ABSTRACT. In this paper we present empirical evidence on how mobile network disruptions are linked to episodes of violence against women. Using high-frequency data for Brazil we show that the outage caused by the transition to 9-digit dialing for all mobile phone numbers had unintended positive effects on hospitalization among women victims of violence. This pattern is consistent with the argument that mobile phones can provide a fast way for women to call for help in risky situations. Results are robust to different specifications and vary according to victims' and municipalities' characteristics. We also show that violence is mainly perpetrated by intimate partners and that unemployed women are the most affected. We further highlight that the effect is tied to places with broader telecommunication infrastructure. Our findings shed some light on the link between mobile technology use and women's safety, supporting the expansion of technology-based policies to deter gender-based crimes.

ABSTRACT. In this paper, we test whether more educated candidates make less corrupt public managers. Leveraging on electoral RDD and on a randomized inspection program, we show that college-degree candidates commit 19% fewer infringements than their less-educated peers. Exploiting data on judicial records, we show that this effect does not stem from differences in corrupt behavior and might be explained by differences in managerial skills. Taking advantage of administrative labor records, we find that educated candidates have more previous experience in the public sector and in high-skill positions, confirming our hypothesis. Finally, we show that there is no differential in the provision of public goods between these two groups.

ABSTRACT. This paper investigates the relationship between grades and social status in Brazil and how it differs between racial groups. Social status is measured using friendship ties among students, assigning higher status to those more central in the network. We find a positive correlation between grades and social status among nonwhite students that is driven by their popularity among white classmates. There is also a positive correlation between grades and same-race social status for whites, but not across races. This differs from the pattern observed in the US, where a negative correlation between nonwhites' grades and their status among racial peers is not compensated by their popularity among white students, which has been linked with "acting white". We also investigate how academic performance is associated with racial identity choice, conditional on skin color. This relationship is is weak across all students. However, among high-school students, females, and in classrooms with smaller share of whites "good grades whiten".

ABSTRACT. Quantile regression models the conditional distribution of a response variable Y given a vector of covariates X at different quantile levels, offering a parsimonious specification for the whole conditional distribution (Koenker and Xiao (2006)). Using quantile regression, as opposed to traditional regres- sion models, provides several advantages, but can add complexity to certain operations. For example, when performing variable selection, there might be a different set of variables selected for each quantile. Frumento and Bottai (2016) propose a parametric modeling of quantile regression coefficient func- tions that allows one to estimate, in a single minimization problem, the co- efficients for all quantile levels in a given grid. Sottile et al. (2020) use this approach to perform LASSO variable selection using information on all quan- tile levels simultaneously. In this work, we propose a method for global vari- able selection and coefficient estimation, similar to the approach presented in Sottile et al. (2020). We introduce group (ada)LASSO penalty, suggested in Yuan and Lin (2006), and apply it to the Quantile Autoregressive Distributed Lag (QADL) model of Galvao, Montes-Rojas and Park (2013). Furthermore, since we are in a time series context, we also evaluate the variable selec- tion penalization applying higher penalties to higher lags, as introduced by Konzen and Ziegelmann (2016). Our simulation suggests that a weighted, lag-penalized approach can provide better results in selecting covariates in models with up to five relevant lags of the regressors, without compromis- ing the estimation of the coefficients. In particular, both LASSO and group LASSO penalization with higher weights for higher lags removed the non- relevant variables more often than the competitors in these scenarios.

ABSTRACT. Using data from criminal cases in the State of São Paulo, Brazil, I analyze whether alternative sentences --- e.g., fines or community services --- decrease recidivism. To do so, I leverage the random assignment of judges within a court district as a source of exogenous variation in the probability of punishment to identify the effect of alternative sentences in comparison to the no-punishment counterfactual. Initially, I show that the usual identification strategy, that uses only the trial judge's sentence, fails to identify the correct treatment effect parameter because the trial judge's decision may misclassify the final sentence due to the appeals process. To avoid this measurement error problem, I follow two approaches. First, I propose a novel partial identification strategy to identify the marginal treatment effect (MTE) with a misclassified treatment. This method explores restrictions on the relationship between the misclassified treatment and the correctly measured treatment, allowing for dependence between the instrument and the potential misclassified treatment variables and the misreporting decision. Second, I collect data on Appeals Court's decisions and estimate the MTE based on the correctly measured final sentence. This last exercise is used as a benchmark for the set identification method that I propose.

