No description available.

While existing off-policy evaluation (OPE) methods typically estimate the value of a policy, in real-world applications, OPE is often used to compare and rank policies before deploying them in the real world. This is also known as the offline policy ranking problem. While one can rank the policies based on point estimates from OPE, it is beneficial to estimate the full distribution of outcomes for policy ranking and selection. This paper introduces Probabilistic Offline Policy Ranking that works with expert trajectories. It introduces rigorous statistical inference capabilities to offline evaluation, which facilitates probabilistic comparisons of candidate policies before they are deployed. We empirically demonstrate that POPR is effective for evaluating RL policies across various environments.

Energy conservation and implementation of effective energy efficiency policies have become imperative to curbing the escalating consumption of energy. The imbalance in the supply and demand of a country's energy has increased the importance of implementing energy efficiency policies. Proper replication of strategic energy efficiency policies that are known to be successful in one country, along with development of new approaches, can be helpful in developing the energy policy portfolio of another country. Some OECD (Organization of Economic Cooperation and Development) countries like Denmark, Finland, France, Germany, Italy, the United Kingdom and the United States have benefited from their energy policies during the most recent energy crisis. The motivation of this research is to provide a tool for developing countries, which are still in the stage of formulating their energy efficiency policies, to compare energy efficiency policy portfolios across countries. These countries can improve their energy efficiency policy portfolios based on lessons learned from the developed countries. The research develops a framework to compare energy efficiency policy portfolios across countries / states. Although this framework can be adopted for any type of energy policy, targeting any sector with few modifications, the current focus is on policies that target the residential building sector to reduce energy consumption. The research begins with identification of the functional domains that influence human behavior–people, economy, environment and technology–followed by identification of the factors affecting household energy consumption. It uses the four functional domains as the evaluation framework's four axes. The various factors affecting household energy consumption are positioned in the framework based on association with the functional domains. The energy efficiency policies implemented in a country are positioned in the same framework based on the pattern of diffusion of each type of policy. In addition, a prototype method is developed to identify the factors targeted by each energy efficiency policy implemented in a country. This evaluation method allows for a uniform assessment process of how energy efficiency policies target specific socio-economic factors that are known to affect energy consumption. The proposed framework will facilitate the work of policy makers and other decision makers with a powerful tool for evaluating and comparing their individual policies, or their complete portfolio of energy efficiency policies, to those from other states or countries, and to benefit from the lessons learned. / Ph. D.

published_or_final_version / Housing Management / Master / Master of Housing Management

Compuesta por tres capítulos: El primero es un estudio sobre la implementación the Sparse Grid métodos para es el estudio de modelos económicos con muchas dimensiones. Llevado a cabo mediante aplicaciones noveles del método de Smolyak con el objetivo de favorecer la tratabilidad y obtener resultados preciso. Los resultados muestran mejoras en la eficiencia de la implementación de modelos con múltiples agentes. El segundo capítulo introduce una nueva metodología para la evaluación de políticas económicas, llamada Synthetic Control with Statistical Learning, todo ello aplicado a políticas particulares: a) reducción del número de horas laborales en Portugal en 1996 y b) reducción del coste del despido en España en 2010. La metodología funciona y se erige como alternativa a previos métodos. En términos empíricos se muestra que tras la implementación de la política se produjo una reducción efectiva del desempleo y en el caso de España un incremento del mismo. El tercer capítulo utiliza la metodología utiliza en el segundo capítulo y la aplica para evaluar la implementación del Tercer Programa Europeo para la Seguridad Vial (Third European Road Safety Action Program) entre otras metodologías. Los resultados muestran que la coordinación a nivel europeo de la seguridad vial a supuesto una ayuda complementaria. En el año 2010 se estima una reducción de víctimas mortales de entre 13900 y 19400 personal en toda Europa.

published_or_final_version / Real Estate and Construction / Doctoral / Doctor of Philosophy

published_or_final_version / Housing Management / Master / Master of Housing Management

