Esta tese é composta por três capítulos: dois dedicados a compras públicas e um à economia política. O primeiro capítulo examina as aquisições realizadas por municípios brasileiros durante a pandemia de COVID-19, quando a legislação flexibilizou os limites para dispensa de licitação. Os dados mostram um aumento significativo no uso de dispensas, especialmente para bens relacionados ao combate da pandemia e em municípios com pior gestão fiscal. Curiosamente, não há evidências de que o aumento nas dispensas tenha levado à seleção de fornecedores associados ao favoritismo ou à corrupção, nem a preços mais altos para produtos relacionados à crise. A maior flexibilidade permitiu que as unidades compradoras acessassem mais fornecedores fora dos limites municipais. Discutimos se essa medida atenuou os impactos na mortalidade em municípios com restrições financeiras. O segundo capítulo analisa uma política brasileira que estabelece a reserva de lotes em leilões exclusivamente para a competição de pequenas e médias empresas (PMEs). A principal conclusão é que o uso de lotes exclusivos para PMEs reduz a concorrência nos leilões. Investigamos também o impacto da política sobre os preços e discutimos a adesão à política. Por fim, no terceiro capítulo, ampliamos tentativas manuais anteriores de classificar relatórios de auditoria de corrupção, utilizando um Modelo de Linguagem de Grande Escala (LLM) para codificar relatórios de municípios brasileiros. Em seguida, utilizamos nossos dados para reavaliar o impacto dos incentivos à reeleição sobre a corrupção. Encontramos algumas evidências de que os incentivos à reeleição reduzem a corrupção, corroborando achados anteriores. No entanto, os efeitos são menores e estatisticamente significativos apenas para uma das três variáveis de corrupção analisadas. Apresentamos explicações alternativas para esses resultados.

No contexto emírico do Brasil, esta tese investiga o impacto de eventos climáticos extremos em três aspectos chave do desenvolvimento: saúde, violência de gênero e educação, focando em como as mudanças climáticas agravam as desigualdades existentes ao afetar desproporcionalmente populações mais vulneráveis. No primeiro capítulo, examina-se a relação entre calor e mortalidade. A análise revela que, enquanto as mortes relacionadas ao calor entre os não idosos são causadas, sobretudo, por doenças cardiorrespiratórias, o risco de morte entre os idosos aumenta para um grupo mais amplo de doenças. Além disso, os resultados indicam que as mortes induzidas pelo calor extremo ocorrem em um intervalo curto de tempo, evidenciando a necessidade de respostas rápidas por parte do sistema de saúde durante ondas de calor. O segundo capítulo explora os efeitos de secas sobre a violência doméstica, comparando áreas urbanas e rurais. Os resultados mostram que as secas de long prazo aumentam significativamente as taxas de violência doméstica em municípios rurais, onde as perdas econômicas provenientes da escassez de água são mais expressivas. Em contraste, em municípios urbanos, onde as secas têm impactos econômicos menores, secas prolongadas não estão associadas a um aumento da violência doméstica. O terceiro capítulo investiga os impactos de curto prazo de chuvas intensas no performance escolar no Sul do Brasil. Os resultados sugerem que episódios curtos de chuvas intensas reduzem as notas dos alunos em uma exame nacional padronizado, com os grupos mais vulneráveis— studantes não-brancos, de origem socioeconômica mais baixa e com pior desempenho escolar prévio—sendo os mais afetados. 

