ABSTRACT. The bayesian additive regression tree (BART) algorithm (Chipman et al., 2010) has recently received considerable interest because of its ability to outperform other machine learning methods in prediction tasks. One of BART’s greatest qualities is its ability to deal with high-dimensional data. This characteristic can be very useful for estimating treatment effects in regression discontinuity designs (RDDs), where the inclusion of covariates might render the derivation of a fully-optimal bandwidth for the commonly employed local polynomial regression infeasible (Calonico et al., 2020). This paper investigates the application of BART to the context of regression discontinuity designs (RDDs). For this purpose, we first investigate how to obtain treatment effect estimates in an RDD with the original BART algorithm and how such estimates compare to the commonly used local linear regression approach. This analysis indicates that BART produces more reasonable results in scenarios with many ‘pre-treatment’ covariates. We then propose an extension of the BART algorithm where we consider a prior which incorporates the RDD assumptions and allows for direct estimation of the treatment effects.

ABSTRACT. In this paper, we explore machine learning (ML) methods to improve inflation fore- casting in Brazil. An extensive out-of-sample forecasting exercise is designed with multiple horizons, a large database of 501 series, and 50 forecasting methods, including new machine learning techniques proposed here, traditional econometric models and forecast combination methods. We also provide tools to identify the key variables to predict inflation, thus helping to open the ML black box. Despite the evidence of no universal best model, the results indicate machine learning methods can, in numerous cases, outperform traditional econometric models in terms of mean-squared error. Moreover, the results indicate the existence of nonlinearities in the inflation dynamics, which are relevant to forecast inflation. The set of top forecasts often includes forecast combinations, tree-based methods (such as random forest and xgboost), breakeven in‡ation, and survey-based expectations. Altogether, these findings offer a valuable contribution to macroeconomic forecasting, especially, focused on Brazilian inflation.

ABSTRACT. Difference-in-Differences (DiD) is a popular method used to evaluate the effect of a treatment. In its most simple version a control group remains untreated at two periods, whereas the treatment group becomes fully treated at the second period. However, it is not uncommon in applications of the method that the treatment rate only increases more in the treatment group. This article presents identification results for the marginal treatment effect (MTE) in such fuzzy designs. We show that we can modify the standard identifying assumptions in DiD designs with covariates to identify the MTE in models with essential heterogeneity. We propose two different procedures for the estimation of the MTE that rely on different assumptions regarding the potential outcomes model and prove their asymptotical normality. Furthermore, we derive a doubly-robust estimator for the local average treatment effect (LATE) which augments the two-way fixed effects regression model with a control function and unit-specific weights that rise from the propensity score. We assert the desirable finite-sample properties through simulation studies of a linear MTE model. Finally, we use our results to investigate heterogeneity on the returns to primary school attendance in Indonesia.

ABSTRACT. We consider the problem of inference in shift-share research designs. The choice between existing approaches that allow for unrestricted spatial correlation involves tradeoffs, varying in terms of their validity when there are relatively few or concentrated shocks, and in terms of the assumptions on the shock assignment process and treatment effects heterogeneity. We propose alternative randomization inference methods that combine the advantages of different approaches. These methods are valid in finite samples under relatively stronger assumptions, while asymptotically valid under weaker assumptions.

ABSTRACT. We propose three mechanisms for two parties to jointly select a group of fixed size. We show that if the parties' preferences over sets are leximin extensions of the parties' preferences over candidates then these mechanisms implement the Unanimity Compromise Set. This work extends the concepts and the results in Barberà and Coelho (2022), in which the parties had to choose a single candidate, to cover a wide class of natural applications.

ABSTRACT. We explore the process of bayesian updating of a Random Choice Rule with a Finite Random Expected Utility representation, as put forth in Ahn and Sarver (2013). We present the necessary and sucient condition, which we call Random Consistency, for a Random Choice Rule to be a update of another after the Decision Maker learns new information and contracts or expands her subjective state spaces. We extend the results in Ahn and Sarver (2013) by characterizing the opposite direction of the unforeseen contingencies representation there proposed, when the subjective states of the Finite Random Expected Utility representation of a Random Choice Rule is contained in the subjective state space of the representation, in the form of Dekel et al. (2001), of a Preference Over Menus. We also discuss the conditions under which a collection of Random Choice Rules represent a partition of a broader Random Choice Rule or of a Preference Over Menus.