Le but de cette thèse est d étendre la théorie du bootstrap aux modèles de données de panel. Les données de panel s obtiennent en observant plusieurs unités statistiques sur plusieurs périodes de temps. Leur double dimension individuelle et temporelle permet de contrôler l 'hétérogénéité non observable entre individus et entre les périodes de temps et donc de faire des études plus riches que les séries chronologiques ou les données en coupe instantanée. L 'avantage du bootstrap est de permettre d obtenir une inférence plus précise que celle avec la théorie asymptotique classique ou une inférence impossible en cas de paramètre de nuisance. La méthode consiste à tirer des échantillons aléatoires qui ressemblent le plus possible à l échantillon d analyse. L 'objet statitstique d intérêt est estimé sur chacun de ses échantillons aléatoires et on utilise l ensemble des valeurs estimées pour faire de l inférence. Il existe dans la littérature certaines application du bootstrap aux données de panels sans justi cation théorique rigoureuse ou sous de fortes hypothèses. Cette thèse propose une méthode de bootstrap plus appropriée aux données de panels. Les trois chapitres analysent sa validité et son application. Le premier chapitre postule un modèle simple avec un seul paramètre et s 'attaque aux propriétés théoriques de l estimateur de la moyenne. Nous montrons que le double rééchantillonnage que nous proposons et qui tient compte à la fois de la dimension individuelle et la dimension temporelle est valide avec ces modèles. Le rééchantillonnage seulement dans la dimension individuelle n est pas valide en présence d hétérogénéité temporelle. Le ré-échantillonnage dans la dimension temporelle n est pas valide en présence d'hétérogénéité individuelle. Le deuxième chapitre étend le précédent au modèle panel de régression. linéaire. Trois types de régresseurs sont considérés : les caractéristiques individuelles, les caractéristiques temporelles et les régresseurs qui évoluent dans le temps et par individu. En utilisant un modèle à erreurs composées doubles, l'estimateur des moindres carrés ordinaires et la méthode de bootstrap des résidus, on montre que le rééchantillonnage dans la seule dimension individuelle est valide pour l'inférence sur les coe¢ cients associés aux régresseurs qui changent uniquement par individu. Le rééchantillonnage dans la dimen- sion temporelle est valide seulement pour le sous vecteur des paramètres associés aux régresseurs qui évoluent uniquement dans le temps. Le double rééchantillonnage est quand à lui est valide pour faire de l inférence pour tout le vecteur des paramètres. Le troisième chapitre re-examine l exercice de l estimateur de différence en di¤érence de Bertrand, Duflo et Mullainathan (2004). Cet estimateur est couramment utilisé dans la littérature pour évaluer l impact de certaines poli- tiques publiques. L exercice empirique utilise des données de panel provenant du Current Population Survey sur le salaire des femmes dans les 50 états des Etats-Unis d Amérique de 1979 à 1999. Des variables de pseudo-interventions publiques au niveau des états sont générées et on s attend à ce que les tests arrivent à la conclusion qu il n y a pas d e¤et de ces politiques placebos sur le salaire des femmes. Bertrand, Du o et Mullainathan (2004) montre que la non-prise en compte de l hétérogénéité et de la dépendance temporelle entraîne d importantes distorsions de niveau de test lorsqu'on évalue l'impact de politiques publiques en utilisant des données de panel. Une des solutions préconisées est d utiliser la méthode de bootstrap. La méthode de double ré-échantillonnage développée dans cette thèse permet de corriger le problème de niveau de test et donc d'évaluer correctement l'impact des politiques publiques. / The purpose of this thesis is to develop bootstrap methods for panel data models and to prove their validity. Panel data refers to data sets where observations on individual units (such as households, firms or countries) are available over several time periods. The availability of two dimensions (cross-section and time series) allows for the identi cation of effects that could not be accounted for otherwise. In this thesis, we explore the use of the bootstrap to obtain estimates of the distribution of statistics that are more accurate than the usual asymptotic theory. The method consists in drawing many ran- dom samples that resembles the sample as much as possible and estimating the distribution of the object of interest over these random samples. It has been shown, both theoretically and in simulations, that in many instances,this approach improves on asymptotic approximations. In other words, the resulting tests have a rejection rate close to the nominal size under the null hypothesis and the resulting con dence intervals have a probability of inclu- ding the true value of the parameter that is close to the desired level. In the literature, there are many applications of the bootstrap with panel data, but these methods are carried out without rigorous theoretical justi fication. This thesis suggests a bootstrap method that is suited to panel data (which we call double resampling), analyzes its validity, and implements it in the analysis of treatment e¤ects. The aim is to provide a method that will provide reliable inference without having to make strong assumptions on the underlying data-generating process. The rst chapter considers a model with a single parameter (the overall expectation) with the sample mean as estimator. We show that our double resampling is valid for panel data models with some cross section and/or temporal heterogeneity. The assumptions made include one-way and two- way error component models as well as factor models that have become popular with large panels. On the other hand, alternative methods such as bootstrapping cross-sections or blocks in the time dimensions are only valid under some of these models. The second chapter extends the previous one to the panel linear regression model. Three kinds of regressors are considered : individual characteristics, temporal characteristics and regressors varying across periods and cross-sectional units. We show that our double resampling is valid for inference about all the coe¢ cients in the model estimated by ordinary least squares under general types of time-series and cross-sectional dependence. Again, we show that other bootstrap methods are only valid under more restrictive conditions. Finally, the third chapter re-examines the analysis of di¤erences-in-differences estimators by Bertrand, Du o and Mullainathan (2004). Their empirical application uses panel data from the Current Population Survey on wages of women in the 50 states. Placebo laws are generated at the state level, and the authors measure their impact on wages. By construction, no impact should be found. Bertrand, Dufl o and Mullainathan (2004) show that neglected heterogeneity and temporal correlation lead to spurious ndings of an effect of the Placebo laws. The double resampling method developed in this thesis corrects these size distortions very well and gives more reliable evaluation of public policies.