A sazonalidade bancária tem sido amplamente examinada na literatura, no entanto, a escassez de dados granulares tem dificultado a análise em nível de agência. Neste estudo, investigo a sazonalidade em nível de agência para identificar picos de crédito, definidos como o mês em que uma agência específica apresenta seu maior volume médio de empréstimos. Notavelmente, agências dentro do mesmo banco em um município dado demonstram diferentes ’meses principais’, sugerindo uma forma de especialização da agência adaptada à sua clientela. No primeiro artigo, exploro o impacto da sazonalidade da agência na transmissão da política monetária, examinando como as flutuações na taxa de juros de referência afetam o volume de empréstimos de forma diferente durante os ’meses principais’ das agências. Os resultados sugerem que a política monetária é menos eficaz durante esses períodos. Há evidências de que o relacionamento com os clientes ou comportamentos semelhantes impedem que as agências transmitam as mudanças na taxa de juros nos meses que são mais importantes para elas. No segundo artigo, examino o papel dos bancos públicos em períodos eleitorais, mostrando evidências de que políticos exploram a sazonalidade das agências para aumentar ainda mais o crédito nas eleições. O uso dos ’meses principais’ adiciona uma nova camada de identificação e sugere que o crédito é realocado de municípios politicamente não atrativos para municípios politicamente atrativos. Finalmente, no terceiro artigo, utilizo aprendizado de máquina para prever e descrever os ’meses principais’ com base em dados detalhados de balanço, para os cinco maiores bancos do Brasil. Os resultados reforçam a ideia de que as agências adaptam suas práticas de empréstimo aos mercados locais. Coletivamente, esses resultados oferecem insights valiosos para formuladores de políticas e profissionais sobre a dinâmica do mercado bancário.

This dissertation consists of three essays concerning inflation forecasting, taking the Brazilian case as an application. In the first essay, we examine the effectiveness of several forecasting methods for predicting inflation, focusing on aggregating disaggregated forecasts. We consider different disaggregation levels for inflation and employ a range of traditional time series techniques, as well as linear and nonlinear machine learning (ML) models that deal with a larger number of predictors. For many forecast horizons, aggregation of disaggregated forecasts performs just as well as survey-based expectations and models generating forecasts directly from the aggregate. Overall, ML methods outperform traditional time series models in predictive accuracy, with outstanding performance in forecasting disaggregates. In our second essay, we investigate the potential benefits of combining individual inflation forecasts by proposing a time-varying bias correction for the average forecast. Our analysis includes estimations using both rolling windows and state-space models that use the recursiveness of the Kalman filter. We achieve good forecast performance for models based on small rolling windows for shorter and intermediate forecast horizons, while a state-space model performs slightly worse than procedures based on rolling windows. In the third essay, we use supervised learning to generate forward-looking indexes based on tweets and news articles for accumulated inflation and investigate whether these indexes can improve inflation forecasting performance. Our results indicate that news-based indexes provide significant predictive gains, particularly for 3- and 12-month-ahead horizons. These findings suggest that incorporating more information sources than just expectations based on experts’ opinions can lead to more accurate forecasts.

This thesis consists of 3 chapters in development economics that relate natural disasters and environmental quality to political engagement and social capital. In the first chapter, we show that natural disasters can lead to punishment of incumbents. Interestingly, such punishment is the result of some heterogeneity in political participation. In local elections, turnout was lower in regions where the incumbent belonged to the party in power at the country level (DPJ), while turnout was higher in regions where the incumbent belonged to the main rival (LDP). As a result, the ruling party suffered a loss in these elections. The possible reason for this heterogeneity lies in the population’s disappointment with the DPJ. In addition, it shows a further heterogeneity in regards to the level of social capital. Whereas it is related to higher political participation, the associated higher community resilience possibly led to different voting behavior. The second chapter proposes a theoretical framework to link the empirical and theoretical literatures on the influence of environmental quality and risk on the emergence of cooperative behavior. Consistent with the empirical literature, it is shown that depending on the relationship between the environment and the club good and individuals’ beliefs about the cooperative behavior of others, the higher the probability of bad times, the greater the propensity of individuals to engage in collective action within a community. Finally, the third chapter examines the impact of natural disasters on the formation of social capital and its long-term persistence. Using data on ancient earthquakes in Japan, it is shown that people living in rural Japanese cities that were strongly hit in the past currently exhibit higher levels of trust and political engagement.

This thesis is composed of 3 chapters in development economics, all of which relate to the role of public policy either in promoting sustainable policies for long-run maintenance of natural resources, or meeting citizens’ needs through efficient service delivery or questioning the accountability of current political institutions. The first paper shows that, by adopting a technology that incorporates sustainable agricultural practices, farmers have positive dynamic effects on productivity and climate resilience.We provide evidence of the mechanism of these dynamic effects: soil improvement. The second one presents the results of an RCT designed to test whether a managerial intervention in health centers in Mozambique can reduce coordination failures, increasing HIV patients’ retention in care and quality of care provided. Finally, the third one look at the large street protests that took place in Brazil in 2013 to analyze whether protests can work as an accountability mechanism for elected politicians. Using Twitter historical data, we create a measure of protest intensity and how noisy protesters’ demands were in each municipality. We present evidence that protests may work as accountability mechanism only if messages sent by protesters are sharp and clear, and politicians face reelection incentives.

