Post by Sapphire Capital on Jul 11, 2008 7:27:46 GMT 4
Operational Risks in Banks: An Analysis of Empirical Data from a Bank
JOHN R. EVANS
University of New South Wales - Australian School of Business
--------------------------------------------------------------------------------
UNSW Australian School of Business Research Paper No. 2008ACTL01
Abstract:
This paper reports the results of an empirical analysis of operational risk in a bank and derives a model to represent the distribution of losses. Comparisons are made with models traditionally used to model operational risk. The paper concentrates more on the severity issue rather than the frequency issue. Several interesting findings are discussed. Several goodness-of-fit techniques are discussed with respect to their ability to assess tail fit. The heavy tailed generalised Pareto distribution (GPD) provides a better fit than the lighter tailed lognormal distribution. This fit is then improved by fitting the body and tail of the data to different distributions. The generalised extreme value distribution (GEV) is shown to provide a good fit to the annual loss distribution, Maximum likelihood and probability weighted moments methods are compared when analysing these models. We found lower VaR estimates for a non-US bank than those reported for US banks in contrast to previous findings.
papers.ssrn.com/sol3/papers.cfm?abstract_id=1133686
JOHN R. EVANS
University of New South Wales - Australian School of Business
--------------------------------------------------------------------------------
UNSW Australian School of Business Research Paper No. 2008ACTL01
Abstract:
This paper reports the results of an empirical analysis of operational risk in a bank and derives a model to represent the distribution of losses. Comparisons are made with models traditionally used to model operational risk. The paper concentrates more on the severity issue rather than the frequency issue. Several interesting findings are discussed. Several goodness-of-fit techniques are discussed with respect to their ability to assess tail fit. The heavy tailed generalised Pareto distribution (GPD) provides a better fit than the lighter tailed lognormal distribution. This fit is then improved by fitting the body and tail of the data to different distributions. The generalised extreme value distribution (GEV) is shown to provide a good fit to the annual loss distribution, Maximum likelihood and probability weighted moments methods are compared when analysing these models. We found lower VaR estimates for a non-US bank than those reported for US banks in contrast to previous findings.
papers.ssrn.com/sol3/papers.cfm?abstract_id=1133686