A comparison of loss aggregation methods for operational risk

Grigory Temnov, Richard Warnung

Abstract

ABSTRACT

In this paper we address the problem of modeling and measuring operational risk. First, we introduce general ideas about operational risk: a brief description of the methodology for selecting models for loss severity and frequency is given. Then we turn to the problem of loss aggregation. Coming from insurance mathematics, there are, in general, three popular and commonly used ways to aggregate operational risk: a Monte Carlo approach, an approach based on the Fourier transformation and a recursion approach. These three methods and their applications are therefore discussed in this paper. For the recursion approach a reinterpretation of the portfolio credit risk model CreditRisk+ is used. The three methods are compared concerning speed and accuracy.