Submitted by Maike Costa (maiksebas@gmail.com) on 2016-04-13T12:31:42Z No. of bitstreams: 1 arquivo total.pdf: 1235998 bytes, checksum: e17b680f3c5ad63d12b8e6e1116cfa50 (MD5) / Made available in DSpace on 2016-04-13T12:31:42Z (GMT). No. of bitstreams: 1 arquivo total.pdf: 1235998 bytes, checksum: e17b680f3c5ad63d12b8e6e1116cfa50 (MD5) Previous issue date: 2015-03-30 / This dissertation encompasses three chapters that study Child Labor. Below are the individual abstracts for each chapter. Chapter 1: Child labor and the wealth paradox: the role of Altruistic Parents Using data from Pakistan and quantile regression techniques, we study the effect of family wealth on the utilization of child labor. We find evidence of a positive relationship between land wealth and child labor only for children in the upper quantiles of the distribution. We hypothesize that the so-called "Wealth Paradox" in child labor documented elsewhere in the literature is driven by parental preferences. Chapter 2: Child labor and household wealth: an analysis for rural Brazil This chapter studies the effect of family wealth on the utilization of child labor in rural areas of Brazil. We using data from PNAD 2012 and the Censored Quantile Instrumental Variable (CQIV) which captures heterogeneity across the distribution of hours worked, and it deals with the problems of censorship and endogeneity in the data. We find evidences of a negative relationship between land wealth and child labor only for children in the lower quantiles of the distribution. On the other hand, at the median and upper quantiles we find a non-linear relationship, supporting the hypothesis of " U-inverted ". We need to highlight that the turning point is bigger at the upper quantile, where families have lower level of altruism. In general, the results indicate that the preferences of the parents are the primary determinant of child labor. Chapter 3: Child labor in urban Brazil: what is the role of the family structure? The aim of this chapter is to investigate the role of single parents on child labor in urban Brazil. We use data provided by Brazilian Demographic Census of 2010 and the models to determine the probability of working (Probit, IV-Probit and Bivariate Probit) and the Yun’s decomposition to capture the differences at the probability of child labor attributed to the difference in behavior between single-parent families, headed by the mother, and two-parent, headed by his father. The results show that boys, age 15 whose parents have low level of education are more likely to work. We also found evidences that children in single-parent homes are more likely to work when compared with children in two-parent households in the father’s responsibility, noting that the most vulnerable scenery for the child is to live in a single parent home with no widowed mother. And the difference in child labor between the two groups of families is mainly due to their unobserved behaviors. / Esta tese compreende 3 capítulos sobre o tema Trabalho Infantil. Abaixo seguem-se os resumos individuais de cada capítulo. Capítulo 1: Trabalho infantil, paradoxo da riqueza e altruísmo: o caso do Paquistão Utilizando dados do Paquistão e a técnicas de regressão quantílica, esse capítulo analisa o efeito da riqueza familiar na utilização de trabalho infantil. As evidências apontam uma relação positiva entre riqueza da terra e trabalho infantil apenas nas crianças que estão no quantil superior da distribuição de horas trabalhadas. Dessa forma, a hipótese derivada desse estudo é de que o chamado “paradoxo da riqueza” do trabalho infantil documentado na literatura é impulsionado por preferências dos pais. Capítulo 2: Trabalho infantil e riqueza familiar: uma análise para o Brasil rural O capítulo investiga o efeito da riqueza familiar sobre o trabalho infantil no meio rural do Brasil. Para tanto, foram utilizados dados da PNAD 2012 e o Censored Quantile Instrumental Variable (CQIV), que permite captar heterogeneidades ao longo da distribuição de horas trabalhadas, e, ainda, lida com os problemas de censura e endogeneidade nos dados. Os resultados mostraram uma relação negativa entre riqueza, medida pelo tamanho da terra, e trabalho infantil no quantil inferior de horas de trabalho infantil, enquanto nos quantis médio e superior, uma relação não linear, corroborando a hipótese do “U invertido”. Destaca-se que o turning point é maior no quantil superior, onde as famílias tem menor nível de altruísmo. Em geral, os resultados apontam as preferências dos pais como principal determinante do trabalho infantil. Capítulo 3: Trabalho infantil no Brasil urbano: qual a importância da estrutura familiar? O capítulo investiga a importância da estrutura familiar na determinação do trabalho infantil no Brasil urbano. Para tanto, foram utilizados dados do Censo Demográfico de 2010 e modelos de determinação da probabilidade de trabalhar (Probit, IV-Probit e Probit Bivariado) e a decomposição de Yun para captar o diferencial de probabilidade de trabalho infantil atribuído à diferença de comportamento entre os tipos de famílias. Os resultados mostraram que meninos, com 15 anos de idade e cujo pai (mãe) não tem instrução são mais propensos à entrada precoce no mercado de trabalho. Também foram achadas evidências que crianças em lares monoparentais têm maior chance de trabalhar quando comparadas com crianças em domicílios biparentais sob responsabilidade do pai, destacando que o cenário de maior vulnerabilidade para a criança é viver em um lar monoparental com mãe não viúva. E, ainda, a diferença de probabilidade de trabalho infantil entre os grupos de análise é explicada principalmente por diferenças de comportamento entre os tipos de família.

