<p>This thesis attempts to describe the essential systems features of a complex real-world domain of operational risk (OR) in banking, by employing general systems theory (GST) as the guiding method. An implementational framework (SAFOR) is presented for operational risk management (ORM), the target of which is to manage and mitigate the risk-around-loss causes. Since reasoning about OR is often scenario based, the framework also includes methods for decision making in addition to Value at Risk (VaR) and Conditional Value at Risk (CVaR). Other computational models that yield prediction intervals are discussed as well. Because the banking industry is one of the most mature sectors when it comes to OR, and contains the most data points, the discussion in this thesis evolves around such institutions. The present state-of-the-art in OR management for banking is surveyed using a systemic-holistic approach and the model framework is presented against this discussion. Tools and concepts from systems theory and systems thinking are employed for assessing systems properties and gaining insights into the interaction of various components. This brings about a number of advantages. This is not in disagreement with current suggestions such as those of the Basle Committee (Basel II), which is doing an excellent job in proving the state-of-the-art in best practice for banking institutions. Rather, this thesis offers a complementary perspective, looking at essentially the same problems but in a broader context and with a differing view.</p><p>OR data has been hard to come by in banking. Confidentiality and difficulties in quantifying OR as well as the short time data has been gathered in a consistent way are some of the reasons for this. Therefore, no case study has been done. Instead, we have chosen to look into a published bank application of an advanced OR model. The application shows that the technique holds as validation of the SAFOR modules.</p>

This thesis attempts to describe the essential systems features of a complex real-world domain of operational risk (OR) in banking, by employing general systems theory (GST) as the guiding method. An implementational framework (SAFOR) is presented for operational risk management (ORM), the target of which is to manage and mitigate the risk-around-loss causes. Since reasoning about OR is often scenario based, the framework also includes methods for decision making in addition to Value at Risk (VaR) and Conditional Value at Risk (CVaR). Other computational models that yield prediction intervals are discussed as well. Because the banking industry is one of the most mature sectors when it comes to OR, and contains the most data points, the discussion in this thesis evolves around such institutions. The present state-of-the-art in OR management for banking is surveyed using a systemic-holistic approach and the model framework is presented against this discussion. Tools and concepts from systems theory and systems thinking are employed for assessing systems properties and gaining insights into the interaction of various components. This brings about a number of advantages. This is not in disagreement with current suggestions such as those of the Basle Committee (Basel II), which is doing an excellent job in proving the state-of-the-art in best practice for banking institutions. Rather, this thesis offers a complementary perspective, looking at essentially the same problems but in a broader context and with a differing view. OR data has been hard to come by in banking. Confidentiality and difficulties in quantifying OR as well as the short time data has been gathered in a consistent way are some of the reasons for this. Therefore, no case study has been done. Instead, we have chosen to look into a published bank application of an advanced OR model. The application shows that the technique holds as validation of the SAFOR modules.

L’intérêt des modèles non-linéaires réside, d’une part, dans une meilleure prise en compte des non-linéaritéscaractérisant les séries macroéconomiques et financières et, d’autre part, dans une prévision plus riche en information.A ce niveau, l’originalité des intervalles (asymétriques et/ou discontinus) et des densités de prévision (asymétriqueset/ou multimodales) offerts par cette nouvelle forme de modélisation suggère qu’une amélioration de la prévisionrelativement aux modèles linéaires est alors possible et qu’il faut disposer de tests d’évaluation assez puissants pourvérifier cette éventuelle amélioration. Ces tests reviennent généralement à vérifier des hypothèses distributionnellessur les processus des violations et des transformées probabilistes associés respectivement à chacune de ces formes deprévision. Dans cette thèse, nous avons adapté le cadre GMM fondé sur les polynômes orthonormaux conçu parBontemps et Meddahi (2005, 2012) pour tester l’adéquation à certaines lois de probabilité, une approche déjà initiéepar Candelon et al. (2011) dans le cadre de l’évaluation de la Value-at-Risk. Outre la simplicité et la robustesse de laméthode, les tests développés présentent de bonnes propriétés en termes de tailles et de puissances. L’utilisation denotre nouvelle approche dans la comparaison de modèles linéaires et de modèles non-linéaires lors d’une analyseempirique a confirmé l’idée selon laquelle les premiers sont préférés si l’objectif est le calcul de simples prévisionsponctuelles tandis que les derniers sont les plus appropriés pour rendre compte de l'incertitude autour de celles-ci. / The interest of non-linear models is, on the one hand, to better take into account non-linearities characterizing themacroeconomic and financial series and, on the other hand, to get richer information in forecast. At this level,originality intervals (asymmetric and / or discontinuous) and forecasts densities (asymmetric and / or multimodal)offered by this new modelling form suggests that improving forecasts according to linear models is possible and thatwe should have enough powerful tests of evaluation to check this possible improvement. Such tests usually meanchecking distributional assumptions on violations and probability integral transform processes respectively associatedto each of these forms of forecast. In this thesis, we have adapted the GMM framework based on orthonormalpolynomials designed by Bontemps and Meddahi (2005, 2012) to test for some probability distributions, an approachalready adopted by Candelon et al. (2011) in the context of backtesting Value-at-Risk. In addition to the simplicity androbustness of the method, the tests we have developed have good properties in terms of size and power. The use of ournew approach in comparison of linear and non-linear models in an empirical analysis confirmed the idea according towhich the former are preferred if the goal is the calculation of simple point forecasts while the latter are moreappropriated to report the uncertainty around them.