This thesis is composed of three papers on financial econometrics. The first two papers explore the relation between intraday equity market returns and implied volatility, represented by the CBOE Volatility Index (VIX). In both papers, we estimate one-minute-ahead forecasts using rolling windows within a day. In the first paper, the estimates indicate that our volatility factor models outperform traditional benchmarks at high frequency time-series analysis, even when excluding crisis periods. We also find that the model has a better out-of-sample performance at days without macroeconomic announcements. Interestingly, these results are amplified when we remove the crisis period. The second paper proposes a machine learning modeling approach to this forecasting exercise. We implement a minute-by-minute rolling window intraday estimation method using two nonlinear models: Long-Short-Term Memory (LSTM) neural networks and Random Forests (RF). Our estimations show that the VIX is the strongest candidate predictor for intraday market returns in our analysis, especially when implemented through the LSTM model. This model also improves significantly the performance of the lagged market return as predictive variable. Finally, the third paper explores a multivariate extension of the FarmPredict method, by combining factor-augmented vector autoregressive (FAVAR) and sparse models in a high-dimensional environment. Using a three-stage procedure, we estimate and forecast factors and its  oadings, which can be observed, unobserved, or both, as well as a weakly sparse idiosyncratic structure. We provide an application of this methodology to a panel of daily realized volatilities. Finally, the accuracy of the stepwise method indicates improvements of this forecasting method when compared to consolidated benchmarks.

Nowcasting in economics is the prediction of the present, the recent past or even the prediction of the very near future of a certain indicator. Generally, a nowcast model is useful when the value of a target variable is released with a significant delay with respect to its reference period and/or when its value gets notably revised over time and stabilizes only after a while. In this thesis, we develop and analyze several Nowcasting methods using high-dimensional (big) data in different contexts: from the forecasting of economic series to the nowcast of COVID-19. In one of our studies, we compare the performance of different Machine Learning algorithms with more naive models in predicting many economic variables in real-time and we show that, most of the time, Machine Learning beats benchmark models. Then, in the rest of our exercises, we combine several nowcasting techniques with a big dataset (including high-frequency variables, such as Google Trends) in order to track the pandemic in Brazil, showing that we were able to nowcast the true numbers of deaths and cases way before they got available to everyone.

This thesis is composed of two articles. In the first one, we propose and implement a new index of vulnerability which is based on a structural time-varying bayesian VAR with a block-exogeneity hypothesis for a given pair of a large economy and a small open economy. The index is based on the sum of the responses of the small open economy to shocks in the large economy over time, thus allowing us to disentangle and measure the source of the shock, impact variables and duration of the co-movement or vulnerability. Our index suggests that the business cycle co-movement is led primarily by country-pair characteristics, but decoupling trends can be observed in a considerable number of country-pairs, specially at long term windows. We provide an application of this approach to a global banks framework - which allows us to measure some yet unmeasured theoretical mechanisms. Using a sample of developed and developing countries, we find no evidence of the prevalence of such mechanisms in business cycle co-movement. In the second article, we study how tax incentives impact the firm’s behavior and choices in the labor market. Do tax incentives affect wages? Do these incentives alter the composition of a firm’s labor force? And what about its size? To answer these questions, we merge RAIS - a linked Brazilian employer-employee dataset - with a novel, firm-level, dataset on two fiscal incentives programs in the state of Espírito Santo - Invest-ES and Compete-ES. By using a differences-in-differences estimator of intertemporal treatment effects, we study the impacts of these programs on municipality-level and, for the first time, firm-level variables. We do not find statistically significant impact of fiscal incentives on any relevant municipality-level variables. On firm-level variables, however, our results point to higher migration from other municipalities in the state of Espírito Santo and temporary growth in the number of jobs. Regarding wages and educational levels of the labor force, no statistically significant impact was documented.

