Review of two Japan Typhoon catastrophe models for commercial and industrial properties

Abstract

Commercially available natural catastrophe (cat) models have significantly improved the decision-making processes utilized by the insurance industry to price and manage cat risk. However, lack of historical loss data, and the need to make a large number of assumptions during the course of development of such models can lead to material biases in their outputs. It is crucial that insurance companies identify these sources of biases and adequately adjust the model outputs. In this study, we present methods that can be utilized for performance evaluation of cat models independent of the underlying peril or region. We illustrate the application of the proposed tests by reviewing two commercially available Japan Typhoon cat models (referred as models A and B) and evaluating their performance in quantifying commercial and industrial property losses. We have identified important limitations in both models including not accounting for storm surge and model specification uncertainty. We observed significant differences between the modeled losses for commercial exposures and uncertainty estimates for the corresponding event losses. Performed sensitivity tests indicate potential inconsistencies in Model B’s assumptions related to the quantification of loss severities across geographic regions, the estimation of contents and business interruption losses, and modeling of inland flooding. Additionally, our comparisons indicate that Model B assumes typhoon landfall frequencies significantly lower than historically observed values. Conducting the proposed tests on Model A also suggests potential underestimation of the losses for both the strongest category of typhoons and typhoons with losses primarily driven by rain-induced flooding. While modeling companies recognize some of these potential limitations and plan to address them in their next updates, it is important that they continue providing increased flexibility in adjusting model parameters and allow insurance companies to develop their own views in management and pricing of cat risks.