ABSTRACT. During the medieval and early modern periods the Middle East lost its economic advantage relative to the West. Recent explanations of this historical phenomenon— called the Long Divergence—focus on Middle Eastern (over-)reliance on religious legitimacy and political centralization. We study these features in a political economy model of the interactions between rulers, secular and clerical elites, and civil society. The model induces a joint evolution of culture and political institutions (delegation of power from rulers to elites) converging to one of two distinct stationary states: a religious and a secular regime. We then map qualitatively parameters and initial conditions characterizing the West and the Middle East (separation between state and religion, initial political power of clerical elites and predominance of religious values in the population) into the implied model dynamics to show that they are consistent with the Long Divergence as well as with several key stylized political and economic facts highlighted in the historical narrative. Most notably, this mapping suggests non-monotonic political economy strategies, in terms of legitimacy and political decentralization, in both regions which indeed characterize their history.

ABSTRACT. The fundamental monetary innovation embedded into cryptocurrencies is accounting coordination. Decentralized management of digital money’s accounting by a network of computers is achieved as a Nash equilibrium of a coordination game among the network’s nodes: the so called miners. Equilibrium analysis demands allowing miners to secretly update their accounting, i.e., to privately build multiple blocks of transactions and to deviate from the longest chain rule. We formalize such reasoning by proposing an accounting coordination game inspired on the Bitcoin design. In particular, by proposing a model that explicitly tells apart mining costs related to energy consumption from those related to computational capacity, we are able to study how symmetric equilibrium existence depends on well known parameters, like the average time for updating accounting records and the rewards collected from mining (accounting) activities. It is shown that the (off-equilibrium) possibility of double spending makes the attractiveness of the equilibrium strategy a decreasing function of the average time for updating accounting records.

ABSTRACT. We investigate how changes in barriers to female labor force participation and in the child penalty affect occupational decisions, fertility and income. We build a general equilibrium model of occupational choice with men and women, human capital investment and fertility. We fit the model to the US and India. Changing gender barriers account for 31% of the US growth between 1960 and 2010 (4.1% for India in 1983-2004). The implications of these barriers for the welfare of female workers with children were even larger, with lower child penalty alone increasing the welfare of this group by 7% in the US.

ABSTRACT. In most countries, the rules governing public and private pension systems are different, and so are hiring procedures, and job contracts. The tenures of government employees are longer and their wages, in general, higher. In this sense, social security reforms will affect not only the decision to leave the labor force, but also the choice of which sector to work. In this article, we study the impact of social security reforms on retirement and occupational behavior. We develop a life-cycle model with three sectors - private formal, private informal and public - and endogenous retirement to evaluate what are the macroeconomic and occupational impacts of social security reforms in an economy with multiple pension systems. In a model calibrated to Brazil, we simulate and quantitatively assess the long-run impact of reforms being discussed and/or implemented in different economies. Among them, the unification of pension systems and the increase of minimum retirement age. These reforms are found to affect the decision to apply to a public job, savings during the life cycle and skill composition across sectors. On the long run, they lead to higher output and capital, less informality and to average welfare gains. They also drastically reduce social security deficit.

ABSTRACT. We investigate the intra-household labor supply dynamics of secondary earners of families - children and mothers - using a household job search approach. We focus on Brazilian youth aged 14 to 29 years living with parents. We develop and estimate a structural household job search model that explicitly considers children as decision-makers in a household job search model. The model distinguishes between unemployment and inactivity of mothers and children, allows for different search behavior and job acceptance depending on the situation of the other member in the labor market and for parental investment in child education. We decompose the contribution of labor market shocks specific to each household member to explain changes in employment and participation between 2003-06 and 2010-13. Our estimation results show that the increase in the employment rate of children in the second period is the result of better labor market conditions, worse job arrival rate for mothers, and an increase in mother's flow out of inactivity. In our counterfactual simulations, we find that most of the observed fall in children participation rate is explained by movements of children in and out of the labor force and an improvement in labor market conditions. We also find evidence for the added worker effect: an increase in father's income reduces employment and participation of children and mothers. Cross-effects are underestimated if we ignore endogenous schooling. Therefore, our results strengthen the argument about the relevance of household search behavior in the labor supply decisions of secondary earners in families, so that using individual job search models to understand aggregate employment and participation can be misleading.

ABSTRACT. We evaluate the effects of violating the intellectual property rights of patents, as established by the generic drug law, on innovation and labor market outcomes in Brazil. Our empirical results suggest that the patent infringement law stimulates innovation, as evidenced by an increase of 71.0-89.6% in patent applications. Our results show that the number of employees increases by 8.1% and the number of inventors increases by 7.7% across innovator firms. Our results also suggest an increase of 8.0% in real average hourly earnings. To reconcile these results, we demonstrate that a country with a high degree of resource misallocation potentiates the positive effects of patent infringement on innovation outcomes. Patent infringements also generate a labor reallocation effect on innovator firms post-intervention.

