Simulating total operational risk using retrospective information and subjective assessments - a Bayesian approach

Abstract

Quantifying operational risk exposure typically involves gathering information from several sources, including historical data as well as subjective assessments. Using historical data one can estimate both an incident frequency distribution, as well as an incident consequence distribution. Based on these two distributions a simulation model can be established. However, by limiting the focus to data related to incidents which may reappear in the future, one is often left with a relatively short incident history. In order to improve the risk quantification, it is often necessary to include subjective risk assessments as well. In the present paper we propose a model for how to combine these two sources of information. The model can be used to represent situations ranging from cases where the two sources are disjoint, overlap completely as well as intermediate cases where the two sources are partially overlapping. The model is illustrated by considering a numerical example. In this example we vary the degree of overlap between the sources of information.