Search

[eng] This thesis provides research catalogued in the area of risk assessment. Specifically, it contributes to the fields of international finance and asset pricing in finance, and risk assessment in energy economics and transportation research. We present in this thesis a generalization of the spillover indexes to analyze interconnectedness at firm level, and define the aggregate influence from a sector and a country on a firm. We also discuss which factors are relevant for predicting conditional quantiles across the distribution of returns and present a method for selecting factors based on the investor interests. We study the performance of quantile regression against quantile time-series models. Finally, we present a regression framework which estimates VaR and CTE ensuring noncrossing conditions for various quantile levels, and discuss results on energy and telematics data. Within the financial contagion literature, we aim to provide a better understanding of international spillovers and a method for visualize which country and sector are its main drivers. We show that not all companies are driven by their own country or sector, which should be considered by investors and risk managers when assessing company risk and managing investments. In this paper we show that a large percentage of firms’ stocks are driven by their country. But contrary to the belief where country is the main driver of a company’s return movements, a part depends mainly on its sector. We note that 1) the financial services and energy companies are positioned at the center of the network, and 2) northern and western Europe are highly interconnected, while eastern and southern Europe present lower spillovers. 3) For the British energy firms British Petroleum (BP) and Royal Dutch Shell, we evidence greater spillovers from France than from Great Britain itself. 4) We identify which non-Russian firms are most influenced by Russia, simulating a risk management analysis in the event of of turmoil distresses such as the recent Ukrainian conflict. 5) We show the improvement on spillover information when using individual firm connectedness and aggregating spillovers afterwards against calculating spillovers directly from indexes. 6) We finally show that eastern Europe has increased interconnectedness with the rest of the continent after the Covid-19 pandemic. Regarding the asset pricing literature, we aim to understand the key elements that predict extreme quantile levels of a stock return. We study which factors for a 7-factor asset pricing specification are more relevant for each part of the distributions’ tail. The 7-factor specification is constituted by the factors size, book-to-market, operating profitability, investment, momentum, market beta and liquidity. We present a method to add more factors depending on the investors’ interests. We use quantile regression models for predicting quantile levels 0.05, 0.25, 0.5, 0.75 and 0.95 of the stock returns using cross-sectional characteristics as covariates from the Open Source Cross-Sectional Asset Pricing Dataset (Chen and Zimmermann, 2021). We observe that the factor size changes from positive to negative sign when predicting lower quantiles to higher quantiles. We show that extreme quantile level estimations perform better than the median in terms of pseudo-R2. Regarding factor significance, the variable investment has lower predictive power than other factors in terms of t-statistics for all tested quantile levels. Liquidity gains significance if quantile levels increase. For book-to-market, profitability, momentum and market beta, median predictions of returns are more significant than extreme quantile level estimations. The opposite happens for size, which presents higher relevance for predicting extreme quantile levels of the returns’ distribution. We observe that during crisis periods, some factors lose significance. This is the case of profitability and momentum for quantile levels 0.05 and 0.5, and size, book-to-market and market beta for quantile level 0.5. We add additional factors individually and compare the weighted average pseudo-R2 obtained across all 5 quantile levels. The weighting depends on the strategy that the investor follows. For all strategies tested, the most relevant factors to add to the 7-factor specification are momentum seasonality and net operating assets. Following, for strategies more interested in predicting losers’ tails (left part of the distribution), adding asset growth is recommended, but if the investor is interested in the winners’ tail (right part of the distribution), the recommended factor to add is enterprise multiple. Within the asset pricing literature, we encourage the use of cross-sectional information against time-series factors to predict extreme quantile levels of the right-hand side of the response distribution during periods of high volatility. By using this methodology, we do not restrict the information on panel-like datasets, which allows us to study more companies, and provide estimates for newly added firms. We use quantile regression specification with cross-sectional characteristics obtained from the Open Source Cross-Sectional Asset Pricing Dataset (Chen and Zimmermann, 2021) and compare results against a CAViaR (Engle and Manganelli, 2004) specification. Fama and French (2020) evidence that the average returns are better explained by using cross-sectional factors than by using time-series factors. We show that this only applies on extreme quantile levels during high volatility periods. We show that individual firm Hits (exceedances above VaR) calculated using time-series models tend to accumulate, while using cross-sectional data we avoid concentrations. We show that cross-sectional information improves the prediction of Value-at-Risk (VaR) and Conditional Tail Expectations (CTE). We finally discuss changes on capital requirements for a firm. In general, by using cross-sectional information, capital requirements should be increased from when time-series information is used. During turmoil periods the opposite happens: capital requirements should decrease compared to when using the CAViaR specification. Inside the area of non-crossing quantiles, we define the non-crossing property for VaR and CTE for several quantile levels. We define a regression framework based on neural networks that creates an environment for predicting VaR and CTE for several quantile levels while asserting non-crossing conditions. The proposed neural network predicts VaR and CTE as positive excesses of the previous VaR and CTE. We prove that this definition satisfies the non-crossing property and show its improvement against the Monotone Composite Quantile Regression Neural Network (Cannon, 2018) and a quantile regression and CTE linear approach on an energy consumption and telematic datasets. We show the estimation improvements on extreme quantile levels of the right part of the distribution against the other tested models by using Murphy diagrams (Ehm et al., 2016). We present examples with crossing predictions to demonstrate the infeasibility of such results in a business context, which we overcome using the proposed model.

