Abstract:

Being still in its early stages, the modeling of operational risk has mainly concentrated on marginal distributions and linear correlations in the Loss Distribution Approach (LDA). Having access to a real--world data set, we analyze the effects of competing strategies for dependence modeling. In particular, we estimate tail dependence both via copulas as well as nonparametric tail dependence estimators, and analyze its effect on aggregate risk capital estimates, providing methodological guidelines for a real-world implementation.

Empirical results show that the presence of tail dependence can lead to larger regulatory capital than Basel II standard approach, pointing out the need of a complete framework for dependence modeling to compute prudential and realistic risk capital estimates.

Bio:
Sandra Paterlini is Assistant Professor in Statistics at the Faculty of Economics, University of Modena and Reggio E. since 2002. She is a member of the ERCIM (European Consortium for Informatics and Mathematics) Working group in Optimization Heuristics in Estimation and Modelling and Vice-chair of IEEE Task Force on Portfolio Optimization. She holds a MSc Financial Mathematics from the University of Warwick (UK) and a PhD in Computational Methods for Financial Forecasting and Economic Decision from the University of Bergamo (IT). She has been visiting scientist at EVALife Group, Dept. of Computer Science, University of Aarhus (DK), at the Chair of Financial Econometrics, Department of Statistics, Ludwig-Maximilians-University Munich and at the School of Mathematics, UMN (USA). Her research interests are on financial times series modelling and forecasting, optimization heuristics for statistical modelling and estimation problems, financial portfolio optimization and risk management.