This dissertation presents three chapters addressing issues pertaining to monetary policy. Chapter 1 evaluates the problem of conducting monetary policy with risky assets in a simple neo-Wicksellian monetary model. I show that monetary policy’s power w.r.t prices and inflation reduces as it can only be conditionally active in the presence of policy-asset risk. Moreover, uncompensated risk premium induces an inflationary bias, as well as default probability and inflation are positively correlated, the same sign of empirical correlations usually found. These results constitute a novel argument in favor of a more hawkish stance in case of a fiscal or political crisis, which helps to explain "monetary policy conservatism" in risky economies. Chapter 2 endogenizes policy-asset risk as a fiscal risk and studies its transmission. I lay out a two-agent New-Keynesian (TANK) model with endogenous fiscal limits in which the central bank operates through defaultable bonds, and then calibrate it to a large emerging economy, Brazil. I find that by ignoring policy-asset risk the central bank reinforces the unpleasant coincidence of higher inflation, real, and nominal interest rates in the equilibrium distribution of the model, what emerges as the result of endogenous expectations of a severe recession in case of default. Additionally, accommodating policy-asset risk induces positive correlation between default risk and inflation. From a policy perspective, these results raise serious concerns about the evaluation of monetary policy stance in default-risky economies, while shed new light on the long-standing discussion about why policy rates have been exceptionally high in Brazil after the Real Plan. Finally, Chapter 3 responds to a recent controversy on the actual presence of a real interest rate transmission channel in New-Keynesian models, as the addition of endogenous capital is consistent with real rates moving in any direction after a monetary shock. I show that this identification problem can be circumvented by the inclusion of another ingredient as prevalent as capital itself in middle-scale models: interest-rate smoothing.

Esta tese estuda a interação entre escolaridade e mercado de trabalho a partir de um fenômeno ainda pouco explorado na literatura: a venda de milhares de diplomas de ensino médio e de graduação no Brasil. O primeiro capítulo descreve a nova base de dados construída, composta por milhares de diplomas de Ensino Médio (EM) da modalidade de Educação de Jovens e Adultos (EJA), expedidos para pessoas que nunca concluíram seus estudos. A análise mostra que a compra do diploma resulta em ganhos salariais consistentes, com rendimentos 5% a 8% maiores em comparação com os indivíduos sem diploma de ensino médio durante o período de análise. Considerando apenas o seu custo pecuniário, estima-se que a compra de um diploma tenha uma taxa de retorno superior a 100%, ou um valor presente entre dois e seis meses de salário. Esses resultados são consistentes com: (i) uma possível sinalização de habilidade em mercados com informações imperfeitas, (ii) aprendizagem assimétrica entre o empregador e outras firmas do mercado, ou ainda (iii) fenômenos de learning-by-doing. O segundo capítulo explora o mesmo banco de dados para investigar a acumulação de capital humano no EM-EJA. A análise mostra que os indivíduos que concluíram de forma regular o EM-EJA desfrutam de certo prêmio salarial em relação aos que não completaram o EM, mas não em relação àqueles que compraram diploma. Tomados em conjunto, os resultados dos dois primeiros capítulos indicam que os benefícios do EMEJA estão relacionados a um efeito-diploma e não à acumulação de capital humano. O último capítulo explora um novo banco de dados de diplomas universitários expedidos entre 2011 e 2016 que foram considerados irregulares e cancelados após investigação judicial. Apresentamos evidências preliminares  sobre ganhos salariais e estabelecemos questões de pesquisa adicionais que podem ser estudadas com este banco de dados. De maneira geral, a tese reforça a importância dos diplomas como dispositivos de sinalização para o mercado de trabalho 

