An Assessment of Operational Loss Data and Its Implications for Risk Capital Modeling

Abstract

A mathematical method based on a special dimensional transformation is employed to assess operational loss data from a new perspective. The procedure, which is formally known as the Buckingham (Pi) Theorem, is used broadly in the field of experimental engineering to extrapolate the results of tests conducted on models to prototypes. When applied to the operational loss data considered in this paper, the approach leads to a seemingly universal trend underlying the resulting distributions regardless of how the data set is divided (e.g., by event type, business line, revenue band). This dominating trend, which appears to also acquire a tail parameter of 1, could have profound implications for how operational risk capital is computed.