This paper performs a meta-regression analysis to derive the role of techniques, data and variable's definition on the effect of the labor-market determinants on interregional migration. We use Spain as a case of study, a country with heterogeneous and even counterintuitive behavior of internal migration flows to its labor-market drivers. We use data from studies released over the last 40 years. The results show that migration flows respond to labor-market differentials in a theoretically consistent way. We find that the vast diversity in the studies' attributes is behind the significant heterogeneity of their estimated effects. Differences in aggregation level, variables measures, model specification, and the national economic context influence the identification of the push and pull effects.

The FeldsteinHorioka1980 study on investment flows through the correlation of domestic saving and investment concluded that liberalization of capital markets does not necessarily lead to a movement of capital looking for a better allocation of resources, as classical theory would suggest. Ever since, literature has been prolific regarding this "puzzle", with arguments for and against this conclusion. This paper aims to analyze the issue from a different perspective. In recent years, the stock markets of Chile, Colombia, Mexico and Peru joined the Latin American Integrated Market through an agreement that allows investors in any of the participating markets to invest in the others as if they were investing locally. Compositional methods are used to assess the hypothesis of a potential flow of capital between markets generated by the creation of the joint market. First, cross-sectional methods for compositional data were used to test the hypothesis. As a result, it was not possible to find a change in the composition of the investment in the four markets produced by the creation of the joint market. Secondly, vector autoregressive models were estimated and tested for structural breaks in the parameters. However, these models were not found to be informative. In conclusion, it was not possible to reject the Felstein-Horioka hypothesis, supporting the idea that liberalization is not enough to generate capital flows between markets.

[spa] Se puede definir el riesgo como la posibilidad de que un evento adverso o perjudicial ocurra. En el mundo de los negocios el riesgo se define como la posibilidad de que haya una pérdida económica inesperadamente grande. Aunque estos eventos son poco frecuentes es de vital importancia que las compañías estén preparadas para tales sucesos y que sean capaces de poder anticiparse a ellos. A pesar de la suma importancia que tiene el análisis del riesgo, no ha sido hasta las últimas décadas en las que ha habido avances con respecto a las metodologías a seguir para realizar este tipo de estudios. El objetivo principal de esta tesis es estudiar que características de los clientes afectan significativamente al número de kilómetros que un conductor conduce por encima del límite de velocidad. En esta tesis se considera que el número de kilómetros conducidos por encima del límite de velocidad permitido es un factor de riesgo que afecta directamente a la probabilidad de que haya un accidente de tráfico. Encontrar que patrones de conducción afectan directamente a los factores de riesgo permitirá a las compañías de seguros detectar los clientes riesgosos y aplicar un aumento o descuento de prima en función de como sean sus patrones de conducción. Esta práctica, a pesar de ser poco común en España, cada vez se aplica en más países i las compañías de seguros están empezando a introducir los datos telemáticos a la hora de modelizar el precio del seguro. Los datos telemáticos son un tipo de datos cuya característica principal es que se obtienen de forma continuada durante un periodo de tiempo determinado. La base de datos con la que se pretende conseguir el objetivo mencionado anteriormente contiene información recogida durante un año, en 2010, sobre los patrones de conducción que tienen 7.691 conductores con una edad de entre 18 y 32 años. Estos conductores contrataron un producto específico que solo se ofreció a conductores jóvenes. En gran parte de esta tesis se aplica la regresión cuantílica como metodología para estudiar el riesgo. Este tipo de regresión permite estudiar los efectos de las variables explicativas en función del cuantil de la variable respuesta. Esta metodología toma como base la regresión lineal, pero da más flexibilidad a la hora de decidir que punto de la distribución de la variable respuesta se estudia ya que, en el caso de la regresión lineal por mínimos cuadrados ordinarios, solo se estudia la media de la variable respuesta. Esta flexibilidad resulta muy interesante a la hora de estudiar el riesgo ya que permite estudiar el comportamiento de las variables en los extremos de la distribución. Como complemento, en uno de los capítulos se estudia otro tipo de metodología relacionado con las distribuciones bivariantes para estudiar el riesgo y tener una visión más amplia del campo. Esta tesis plantea un problema de un tema actual que es la escasez de metodologia para estudiar el riesgo y lo enfoca en el mundo del seguro. Se desarrollan y estudian soluciones para mejorar las existentes metodologías del estudio del riesgo y se abren nuevos frentes de investigación que permitirán avances en el futuro.

Do independent local parties make different decisions on municipal finances compared to their national counterparts? In this paper, we empirically analyze whether independent local parties affect public finances in Dutch municipalities. Using a matching strategy, we compare municipalities that are similar in their observable characteristics except for the presence of an independent local party majority in the municipal council. We provide evidence that shows that municipalities with independent local majorities indeed differ in terms of local spending, specifically they spend more on categories of Local Public Administration, Public Health and Environment and Culture and Recreation which are arguably more local-oriented. We extend our analysis by looking at the local effects of local independent majorities. Using a regression kink design, we find consistent results if we look at the changes that take place once the majority share of the seats in the municipal council has been reached by independent local parties.