ABSTRACT. This paper explores the complementarity between traditional econometrics and machine learning and applies the resulting model - the VAR-teXt - to central bank communication. The VAR-teXt is a vector autoregressive (VAR) model augmented with information retrieved from text, turned into quantitative data via a Latent Dirichlet Allocation (LDA) model, whereby the number of topics (or textual factors) is chosen based on their predictive performance. A Markov chain Monte Carlo (MCMC) sampling algorithm for the estimation of the VAR-teXt that takes into account the fact that the textual factors are estimates is also provided. The approach is then extended to dynamic factor models (DFM) generating the DFM-teXt. Results show that textual factors based on Federal Open Market Committee (FOMC) statements are indeed useful for forecasting.

ABSTRACT. We study how revenue-dependent taxation can lead to output and welfare effects, as well as impacts on firm informality. First, we modify an equilibrium entrepreneurial choice model to incorporate a revenue-dependent tax regime. We then bring our model to the data in order to assess the effects of the Brazilian Simples tax regime, a simplified tax scheme that combines nine different taxes into a lower-rate (vis-à-vis the standard tax regime) revenue tax. Firms are eligible to Simples if their revenues do not exceed a yearly threshold, meaning that, inasmuch as Simples may reduce firm informality (one of its original goals), it may also hinder firm growth, as some firms have an incentive to withhold production and pay lower taxes. We structurally estimate the parameters of our model using a combination of survey and administrative data on Brazilian firms. This approach allows us to analyze the effects of Simples on the size and productivity distribution of firms – as well as the effects of counterfactual policies such as a reduction in the tax burden gap between formal firms. We find that the Simples tax regime increases the average output and productivity of the Brazilian economy, and also encourages the formalization of informal firms. This comes at the cost of reduced tax collection, as well as a reduction of output from formal firms that withhold production to pay lower taxes; and a higher wage penalizing productive formal firms. The latter effects dominate, and consequently welfare decreases by 1.5\% in an equilibrium with the Simples regime. Finally, we show that we may improve welfare under the Simples regime if we introduce a tax scheme with a lower tax rate gap between the simplified and standard formal sector.

ABSTRACT. We merge two important Brazilian datasets (RAIS and PIA) to produce firm-level total factor productivity estimates that control for workers’ human capital. Then we investigate the correlation between managers’ education and firms’ TFP considering different levels of industry disaggregation. We find a positive, albeit small correlation for the manufacturing sector as a whole, and much higher and more statistically significant correlations for some 2-digit industries. Also at the 2-digit level, we find that the positive correlation between firm TFP and manager schooling is higher for more R&D intensive industries, and lower for industries more dependent on external finance.

ABSTRACT. Life expectancy has risen steadily around the world in recent decades, as well as health expenditures as a share of GDP. Technological progress in medicine is related to these trends, but faced with the difficulties in measuring productivity in the health sector, how can we assess the relative importance of technology and preferences in determining longevity and health spending increases? Based on two stylized facts on human aging, the Gompertz curve and the Strehler-Mildvan correlation, we propose a longevity production function in which productivity gains are identified from parametric changes in these empirical regularities. We construct an overlapping generations model with structural transformation related to the health sector, promoted by sector heterogeneity in technological characteristics and the non-homothetic preferences for longevity improvements given by Rosen (1988) and Hall and Jones (2007). After estimating the longevity production parameters in a panel regression with data for 96 countries from 1960 to 2010, we use the model to simulate optimal trajectories of life expectancy at age 30 and health expenditures as a share of GDP in Brazil, France, and the US. We find that productivity growth in longevity production explains 70% to 80% of life expectancy gains in all cases, but actually the increase in health expenditures as a share of GDP is driven by preferences.