Cette thèse de doctorat porte sur la dynamique de l’emploi dans les entreprises et sur la relation entre la dynamique de l’emploi et l’innovation, avec une attention particulière portée sur les entreprises nouvellement créées. Cette thèse conceptualise théoriquement et analyse empiriquement les différents aspects de l’interaction complexe entre le changement technologique et la dynamique de l’emploi, en se concentrant sur les effets hétérogènes des différents types d’innovation sur la croissance de l’emploi. Compte tenu le rôle primordial joué par les nouvelles et jeunes entreprises dans le processus de destruction créatrice et leur apport à la création globale de l’emploi, cette thèse fournit une caractérisation de la contribution nette d’emplois des nouvelles entreprises dans un nombre important de pays, en utilisant des données micro-agrégées issues d’une nouvelle base de données. En outre, elle analyse comment un certain nombre de caractéristiques institutionnelles affectent la création nette d’emplois dans les start-ups, en se concentrant sur les effets hétérogènes des politiques sur les nouvelles entreprises et les entreprises déjà existantes. Cette thèse étudie enfin une caractéristique particulière des lois de distribution des taux de croissance de l’emploi, c’est-à-dire la volatilité de la croissance de l’emploi, que non seulement se révèle être une médiation cruciale des effets des politiques sur la création nette d’emplois, mais a aussi d’importantes implications à la fois micro- et macroéconomiques. / This doctoral thesis focuses on employment dynamics in firms, and on the relationship between employment dynamics and innovation, with a particular focus on the entry process. It conceptualizes theoretically and analyses empirically different aspects of the complex interaction between technical change and employment dynamics, focusing on the heterogeneous effects of different types of innovation on employment growth. In the light of the prominent role of newly-born firms in shaping the creative destruction process and contributing to overall job creation, this thesis provides a characterization of the net job contribution by surviving entrants across a significant number of countries. Using newly collected representative micro-aggregated data, it further analyses whether and how a number of institutional characteristics affect start-ups’ net job creation, focusing on the heterogeneous effects of policies on entrants and incumbents. This thesis finally characterizes a particular feature of the employment growth distributions – employment growth volatility – that not only proves to be crucially mediating the effects of policies on entrants’ net job creation, but also has important micro and macroeconomic implications.