ABSTRACT. Siblings share a bond that makes their relationship unique and, thus, probably influence each other’s behavior and decisions. In this paper, we examine siblings’ spillovers in elementary school. We explore the cutoff entry rule as an exogenous variation to school starting age and estimate spillovers causal effects using regression discontinuity design. Using data from state schools in São Paulo, our results show that entering school being older in the cohort generates negative spillovers in younger sisters’ proficiency levels. These estimates are particularly significant in less affluent families with close-in-age siblings in the early years of elementary school. However, the results are quite different when analyzing spillovers from younger to older siblings. A younger student entering school later has positive spillovers on their older siblings’ test scores. These results are driven by more affluent families with similar siblings (close in age and same gender). Our results provide empirical evidence that we should have externalities between siblings in the educational context. Moreover, these effects can diverge when considering the families’ socioeconomic and demographic characteristics.

ABSTRACT. This paper studies the unintended long-run effects of a permanent agricultural shock led by agro-terrorism in Brazil on the education and labor market. We explore the witch broom outbreak in cocoa farms in the world's second most important cocoa production region until 1989, the southeast of Bahia's state in the northeast of Brazil. Although the introduction of the plague had political motivations, it had unintended effects on poor people's lives. To assess the impact of witch broom disease, we leverage information about people born in cities affected and not affected by the disease and explore the difference in educational attainments between cohorts older and younger than eighteen years old at the time of the witch broom outbreak. The main results show that the witch broom outbreak negatively affected the long-term education and earnings of individuals living in affected cities. The results are very similar by gender and race. We show a piece of evidence that the increase in child labor may drive our results. The negative effects on young cohorts are consistent with the known relation between child labor and cocoa production and the literature about the long-term effects of economic shocks.

ABSTRACT. The humanitarian crisis triggered by Venezuela’s political and socioeconomic situation in mid-2010s has led to the most significant movement of refugees and migrants in the recent history of Latin America and the Caribbean. We explore this episode as a natural experiment to investigate the effects of Venezuelan influx to Brazil on students’ academic performance in grade 5. Using a difference-in-difference approach at school level, we found larger increases in average math test scores in schools that experienced larger increases in the share of Venezuelan students. No impact on average for language scores was found, but there was a positive impact on the median and a reduction in language test score inequality. The heterogeneity analysis shows a higher impact on schools presenting worse school quality indicators in the pre-immigration period.

ABSTRACT. The objective of this paper is to investigate whether higher political competition induces the expansion of public early childhood education and care (ECEC) services using Brazil as a case study. Public ECEC services are provided mainly by Brazilian municipalities and although enrolment for children aged 0 to 3 is not mandatory, it has increased continuously in the last two decades, with large heterogeneity across municipalities. In addition, electoral rules in Brazil establishes a dual-ballot system only for municipalities with more than 200,000 registered voters. This provides an exogenous variation in political competition that enables us to evaluate its impact on ECEC provision through a regression discontinuity design. We find that average estimated municipal crèche net enrolment rates are around 3 percentage points larger in municipalities with the dual-ballot. ECEC expenditure levels are also higher in these municipalities.

ABSTRACT. Salience often covary with information. Hence, empirically showing that the salience of a stock, in itself, affects retail investors trading decisions is challenging. We document that living in a small city that has a local store of a brick-and-mortar firm more than doubles the chances of individuals picking the stock of that firm to day-trade. This suggests a direct relation between salience and retail investors’ trading decisions: a local store in a small city i) increases the visual salience of the firm for the city residents but ii) does not provide any useful information for day-trading, which depends exclusively on high-frequency indicators. We explore the granularity of our dataset to control for indirect channels that can make retail day-trading correlate with local stores.

ABSTRACT. Hedging maturity, i.e., how far out in time hedging activities stretch, is an important yet under-theorized aspect of corporate risk management. In this article, we analyse firms’ hedging maturity decision and carry out a comprehensive empirical analysis. We develop three hypotheses to explain hedging maturity. The collateral hypothesis states that longer maturities are predicated on the availability of financial resources. The flexibility hypothesis holds that the ability to change operations or investment strategies at low cost is conducive to shorter maturities. The matching hypothesis argues that firms match their hedging maturity with the maturity of their debt and investment portfolios. Using handcollected data on derivative positions in the oil and gas industry, we find evidence consistent with all three hypotheses.

ABSTRACT. We obtain risk-neutral probabilities of the Brexit referendum using data from both the options and prediction markets. We then provide a risk-corrected measure of these probabilities using both non-parametric and parametric methods. While former correction marginally changes the risk-neutral probability, the effect of the latter depends on relative wealth calibration and risk preferences. We estimate subjective Brexit probabilities from past opinion polls and provide daily estimates of voting intention to leave from the BES survey. By comparing the subjective probabilities with our risk-corrected measures, our results show that both FX option and prediction market participants are likely to reveal moderate risk seeking preferences before the Brexit referendum.

