This dissertation contains three essays on firm-sponsored training. Paper 1 develops a general theoretical framework in a frictional labour market to investigate how firms decide to sponsor how much general as well as specific training to workers assuming complementarity between the two types of training as well as education. It shows that firms’ profit maximizing decisions provide firms with an incentive to provide more training, general as well specific, to the more educated workers, more training for more educated workers may lead to low turnover rate, and the resulting life-time profile of firm-sponsored training is U-shaped or decreasing. The policy implications are that governments can subsidize both education and training to improve efficiency. Paper 2 and paper 3 try to provide empirical evidence from different perspectives, respectively determinants and effects of three types of firm-sponsored training, i.e., class-room training, on-the-job-training, and career-related but not job directly related training based on Statistics Canada’s Worker Place and Employee Survey (WES) of 2003/2004. The major empirical findings arising from our estimation results are: (1) Education is positively and significantly associated with the incidence of all three types of training, and significantly positively correlated with the intensity of on-the-job training. (2) Workers in larger firms are more likely to obtain classroom training and on-the-job training than workers in smaller firms. (3) Job tenure is significant and negative for the intensity of classroom training or on-the-job training. (4) Classroom-training and on-the-job training increases the average earnings of workers but less than average resultant firm-level productivity growth. Firm sponsored career related training has no significant impact on a worker’s earnings but increases the firm’s productivity significantly. All these findings by and large are consistent with the theory developed in first paper.

This dissertation contains three essays on firm-sponsored training. Paper 1 develops a general theoretical framework in a frictional labour market to investigate how firms decide to sponsor how much general as well as specific training to workers assuming complementarity between the two types of training as well as education. It shows that firms’ profit maximizing decisions provide firms with an incentive to provide more training, general as well specific, to the more educated workers, more training for more educated workers may lead to low turnover rate, and the resulting life-time profile of firm-sponsored training is U-shaped or decreasing. The policy implications are that governments can subsidize both education and training to improve efficiency. Paper 2 and paper 3 try to provide empirical evidence from different perspectives, respectively determinants and effects of three types of firm-sponsored training, i.e., class-room training, on-the-job-training, and career-related but not job directly related training based on Statistics Canada’s Worker Place and Employee Survey (WES) of 2003/2004. The major empirical findings arising from our estimation results are: (1) Education is positively and significantly associated with the incidence of all three types of training, and significantly positively correlated with the intensity of on-the-job training. (2) Workers in larger firms are more likely to obtain classroom training and on-the-job training than workers in smaller firms. (3) Job tenure is significant and negative for the intensity of classroom training or on-the-job training. (4) Classroom-training and on-the-job training increases the average earnings of workers but less than average resultant firm-level productivity growth. Firm sponsored career related training has no significant impact on a worker’s earnings but increases the firm’s productivity significantly. All these findings by and large are consistent with the theory developed in first paper.

Los sistemas online para compartir fotos proporcionan una valiosa fuente de contenidos generado por el usuario (UGC). La mayor a de los sistemas de re- cuperaci on de im agenes Web utilizan las anotaciones textuales para rankear los resultados, sin embargo estas anotaciones no s olo ilustran el contenido visual de una imagen, sino que tambi en describen situaciones subjetivas, espaciales, temporales y sociales, que complican la tarea de b usqueda basada en palabras clave. La investigaci on en esta tesis se centra en c omo mejorar la recuperaci on de im agenes en sistemas de gran escala, es decir, la Web, combinando informaci on proporcionada por los usuarios m as el contenido visual de las im agenes. En el presente trabajo se exploran distintos tipos de UGC, tales como anotaciones de texto, anotaciones visuales, y datos de click-through, as como diversas t ecnicas para combinar esta informaci on con el objetivo de mejorar la recuperaci on de im agenes usando informaci on visual. En conclusi on, la investigaci on realizada en esta tesis se centra en la impor- tancia de incluir la informaci on visual en distintas etapas de la recuperaci on de contenido. Combinando informaci on visual con otras formas de UGC, es posible mejorar signi cativamente el rendimiento de un sistema de recuperaci on de im agenes y cambiar la experiencia del usuario en la b usqueda de contenidos multimedia en la Web. / Online photo sharing systems provide a valuable source of user generated content (UGC). Most Web image retrieval systems use textual annotations to rank the results, although these annotations do not only illustrate the visual content of an image, but also describe subjective, spatial, temporal, and social dimensions, complicating the task of keyword based search. The research in this thesis is focused on how to improve the retrieval of images in large scale context , i.e. the Web, using information provided by users combined with visual content from images. Di erent forms of UGC are explored, such as textual annotations, visual annotations, and click-through-data, as well as di erent techniques to combine these data to improve the retrieval of images using visual information. In conclusion, the research conducted in this thesis focuses on the impor- tance to include visual information into various steps of the retrieval of media content. Using visual information, in combination with various forms of UGC, can signi cantly improve the retrieval performance and alter the user experience when searching for multimedia content on the Web. 1

CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior / The objective of this paper turns to evaluation of the factor that may influence the chances of approval in an aptitude test for admission in public jobs. Using a database obtained from a aptitude test realized by the Banco do Nordeste do Brasil, occurred in 2003 to mid-level positions, and from Pesquisa Nacional por Amostra de DomicÃlio (PNAD) database, also 2003, sought to draw a map of the determinats to approval in the aptitude test, however the hypothesis of sample selection occurrence was raised, since the enrollment process in the aptitude test was coupled with certain socioeconomic characteristics needing so, be considered in the analysis of the approval factors, or in other way, we could obtain biased results, if the problem (sample selection) occurs. To obtain success in our demand, we use an adaptation of the methodology proposed by Heckman, with an addition that was employed in a discrete choice model, because ours dependent variables are dummies, being to the enrollment process, if he enrolls or not and for the aptitude process, if he has succeeded or not. We estimated three models, being two models, considering the sample selection problem, and another without the selectivity. The results found pointed to nonexistence of the selectivity problem, confirming other papers on the same theme. / O objetivo do presente trabalho volta-se para a avaliaÃÃo de fatores que possam influenciar as chances de aprovaÃÃo em um concurso pÃblico para o provimento de cargos pÃblicos. Com a utilizaÃÃo de uma base de dados obtida a partir de um concurso realizado pelo Banco do Nordeste do Brasil, ocorrido em 2003 para cargos de nÃvel mÃdio e da Pesquisa Nacional por Amostra de DomicÃlio (PNAD) tambÃm de 2003, buscou-se traÃar um mapa dos fatores que influenciam a aprovaÃÃo no concurso, contudo levantou-se a hipÃtese da ocorrÃncia de viÃs de seleÃÃo amostral, posto que o simples processo de inscriÃÃo no concurso estaria atrelado a certas caracterÃsticas sÃcio econÃmicas devendo portanto ser levado em consideraÃÃo quando da anÃlise dos fatores para aprovaÃÃo, pois de outra forma obterÃamos resultados enviesados caso tal problema (viÃs de seleÃÃo) ocorra. Para lograr Ãxito em nossa demanda utilizamos uma adaptaÃÃo da metodologia proposta por Heckman, com a diferenÃa que empregamos modelos de escolha binÃria, pois nossas variÃveis dependentes tratavam-se de variÃveis dummies, sendo para o processo de inscriÃÃo se o indivÃduo inscreveu-se ou nÃo e para o processo de seleÃÃo se o candidato foi aprovado ou nÃo. Foram realizadas trÃs estimaÃÃes sendo duas que levam em consideraÃÃo a problemÃtica do viÃs de seletividade e uma terceira em que ignora-se tal possibilidade. Os resultados encontrados apontam para a inexistÃncia do viÃs corroborando estudos realizados anteriormente relativos ao mesmo tema.