全球金融危機之後，在許多國家嚴重赤字之下，各界皆相當關心財政整頓效果。本文參考 Lin(2016) 的三階段方法估計動態追蹤資料分量迴歸，並使用敘事法 (narrative approach) 所建構的財政整頓衝擊變數，分析 17 個 OECD 國家 1978-2007 年跨國追蹤資料，探討財政整頓對經濟成長的影響效果。實證結果顯示財政整頓對經濟成長為負向影響，實施財政整頓會抑制產出，對一國經濟造成損害，且在經濟情勢不佳時，該負向效果更明顯。

Esse trabalho apresenta um estudo da causalidade de Granger em Risco bivariado aplicado a séries temporais financeiras. Os eventos de risco, no caso de séries financeiras, estão relacionados com a avaliação do Valor em Risco das posições em ativos. Para isso, os modelos CaViaR, que fazem parte do grupo de modelos de Regressão Quantílica, foram utilizado para identificação desses eventos. Foram expostos os conceitos principais envolvidos da modelagem, assim como as definições necessárias para entendê-las. Através da análise da causalide de Granger em risco entre duas séries, podemos investigar se uma delas é capaz de prever a ocorrência de um valor extremo da outra. Foi realizada a análise de causalidade de Granger usual somente para como comparativo. / Quantile Regression, Value at Risk, CAViaR Model, Granger Causality, Granger Causality in Risk

A tese abordou questões relacionadas à economia da saúde, particularmente à visão econômica dos transplantes renais. Foi conduzida uma análise de custo-utilidade para verificar qual tratamento, transplante renal ou hemodiálise, possui menor razão de custo por anos de vida ajustados por qualidade. O resultado obtido corrobora as evidências internacionais, as quais indicam o transplante renal como estratégia mais custo-efetiva. No caso deste estudo, a razão de custo-utilidade para o transplante renal e hemodiálise foi de R$ 18.161,00/AVAQ e R$ 40.872,00/AVAQ, respectivamente. Apesar de o transplante renal ser uma estratégia dominante, a escassez de órgãos impede que essa estratégia seja amplamente utilizada, reduzindo, assim, os ganhos de eficiência na alocação dos recursos escassos. Nesse contexto, também foi alvo desta tese a questão da escassez de órgãos. Pelo levantamento feito, há tendência de aumento do desequilíbrio entre demanda e oferta de órgãos, visto que a primeira cresce rapidamente, enquanto a segunda mostra pequena tendência de crescimento. Assim, alternativas para contornar esse problema foram analisadas, especialmente as relacionadas a mudanças institucionais na lei de doação de órgãos. Entre elas, foi argüido que a lei de consentimento presumido seria a opção mais factível, por não ferir o pressuposto do altruísmo. Objetivando estimar quanto seria o eventual incremento na doação de órgãos por doador cadáver devido à lei de consentimento presumido, fez-se uso do ferramental da econometria da saúde, aplicando, para uma amostra de 34 países ao longo de cinco anos, o método de regressão quantílica para dados de painel. Os resultados obtidos nessa aplicação indicam que há benefício na adoção da lei de consentimento presumido, que tem um efeito positivo sobre a taxa de doação de órgãos, em torno de 21-26%, comparada à lei de consentimento informado. / The thesis broaches questions related to health economics, particularly the economic vision of renal transplants. A cost-utility analysis was conducted to assess which treatment, renal transplant or hemodialysis, has a lower cost rate per quality-adjusted life years. The result obtained corroborates the international evidence, which indicates renal transplant as the most cost effective strategy. In the case of this study, the cost-utility ratio for renal transplant and hemodialysis was US$ 11,157/QALY and US$ 25,110/QALY, respectively. In spite of renal transplant being the dominant strategy, the scarcity of organs hinders this strategy to be widely used, reducing in this way, the efficiency gain in the allocation of scarce resources. Within this context, the organ shortage was also a target issue of this thesis. Through the survey performed, there is a tendency towards the increase of unbalance between the demand and supply of organs, being that the first grows rapidly while the second shows small tendency towards growth. Within this context, the investigation target of this thesis was to look into possible alternatives to by-pass this problem, especially those related to institutional changes in the organ donation law. Among them, it was argued that the law of presumed consent would be the most feasible option, since it does not harm the presupposition of altruism. With the object of estimating what would be the eventual increase in organ donation, per cadaveric donor, due to the law of presumed consent, the health econometric tool of quantile regression method for panel data was used, applied to a sample of 34 countries during a five-year period. The results obtained in this application indicate that there is benefit in adopting the law of presumed consent, which has a positive effect on the organ donation rate, around 21 – 26%, compared to the law of informed consent.