ABSTRACT. This paper assesses the direction-of-change forecasts based on conditional variance from the Dynamic Nelson-Siegel model. Although the literature focuses on forecasting the level of yield curves, which is a difficult task, we propose forecasts for the direction-of-change of the yield curve returns. The results suggest that models with information of skewness and kurtosis of returns outperform the benchmark model, mainly in long maturities and short horizons of forecasts.

ABSTRACT. This paper develops a tractable endogenous network formation model of the interbank market. Due to liquidity shocks, banks face a trade-off between investing their resources in a liquid asset and a high-yield illiquid asset. The interbank market is modeled as a network. A link extended by one bank to another is interpreted as a credit line that the former bank can use to cover liquidity outflows. The central bank, by means of its standing facilities, lends resources to banks that are short in liquidity and borrows from institutions with liquidity surpluses at predetermined rates. These rates establish a corridor in which the interbank rate must lie. In this setting, we characterize the unique equilibrium of banks' liquidity holdings for any network. We then endogenize the network, via banks' decision of credit lines, and provide a sharp equilibrium characterization: every equilibrium network is a complete core-periphery graph. This characterization is consistent with empirically observed networks. Moreover, we introduce a trade-off for central banks when choosing the corridor rate: a narrower corridor implies more precise targeting of the interbank rate, which is important for the conduct of monetary policy. However, if we account for banks' linking decisions, this may lead to an equilibrium with a sparser network, where total liquidity holdings are higher, incurring an implicit cost since these funds could be invested in the more productive illiquid asset instead. We then provide conditions such that the central bank does not find optimal to minimize the interbank rate variance and study how it should react during moments of financial crises.

ABSTRACT. We ask whether the decisions of a rational and impartial judge can be distinguished from a coin toss. The question is inspired by the logic of the selection effect: Cases that have a clear outcome tend to be settled out of court. But that means that the cases that tend to go to court are often decided by small shocks on perceptions, the equivalent of a coin toss. That conclusion changes when judges are biased in their prior beliefs. In that event, outcomes will be determined exclusively by the judge's prior beliefs. Either way the outcome of the case tend to be decided as if it is unaffected by the merits of the arguments presented in court. However, taken to its logical conclusion, the selection effect leads to something we refer to as the paradox of open methods: if rational judges decide to evaluate the merits of the cases that go to court, they will come to regret this as wasteful effort and strictly prefer to ignore the merits of the case. But if judges ignore the merits of the case, and this becomes generally known and taken into account by the litigants, judges will come to regret doing so and end up strictly preferring to attend to the merits of cases. This paradoxical seesaw can only be resolved in a full game-theoretical model of strategic interaction between the judge and the litigants. We refer to this game as the Judging Game. The strength of the selection effect and the fraction of judges who evaluate the merits of cases can thus be determined by the equilibrium of the Judging Game.

ABSTRACT. This paper investigates how static distortions present in the sectoral goods market affect growth incentives in open economies. In the model, capital accumulation and exogenous technology adoption jointly generate output growth. Static distortions distance the economy from the actual productivity profile across sectors changing the country specific real rate of return on capital accumulation in the world balanced growth path. We calibrate the model for the Mexican economy between 1995-2011, a period of stagnation of per capita income. Using the World Input-Output Database we retrieve distortions directly from data through statistics implied by the model. Counterfactual exercises show that aggregate losses could be as high as 54%.

ABSTRACT. What is the optimal policy regarding patents? To answer this question, we develop a model of dynamic competition and incorporate this model into a tractable endogenous growth model. Within this framework, patent laws function as an institutional means to manipulate the degree of competitiveness in industries. Patents reward the creation of novel technologies, which might increase the equilibrium stock of technologies. Still, they slow down growth in process productivity in each technology, which is driven by competition. Depending on the parametrization, the optimal policy might or might not feature patent protection at all. However, when we calibrate the model with parameters consistent with the empirical evidence, the model suggests that the optimal policy is that patents should last approximately fifteen to twenty years.

ABSTRACT. We study how a network structure can determine the evolution of an epidemic. For that, we use a Susceptible-Infected-Recovered (SIR) macroeconomic model in the presence of a network environment. Network models have been important in the job search discussion. In an epidemiological model, the network structure is one of the main causes of the spread of the disease. Intuitively, more connected people in the social circle are the main vector of the virus. On the other hand, those people with few connections should be less exposed to the disease. We study the behavior of the pandemic for different types of network, from a low connected one to a high connected one. We find exactly the expected relationship: because more connected economies (economies with a higher average number of links) spread the virus faster, they face harder consequences in a pandemic scenario, such as a greater fall on aggregate consumption and hours worked due to both the higher number of deaths and the susceptible agents’ higher attempt to stay at home and avoid physical contacts. Susceptible agents are more cautious in regard to the decision of their level of consumption and hours worked as the economy becomes more socially connected, once the consequences of leaving home to consume or to work are harder in the higher connected economy because of its higher number of infected people.