Berlin 2008 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

AKSOE: Arbeitskreis Physik sozio-ökonomischer Systeme

AKSOE 13: Poster Session (posters on display 10:00-19:00)

AKSOE 13.16: Poster

Mittwoch, 27. Februar 2008, 17:30–19:00, Poster G

Correlation problem in economic capital issues of operatioal risk — •Chitro Majumdar — i-flex Inc.

In Operational Risk we need to estimate loss distributions for 56 Business-Event type combinations (7*8 matrix). Loss Distribution is a combination of frequency dist and severity dist. Each of the 56 cells will have their own frequency and severity dist. Now the problem is to aggregate the Loss Dist across different cells. Currently Basel II recommends simple addition but this is too conservative. So the problem is to determine correlation across Frequency and Severity dist. Currently in the industry there are no standard methods for severity dist aggregation. The practice is to use frequency dist aggregation. Aggregation of frequency is done using copulas. Gaussian/Frank/Gumbel and Clayton are some of the possibilities. But all of these would require estimation of some form of correlation. So the problem is how to find out this frequency correlation? volatility measures the uncertainty of returns, beta measures how much an individual asset is likely to move with the general market and Value at Risk, which is a recent innovation, measures the maximum loss (in the probabilistic sense) that is likely to be occurred in the immediate future. Given the distribution of the risk factors, their Tail Correlations and the Functional Relationship between Loss Metric for the Cluster and underlying factors, we perform a Monte Carlo simulation using Cholesky Factorization, to include correlation effects, to generate the Loss Distribution of the cluster.