Dans cette thèse, nous étudions des tests du type : (H0) : E [U | X] = 0 p.s. contre (H1) : P {E [U | X] = 0} < 1 où U est le résidu de la modélisation d'une variable Y en fonction de X. Dans ce cadre et pour plusieurs cas particuliers – significativité de variables, régression quantile, données fonctionnelles, modèle single-index –, nous proposons une statistique de test permettant d'obtenir des valeurs critiques issues d'une loi asymptotique pivotale. Dans chaque cas, nous donnons également une méthode de bootstrap appropriée pour les échantillons de petite taille. Nous montrons la consistance envers des alternatives locales – ou à la Pitman – des tests proposés, lorsque ce type d'alternative ne tend pas trop vite vers l'hypothèse nulle. À chaque fois, nous vérifions à partir de simulations sous l'hypothèse nulle et sous une séquence d'hypothèses alternatives que les résultats théoriques sont en accord avec la pratique. / In this thesis, we study test statistics of the form : (H0) : E [U | X] = 0 p.s. contre (H1) : P {E [U | X] = 0} < 1 where U is the residual of some Y modeling with respect to covariates X. In this setup and for several particular cases – significance, quantile regression, functional data, single-index model –, we introduce test statistics that have pivotal asymptotic critical values. For each case, we also give a suitable bootstrap procedure for small samples. We prove the consistency against local – or Pitman – alternatives for the proposed test statistics, when such an alternative does not get close to the null hypothesis too fast. Simulation studies are used to check the effectiveness of the theoretical results in applications.

The thesis examines impact of online user reviews, expressed in 5-star rating system, on demand for PC and video games. Used weekly data are collected from www.amazon.co.uk for period 2000 - 2012. The thesis contributes to discussion which takes place between supporters of superstar theory and supporters of long tail theory. Panel structure of the dataset referred to application of fixed effects model for estimation of basic and extended model. Both estimates proved negative influence of assigned number of stars on sales rank. Basic model estimated influence of one unit change of stars on rank by 9 positions, extended model estimated this change to be 10 positions. Quantile estimate also proved negative influence of stars on sales rank. The influence is strongest for the lowest quantile of dependent variable (tau = 0,05) representing 5 % best-selling titles, where one unit change of stars causes change of 35 positions in opposite direction. With increasing quantiles this influence decreases to zero. It proves that within the market for PC and video games online user reviews contribute to superstar effect.

En prévision numérique du temps, les modèles de prévision d'ensemble sont devenus un outil incontournable pour quantifier l'incertitude des prévisions et fournir des prévisions probabilistes. Malheureusement, ces modèles ne sont pas parfaits et une correction simultanée de leur biais et de leur dispersion est nécessaire.Cette thèse présente de nouvelles méthodes de post-traitement statistique des prévisions d'ensemble. Celles-ci ont pour particularité d'être basées sur les forêts aléatoires.Contrairement à la plupart des techniques usuelles, ces méthodes non-paramétriques permettent de prendre en compte la dynamique non-linéaire de l'atmosphère.Elles permettent aussi d'ajouter des covariables (autres variables météorologiques, variables temporelles, géographiques...) facilement et sélectionnent elles-mêmes les prédicteurs les plus utiles dans la régression. De plus, nous ne faisons aucune hypothèse sur la distribution de la variable à traiter. Cette nouvelle approche surpasse les méthodes existantes pour des variables telles que la température et la vitesse du vent.Pour des variables reconnues comme difficiles à calibrer, telles que les précipitations sexti-horaires, des versions hybrides de nos techniques ont été créées. Nous montrons que ces versions hybrides (ainsi que nos versions originales) sont meilleures que les méthodes existantes. Elles amènent notamment une véritable valeur ajoutée pour les pluies extrêmes.La dernière partie de cette thèse concerne l'évaluation des prévisions d'ensemble pour les événements extrêmes. Nous avons montré quelques propriétés concernant le Continuous Ranked Probability Score (CRPS) pour les valeurs extrêmes. Nous avons aussi défini une nouvelle mesure combinant le CRPS et la théorie des valeurs extrêmes, dont nous examinons la cohérence sur une simulation ainsi que dans un cadre opérationnel.Les résultats de ce travail sont destinés à être insérés au sein de la chaîne de prévision et de vérification à Météo-France. / In numerical weather prediction, ensemble forecasts systems have become an essential tool to quantifyforecast uncertainty and to provide probabilistic forecasts. Unfortunately, these models are not perfect and a simultaneouscorrection of their bias and their dispersion is needed.This thesis presents new statistical post-processing methods for ensemble forecasting. These are based onrandom forests