This article proposes a framework to study how the existing knowledge portfolio of regional economies affects the emergence and occurrence of breakthrough technologies. The study discusses the relevance of cognitive distance between the technology of a breakthrough invention and the existing technological base in their geographic vicinity. Theoretically, it introduces the idea that both relatedness and unrelatedness between the technologies in breakthrough inventions and the regional portfolio of technologies can positively influence the appearance of these breakthroughs, but especially relatedness. Empirically, it investigates a sample of 277 NUTS2 regions in Europe in the period 1981 to 2010 and reveals that, by far, most combinations breakthrough inventions make are between related technologies: almost no breakthrough patent combines unrelated technologies only. Regressions show that the occurrence of breakthrough patents in a technology in a region is positively affected by the local stock of technologies that are related to such technology, but such an effect for the local stock of unrelated technologies is not found. However, the region’s ability to enter new breakthrough inventions in a technology relies on the combination of knowledge that is both related and unrelated to such technology.

Modeling bivariate (or multivariate) count data has received increased interest in recent years. The aim is to model the number of different but correlated counts taking into account covariate information. Bivariate Poisson regression models based on the shock model approach are widely used because of their simple form and interpretation. However, these models do not allow for overdispersion or negative correlation, and thus, other models have been proposed in the literature to avoid these limitations. The present paper proposes copula-based bivariate finite mixture of regression models. These models offer some advantages since they have all the benefits of a finite mixture, allowing for unobserved heterogeneity and clustering effects, while the copula-based derivation can produce more flexible structures, including negative correlations and regressors. In this paper, the new approach is defined, estimation through an EM algorithm is presented, and then different models are applied to a Spanish insurance claim count database.

This paper deals with the selection of the training dataset in kernel-based methods for function reconstruction, with a focus on kernel ridge regression. A functional analysis is performed which, in the absence of noise, links the optimal sampling distribution to the one minimizing the difference between the kernel matrix and its low-rank Nyström approximation. From this standpoint, a statistical passive sampling approach is derived which uses the leverage scores of the columns of the kernel matrix to design a sampling distribution that minimizes an upper bound of the risk function. The proposed approach constitutes a passive method, able to select the optimal subset of training samples using only information provided by the input set and the kernel, but without needing to know the values of the function to be approximated. Furthermore, the proposed approach is backed up by numerical tests on real datasets. This work has been funded by the Ministerio de Ciencia e Innovación (MICINN) of the Spanish Government and by the Agencia Estatal de Investigación (AEI/10.13039/501100011033) and ERDF funds (PID 2019-104958RB-C41/C43, RED2018-102668-T); and by the Catalan Government (2017 SGR 578).

Purpose: Housekeeping is an important hotel process from the point-of-view of the number of work hours it takes and its impact on customer satisfaction. However, few previous scientific studies have addressed this topic or the variables that are determinants of the time required to clean a room. Design/methodology/approach: A stopwatch Time Study has been performed in a 4-star-hotel. Additionally, data on several variables that could affect cleaning time have been collected and subjected to regression analysis. Findings: Results show that only the task-related variables have a statistically significant influence on total cleaning time. None of the analyzed employee-related variables have any effect on cleaning time. Moreover, five tasks represent over 2/3 of the total cleaning time. Originality/value: In addition to empirically identifying the variables that influence cleaning time and to what extent, this study demonstrates the advantages of using stopwatch time studies to establish cleaning times.

A fast and precise prediction of stock market crashes is an important aspect of economic growth, fiscal and monetary systems because it facilitates the government in the application of suitable policies. Many works have examined the behavior of the fall of stock markets and have built models to predict them. Nevertheless, there are limitations to the available research, and the literature calls for more investigation on the topic, as currently the accuracy of the models remains low and they have only been extended for the largest economies. This study provides a comparison of quantum forecast methods and stock market declines and, therefore, a new prediction model of stock market crashes via real-time recession probabilities with the power to accurately estimate future global stock market downturn scenarios is achieved. A 104-country sample has been used, allowing the sample compositions to take into account the regional diversity of the alert warning indicators. To obtain a robust model, several alternative techniques have been employed on the sample under study, being Quantum Boltzmann Machines, which have obtained very good prediction results due to their ability to remember features and develop long-term dependencies from time series and sequential data. Our model has large policy implications for the appropriate macroeconomic policy response to downside risks, offering tools to help achieve financial stability at the international level.

With the outbreak of COVID-19 in spring 2020, small, medium, and large companies were forced to cope with the unexpected circumstances. Faced by this health emergency, it was necessary to ensure that staff remained motivated and that they could continue to carry out their duties despite the obstacles. The main goal of this exploratory research was to characterize employees who teleworked and who did not, and their motivation during the lockdown. A total of 11,779 workers from different-sized companies in various sectors answered an ad hoc questionnaire. By using non-parametric comparisons and Classification and Regression Trees (CRTs), the results show differences in both the assessment of strategies put into practice by the companies and the level of motivation of teleworkers and non-teleworkers, with the latter being more highly motivated. Nonetheless, teleworkers assessed their companies’ strategies and the role of their managers and colleagues more positively. This research helps to understand how different sectors have dealt with the crisis, according to the degree of teleworking implemented in each sector, and to what extent the motivation of the employees has been affected. The analysis of the large amount of data obtained confirms the importance of the role of managers in sustaining the motivation of their subordinates in times of crisis. In this sense, it is necessary to develop managers’ competencies in order to develop and maintain relations of trust and support with their coworkers. On the other hand, it is necessary to foster employees’ sense of meaningfulness and responsibility at work in order to keep them motivated.