ABSTRACT. This paper addresses two related research questions: first, how firms’ dynamics evolve after exposure to immigrant influx shocks in a given region? Second, how the labor market outcomes of natives are affected in those regions? While most of the migration literature focuses on native individual outcomes aggregated at some geographical level, hiring an immigrant is a firm-level decision that might change firm’s dynamics and might affect natives within firms or across different firms operating in the same labor market area. In this draft, we use a matched employer-employee dataset from 2000 to 2017 to analyse firm and native worker outcomes aggregated at the regional level. Our empirical strategy combines past settlements of immigrants and push factors (e.g. environmental disasters and political instability) at the country of origin, which are plausibly uncorrelated with local labor market outcomes in Brazil. We find evidence that immigration had a negative effect on the number of firms hiring formally in a region and also on native labor market outcomes (decrease in wages and increase in layoffs). We aim to shed light on the mechanisms by analyzing firm-level and worker-level outcomes within and across firms.

ABSTRACT. Government agencies prefer to choose their suppliers directly. Exploiting a rule in Brazil that waives competitive bidding for small-value purchases in public procurements, I find there are 7 times more procurements just below the small-value purchase threshold. Products are also 20 percent more expensive under bid waivers than under auctions. While these results raise suspicion of corruption, higher quality products under bid waivers explain 50 percent of the difference in prices. Healthcare, public security, and armed forces agencies benefit the most from this purchase of premium products. Overall, discretion allows the government to choose better quality goods and benefit society.

ABSTRACT. In this paper we investigate the spatial extent of agglomeration economies on the wage earnings within the Brazilian cities. For this, we use a geocoded employer-employee panel data 2006-2014, exogenously determined grid and different geographical distance bands. We deal with the spatial sorting and endogeneity in the wage-agglomeration economies relationship with controls for observable and unobservable individual and establishment characteristics and instrumental variables based on the exogenous expansion of higher education in Brazil in period 1991-2004. Our main results indicate that distance seems to be still more important in developing countries context. Adding 1,000 college-educated workers up to 1 km from the individual's workplace would increase the wages of workers on average by 6.78 percent. On the other hand, if the same number of workers are added to the 1-5 km or 5-10 km range, the wages of workers would increase, on average, by 1.95 and 1.27 percent, respectively. This evidence is robust to different specifications (e.g., including worker-plant or worker-city matches) and shows that the speed of decay of agglomeration economies is higher in Brazil than observed in developed countries.

ABSTRACT. This paper examines how religious subsidies contributed to the rise of Evangelicals and the decline of the Catholic church, a commonly observed trend in the Christian world. Using the Brazilian experience as a showcase, we build and estimate a dynamic game of church entry using temple entry/exit data across municipalities for 1991-2018. Our counterfactual analysis shows that Evangelicals gain market share from Catholics as tax exemptions increase entry of smaller churches. By combining DiD estimates and our counterfactual scenario, we show that the vote share of the Congressional Evangelical caucus would have been 20% lower if state subsidies had been removed.

ABSTRACT. An enormous literature estimates the effects of minimum wage policies on employment and wages as well as on earnings inequality. Yet whether such policies can affect children of covered workers over the long run remains an open question. Previous studies tend to focus mainly on the contemporary impacts on individuals making high school dropout or college enrollment decisions, but retrospective exposure during critical periods of child development may be equally or even more important. In this paper, we study the long-run consequences of the unprecedented minimum wages expansion caused by the 1966 Amendments to the Fair Labor Standards Act. We exploit geographic variation in the baseline size of the population gaining below the new minimum wage floor just prior to the reform and the timing of policy adoption for identification. The results indicate that childhood exposure to the reform increased adult human capital and adult economic self-sufficiency. In an attempt to understand the mechanisms behind these effects, we find that the reform led to an immediate and persistent increase in parental income, especially for mothers and those at the bottom of the distribution. At the same time, we observe no meaningful changes in parental employment, fertility, or marital outcomes in the immediate aftermath of the reform. Taken in their entirety, these findings suggest that minimum wage policies could constitute an effective welfare tool to reduce inequality by improving the early life conditions of the poor.