Modeling Fundamentals: Advanced Loss Modification in Touchstone Re
Oct 16, 2019
Gone are the days when analysts simply fed data into their catastrophe models and accepted the results at face value. Today’s model users expect to understand what goes on “under the hood”—to truly understand the risk analysis tools they use and the output they produce. They are more engaged with the modeling process and want the “knobs and dials” that give them the flexibility to modify model output to form their own view of the risk.
There are many reasons a model user may wish to modify model results to obtain a different view of the risk, but the following are a few of the main reasons. One, based on claims experience or internal research, a company may believe that a model is overstating or understating risk for certain types of events, specific regions, lines of business, or construction classes. Two, a company may want to capture a source of loss such as a secondary peril or a coverage that is not explicitly modeled. Three, government hazard maps may get updated after a model has been released and before it has been updated to account for the change in hazard, so a company may want to the ability to modify the hazard in the interim. Ultimately, companies seek to use their unique insight and experience for their own risk management to make decisions that make sense for their business so that they can truly own their risk.
The Advanced Loss Modification (ALM) module in Touchstone Re, the successor to and next generation of the Exposure Management Module (EMM) in CATRADER™, enables companies to analyze global reinsurance treaties beyond the scope covered by AIR’s traditional suite of models and offers clients the flexibility to modify modeled losses and perform advanced sensitivity testing, all within the same reinsurance platform they are using today.1
The module enables users to examine the drivers of risk and adjust AIR modeled losses by event, line of business, or program to reflect a company’s own knowledge and claims experience. Users can analyze non-modeled sources of loss to enable sophisticated risk selection anywhere in the world.
Within the separately licensed ALM module is a component that allows the application of adjustment factors to event-level losses, or Event-Based Adjustment Factors (EBAF). This functionality is discussed in the next section to illustrate just one of the many types of things ALM allows clients to do.
Event-Based Adjustment Factors
Event parameters can’t be changed, but results can be adjusted by user-defined factors. So while you can’t change an M7.5 earthquake into an M8.0, if your experience suggests, for example, that earthquake losses are 5% higher than the AIR model indicates, you can apply an adjustment factor of 5% to your results. Conversely, if you perceive AIR’s losses to be too high, you can lower them to reflect your own assumptions.2 At a more granular level, the geographic and line of business (LOB) flexiblity allows the scaling of losses up or down by region or LOB. In addition, losses can be zeroed out to reduce frequency or scaled up to effectively increase the frequency of high severity events. The scaling is applied directly to company losses before the application of any terms, factors, or filters. A 20% reduction does not therefore necessarily mean a 20% reduction in reinsurance contract losses.
These adjustments can be used to:
Account for model deviation from historical experience revealed in validation. Historical experience incorporated in the AIR model, for example, did not anticipate the September 2010 M7.0 Darfield earthquake; its major aftershock, the February 2011 M6.2 Christchurch earthquake; or the November 2016 M7.8 Kaikoura earthquake. EBAFs could be used to tweak assumptions pending updates to an AIR model prompted by such events. And if you want to account for an event not in the historical experience but still scientifically plausible—say a strong Category 4 storm striking New England, for example—you can do so by applying EBAFs to the existing hurricane events that impact the Northeast in AIR’s stochastic catalog to reflect the associated increase in damage that would come with these higher wind speeds.
Investigate the sensitivity of modeled results and their effects on reinsurance treaties. EBAFs enable you to apply different adjustments and see how your view of losses changes when reinsurance terms are applied with and without those adjustments.
Account for secondary perils or other sources of loss not captured by the model. EBAFs can be used, for example, to scale results to account for loss adjustment expenses, portfolio growth/inflation, policy leakage effects, evacuation costs, looting, and vandalism, or special coverages. For instance, you could employ them to modify modeled losses by a user-defined factor to account for fire following earthquakes in urban regions for models that do not yet cover this sub-peril.
Apply loads to large event losses to obtain a conservative view of the risk. Sometimes, to obtain an upper bound for potential losses for a specific peril or in a selected region, you may want to simulate the impact of events that are more extreme than those accounted for in a model’s 10,000-year catalog. By applying factors to events at the tail of the standard 10K catalog, you can effectively assess the scale of losses that a substantially more severe event might cause, thereby establishing an upper limit to inform a more conservative view of the risk.
Incorporate new scientific findings prior to model releases/updates. Prior to the 2011 Tohoku earthquake, the scientific consensus did not anticipate that the region would experience such a large multi-segment rupture resulting in a devastating earthquake and tsunami. And when Geoscience Australia (GA) published an updated national seismic hazard model in 2019, for example, it included many changes compared to previous versions and the seismic hazard was reduced by more than 60%, on average. The new seismic model—NSHA18—will thus have a major impact on modeled losses when risk modeling firms update their models to incorporate these new findings. In cases such as these, EBAFs can be used to scale losses appropriately upward or downward to reflect the impact of a revised appreciation of risk before the updated model is released.
Undoubtedly, the ability to modify losses gives users the power to apply their own adjustments and provides yet another avenue for sensitivity testing. When exploiting Touchstone Re's loss adjustment flexibility, however, users should keep in mind the importance of maintaining a scientifically credible, internally consistent view of risk.
It is important to bear in mind that AIR spends half the person hours required for model development on validation; AIR model components and configurations have been carefully calibrated to effectively align with observed loss output for historical events. A modified view of risk must be clearly documented and justified when model results are shared with both internal and external stakeholders, including regulators who will undoubtedly ask ever harder questions. In all cases, the AIR view of risk will be preserved and provided as a benchmark alongside the modified view, but the responsibility of validating modified model results is borne by the model user.
AIR's mission is to empower our clients to truly own their risk. We are committed to delivering the flexibility that helps each company use their unique insight and experience in making decisions that make sense for their business. We want to provide the tools to empower you to develop a customized view of risk—responsibly.
1 Clients licensing the ALM module in CATRADER will be able to port program EP adjustments and custom models and LOBs when migrating to Touchstone Re. EBAFS will need to be reimported and an adjustment factor file set up to use the Event-Based Adjustment feature.
2 Users are responsible for the adjustments made and their validation and should use their own experience and/or research to ensure their reasonableness.