The literature presents mixed findings regarding the economic conditions under which cartels form and collapse, and regarding how stable they are across firm-specific and industry-wide business cycles. The relationship between cartel life cycles and business cycles has been insufficiently analyzed to date. In this paper, we study in depth whether collusion is pro-cyclical or counter-cyclical. We analyze the relationship between cartel start-ups/break ups and economic cycles using a dataset of sanctioned cartels by the European Commission (EC) that were active between 1997 and 2018, after the leniency program had already been introduced. We also double check whether this relationship has changed with respect to the pre-leniency period from 1991 to 1996. Our results show that cartels are more likely to be formed when the business has evolved positively in the previous months, and cartels are less likely to collapse when the business has evolved positively, and managers expect prices to decline. The EC’s sanctioning activity has been an effective deterrent and has had a destabilizing effect on cartels. However we found no evidence that managers’ expectations on prices affect cartel formation. All these results are an important issue for anti-cartel policy enforcement since knowing when cartels are more prone or less likely to occur would help authorities prevent their formation or their early detection.

En este artículo se estiman los modelos de Lee-Carter, Renshaw-Haberman y Cairns-Blake-Dowd, expuestos en el marco de modelos de mortalidad estocástica edad-periodo-cohorte generalizados, con datos de mortalidad de Chile. El modelo de Lee-Carter con distribución binomial resulta ser el más adecuado para describir la evolución de la mortalidad en Chile. Se muestra una aplicación de la construcción de tablas de mortalidad para el cálculo de primas en seguros de vida, y se comparan con las obtenidas utilizando la tabla de mortalidad M95 (publicada por el regulador en Chile). Se concluye que las tablas M95 parecen adecuadas para garantizar la solvencia. Finalmente, se calcula la esperanza de vida al nacer en el año 2019 y en el año 2050 por sexo.

Most classical econometric methods and tree boosting based algorithms tend to increase the prediction error with binary imbalanced data. We propose a synthetic penalized logitboost based on weighting corrections. The procedure (i) improves the prediction performance under the phenomenon in question, (ii) allows interpretability since coefficients can get stabilized in the recursive procedure, and (iii) reduces the risk of overfitting. We consider a mortgage lending case study using publicly available data to illustrate our method. Results show that errors are smaller in many extreme prediction scores, outperforming a number of existing methods. Our interpretations are consistent with results obtained using a classic econometric model.

Logistic regression as a modelling technique of rare binary dependent variables with much fewer events (ones) than non-events (zeros) tends to underestimate their probability of occurrence. The vast literature devoted to the prediction of rare binary data identifies several ways to improve predictive performance by making modifications to the likelihood estimation. We propose two weighting mechanisms for incorporation in a pseudo-likelihood estimation that improve the predictive capacity of rare binary responses in data collected in complex surveys. We multiply sampling weights by specific correctors that lead to lower root mean square errors for event observations in almost all deciles. A case study is discussed where this method is implemented to predict the probability of suffering a workplace accident in a logistic regression model that is estimated with data from a survey conducted in Ecuador.

The Lee-Carter (LC) stochastic mortality model has been widely used for making future projections of mortality rates. In the framework of the LC model, the response function is non-linear in parameters. Here, we adapt this LC framework to compute conditional quantiles. The LC quantile model can be defined as quantile non-linear regression conditioned to age and the calendar year. Two strategies for estimating coefficients based on interior-point methods are described. We show that the LC quantile model provides additional information to that furnished by the traditional LC conditional mean. An application to Spanish mortality data is reported.

[spa] El objetivo de esta investigación es el diseño y validación de un modelo de Employer Branding (EB), como facilitador de la gestión estratégica del talento y su vinculación con el compromiso organizacional. Para ello se parte conceptualmente de las 4 Ps o Marketing Mix (Mc Carthy, 1956), así como de los principios del Marketing aplicados a las actividades de Recursos Humanos (Edwards, 2010). Para alcanzar dicho objetivo, el estudio se ha llevado a cabo en dos fases. En la primera se identifican, a partir de la literatura científica, los elementos configuradores del modelo, desarrollándose así un nuevo modelo, el EB Mix. Este, inicialmente, fue validado a partir del ajuste interjueces, administrando un cuestionario a 30 expertos que valoraron tanto la pertinencia, con un criterio de V, [eng] The objective of this research is the design and validation of an Employer Branding (EB) model, to facilitate the strategic management of talent and its links to organizational commitment. In order to do so, we based our conceptual foundation on the 4Ps, also known as Marketing Mix (Mc Carthy, 1956), as well as on the principles of Marketing as they apply to Human Resources (Edwards, 2010). In order to achieve our goal, the study has been carried out in two phases. In the first phase, the configuring elements of the model are identified based on the available scientific literature, and used to develop a new model: EB Mix. This was initially validated through the method of inter-rater adjustment, by providing a questionnaire to 30 experts in which they evaluated both relevance, with a criterion of V

Treballs Finals del Doble Grau d'Administració i Direcció d'Empreses i de Matemàtiques, Facultat d'Economia i Empresa i Facultat de Matemàtiques i Informàtica, Universitat de Barcelona, Curs: 2021-2022, Tutor: Josep Vives i Santa Eulàlia i F. Javier Martínez de Albéniz, [en] Time series analysis allows complex processes to be expressed in simple terms to understand how these processes were generated and to predict future values. SARIMA models assume that the observations of a process depend on the previous observations and the variation between them to give an expression of the underlying data generating process. To find the SARIMA model that better fits our data we introduce the Box and Jenkins method, based on three iterative steps: model identification, parameter estimation and fitness check. Once we have identified the most appropriate fitting model, we use it to forecast future values. We have followed this methodology to find the model that best fits the Spanish unemployment series from 2002 to the first quarter of 2022 and to forecast the next 8 observations.

