College Access Spillover on High School Performance: Evidence from College Openings, September 2025, with T. Rands. Resubmitted to Economics of Education Review.
We investigate the effects of an expansion of a tuition-free technical college on local high-school academic performance. Using a difference-in-differences strategy, we exploit the staggered opening of new technical colleges in the state of São Paulo, Brazil, which we compare with student grades on a no-stakes test taken in the last year of high school. We document that the opening of one such college has a positive effect of 0.106 of a standard deviation on the performance in the exam. Our findings are consistent with the hypothesis that local access to tuition-free colleges increases high-school students’ incentives to learn.
Paths that Led to Gold: Historical Roads, Trade, and Persistence, April 2025, with A. Portugal. Revision requested by Regional Science and Urban Economics.
We investigate the heterogeneity in the long-run effects of two colonial roads in Brazil. These roads connected the coast to a thriving gold mining region in the highlands. The oldest road was heavily used at the beginning of the gold rush, but tax collection concerns led colonial authorities to commission the construction and favor the use of an alternative road. By the middle of the gold rush, trade flows through the newer road vastly surpassed those through the older road. We combine georeferenced data on the colonial roads with spatial data on modern economic activity, modern infrastructure, and historical settlements. By comparing the patterns of how density increases with proximity to each of these colonial roads, we estimate that the persistent effects of the newer colonial road are larger than the persistent effects of the older colonial road. Our results suggest that the past importance for trade of a historic road determines whether the effects of the road persist in the long run.
Do Landslides Impact Urbanization Patterns? Evidence from Brazilian Cities, May 2025, with P.J. Alves, R.C.A. Lima and F. Cavalcanti. Revision requested by Journal of Urban Economics (JUE Insights).
This paper investigates the impact of landslides on the subsequent urbanization patterns of Brazilian cities. Combining satellite data on urbanization with detailed landslide records and adopting a staggered difference-in-differences approach, we show that municipalities exposed to landslides experience an average reduction of 5.8% in the size of their urban area. Landslides also increase the fragmentation of the urban area. Both effects persist over time. Guided by a simple monocentric city model, we show that negative effects on local labor markets plausibly explain these effects by making affected cities less attractive.
Modern deforestation in the Brazilian Amazon is higher in areas close to historical Indigenous extinctions, April 2025, with A. Ferreira.
Amazon deforestation is a major environmental problem. While short-term determinants of deforestation have been extensively studied, historical determinants remain little understood. Violence, land disputes and demographic changes characterize the history of human occupation in the Amazon. Although most deforestation is recent, it is potentially influenced by historical events. Here, we provide evidence for the Brazilian Amazon that modern deforestation is significantly higher in areas historically inhabited by extinct Indigenous groups. We compile a database of historical Indigenous groups, where we locate those extinct before 1944, and combine it with remote-sensed deforestation measures. Each extinct historical Indigenous group in a 20-km neighborhood of an area is associated with a 29% increase in accumulated deforestation by 2010. The relationship with deforestation is strongest for extinctions likely due to disease or violence, which usually led to demographic voids, but absent for extinctions likely due to cultural assimilation, where the presence of traditional communities may have persisted. Our findings support the hypothesis that historical Indigenous extinctions opened the way for deforestation by removing deterrence against non-Indigenous land occupation. Ethnohistorical Data. Data on Extinction Circumstances. Replication Data.
Optimal Provision of Urban Infrastructure: Theory and Evidence from Brazilian Cities, August 2025, with T.L. Padovani and V.P. Neto.
We investigate how urban infrastructure is, and should be, distributed within cities. Using high-resolution data on population, housing prices, and infrastructure, including sanitation access and street characteristics, for Brazilian cities, we document that infrastructure provision consistently declines with distance from the city center. Moreover, unconditional housing price and population density gradients are steeper than when conditioning on infrastructure. To interpret these stylized facts, we consider a monocentric city model in which infrastructure is a local public good and we derive a sufficient-statistics rule for its optimal spatial allocation. This rule implies that, under certain conditions, population density and land prices should fully explain spatial differences in infrastructure. We find systematic deviations from this prediction, with infrastructure biased toward central neighborhoods. While aggregate losses in welfare or city size are modest, this bias shapes city structure, implying city centers 21% denser than optimal.