Connus sous le nom de Systèmes d’Alerte Avancés, ou Early Warning Systems (EWS), les modèles de prévision des crises financières sont appelés à jouer un rôle déterminant dans l’orientation des politiques économiques tant au niveau microéconomique qu’au niveau macroéconomique et international. Or,dans le sillage de la crise financière mondiale, des questions majeures se posent sur leur réelle capacité prédictive. Deux principales problématiques émergent dans le cadre de cette littérature : comment évaluer les capacités prédictives des EWS et comment les améliorer ?Cette thèse d’économétrie appliquée vise à proposer (i) une méthode d’évaluation systématique des capacités prédictives des EWS et (ii) de nouvelles spécifications d’EWS visant à améliorer leurs performances. Ce travail comporte quatre chapitres. Le premier propose un test original d’évaluation des prévisions par intervalles de confiance fondé sur l’hypothèse de distribution binomiale du processus de violations. Le deuxième chapitre propose une stratégie d’évaluation économétrique des capacités prédictives des EWS. Nous montrons que cette évaluation doit être fondée sur la détermination d’un seuil optimal sur les probabilités prévues d’apparition des crises ainsi que sur la comparaison des modèles.Le troisième chapitre révèle que la dynamique des crises (la persistance) est un élément essentiel de la spécification économétrique des EWS. Les résultats montrent en particulier que les modèles de type logit dynamiques présentent de bien meilleurs capacités prédictives que les modèles statiques et que les modèles de type Markoviens. Enfin, dans le quatrième chapitre nous proposons un modèle original de type probit dynamique multivarié qui permet d’analyser les schémas de causalité intervenant entre différents types crises (bancaires, de change et de dette). L’illustration empirique montre clairement que le passage à une modélisation trivariée améliore sensiblement les prévisions pour les pays qui connaissent les trois types de crises. / Known as Early Warning Systems (EWS), financial crises forecasting models play a key role in definingeconomic policies at microeconomic, macroeconomic and international level. However, in the wake ofthe global financial crisis, numerous questions with respect to their forecasting abilities have been raised,as very few signals were drawn prior to the starting of the turmoil. Two questions arise in this context:how to evaluate EWS forecasting abilities and how to improve them?The broad goal of this applied econometrics dissertation is hence (i) to propose a systematic model-free evaluation methodology for the forecasting abilities of EWS as well as (ii) to introduce new EWSspecifications with improved out-of-sample performance. This work has been concretized in four chapters.The first chapter introduces a new approach to evaluate interval forecasts which relies on the binomialdistributional assumption of the violations series. The second chapter proposes an econometric evaluationmethodology of the forecasting abilities of an EWS. We show that adequate evaluation must take intoaccount the cut-off both in the optimal crisis forecast step and in the model comparison step. The thirdchapter points out that crisis dynamics (persistence) is essential for the econometric specification of anEWS. Indeed, dynamic logit models lead to better out-of-sample forecasting probabilities than those oftheir main competitors (static model and Markov-switching one). Finally, a multivariate dynamic probitEWS is proposed in the fourth chapter to take into account the causality between different types of crises(banking, currency, sovereign debt). The empirical application shows that the trivariate model improvesforecasts for countries that underwent the three types of crises.