When datasets present long conditional tails on their response variables, algorithms based on Quantile Regression have been widely used to assess extreme quantile behaviors. Value at Risk (VaR) and Conditional Tail Expectation (CTE) allow the evaluation of extreme events to be easily interpretable. The state-of-the-art methodologies to estimate VaR and CTE controlled by covariates are mainly based on linear quantile regression, and usually do not have in consideration non-crossing conditions across VaRs and their associated CTEs. We implement a non-crossing neural network that estimates both statistics simultaneously, for several quantile levels and ensuring a list of non-crossing conditions. We illustrate our method with a household energy consumption dataset from 2015 for quantile levels 0.9, 0.925, 0.95, 0.975 and 0.99, and show its improvements against a Monotone Composite Quantile Regression Neural Network approximation.

Using a high-frequency framework, we show that the Auroba-Diebold-Scotti (ADS) daily business conditions index significantly increases the accuracy of U.S. unemployment nowcasts in real-time. This is of particular relevance in times of recession, such as the Global Financial Crisis and the Covid-19 pandemic, when the unemployment rate is prone to rise steeply. Based on our results, the ADS index presents itself as a better predictor than the financial indicators widely used by the literature and central banks, including both interest and credit spreads and the VXO.

Treballs Finals del Màster de Recerca en Empresa, Facultat d'Economia i Empresa, Universitat de Barcelona. Curs: 2021-2022 ; Tutor: Paloma Miravitlles Matamoros, Development of the global economy makes a deeper level of interaction between the domestic and international markets in recent years. For this reason, more and more firms choose to internationalize part of their value chain in foreign countries, including research and development (R&D) activities. This paper aims to investigate the impact of R&D internationalization of Chinese MNEs on their corporate’s innovation performance in terms of intensity and diversity. Using data of the Top 100 MNEs from China in 2021, I perform an Ordinary Least Square (OLS) regression to test the model. The results indicate that there is a U-shaped relationship between the intensity of R&D internationalization and firm’s innovation performance and an inverted U-shaped relationship for the diversity of R&D internationalization. This paper expands the literature on innovation outcomes of R&D internationalization in the context of emerging economies and also provides guideline for those managers in MNEs to realize the threshold in the process of R&D internationalization to maximize their benefits.

Treballs Finals de Grau en Estadística UB-UPC, Facultat d'Economia i Empresa (UB) i Facultat de Matemàtiques i Estadística (UPC), Curs: 2021-2022, Tutor: Laura Marquès Padreny, Hoy en día tenemos mucha información antes de realizar el alquiler de una vivienda, toda esta información puede contener las claves que ayuden al éxito de sus propietarios, aunque estos no lo puedan ver. Utilizando herramientas de Machine Learning, intentaremos predecir el éxito de las diferentes viviendas y de esta manera podremos ayudar a sus propietarios a tener unas condiciones que resulten más atractivas para el público. Esto implicara obtener una base de datos, realizar las distintas transformaciones necesarias en esta, aplicar un modelo de regresión y finalmente utilizar los resultados obtenidos para hacer una función que pueda ser utilizado por un público no especializado.

[Publicació online. Pendent publicació definitiva (04/11/22)], This paper evaluates the dynamic response of economic activity to shocks in agents' perception of uncertainty. The study focuses on the comparison between manufacturers' and consumers' perception of economic uncertainty, gauged by a geometric discrepancy indicator to quantify the proportion of disagreement in eleven European countries and the Euro Area. A vector autoregressive framework is used to estimate the impulse response functions to innovations in disagreement, both for manufacturers and consumers. The effect on economic activity of shocks to the perception of uncertainty is found to differ markedly between both types of agents. On the one hand, shocks to consumer discrepancy tend to be of greater magnitude and duration than those to manufacturer discrepancy. On the other hand, innovations in disagreement between the two collectives have an opposite effect on economic activity: shocks to manufacturer discrepancy lead to a decrease in economic activity, as opposed to shocks to consumer discrepancy. This finding is of particular relevance to researchers when using cross-sectional dispersion of survey-based expectations for approximating and assessing economic uncertainty, since the effect on economic growth of shocks to disagreement may be dependent on the type of agent and the way in which expectations have been elicited.

Quantile regression provides a way to estimate a driver’s risk of a traffic accident by means of predicting the percentile of observed distance driven above the legal speed limits over a one year time interval, conditional on some given characteristics such as total distance driven, age, gender, percent of urban zone driving and night time driving. This study proposes an approximation of quantile regression coefficients by interpolating only a few quantile levels, which can be chosen carefully from the unconditional empirical distribution function of the response. Choosing the levels before interpolation improves accuracy. This approximation method is convenient for real-time implementation of risky driving identification and provides a fast approximate calculation of a risk score. We illustrate our results with data on 9614 drivers observed over one year.