Esta tese é composta por três capítulos que enfocam o crescimento da China como um experimento quasi-natural de forma a avaliar os efeitos de choques de comércio exterior sobre a economia política da política comercial e sobre a dinâmica do mercado de trabalho e desigualdade de salários no Brasil. No primeiro capítulo, utilizamos evidência sobre diferenciais de exposição a esse choque da China entre mercados de trabalho locais para estimar seu efeito em indicadores do mercado de trabalho brasileiro, em particular em medidas de desigualdade de rendimentos. Primeiro, encontramos que o choque de demanda por exportações diminuiu a desigualdade de salários no setor de bens comercializáveis, sobretudo por meio do componente entre firmas da dispersão salarial, e apresentamos evidências de que essa redução parece causada por uma mudança no comportamento das firmas, e pode estar relacionado com uma redução no prêmio salarial de firmas exportadoras. Em seguida, estimamos um modelo baseado em Helpman et al. (2016), e exploramos diferenças setoriais no choque de demanda externa para realizar exercícios contrafactuais que corroboram a hipótese de que esse choque pode explicar parte da redução agregada no prêmio salarial de firmas exportadoras e na dispersão de salários. No segundo capítulo, desenvolvemos uma versão do modelo dinâmico de Caliendo et al. (2019) de modo a estimar os efeitos do duplo choque da China na dinâmica setorial do emprego no Brasil. Mostramos que ambos os choques levam à contração da maioria dos setores de manufaturas, e expansão da maioria dos setores de serviços, mas os efeitos de equilíbrio geral dos choques são modestos, especialmente quando comparados a um contrafactual alternativo no qual a produtividade brasileira nos setores primários aumenta. Estendemos o modelo para incluir dois tipos de trabalho, de alta e baixa qualificação; resultados apontam para efeitos distributivos pequenos, mas consistentes com resultados em forma reduzida obtidos no primeiro capítulo. No capítulo final, construímos uma base de dados inédita sobre características de associações setoriais brasileiras, com o intuito de investigar se os setores com maior capacidade de organização política são capazes de obter maior proteção contra competidores estrangeiros. Usamos variação na penetração de importações como uma medida da necessidade de proteção comercial, e para lidar com a endogeneidade nessa medida usamos um instrumento baseado no choque de importações da China. A evidência sugere que setores com maiores sindicatos patronais são capazes de obter maior proteção comercial, em particular por meio de licenciamento não-automático; as estimativas sugerem que esse efeito é mais alto quando a penetração de importações aumenta mais intensamente, o que é interpretado como um aumento na necessidade de medidas de proteção.

Sequential learning problems are common in several fields of research and practical applications. Examples include dynamic pricing and assortment, design of auctions and incentives and permeate a large number of sequential treatment experiments. In this essay, we extend one of the most popular learning solutions, the t-greedy heuristics, to high-dimensional contexts considering a conservative directive. We do this by allocating part of the time the original rule uses to adopt completely new actions to a more focused search in a restrictive set of promising actions. The resulting rule might be useful for practical applications that still values surprises, although at a decreasing rate, while also has restrictions on the adoption of unusual actions. With high probability, we find reasonable bounds for the cumulative regret of a conservative high-dimensional decaying t-greedy rule. Also, we provide a lower bound for the cardinality of the set of viable actions that implies in an improved regret bound for the conservative version when compared to its non-conservative counterpart. Additionally, we show that end-users have sufficient flexibility when establishing how much safety they want, since it can be tuned without impacting theoretical properties. We illustrate our proposal both in a simulation exercise and using a real dataset. The second essay studies deterministic treatment effects when the assignment rule is both more complex than traditional ones and unknown to the public perhaps, among many possible causes, due to ethical reasons, to avoid data manipulation or unnecessary competition. More specifically, sticking to the well-known sharp RDD methodology, we circumvent the lack of knowledge of true cutoffs by employing a forest of classification trees which also uses sequential learning, as in the last essay, to guarantee that, asymptotically, the true unknown assignment rule is correctly identified. The tree structure also turns out to be suitable if the program’s rule is more sophisticated than traditional univariate ones. Motivated by real world examples, we show in this essay that, with high probability and based on reasonable assumptions, it is possible to consistently estimate treatment effects under this setup. For practical implementation we propose an algorithm that not only sheds light on the previously unknown assignment rule but also is capable to robustly estimate treatment effects regarding different specifications imputed by end-users. Moreover, we exemplify the benefits of our methodology by employing it on part of the Chilean P900 school assistance program, which proves to be suitable for our framework.

