that maximize electoral returns and continue to test these predictions using a novel dataset of presidential visits from the 2012 and 2016 election at the county level.

Bandwagon Effects in Caucuses 

With Gaurav Bagwe and Ben Solow 

In multi-candidate elections, past electoral outcomes may serve as a coordination device for strategic voters and as a bandwagon for behavioral voters who have a preference for voting for the winner. Both motivations working in tandem can amplify the effect of close wins on future electoral outcomes (Pons and Tricauld 2020). We use the setting provided by the Iowa Democratic presidential caucuses to test this hypothesis using a regression discontinuity design. We find that candidates who barely ranked first in the first round get a higher vote share (5.6 pp) in the second round and are more likely to win (12.5 pp), compared to those who barely ranked second. We find a similar increase (7.3 pp) in the vote share for candidates who barely ranked second in the first round compared to those who barely ranked third. The unique electoral rules that govern caucuses allow us to rule out alternative channels such as a change in the composition of the electorate, endogenous responses by candidates, and media coverage of winners. The results suggest that both coordination and behavioral motivations could play a significant role in determining electoral outcomes even in the context of high-stakes elections.

Technical Notes

Implementar mudanças curriculares é uma tarefa desafiadora para as redes de ensino. Fazer essa mudança em um momento de transição para o Ensino Médio, onde são identificadas as maiores taxas de evasão escolar, torna o processo duplamente desafiador. O Guia para sua Jornada no Novo Ensino Médio Capixaba foi estruturado como uma ferramenta interativa e informacional, partindo da premissa de que estudantes bem informados escolhem escolas e trajetórias formativas mais alinhadas a suas preferências. Estratégia inédita, o Guia coletou as preferências dos estudantes de forma lúdica. Por trás da ferramenta, um experimento com mais de 10 mil participantes baseado em nudges e na teoria da economia comportamental investigou-se estimular a escolha por educação integral geraria um aumento significativo na candidatura às escolas dessa modalidade. Não foram verificados impactos estatisticamente significantes desse experimento (RCT - randomized control trial), mas observamos outros fatores, não relacionados aos nudges. Um exemplo: estudantes de nível socioeconômico mais elevado e provenientes de escolas de tempo integral costumam procurar essas escolas com mais frequência, o que indica um efeito de permanência. Ao longo da intervenção e com base no uso da ferramenta pelos estudantes, documentamos os aprendizados relacionados à política pública de educação: eles podem servir como referência e aprendizado para outros estados na implantação do Novo Ensino Médio e como estratégias para informar os estudantes.

Teachers are the most important and expensive input when it comes to improving the quality of education. In many Latin American countries, the decentralized way in which teachers are assigned is extremely costly and inefficient and lacks objectivity. This document advocates the implementation of digital centralized teacher assignment systems as more efficient, equitable, objective, and cost-effective. Using administrative data from Ecuador, this study estimates that the implementation of digital centralization could have an annual impact of over USD 17 million. This figure takes into account the direct implementation costs, the savings generated by the transition, and the learning gains achieved by improving the assignment of better-quality teachers and reducing the number of empty vacancies. These results confirm that this policy is not only efficient, but also cost-efficient.

Traditional (decentralized) student assignment systems are characterized by their inefficiency, lack of transparency, and inequity of distribution. In response, ever more places have begun to implement digital centralized assignment systems. Although the literature has endorsed these systems assignment efficiency, this document is the first to analyze their cost-efficiency and discuss traditionally overlooked benefits: monitoring, transparency, and planning tools in the market. Using administrative data from Chile, the analysis estimates that the implementation of digital centralization could generate a net impact of USD 13 million, considering the direct implementation costs, the savings generated during the transition, and the resulting efficiency gains. Our results confirm that this policy is not only efficient but also cost-efficient.

During the COVID-19 pandemic, 165 million students in Latin America and the Caribbean were abruptly disconnected from education. By the end of 2021, schools had been closed for an average of 237 days, more than anywhere else in the world. Although the COVID-19 pandemic has affected the lives of many people around the world, it has not done so uniformly. Perhaps one of the most devastating legacies of the pandemic is its effect on young people. During one of life's most critical developmental periods, large numbers of young people have been deprived of education and exposed to increased levels of social, health, economic insecurity, domestic violence and abuse. Despite the efforts of the entire educational community to ensure that students maintain some level of learning, the evidence collected in this report shows that many schoolchildren did not participate in meaningful learning activities and that the slowdown in the accumulation of skills and human capital it will have immediate and long-term consequences for the well-being of countries. Why? It wasn't just the pandemic. It was the fact that the region and its educational systems were not well prepared to withstand a shock of this magnitude. The short and long-term effects of the health crisis cannot be understood if the story does not include the starting conditions. The purpose of this report is to provide those responsible for educational policy throughout the region with an idea of the magnitude of the damage, its consequences if immediate, substantial and effective measures are not taken, and the priorities of educational policy taking into account the starting point and the effects of the pandemic. More importantly, the report also provides a compass for the education sector to develop evidence-based responses to the immediate needs of young people, as well as the medium-term measures needed to rebuild our education systems so that be more resilient, equitable, and efficient in developing effective learners throughout life. If we do nothing, we will leave an entire generation behind. Governments have to use all the levers at their disposal to recover, and education is key in that process.

El COVID-19 afectó a los sistemas educativos de todo el mundo. América Latina ya enfrentaba problemas importantes en términos de calidad educativa y se espera que el cierre prolongado de las escuelas haya afectado aún más el aprendizaje y las trayectorias de los estudiantes. Este documento contribuye a la discusión al analizar el caso de Colombia, uno de los países donde más duró el cierre de las escuelas, no obstante, mantuvo en 2020 la aplicación de las pruebas estandarizadas Saber 11 que rinden los estudiantes de último grado de bachillerato. Los resultados indican que los colegios con mayores pérdidas proporcionales de evaluados entre 2019 y 2020 vieron un aumento en sus puntajes en el examen, lo cual sugiere que los estudiantes de menores habilidades fueron los que dejaron la escuela y, por ende, no presentaron la prueba. Los datos de 2021 muestran que, a pesar de un incremento general en el número de evaluados con respecto a 2020, la participación de los estudiantes de menor nivel socioeconómico no regresó a los niveles prepandemia y esto se sigue reflejando en mejores promedios en los puntajes de sus escuelas. Las estimaciones del posible impacto de la pandemia en los resultados de aprendizaje sugieren un efecto de 0,05 desviaciones estándar, por debajo de lo encontrado en otros países de la región que han medido impactos en estudiantes de grados inferiores. Los cambios en la composición de la muestra y el tipo de evaluación pueden ayudar a explicar las discrepancias entre los resultados de esta investigación y los del resto de la región. Finalmente, se presenta una discusión acerca de las precauciones que se deben tener al momento de analizar resultados de aprendizaje usando evaluaciones de corte transversal.