Aquest projecte tracta sobre el desenvolupament de tècniques estadístiques multivariants per a la predicció de l'ús d'energia, basades en els principals factors que influeixen en el consum d'energia. S'estudiaren diverses estratègies per elaborar models que siguin capaços de pronosticar la variable dependent (ús d'energia) utilitzant diverses variables independents. Els models de predicció proposats són relativament simples i fàcils dimplementar, per tant, poden jugar un paper important en el manteniment del consum d'energia de les instal·lacions i ser aplicats com un mètode universal per al món industrial. Una preocupació principal del treball és que les dades sense processar com a dades de sèries de temps i contenien una gran quantitat de variables independents, com a resultat, es va utilitzar l'anàlisi de components principals (PCA) per reduir la quantitat de paràmetres i simplificar el model. Després es va realitzar la implementació de dos tipus de models de regressió lineal: Model de Regressió Lineal Múltiple (MLR) i Model Lineal Generalitzat (GLM). Per determinar els paràmetres ubicats en aquests dos models es van utilitzar els Mínims Quadrats Ordinaris (OLS) i l'Estimació de Màxima Versemblant (MLE) i se'n van estudiar les diferències. A més, també es va investigar un algorisme de machine learning anomenat Random Forest (RF) per fer una comparació entre els mètodes clàssics i el model de machine learning. Els models proposats presenten les característiques següents: simplicitat, gran aplicabilitat, bon ajustament amb les dades obtingudes i fàcil implementació. Este proyecto trata sobre el desarrollo de técnicas estadísticas multivariantes para la predicción del uso de energía, basadas en los principales factores que influyen en el consumo de energía. Se estudiaron varias estrategias para elaborar modelos que sean capaces de pronosticar la variable dependiente (uso de energía) utilizando varias variables independientes. Los modelos de predicción propuestos son relativamente simples y fáciles de implementar, por lo tanto, pueden jugar un papel importante en el mantenimiento del consumo de energía de las instalaciones y ser aplicados como un método universal para el mundo industrial. Una preocupación principal del trabajo es que los datos sin procesar como datos de series de tiempo y contenían una gran cantidad de variables independientes, como resultado, se utilizó el análisis de componentes principales (PCA) para reducir la cantidad de parámetros y simplificar el modelo. Luego, se realizó la implementación de dos tipos de modelos de regresión lineal: Modelo de Regresión Lineal Múltiple (MLR) y Modelo Lineal Generalizado (GLM). Para determinar los parámetros ubicados en estos dos modelos se utilizaron los Mínimos Cuadrados Ordinarios (OLS) y la Estimación de Máxima Verosimilitud (MLE) y se estudiaron sus diferencias. Además, también se investigó un algoritmo de machine learning llamado Random Forest (RF) para hacer una comparación entre los métodos clásicos y el modelo de machine learning. Los modelos propuestos presentan las siguientes características: simplicidad, gran aplicabilidad, buen ajuste con los datos obtenidos y fácil implementación. This project concerns the development of multivariate statistical techniques for the prediction of energy use, based on the main factors that influence energy consumption. Several strategies were studied to elaborate models that are capable to forecast dependent variable (energy use) using various independent variables. The proposed prediction models are relatively simple and uncomplicated to implement, therefore, can play an important role in the maintenance of facilities' energy consumption and be applied as an universal method for industrial world. A primary concern of the work is that the raw data as time series data and contained huge amount of independent variables, as a result, Principal Component Analysis (PCA) was used to reduce the number of parameters to make the model more simple. Then, implementation of two types of linear regression models was performed: Multiple Linear Regression model (MLR) and Generalized Linear Model (GLM). In order to determine the parameters located in these two models, the Ordinary Least Squares (OLS) and Maximum likelihood estimation (MLE) were utilized and their differences were studied. In addition, a machine-learning algorithm named Random Forest (RF) was also investigated to make an comparison between classical methods and machine-learning model. The proposed models present the following characteristics: simplicity, large applicability, good match with obtained data and easy deployment.

Bilinear regression models involving a nonlinear interaction term are applied in many fields (e.g., Goodman’s RC model, Lee-Carter mortality model or CAPM financial model). In many of these contexts data often exhibit extreme values. We propose the use of bilinear models to estimate the median of the conditional distribution in the presence of extreme values. The aim of this paper is to provide alternative methods to estimate median bilinear models. A calibration strategy based on an iterative estimation process of a sequence of median linear regression is developed. Mean and median bilinear models are compared in two applications with extreme observations. The first application deals with simulated data. The second application refers to Spanish mortality data involving years with atypical high mortality (Spanish flu, civil war and HIV/AIDS). The performance of the median bilinear model was superior to that of the mean bilinear model. Median bilinear models may be a good alternative to mean bilinear models in the presence of extreme values when the centre of the conditional distribution is of interest. &nbsp

Evaluating value at risk (VaR) for a firm's returns during periods of financial turmoil is a challenging task because of the high volatility in the market. We propose estimating conditional VaR and expected shortfall (ES) for a given firm's returns using quantile regression with cross-sectional (CSQR) data about other firms operating in the same market. An evaluation using US market data between 2000 and 2020 shows that our approach has certain advantages over a CAViaR model. Identification of low-risk firms and a reduction in computing times are additional advantages of the new method described.

