The Modelling of Operational Risk : Experience with the Analysis of the Data Collected by the Basel Committee

Since 2001, the Risk Management Group of the Basel Committee has been performing specific surveys of banks' operational loss data. The second loss data collection was launched in the summer of 2002: The 89 banks participating in the exercise provided the Group with more than 47,000 observations, grouped by eight standardised Business Lines and seven Event Types.The objective of this paper is to increase the level of understanding of operational risk within the financial system, by exploring the statistical behaviour of those data and going on to derive bottom-up capital charge figures and Gross Income related coefficients for the Business Lines envisaged in the revised framework of the Capital Accord.The work aims, first of all, to compare the sensitivity of conventional actuarial distributions and models stemming from the Extreme Value Theory in representing the highest percentiles of the data sets: The exercise shows that the extreme value model, in its Peaks Over Threshold representation, explains the behaviour of the operational risk data in the tail area well. Then, measures of severity and frequency of the large losses are gained and, by a proper combination of these estimates, a bottom-up operational risk capital figure is computed for each Business Line. Finally the contributions of the expected losses to the capital figures are evaluated and the relationships between the Business Lines capital charges and the corresponding average level of Gross Incomes are determined and compared with the current coefficients envisaged in the simplified approaches of the regulatory framework.To the end of promoting the development of a pragmatic and effective dialogue on the operational risk measurement issues between regulators and institutions in the course of the implementation of the new Capital framework, feedback from academics and practitioners on the topics dealt with in the paper is welcome. Specifically, the statistical methodologies implemented to model the data and the outcomes of the analysis, in terms of the capital charges and coefficients of the Business Lines, are the topics on which comments should be focused.For assistance in the feedback process and in order to submit comments, please refer to the procedures set by your country's competent national supervisory authority or central bank. Comments should also be sent, preferably via e-mail, to the author