A Positive View of Entry Costs, work in progress (draft soon!), with B. Camargo and G. Stein.
Capital as an Anchor of Agricultural Employment: Evidence from the 1975 Frost, Regional Science and Urban Economics, 108 (2024).
How do temporary shocks affect the spatial distribution of employment in agriculture? I investigate this question by examining the 1975 frost that damaged coffee trees in Brazil. I find that the frost persistently affected the spatial distribution of employment in agriculture. To identify the effects of the capital destruction from the frost, I compare changes in agricultural employment across local economies that had different coffee tree densities right before the frost and that were differently affected by the extreme weather. The frost resulted in a persistent decline in agricultural employment. The findings are consistent with a history versus expectations model in which fixed and specific capital (such as coffee trees) prevents multiple equilibria despite strategic complementarities in crop choice. Working paper version.
Road Endpoints and City Sizes, Review of Economics and Statistics, forthcoming.
I examine the long-run effects of the timing of railroad construction on city sizes. I first present a stylized model which predicts that towns that are railroad endpoints for longer become persistently larger. I then show that, in a sample of Brazilian railroad towns, time as endpoint strongly predicts town size: each additional year that a town was a railroad endpoint in the past is associated with a town population 0.107 log points larger in 2010. Additional testable implications of the model and an instrumental variable approach suggest that such association reflects a causal effect. Working paper version. Replication data.
Signaling in Dynamic Markets with Adverse Selection, Journal of Economic Theory, 206 (2022), with B. Camargo.
We study trade in dynamic decentralized markets with adverse selection. In contrast to the literature on the topic so far, we assume that the informed sellers make the offers so that signaling through prices is possible. We establish basic properties of equilibria, discuss the standard two-type case in detail, and then analyze the general finite-type case. We prove that market efficiency, measured by the maximum gains from trade in equilibrium, is invariant to trading frictions. Our analysis shows that screening and signaling lead to markedly different trading outcomes. Longer working paper version.
Cities on Pre-Columbian Paths, Journal of Urban Economics, 122 (2021).
Does the geographic distribution of pre-Columbian societies determine the location of New World cities? This paper provides evidence that a pre-colonial indigenous trail influenced the location of modern cities in southern Brazil. To distinguish the causal effects of historical settlement near the trail from the effects of geographic fundamentals that could correlate with it, I compare how population density and urbanization change with proximity to the trail in two different regions. The first region has been settled by Europeans since the 16th century, while European settlement in the second region was interrupted after a 17th-century slave raid. Proximity to the indigenous path is associated with higher population density and urbanization in the first region, but not in the second. These findings suggest a path dependence that goes back to the pre-colonial past. Working paper version.
Employment Resilience to Transit Interruption: Evidence from São Paulo, July 2024, with G. Navarro.
This paper estimates null effects of a temporary interruption of transit on the spatial distribution of employment. We employ a difference-in-differences approach, exploiting a plausibly exogenous six-month shutdown of service in a line of the São Paulo metropolitan rail network. Using administrative data on formal employment, we find that the interruption did not change employment flows between affected neighborhoods and the central business district. From a monthly labor market survey, we find null effects on unemployment, earnings, and formality, but negative effects on self-employment activity. We interpret our findings as evidence that employment matches are sufficiently valuable so they are resilient to large temporary increases in commuting costs.
Teaching: I taught the following classes in recent years
Economic Geography (graduate elective, syllabus)
Statistics for Economics (undergrad, syllabus, in Portuguese)
Development Economics (graduate elective, syllabus)
Microeconomics (third class in graduate micro sequence, syllabus)
Post-Crisis Microeconomics (executive, syllabus, in Portuguese)
Ethnohistorical data: we published our work in cataloging the Ethnohistorical Map of Curt Nimuendajú (2017), used in Barsanetti and Ferreira (2024), as a publicly available dataset for other researchers. If you use them, please cite these data, Barsanetti and Ferreira (2024), and the references for the original map. This version: updated on July, 2024.
Second Annual LAUrban Meeting: I co-organized this workshop, which took place at FGV EPGE on December 1 and 2, 2023. See: call for papers, program, website.