Operational risk is being considered as an important risk component for financial institutions as evinced by the large sums of capital that are allocated to mitigate this risk. Therefore, risk measurement is of paramount concern for the purposes of capital allocation, hedging, and new product development for risk mitigation. We perform a comprehensive evaluation of commonly used methods and introduce new techniques to measure this risk with respect to various criteria. We find that our newly introduced techniques perform consistently better than the other models we tested.
This paper was revised in March 2007 to include corrections to citations, formulas, typographical errors, and technical points based on numerous comments. All of the tables, analyses, results, and conclusions have remained entirely consistent with the previous draft as no new analysis was performed. The authors wish to thank those who have provided comments in an effort to improve the quality and readability of the paper.
This paper was revised in April 2007.
Keywords: exploratory data analysis, operational risk, g-and-h distribution, goodness-of-fit, skewness-kurtosis, risk measurement, extreme value theory, peak-over-threshold method, generalized Pareto distribution, truncated lognormal distribution, loglogistic distribution
JEL Classifications: G10, G20, G21, D81
PDF version of Appendix
Tables with different exposure indicators used to scale the data (gross income and total assets) and that include some business lines and event types excluded from the original paper.