A semi-parametric approach to estimating the operational risk and Expected Shortfall |
| |
Authors: | Ainura Tursunalieva |
| |
Institution: | Department of Econometrics and Business Statistics, Monash University, Caulfield East, VIC 3145, Australia |
| |
Abstract: | In this article, we propose improvements to the peak-over-threshold (POT) method and apply this improved method for modelling US business operational losses and estimating operational risks (ORs). In the widely used traditional POT method, the generalized Pareto distribution (GPD) is fitted to severity losses, while an empirical distribution is fitted to small to medium losses. Then, the Expected Loss and the 99.9% operational value-at-risk (OpVaR) are estimated. Additionally, the Expected Shortfall (ES) – a coherent risk measure – is estimated in this article as an alternative to OpVaR. These risk measures constitute the levels of regulatory and economic capitals to cover risks. With the improved POT method, the risks can be estimated more accurately than with the traditional POT method. The results indicate that the OpVaR are much lower than the ES and that the larger the tail losses the greater the difference between these two risk measures. Our findings imply that the ES would provide higher levels of capitals to cover risks than would the OpVaR, particularly during crises, and they have implications for the efficient OR management and regulators. |
| |
Keywords: | heavy-tailed distribution tail losses generalized Pareto distribution OpVaR intervals Expected Shortfall |
|
|