The modeling of emerging and casualty catastrophe risks remains challenging and the models continue to vary in their approach, level of development and industry acceptance. With the potential scenarios numerous, diverse and constantly changing, there is no single model or approach that could contemplate all of them. Unlike their fully probabilistic property counterparts, the various perils and scenarios also need to be adapted and structured according to each carrier's specific exposure. Furthermore, the various disaster scenarios with which carriers are being increasingly confronted need to be prioritized and synthesized within their enterprise risk management (ERM) framework. By their very definition, there may be limited data on hand on which to base any modeling. As a result, much of the industry continues to rely on multiple models and actuarial approaches that encompass: model applications, probable maximum loss (PML) estimates, realistic disaster scenarios, experience and exposure ratings to create a broad set of scenarios and deterministic views.
In addition to peril and scenario based commercially available catastrophe models, niche data best practices and models are being developed to meet the demand in varying degrees within the crystallization and aggravating categories. Here, ample data and modeling applications are being synthesized and adapted within existing model frameworks allowing carriers to get their underwriting grip on these scenarios. Examples of these efforts within Guy Carpenter include the various flood models we have developed specifically for Asia and Central Europe. Other applications involve identifying and quantifying emerging "aggregating" exposure concentrations such as those in Asian industrial parks. The losses emanating from the Thailand floods were more about the insureds and the industry taking its eye off of exponential exposure growth and changes in global supply chain dynamics - in lieu of the size of loss and return period dynamics - after they were flooded. Other niche models, such as Guy Carpenter's MetaRisk® ReserveTM can focus on various "crystalizing" emerging threats emanating from the accumulation of systemic reserves over multiple years.
The Oasis Loss Modeling platform, of which Guy Carpenter is a member and supporter, will help facilitate further development of additional niche catastrophe models by allowing independent developers to create and input various hazards, vulnerability and exposure elements. We believe that open-source platforms such as Oasis will lower the barrier of entry for academics and small specialist teams on innovating and developing models that will create more views of overall risk and the ever-increasing number of emerging perils and cat risks. Access to these additional views should prove instrumental to our clients where credible views of particular emerging risks may not currently exist and should be analyzed from a variety of different modeling perspectives.
The mapping and deterministic modeling of emerging risk scenarios has and will continue to play an important role in this area. Lloyd's approach to emerging liability risks in some ways has been no different than what has been required of their syndicates to report on for well-established property risks. Specific realistic disaster scenarios (RDS) are required to quantify and model for specific earthquake, windstorms and even terrorism event footprints through a combination of licensed software (AIR, EQECAT or RMS), internally modeled or via maximum line estimates. With a relative shortage of these options and data available for professional, non-professional as well as multiple public and products-based liability RDS losses, a reliance on simpler market share or premium derived PMLs based on de minimis approaches has been typically required. However, as the level of sophistication and tools for deterministic modeling capabilities here increases, the next question that arises involves the more challenging leap towards a more fully probabilistic and holistic model approach. It is important to note that although probabilistic terrorism models have been available in several countries including the United States for nearly a decade, the A.M. Best rating agency continues to rely on deterministic loss scenario modelings due to what they continue to view as a lack of historical and credible probabilistic based events.
The availability of essential insured-level data on emerging and casualty catastrophe risks remains an important challenge that many carriers continue to work towards improving. Property catastrophe models that were developed during the 1980s contemplate highly granular and sophisticated geo-coded data that is readily available today and get interfaced with very specific and robust building construction and historical event sets. Casualty catastrophe modeling similarly also requires highly granular exposure data related to the particular industry(ies) covered within the portfolio. Models that are beginning to emerge in this area differ according to the data they require, the approach taken as well as the specific scenario set(s) the development is focused on. Some are taking a highly granular data intensive bottom-up approach whereas others may be contemplating a more general top down approach to the exposure data required. Some are loss experience-based and are contemplating an integrated historical event set, yet others are much more exposure-based. The exposure-based models tend to be highly ingrained in generally accepted scientific and mass tort data and operate under the fundamental assumption that past losses and patterns may not necessarily be indicative and directly applicable to future emerging threats. And as a result they tend to focus predominantly on products-based liability scenarios and their latent impact on bodily injury.