Workshop
One-day Workshop


Download details 		Prof. Dr. Wolfgang Härdle (Humboldt University, Berlin) and Dr. Gerhard Stahl (Talanx, Hannover) will present a one-day workshop on

"Risk Management Theory and Applications: Extremes, Joint extremes, Copulae"


On 1 July, 2008, immediately preceding ISBIS-2008.
Location Andels Hotel Prague - SAPPHIRE Room
http://www.andelshotel.com
Intended audience The workshop is intended for practitioners in banks, UCITS and insurance companies. Furthermore, academics working on practical applications of statistics will benefit from attending the workshop.
Workshop Fees
Before May 1, 2008: CZK 13250 / EUR 500
After May 1, 2008: CZK 14575 / EUR 550
On-site (after June 26, 2008): CZK 15900 / EUR 600
Intended audience The workshop is intended for practitioners in banks, UCITS and insurance companies. Furthermore, academics working on practical applications of statistics will benefit from attending the workshop.
About the presenters Wolfgang Härdle has worked with the finance industry for nearly 15 years as a consultant, specialising in computer packages and advanced numerical methods. He has given many lectures and training workshops on quantitative finance to industry practitioners around the world.

Currently, he is Professor of Statistics in the School of Economics and Business at the Humboldt University, Berlin. He has extensive cooperative research and consultancy engagements with the European banking industry and with the Treasury in Frankfurt.

He is an author of several books on statistical methods in finance and insurance. Recent consultancies and research have included CAT bond pricing, Value at Risk Calculations with GH distributions, and Independent Component Analysis and Recursive portfolio optimization with CART - Classification and Regression Trees. He currently develops a copula based model for market risk management and CDO pricing.

Gerhard Stahl is Deputy Chief Risk Officer at the Talanx Group, Hannover. He is involved with the implementation of an ERM. He also heads the Quantitative Risk Management unit of the Talanx.

Gerhard was Senior Director and Head of the Risk Modelling Group at the BaFin (German Federal Financial Supervisory Authority) based in Bonn. He has practised regulation, supervision and auditing of internal models at banks since 1995. His group has been responsible for the supervision of internal models across all financial sectors in Germany. His responsibilities cover market, credit, liquidity and operational risk as far as stochastic models are involved.

Gerhard holds an honorary doctoral degree from the University of Bamberg for his scientific contributions to financial risk management. Two of his scientific research papers have been awarded "best of" in Risk Magazine and Journal of Risk and both have been reprinted in special Risk Books.
Overview The simple procedure of approximating risk factors with multivariate normal distributions underestimates the prevailing risk level and fails to mimic the stylized facts observed in financial markets. We present modern risk management methodologies that rely on a fine tail analysis of the statistical data features. A joint analysis of the portfolio elements leads us to copulae and the tail Value-at-Risk (VaR). This w is practically oriented with integrated real life software examples that will demonstrate the improved accuracy of modern VaR analysis. Applications will also cover model risk in credit and economic scenario generation.
Statistical Fundamentals I (Härdle) Modelling the returns of a single asset using a GARCH process with normally distributed innovations yields a first approximation to the observable stylized facts. The heavy-tailed distributional properties of the returns though, is missing, even though the GARCH process models the volatility's heteroscedasticity. Thus heavy-tailed distributions have attracted much attention. The probability of extreme values largely depends on how slowly the probability density function of the innovations goes to zero as the values of observations approach infinity. The rate at which it diminishes must be estimated from the data. Since extreme observations are rare, this produces a difficult estimation problem. Even large data sets contain only limited information on the true probability of an extreme loss (profit). In such a situation methods from extreme value statistics produce a more realistic estimate of the risk. We will give an introductory overview of the basic ideas and several of the latest applications as well.
Statistical Fundamentals II (Härdle) A generalized representation of the dependence of the risk factors can be obtained using a Copula. We will concentrate on the representation and concepts of copulae and discuss their applications in calculating the Value-at-Risk (VaR). Additionally we will show the concept of tail dependence, which refers to the degree of dependence in the corner of the lower-left quadrant or upper right quadrant of a bivariate distribution. Tail-dependent distributions are of interest in the context of VaR estimation for asset portfolio since these distributions, contrary to multinormal distributions, can model dependence of large loss events between different assets.
Applications I (Härdle) In the first application session we will discuss the recent advances in risk management methodology. First we will show a realistic and fast method, GHICA, which overcomes the limitations in multivariate risk analysis. The idea is to first retrieve independent components (ICs) out of the observed high-dimensional time series and then individually and adaptively fit the resulting ICs in the generalized hyperbolic (GH) distributional framework. Second, we will show the usage of copulae functions with adaptively estimated time varying parameters (ADACOP) for modelling the distribution of returns, free from the usual normality dependence assumptions. Both methods are applied for VaR calculations of high dimensional stock portfolios.
Applications II (Stahl) The first part of session II is devoted to the topic of Model Risk in credit and operational risk modelling. The presentation covers an indebt study of real life examples from German experience of auditing internal models for these risk categories. For both cases the impact of data quality is highlighted and analysed. This highlights to the changing role of auditors for such models.

The second part of session II is devoted to the problem of economic scenario generators. This is done in the context of the future Solvency II project, where insurance companies will use these generators in order to catch their market risks in the asset side. An analysis and description of models is given which are used in practice. They cover advanced volatility models with jumps and rather advanced models for interest rate curves. Again the results are compared to models from other providers and show the importance of an indebt understanding of all the bells and whistles in such a model.