Catastrophe Risk

Losses due to natural catastrophes (CAT losses) are rare events with very high impact. The CAT loss distributions exhibit highly non-normal (fat-tailed) behavior. The company specific distributions depend on the geography of the insured properties and businesses. As a result, such losses are notoriously difficult to model reliably, and the distributions are provided by just a few vendors who specialize in CAT risk. When the distribution is available however, it is relatively easy to include the CAT risk into the overall framework of ERM. Within the simulation model, we need to add catastrophic losses drawn from the CAT distribution to the losses generated in our regular model (3.2.15). This procedure will result in adjusted ai (and possibly γ k , if we assume that CAT loss development is different from that of the regular losses) in such catastrophic scenarios. In addition, the CAT scenarios will have significant reinsurance receivables, hence additional credit risk (see below). The real difficulty with catastrophic risk within a Monte Carlo approach stems from the necessity to estimate the risk measures dominated by rare events. In Appendix D:, we show how application of importance sampling and robust estimators can overcome these problems.