algorithms, which are non-parametric.Contrary to state of the art procedures, random forests can take into account non-linear features of atmospheric states. They easily allowthe addition of covariables (such as other weather variables, seasonal or geographic predictors) by a self-selection of the mostuseful predictors for the regression. Moreover, we do not make assumptions on the distribution of the variable of interest. This new approachoutperforms the existing methods for variables such as surface temperature and wind speed.For variables well-known to be tricky to calibrate, such as six-hours accumulated rainfall, hybrid versions of our techniqueshave been created. We show that these versions (and our original methods) are better than existing ones. Especially, they provideadded value for extreme precipitations.The last part of this thesis deals with the verification of ensemble forecasts for extreme events. We have shown several properties ofthe Continuous Ranked Probability Score (CRPS) for extreme values. We have also defined a new index combining the CRPS and the extremevalue theory, whose consistency is investigated on both simulations and real cases.The contributions of this work are intended to be inserted into the forecasting and verification chain at Météo-France.

Transportation via train is considered the most environmentally friendly way of traveling and is widely seen as the future of transportation. Canceled and delayed trains worsen customer satisfaction; thus, punctual trains are crucial for railway companies. One reason for canceled and delayed trains is the shortage of employees due to sickness or care of relatives, known as short-term absences. Therefore, it is important for railway companies to have reliable predictions of these. This thesis is in collaboration with SJ, the largest railway company in Sweden which offers trips all over Sweden and some other parts of northern Europe. The thesis predicts short-term absences with data provided by SJ, by using the machine learning methods random forest and extreme gradient boosting (XGBoost). The aim is to investigate if SJ can use machine learning algorithms and statistical analysis in their absence predictions and if it can yield better results than their current absence prediction methodology. Furthermore, the thesis identifies which factors are most important for the predictions. In addition to this, quantile regression is implemented for both methods since overestimating absenteeism could be better for avoiding employee shortage.  Two different datasets are used for two different tasks; one regression task to predict the number of absent employees on each date and one classification task to predict the probability of an absent employee on a specific duty, and then adding the probabilities to achieve the total predicted number of absent employees on each date. Both task formulations yielded good absence prediction results. XGBoost resulted overall in lower errors than random forest, meaning it was a slightly better model to implement for this task. When comparing the results, the performance for the developed models was better than the current predictions at SJ, meaning machine learning models could benefit SJ's prediction work. / Tågresor anses vara det mest miljövänliga sättet att resa på och betraktas av många som framtidens transportmedel. SJ är Sveriges största järnvägsföretag och erbjuder resor över hela Sverige och delar av norra Europa. Punktliga tåg är en mycket viktig faktor för järnvägsföretag, för att inte ha inställda och försenade tåg som försämrar kundnöjdheten. En orsak till inställda och försenade tåg är brist på personal på grund av sjukdom eller vård av anhöriga, så kallad korttidsfrånvaro. Därför är det viktigt för SJ att ha tillförlitliga prognoser gällande detta. Detta examensarbete försöker förutspå korttidsfrånvaro med data från SJ, genom att använda maskininlärningsmetoderna random forest och extreme gradient boosting (XGBoost). Syftet är att undersöka om SJ kan använda maskininlärningsalgoritmer och statistisk analys i sina frånvaroprognoser och om det kan ge bättre resultat än deras nuvarande prognoser. Vidare identifierar arbetet vilka faktorer som är viktigast för en pålitlig prognos. Utöver detta implementeras kvantilregression för båda metoderna eftersom överskattningar av frånvaro kan vara bättre för att undvika personalbrist. Två olika datamängder används för två olika uppgifter; en regressionsuppgift för att förutspå antalet frånvarande personal varje dag och en klassificeringsuppgift för att förutspå sannolikheten av en frånvarande personal under ett visst arbetspass. Modellen adderar sedan sannolikheterna för att få en prognos av det totala antalet frånvarande personal under varje dag. Båda uppgiftsformuleringarna resulterade i bra sjukprognoser. XGBoost resulterade totalt sett i lägre fel än random forest, vilket betyder att den var en något bättre modell att implementera för detta arbete. Vid en jämförelse av resultaten var prestationen för de utvecklade modellerna bättre än de nuvarande prognoserna hos SJ, vilket innebär att maskininlärningsalgoritmer kan gynna SJ:s prognosarbete.