Within the much broader framework of global interest, the dilemma concerning the real impact of mode of transport on the spread of COVID-19 has been a priority for transport stakeholders and policy-makers. How dangerous is it to move around a certain territory? Does the danger depend on the mode of transport? By considering a novel and detailed dataset at the level of local labour markets, we analysed the spatial association between the pre-pandemic propensity to use public transport and excess mortality in Italy attributable to the initial spread of COVID-19. We found that places characterised by larger commuting flows exhibit higher excess mortality during the first wave of the pandemic, but observed no significant spatial association between excess mortality and transit usage. Our results were obtained by considering a wide range of heterogeneity in the estimation of quantile regressions across a variety of specifications. Although we do not provide a definitive answer concerning the risk associated with transit use, our analysis suggests that mobility, not modal choice, should be considered a main driver of the initial contagion.

Treballs Finals del Grau d'Economia i Estadística. Doble titulació interuniversitària, Universitat de Barcelona i Universitat Politècnica de Catalunya. Curs: 2020-2021. Tutor: Montserrat Guillen i Estany, Els objectius d’aquest treball són construir un model de regressió lineal múltiple per tal de determinar el valor de mercat dels futbolistes professionals a partir de mètriques de rendiment futbolístic i construir un model de predicció de gol que servirà com a base de l’Expected Goals Method. Les dades que s’han utilitzat fan referència a les 5 lligues europees principals (lligues anglesa, espanyola, alemanya, italiana i francesa) de la temporada 2017/2018 proporcionades per l’empresa Wyscout i la web Transfermarkt. La part inicial del treball està formada per la introducció, la revisió de la literatura i l’apartat de metodologia i dades. En aquests apartats introductoris es posa en context al lector i es justifica el tema, es revisen els treballs més recents d’altres autors i s’explica l’estructura de les dades per a la realització del treball i la seva extracció, i la metodologia emprada. Arribats a aquest punt, en la primera part dels resultats es construeix el model de predicció de gol, on s’avaluen un total de 40.461 xuts. S’ha trobat evidència estadística que la distància i l’angle de tir, el rol del jugador, la cama o part del cos amb la qual s’efectuï el tir, que la jugada procedeixi d’un contra-atac o d’una centrada són factors que determinen la probabilitat que un xut acabi essent gol. Finalment, pel que fa a la segona part, s’ha mesurat el valor de mercat de 2.662 jugadors, on s’ha fet una distinció segons el rol de joc (davanters, mig-campistes, defenses i porters). Per a cadascun d’aquests s’ha pogut elaborar un model de regressió per estimació robusta on les variables explicatives són mètriques de rendiment futbolístic específiques depenent del rol que es tracti.

Based on recent developments in joint regression models for quantile and expected shortfall, this paper seeks to develop models to analyse the risk in the right tail of the distribution of non-negative dependent random variables. We propose an algorithm to estimate conditional tail expectation regressions, introducing generalized risk regression models with link functions that are similar to those in generalized linear models. To preserve the natural ordering of risk measures conditional on a set of covariates, we add extra non-negative terms to the quantile regression. A case using telematics data in motor insurance illustrates the practical implementation of predictive risk models and their potential usefulness in actuarial analysis.

The question of whether private organizations can outperform public ones in public service delivery has been a major topic of interest over the last few decades. However, the empirical evidence does not systematically support the hypothesis of lower costs and higher efficiency when private organizations deliver public services. To better understand the cost and efficiency differences between public and private organizations, we have conducted a meta-regression analysis of econometric studies relating to hospital ownership and performance. We have analyzed 61 estimates extracted from health studies, using public versus private hospital costs and efficiency as an independent variable. Our analysis shows a genuine true effect in favor of public sector hospitals. Although we found evidence that the public sector may provide public health services more cheaply than the private sector, the latter achieves better productive efficiency. We discuss how this divergence in results is affected by factors such as country, year, the use of panel data in a study, the extent to which performance is measured by examining financial costs, and the inclusion of not-for-profit hospitals.

Banking is becoming branchless, contemporary and digital at a very fast pace. As banks compete to gain a competitive advantage, the need to manage large amounts of data and analytics becomes more relevant as well as the optimization of resources. Within this sector, one of the things they want to optimize is the volume of cash that different tellers will need to handle customer demand. And it is precisely on this aspect that this thesis will be based: to predict the volume of money needed to cover all the withdrawals demanded by customers at the ATMs using different machine learning models.

Treballs Finals de Grau de Matemàtiques, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2021, Director: Josep Fortiana Gregori, [en] Lasso (Least Absolute Shrinkage and Selection Operator) is a regression method that performs both regularization and variable selection, improving prediction accuracy and interpretability of the resulting model. In this project we follow evolution from the plain linear model, through Ridge regression, for many years the most popular technique to improve the precision of predictions, to Lasso. We delve into numerical procedures for calculating Lasso solutions: coordinate descent and LARS. We see some extensions of Lasso such as Elastic Net regression, a neat improvement when optimality fails. We illustrate these methods with several real data examples using the R programming language (see notebooks and HTML files in appendices to the main text).