This thesis is composed by two chapters. The first chapter shows that the presence of lead-lag effects in the US equity market is a broader phenomenon than previously found in the literature and is associated with the existence of a strong one-day factor momentum. Lead-lag effects are present whenever stocks are exposed to the same common risk factor, holding for almost 100 factors on a daily frequency. This phenomenon is not explained by the previously reported industry, large-cap to small-cap and other lead-lag effects. One-day factor momentum is directly related to the existence of factor-based stock cross-autocovariance and is present both in the cross-section and the time series. One-day factor momentum is profitable after trading costs and does not present crashes. One-month factor momentum is subsumed by one-day factor momentum with negative alpha in spanning tests. The relevance of the one-day effect is confirmed with machine learning techniques. Short-term reversals in stocks also become stronger after we control for this factor-based cross-autocovariance pattern. The second chapter shows how factor momentum impacts the performance of standard short-term single-equity reversal strategies in the US equity market. Significant benefits in performance can be achieved if the effects of factor momentum is considered in the construction of reversal strategies. Standard short-term reversal strategies have a negative exposure to factor momentum since they sell winner stocks that on average are more exposed to the winner factors and buy loser stocks that on average are more exposed to loser factors. The best way to neutralize this effect that drags down short-term reversal performance is to hedge stocks exposures simultaneously to a very large set of factors. For instance, hedging only with the 3 Fama-French factors does not eliminate the exposure to factor momentum. Sorting stocks using residual returns is not as efficient as sorting on total returns as it does not completely neutralize the negative exposure to factor momentum. We propose a fully-hedged reversal strategy that, differently from conventional short-term reversal strategies, is profitable after trading costs, that do not present crashes, that has Sharpe ratio 2.5 times higher than the conventional reversal strategies and that is profitable even if we restrict our sample to only large-cap stocks.

This doctoral thesis, is comprised of three chapters. In the first chapter, we explore a discontinuity in eligibility for financial aid for private higher education in Brazil to investigate: (i) if students in the lower end of the ability distribution change their behavior in response to a policy designed to subsidize access to private higher education, and (ii) how students react when they are given the chance to choose between public tuition-free institutions and subsidized access to private higher education. We show that eligibility for financial aid increases students’ likelihood of enrolling and persisting in higher education. We also find that eligible students are less likely to enroll in public tuition-free institutions, with no clear impact on the choice of quality at the program level. In the second chapter, we develop a structural discrete choice model of demand to investigate the determinants of major choice. We use this model and the expansion of a government-funded student credit program in Brazil to evaluate how the availability of credit impacts major choice when — as in Brazil — tuition varies at the major level. We find that tuition and expected labor market returns are the main factors determining students’ choice between different majors.We also find that, when student credit is available, students — especially lower income — are less sensitive to price variations at the major level. In the final chapter, we investigate whether water shortages caused by an extreme climate event impact educational performance. We answer this question exploring the consequences of a water rationing policy that affected some neighborhoods in Brazil’s Distrito Federal. Comparing the academic performance of students enrolled in schools located in neighborhoods affected by the rationing against the performance of students enrolled in non-affected neighborhoods, we find that water rationing has a negative and significant impact on students’ performance. In particular, we show that the impact is significantly stronger for students enrolled in schools with poor infrastructure.

This dissertation explores the impact of transportation infrastructure on economic development and the effects of agrarian elites’ political power on investments in education. The first chapter documents the impact of the Brazilian railway network on structural transformation between the late nineteenth and middle twentieth century. We exploit variation induced by geographic location, where municipalities near the least-cost routes were more likely to be connected to railway system, to identify the causal effects of railroads on structural transformation. We show that the expansion of the transportation infrastructure shifts the workers from agriculture to manufacturing. We provide evidence that market integration and the adoption of new technologies by manufacturing firms were two important mechanisms that explain the change in the occupational structure of the economy. The second chapter explores the persistence of the impact of Brazil’s railway network on long-run economic development between 1950 and 2010. We exploit variation induced by geographic location, where municipalities near the least-cost routes were more likely to be connected to railway system, to identify the causal effects of railroads on economic activity. We show that, despite the decline of the railway network post-1950s, the effect on economic development persisted. Our findings suggest that agglomeration and urbanization were key drivers of economic activity persistence. In the third chapter, we study the relationship between the political power of agrarian elites and the spread of mass schooling in the early 20th century in Brazil. We use a novel dataset on the occupational structure of the voting elites in 1905 and historical censuses to test whether places, where more voters belonged to the agriculture elite, invested less in schooling. We find that municipalities with a higher share of agriculture voters have a lower literacy rate in 1920 and these effects persist until the 1970s. In the long-run, municipalities that had a higher share of agriculture voters have fewer years of schooling and lower income per capita. We provide evidence that the supply of educational inputs is the main mechanism that explains the long-term persistence.