Diese Dissertation behandelt verschiedene Aspekte moderner Ökonometrie und Machine Learnings. Kapitel 2 stellt einen neuen Schätzer für die Regressionsparameter in einem Paneldatenmodell mit interaktiven festen Effekten vor. Eine Besonderheit unserer Methode ist die Modellierung der factor loadings durch nichtparametrische Funktionen. Wir zeigen die root-NT-Konvergenz sowie die asymptotische Normalverteilung unseres Schätzers. Kapitel 3 betrachtet die rekursive Schätzung von Quantilen mit Hilfe des stochastic gradient descent (SGD) Algorithmus mit Polyak-Ruppert Mittelwertbildung. Der Algorithmus ist rechnerisch und Speicher-effizient verglichen mit herkömmlichen Schätzmethoden. Unser Fokus ist die Untersuchung des nichtasymptotischen Verhaltens, indem wir eine exponentielle Wahrscheinlichkeitsungleichung zeigen. In Kapitel 4 stellen wir eine neue Methode zur Kalibrierung von conditional Value-at-Risk (CoVaR) basierend auf Quantilregression mittels Neural Networks vor. Wir modellieren systemische Spillovereffekte in einem Netzwerk von systemrelevanten Finanzinstituten. Eine Out-of-Sample Analyse zeigt eine klare Verbesserung im Vergleich zu einer linearen Grundspezifikation. Im Vergleich mit bestehenden Risikomaßen eröffnet unsere Methode eine neue Perspektive auf systemisches Risiko. In Kapitel 5 modellieren wir die gemeinsame Dynamik von Kryptowährungen in einem nicht-stationären Kontext. Um eine Analyse in einem dynamischen Rahmen zu ermöglichen, stellen wir eine neue vector error correction model (VECM) Spezifikation vor, die wir COINtensity VECM nennen. / This thesis focuses on different aspects of the union of modern econometrics and machine learning. Chapter 2 considers a new estimator of the regression parameters in a panel data model with unobservable interactive fixed effects. A distinctive feature of the proposed approach is to model the factor loadings as a nonparametric function. We show that our estimator is root-NT-consistent and asymptotically normal, as well that it reaches the semiparametric efficiency bound under the assumption of i.i.d. errors. Chapter 3 is concerned with the recursive estimation of quantiles using the stochastic gradient descent (SGD) algorithm with Polyak-Ruppert averaging. The algorithm offers a computationally and memory efficient alternative to the usual empirical estimator. Our focus is on studying the nonasymptotic behavior by providing exponentially decreasing tail probability bounds under minimal assumptions. In Chapter 4 we propose a novel approach to calibrate the conditional value-at-risk (CoVaR) of financial institutions based on neural network quantile regression. We model systemic risk spillover effects in a network context across banks by considering the marginal effects of the quantile regression procedure. An out-of-sample analysis shows great performance compared to a linear baseline specification, signifying the importance that nonlinearity plays for modelling systemic risk. A comparison to existing network-based risk measures reveals that our approach offers a new perspective on systemic risk. In Chapter 5 we aim to model the joint dynamics of cryptocurrencies in a nonstationary setting. In particular, we analyze the role of cointegration relationships within a large system of cryptocurrencies in a vector error correction model (VECM) framework. To enable analysis in a dynamic setting, we propose the COINtensity VECM, a nonlinear VECM specification accounting for a varying system-wide cointegration exposure.