Approved for public release; distribution is unlimited. / This thesis investigates the effect of marital and family status on the performance and job productivity of male U.S. Marine Corps officers. The analysis includes evaluation of fitness reports, retention, and promotion to O-4 and O-5 ranks as performance measures. The primary goal is to examine the existence of any marriage premium on officers' performance and productivity and to investigate potential causal hypotheses. The personnel database used for the analysis includes more than 27,000 male Marine officers who entered the Marine Corps between FY 1980 and 1999. After controlling for selection, estimating fixed effects and using panel data in order to capture timely-varying effects, this study finds that there is a marriage premium for all performance measures. The thesis rejects the explanation that such premiums are due to supervisor favoritism. Moreover, married male officers obtain higher fitness report scores, higher promotion probabilities, and higher retention probabilities than single officers. Each additional year spent in marriage increases fitness report scores and retention probabilities. Having additional non-spousal dependents increase fitness report scores and retention probabilities. On the other hand, being a currently single but "to-be-married" officer yields higher premium, as married officers, for all productivity and performance indicators. This supports selectivity into marriage as a partial explanation of the source of the marriage premium. / Lieutenant Junior Grade, Turkish Navy / First Lieutenant, Turkish Army

This dissertation consists of four papers, each with an application of a discrete dependent variable model, censored regression or duration model to a credit market phenomenon or monetary policy question. The first three essays deal with bank lending policy, while the last one studies interest rate policy by Central Banks. In the first essay, a bivariate probit model is estimated to contrast the factors that influence banks’ loan granting decision and individuals’ risk of default. This model is used as a tool to construct a Value at Risk measure of the credit risk involved in a portfolio of consumer loans and to investigate the efficiency of bank lending policy. The second essay takes the conclusions from the first paper as a starting point. It investigates if the fact that banks do not minimize default risk can be explained by the existence of return maximization policy. For this purpose, a Tobit model with sample selection effects and variable censoring limits is developed and estimated on the survival times of consumer loans. The third paper focuses on dormancy, instead of default risk or survival time, as the most important factor affecting risk and return in bank lending. By means of a duration model the factors determining the transition from an active status to dormancy are studied. The estimated model is used to predict the expected durations to dormancy and to analyze the expected profitability for a sample loan applicants. In the fourth paper, the discrete nature of Central Bank interest rate policy is studied. A grouped data model, that can take the long periods of time without changes in the repo rate by the Central Bank into account, is estimated on weekly Swedish data. The model is found to be reasonably good at predicting interest rate changes. / Diss. (sammanfattning) Stockholm : Handelshögsk.