The thesis is composed of two papers on empirical finance. The first paper focuses on FX markets. It constructs measures of term structure slope changes for the US and other G10 countries using short-term interest futures at different horizons. These changes in term structure slopes have immediate impact on currency returns but also a strong delayed effect over the following weeks, implying that currencies are predictable. The paper presents strong evidence of out-of-sample short-term predictability for both individual currencies and currency portfolios. These findings represent empirical evidence of delayed currency market reaction to information in interest rates markets. The second paper focuses on equity returns in the US.  It proposes a novel forecasting measure for aggregate market returns that solely uses cross-sectional information on CAPM betas simple dispersion measures. The latent factor structure of portfolio or individual stock betas has strong predictive power both in and out-of-sample, and are robust to different estimation windows. Unlike most measures in the literature, it is not a price or valuation-based ratio. This novel measures also vary with the business cycle and correlate with other commonly used forecasting variable such as dividend-to-price or consumption-to-wealth ratios, but provide explanatory power above and beyond these standard predictors.

This thesis is comprised of three essays. The first two investigate the relationship between households per capita income and sectoral expenditure shares both in times series and in cross-section in the postwar US. The first uses a partial approach to estimate the rise of consumption (income) dispersion and income effects in the US from 1980 to 2010. We show that income effects are heterogeneous across households grouped by income quintiles and then consumption dispersion correlates the two main driving forces of structural change (price and income effects) in accounting for the magnitude of structural change in the shares of consumption expenditure in the US over this period. The second extends a canonical Bewley-Aiyagari model in continuous time embedded with a two-sector environment to depict quantitatively three empirical regularities in the postwar US (relative price of goods falls and expenditure shares of goods falls systematically with per capita income, both in times series and in cross-section) without departing from benchmark Stone- Geary preferences. We assess the importance of changes in income and relative prices for structural change in the shares of consumption expenditure in the postwar US and conclude they are nearly equivalent forces. We reinforce that reconciling these three main empirical regularities in the postwar US calls for a growth theory that accommodates long-run demand and supply drivers of structural change. Finally, the third essay uses a unique panel dataset with individual-level administrative records of credit transactions, program benefits, individual demographics and features of labor contracts to study how consumers respond to a liquidity shock arising from withdrawals releases from inactive accounts of the Guarantee Fund for Time of Service (FGTS) in Brazil in 2017. Using a difference-in-differences identification design, we find consumption rose and total debt declined after the announcement: during up to twelve subsequent months, for each $1 of program benefit, consumers on average increased consumption spending by $ 0.53 - 25 percent of which occurs during the announcement window - and total debt declined by $0.07, specially in payroll debt. Consumption response occurred mostly via credit card spending, but evidence of debt-financed durables was also found. Indebted consumers used short-term liquidity in debt modalities (overdraft debt and credit card debt) in addition to credit card spending to smooth consumption. Constrained consumers, measured as young or old, showed stronger consumption responses.

The thesis is composed of three essays. The first designs a DSGE model based on Gertler and Karadi (2011) to study the effects of the adoption of negative interest rate policies along with liquidity intervention, in a scenario where the ZLB is transferred to private banks instead of central banks. We show that, during a recession, if banks do not pass along negative rates to depositors in an environment of heavy liquidity injection by the CB, the recovery is slower. The second essay uses the same model in a simpler setting to study how the adoption of central bank digital currencies (CBDCs) might reestablish the traditional monetary policy transmission under negative interest rates, and study the responses of the economy under such a regime to monetary policy shocks. The third essay tries different models for the forecast of medium-term output growth. We use new methods such as adaLASSO and Random Forest, along with a very large data set of regressors, in order to improve accuracy over traditional model long term forecasting such as autoregressions and DSGE models, which have a very good track record. We show that Random Forest is able to better predict output growth over que two year horizon, but has mixed results in forecasting trend GDP growth and the output gap.