From: GuyCarpenter
Modeling methodologies have been continually refined and updated relative to the peril of terrorism. However, quantifying the economic and human losses from an act of terrorism continues to pose major challenges for insurers and reinsurers. A variety of approaches exist for insurers to model terrorism risk. Most models involve three techniques:
1. Producing probabilistic loss estimates.
2. Conducting exposure-concentration analysis.
3. Generating deterministic loss estimates.
• Probabilistic modeling, also known as CAT modeling, estimates losses based on a large number of events. A key factor is the estimated frequency a modeler applies to all the possible events that could occur. The industry and rating agencies continue to question the credibility of probabilistic terrorism modeling as it requires predictions of human behavior. As a result, unlike hurricane and earthquake CAT modeling, little consideration is placed on probabilistic terrorism modeling.
• Exposure-concentration analysis, also known as accumulation assessment, identifies and quantifies concentrations of exposures around potential terrorist targets as defined by the modeler. Target-based accumulation assessment locates potential targets-typically with high economic, human, and/or symbolic value-and aggregates an insurer’s exposures in and around various distances from these targets.
An important variation of this analysis looks at an insurer’s largest exposure concentrations, independent of what any particular source defines as a target. Therefore, the scanning of clusters of multi-line exposure exceeding an economic threshold within a portfolio-irrespective of these perceived and defined targets-is essential. According to A.M. Best’s “terror charge” methodology, it is these largest of insured locations (differentiated by city) that can be potentially stress tested against published BCAR, regardless of their proximity to landmarks.
• Deterministic modeling represents a compromise between the lack of accuracy in accumulation analysis and the vast uncertainty surrounding probabilistic models. By imposing an actual event’s damage “footprint” at a specified target, a specific-yet hypothetical-scenario can be analyzed with some certainty. Major modeling firms offer an array of deterministic-analysis tools for conventional and NCBR attacks at defined target and non-target locations. This approach can be effective where coarse screening studies show that exposures for an area or event could be high, and a detailed assessment may reduce uncertainties and help decision making.
Relative to natural perils such as hurricane or earthquake, terrorism modeling is still young and untested. Insurers, reinsurers, and modeling companies are constantly learning, assessing and refining their models and the assumptions that underlay those models, thereby increasing their ability to manage terrorism risk in an educated and more quantitative fashion. Currently, deterministic, scenario-based testing is the most common tool used by insurers to assess their vulnerability to terrorism. Given the human and societal nature of the risk, it cannot be expected that the probability of terror events can be determined with the certainty needed to make critical risk management decisions.
Dr. Gordon Woo, Catastrophist at RMS' Emerging Risks Solutions dept. countered this point beautifully by stressing that in modeling terrorism risk, we are not modeling human behavior but social networks... which are more or less quantifiable. More on this here. Using network theory, Woo was able to incorporate the fact that "too many terrorists spoil the plot" into RMS' probabilistic terrorism model. Woo also authored a beutiful paper on game theory & terrorism found here.