One of the main problems in machine learning is to train a predictive model from training data and to make predictions on test data. Most predictive models are constructed under the assumption that the training data is governed by the exact same distribution which the model will later be exposed to. In practice, control over the data collection process is often imperfect. A typical scenario is when labels are collected by questionnaires and one does not have access to the test population. For example, parts of the test population are underrepresented in the survey, out of reach, or do not return the questionnaire. In many applications training data from the test distribution are scarce because they are difficult to obtain or very expensive. Data from auxiliary sources drawn from similar distributions are often cheaply available. This thesis centers around learning under differing training and test distributions and covers several problem settings with different assumptions on the relationship between training and test distributions-including multi-task learning and learning under covariate shift and sample selection bias. Several new models are derived that directly characterize the divergence between training and test distributions, without the intermediate step of estimating training and test distributions separately. The integral part of these models are rescaling weights that match the rescaled or resampled training distribution to the test distribution. Integrated models are studied where only one optimization problem needs to be solved for learning under differing distributions. With a two-step approximation to the integrated models almost any supervised learning algorithm can be adopted to biased training data. In case studies on spam filtering, HIV therapy screening, targeted advertising, and other applications the performance of the new models is compared to state-of-the-art reference methods. / Eines der wichtigsten Probleme im Maschinellen Lernen ist das Trainieren von Vorhersagemodellen aus Trainingsdaten und das Ableiten von Vorhersagen für Testdaten. Vorhersagemodelle basieren üblicherweise auf der Annahme, dass Trainingsdaten aus der gleichen Verteilung gezogen werden wie Testdaten. In der Praxis ist diese Annahme oft nicht erfüllt, zum Beispiel, wenn Trainingsdaten durch Fragebögen gesammelt werden. Hier steht meist nur eine verzerrte Zielpopulation zur Verfügung, denn Teile der Population können unterrepräsentiert sein, nicht erreichbar sein, oder ignorieren die Aufforderung zum Ausfüllen des Fragebogens. In vielen Anwendungen stehen nur sehr wenige Trainingsdaten aus der Testverteilung zur Verfügung, weil solche Daten teuer oder aufwändig zu sammeln sind. Daten aus alternativen Quellen, die aus ähnlichen Verteilungen gezogen werden, sind oft viel einfacher und günstiger zu beschaffen. Die vorliegende Arbeit beschäftigt sich mit dem Lernen von Vorhersagemodellen aus Trainingsdaten, deren Verteilung sich von der Testverteilung unterscheidet. Es werden verschiedene Problemstellungen behandelt, die von unterschiedlichen Annahmen über die Beziehung zwischen Trainings- und Testverteilung ausgehen. Darunter fallen auch Multi-Task-Lernen und Lernen unter Covariate Shift und Sample Selection Bias. Es werden mehrere neue Modelle hergeleitet, die direkt den Unterschied zwischen Trainings- und Testverteilung charakterisieren, ohne dass eine einzelne Schätzung der Verteilungen nötig ist. Zentrale Bestandteile der Modelle sind Gewichtungsfaktoren, mit denen die Trainingsverteilung durch Umgewichtung auf die Testverteilung abgebildet wird. Es werden kombinierte Modelle zum Lernen mit verschiedenen Trainings- und Testverteilungen untersucht, für deren Schätzung nur ein einziges Optimierungsproblem gelöst werden muss. Die kombinierten Modelle können mit zwei Optimierungsschritten approximiert werden und dadurch kann fast jedes gängige Vorhersagemodell so erweitert werden, dass verzerrte Trainingsverteilungen korrigiert werden. In Fallstudien zu Email-Spam-Filterung, HIV-Therapieempfehlung, Zielgruppenmarketing und anderen Anwendungen werden die neuen Modelle mit Referenzmethoden verglichen.

No description available.

xi, 108 p. A print copy of this thesis is available through the UO Libraries. Search the library catalog for the location and call number. / Micro-level survey data are widely used in applied economic research. This dissertation, which consists of three empirical papers, demonstrates challenges in empirical research using micro-level survey data, as well as some methods to accommodate these problems. Chapter II examines the effect of China's recent public health insurance reform on health utilization and health status. Chinese policy makers have been eager to identify how this reform, characterized by a substantial increase in out-of-pocket costs, has affected health care demand and health status. However, due to self-selection of individuals into the publicly insured group, the impact of the reform remains an unresolved issue. I employ a Heckman selection model in the context of difference-in-difference regression to accommodate the selection problem, and provide the first solid empirical evidence that the recent public health insurance reforms in China adversely affected both health care access and health status for publicly insured individuals. Chapter III examines the construct validity of a stated preference (SP) survey concerning climate change policy. Due to the fact that the SP survey method remains a controversial tool for benefit-cost analysis, every part of the survey deserves thorough examination to ensure the quality of the data. Using a random utility approach, I establish that there is a great deal of logical consistency between people's professed attitudes toward different payment vehicles and their subsequent choices among policies which vary in the incidence of their costs. Chapter IV employs the same survey data used in Chapter III, but demonstrates the potential for order effects stemming from prior attitude-elicitation questions. In addition, it considers the potential impact of these order effects on Willingness to Pay (WTP) estimates for climate change mitigation. I find the orderings of prior elicitation questions may change people's opinions toward various attributes of the different policies, and thereby increase or decrease their WTP by a substantial amount. Thus, this chapter emphasizes the significance of order effects in prior elicitation questions, and supports a call for diligence in using randomly ordered prior elicitation questions in stated preference surveys, to minimize inadvertent effects from any single arbitrary ordering. / Adviser: Trudy Ann Cameron

No description available.