O presente trabalho trata da avaliação de medidas voltadas à distribuição urbana de cargas considerando ser essa uma etapa fundamental do processo de tomada de decisão, pois permite a análise dos possíveis resultados acerca dos objetivos pretendidos, os quais geralmente estão relacionados à diminuição dos congestionamentos de veículos e das emissões de gases poluentes. Considerando que grande parte dos problemas decorrentes da distribuição urbana ocorrem localmente, no nível desagregado, propõe-se um procedimento que dê subsídios ao desenvolvimento de uma metodologia para a avaliação de medidas através da microssimulação de tráfego. Isso porque a microssimulação permite um alto nível de detalhe na modelagem e na observação do comportamento dos veículos, o que é essencial para a quantificação dos possíveis impactos locais gerados pelos veículos de carga. A integração entre as coletas de dados e os modelos que compõem o procedimento forneceu os dados necessários à representação dos movimentos dos veículos de carga no VISSIM, onde foram simulados três diferentes cenários, cada um representando alterações impostas por diferentes medidas em avaliação. A aplicação do procedimento proposto, a modelagem no VISSIM e os resultados da microssimulação permitiram a ponderação sobre o uso dessa técnica para a avaliação de medidas voltadas à distribuição urbana de cargas, considerando suas vantagens e limitações. / The present research deals with the evaluation of urban freight policies, considering this is fundamental for the decision making process of a policy implementation. Evaluation allows an analysis of the possible policy effects compared to its intended objectives, which are generally related to reductions in traffic congestion and pollutant emissions. Whereas most of the problems caused by urban distribution occur locally, in the disaggregated level, we propose a procedure that supports the development of a methodology for evaluating freight policies through traffic microsimulation. That\'s because this technique allows a high level of detail in modeling and observation of vehicles\' behavior, which is essential for quantifying the likely local impacts generated by freight vehicles. The procedure is based on a set of integrated data collections and models, which provided the data for representing freight vehicles movement in VISSIM, were three different scenarios, each one representing changes imposed by different freight policies, were simulated. The application of the proposed procedure, the VISSIM\'s modeling process and the microsimulation results allowed the weighing for the use of this technique in evaluating urban freight policies, considering its advantages and limitations.

O Estado brasileiro, seguindo as tendências mundiais, vem desenvolvendo e implementando, de forma cada vez mais sistematizada, aparatos institucionais e políticas públicas de estímulo à inovação. Entretanto, tais políticas apresentam um longo histórico de falhas e de descontinuidades nas estratégias do governo e nos planos que incentivam o desenvolvimento da ciência, tecnologia e inovação no país. Tendo em vista esse contexto, o presente trabalho busca analisar em detalhes, por meio de um estudo de casos, como o governo formula e implementa seus programas de estímulo a inovação. Para tanto, foram escolhidos dois programas para serem comparados como estratégias de investigação, distantes no tempo e nos alicerces políticos e institucionais em que foram concebidos. O primeiro deles foi uma ação da década de oitenta, intitulada Programa Inovação Tecnológica (PIT), que objetivava a implantação de Núcleos de Inovação Tecnológica em universidades e institutos de pesquisas. O segundo e mais recente é o Programa Primeira Empresa Inovadora (PRIME), que visa o apoio financeiro, com recursos não reembolsáveis para empresas nascentes inovadoras. Verificou-se que houve avanços significativos na forma como são concebidos e implementados, todavia, alguns aspectos críticos das heranças culturais e políticas brasileiras continuam ameaçando a continuidade desse tipo de ação, com destaque para a excessiva politização dos altos escalões das entidades responsáveis por executar essas ações e a intensa corporatização dos atores centrais nos movimentos de tentativas de estímulo a inovação no Brasil. / The Brazilian State, following the worldwide trends, has been developing and implementing, in increasingly systemized way, public policies to stimulate innovation. However, such policies present a long description of imperfections and discontinuities in the government strategies, and plans that stimulate the development of science, technology and innovation in the country. In this context, this work aims to analyze at great length, through a study of cases, the way that the government formulates and implements its programs to stimulate innovation. Therefore, two programs had been chosen to be comparing as inquiry strategies, distant in time as well as in the institutional and politicians foundations where they had been conceived. The first one was an action of the eighties, entitled Programa Inovação Tecnológica (PIT), which aimed to implement the Core of Technological Innovation in universities and research institutes. The second and most recent is the Programa Primeira Empresa Inovadora (PRIME), which aims to support innovative emerging companies with non-refundable grant. It was verified that they had significant advances in the way that they are conceived and implemented, however some critical aspects of the cultural inheritances and Brazilian policies keep threatening the continuity of those kind of action, with prominence for the excessive politicalization of the high steps of the responsible entities for executing those actions and the intense corporatization of the central actors in the movements that attempts to stimulate innovation in Brazil.