Treballs Finals de Grau de Matemàtiques, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2021, Director: Josep Vives i Santa Eulàlia i Carmen Florit i Selma, [en] In this project we will study some methods of multivariate analysis and give some interesting examples. In particular, we will treat the following methods: • Multiple linear regression model: which will allow us to relate a random variable to a random vector and analyze its most important applications in Econometrics. • Canonical correlation analysis: which will allow us to relate two random vectors. • Principal component analysis, which allows us to transform a random vector of dimension n to dimension 2 or 3 so that it can be represented graphically. • Discriminant analysis, which given two populations, tells us where and individual belongs to. All of this will allow us to have a very strong idea about the different methods of multivariate analysis.

Treballs Finals de Grau d'Enginyeria Informàtica, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2021, Director: Simone Balocco, [en] The goal of this study is to demonstrate if failures reported in an aircraft can be related to the environmental conditions during operation time. The current study is the first step of a long-term predictive maintenance project driven by the company DMD Solutions. First of all, the concepts of reliability and predictive maintenance are introduced. Furthermore, the fundamentals of machine learning and the state of the art are detailed. Gathering quality data was a complex process, since the available data was incomplete, noisy and unbalanced. The analysis proposes and compares several solutions. Two different approaches were carried out: the first one consisted of the prediction of failure (binary classification), and the second one, more ambitious, the prediction of the time before the next defect using time intervals (multi-class classification). Both approaches were designed using an iterative process that improved quality of both models and data at each stage of the study. The obtained results were promising and encourage further research.

Treballs Finals de Grau d'Enginyeria Informàtica, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2021, Director: Jordi Vitrià i Marca, [en] The appearance of the COVID-19 pandemic has urged governments from around the world to resort to the scientific community in order to predict how the spread of the virus would evolve. In Catalonia, the need to compare the two most used models arose to decide on whether restrictions should be applied or not. The Cata- lan government has mainly used the Gompertz model to predict the evolution of the infected population but some scientists have shown concern for this model as compared to the more traditional deterministic compartmental models. This created the need to compare the effectiveness of these two models and determine which of them is more useful for predicting future pandemics. In this project, we will present and analyse both types of models. In the case of compartmental models we will summarise only the three most basic types, SIR,SEIR and SEIRS. We will determine which of these models is most effective in predicting the spread of the virus by retrospectively comparing the predicted values and the real ones obtained. We will use data from the database provided by the Catalan government, Dades Obertes, and train our models with data from 30 days in order to predict two weeks in advance. Our results have shown that SEIR is the best of the initially proposed methods but is clearly affected by noise in our data, especially after restrictions were applied. A clear improvement was shown by averaging the values from the Gompertz model together with the values of the SEIR model. We clearly obtained the best possible model from combining both of them as it is smoother than predicting with SEIR and yields better results than the Gompertz model. In future work, we plan to include more complex compartmental models, such as the ones including geographical and transportation factors, given the necessary data. In addition, a larger average of models can be used, in which more than 2 models is applied, for an even more precise prediction of the evolution of pandemics.

Treballs Finals de Grau d'Enginyeria Informàtica, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2021, Director: Santi Seguí Mesquida, [en] In this work, a solution is presented for the real-estate sector, which can be implemented not only for professionals but also for individuals. Due to the difficulty that entails defining the rental price of a property and the fact that the criteria used to this end are always variable, a technological solution was provided in order to facilitate this challenge. The main aim of this work remains on establishing a system that will be able of predicting, through its features, the rental price of a house located in Barcelona. On account of the criteria already mentioned, which fluctuates depending on the other houses, it was intended to elaborate a system that could be able to be fed consistently with the most recent examples from the market. For this purpose, it has been followed the typical processes of data science discipline and machine learning, as well as the techniques for the data mining contained in a web environment. Obtaining information has been achieved using a program based on “web scraping”, capable of surfing the website of a real estate portal and drawing all the attributes of the several published advertisements. The analysis and the data manipulation, along with the implementation of the predictive models, has been performed with the “Jupyter Notebook” environment. For the code development of all the described functionalities, it has been used the “Python” language, together with various libraries targeted towards these objectives. Concerning the data analysis, it has been explained in this document all considerations about data cleansing, data transformation and other modifications that have been required for the purpose of obtaining data quality. Furthermore, there is also an explanation about the different predictive models which have been created, and the results returned. At the end of this document, the results obtained are compared and the conclusions are provided, in which it is considered whether these results fulfil the ultimate purpose of the work.

Treballs Finals de Grau de Matemàtiques, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2021, Director: F. Javier Martínez de Albéniz, [en] In this work we study the competition between two retailers that have to choose which of the available products to sell and at which price, knowing that each customer relates a product and its price with a value and will only acquire the product with highest value. In the first part of this project we present our framework by defining some of the basic concepts of Game Theory and introducing new terminology. Moreover, we also review two of the pioneering models in duopolistic competition: the Cournot and Bertrand model. The second part of the project formally defines and models the competitive situation. We formally establish the actions available for both retailers and determine how each customer ranks all products for sale. Moreover, in this chapter we also study the effects that this competition has, the possible equilibria of this competitive market and under which circumstances both retailers can reach such equilibria. The third part of the project further studies the relationship between the price a retailer sets for his products and whether the product is common or exclusive. We attempt to model this relationship with a linear regression model and we study the results with data from clothing retailers Gucci and Farfetch.