While insurers’ use of models to analyze and quantify risk is well established in the property space, it’s becoming increasingly important in the casualty space. The use of models in this area has grown significantly in the past five years, as companies recognize the looming specter of large, unforeseen losses that may threaten insurers’ balance sheets and future earnings.
Executive SummaryOffering analogies to property-catastrophe modeling, Verisk SVP and Actuary Eric Gesick explains the basics of developing a framework for liability-catastrophe modeling. In the liability world, perils can be corporate activities, product flaws or operational losses instead of storms and earthquakes, and more specific named perils can include defective auto parts or contaminated foods rather than Atlantic hurricanes or West Coast wildfires.
The framework that Gesick defines can be used to classify both systemic and emerging risks.
In Part 2 of this two-part article, Gesick will demonstrate how this framework described here in Part 1 can be used to classify different types of emerging risks.
Part 1 of 2
“Liability catastrophes”—large-scale events triggered by a common underlying cause that can affect multiple organizations and corporations, industries, coverages and policy years—can ripple through global supply and distribution chains and result in widespread liability losses across multiple divisions within insurance companies and across insurers.
Unlike natural catastrophe risks, which are predominantly driven by well-understood natural processes, liability risks depend on complex human dynamics and how they interact with a constantly evolving legal, social, technological, financial and economic landscape. As a result, the liability catastrophe event space is inherently volatile, which has heretofore limited the ability of models to accurately capture the full breadth of the risk.
In addition, liability events can transpire over years or even decades, and typically manifest as reserve inadequacies when management recognizes that previously established reserves are insufficient to cover the developing liability claims that are far beyond initial expectations.
For these models to take hold, they need to capture both the event space itself and its volatility within a consistent framework. The goal is not only to quantify losses from known, expected risks but also to enable organizations to accurately estimate potential future claims from events that already have occurred but are unrecognized, as well as from new systemic risks that can emerge and evolve over time. So, developing a flexible modeling framework that can simulate the impact from a wide array of systemic and emerging risks is the key to more widespread usage of liability risk models.
A Foundation: Developing a System to Classify Risks
The first step to modeling and ultimately quantifying liability risk is to develop a framework to understand and organize all the different types of liability catastrophes. Within the natural catastrophe world, this would be akin to developing an organizing principle capable of estimating losses from multiple different unrelated types of perils, including hurricanes, earthquakes, wildfires and floods.
Just as with natural catastrophes, when we look back at historical liability events, we observe that the liability catastrophe event space can also be segmented into different types of “perils.” These peril types are defined by the common patterns of the underlying relationships and activities within and across organizations. These patterns do not just reveal inter-organizational relationships, they also provide the foundation for how potential systemic liability for harm spreads across organizations, industries and supply chains.
These underlying patterns create similar “footprints” within a peril type that map the networks of relevant relationships. In the natural-catastrophe world, these relationships are typically defined by geographic proximity since the impact of most natural catastrophes are concentrated within a limited physical area. Natural perils create a geographic footprint across a geographic map, establishing the correlation structures between exposures that allow for modeling the peril, or identifying the common cause of loss.
In the liability world, on the other hand, these relationships are typically defined by the economic, contractual, social or geographic relationships between organizations and industries. We need to look at how liability perils unfold across an economic map (creating the liability footprint) rather than a geographic map. The relationships between entities within the economic map creating the footprint establish the correlation structures that allow for modeling liability perils.
We’ve defined three types of perils that create similar footprint characteristics within each type—not unlike a set of Atlantic Basin hurricanes that take different paths but are ultimately governed by a common, independent set of atmospheric conditions. The three types are:
- Corporate: This peril type includes events from potential harms resulting from corporate activities, e.g., a financial wrongdoing, bankruptcy, fraud or securities class actions, sometimes resulting in the implosion of the company perpetrating it and potentially involving its professional advisers and other related institutions.
- Products/components: This peril type includes events from potential harms resulting from contaminated or faulty product components or ingredients that can cascade through the supply and distribution chain, impacting numerous finished goods and potentially implicating a wide number of insureds in a wider number of industries. This peril also includes faulty finished products.
- Operational: This peril type includes events arising from potential harm resulting from business operations that cause infrastructure accidents or operational losses, including those arising from services provided to customers, such as entertainment or education, and from the operations and responsibilities of contractors or owners of operations and their suppliers.
We can segment the liability space even further to account for more nuanced differences between event types. Within each of these three perils, there are specific “named perils” that differ from one another in frequency, severity and how they spread through the network.
For the product peril type, consider the differences between a defective auto part and a contaminated type of food. Both product perils share a common structure of relationship networks (products being disseminated through supply and distribution chains) as well as a set of responsibilities and duties owed to third parties. Obviously, the industries, organizations, and supply and distribution chains will be different between the two events. We would also expect there to be differences in the severities and frequencies of these events, as well as in their liability insurance impacts. These differences in the event characteristics establish unique parameters for each named peril type.
The segmentation of liability catastrophes into perils and named perils creates a framework flexible enough to enable us to capture many different types of liability catastrophe scenarios. This framework forms the basis for developing a stochastic catalog of each of these peril types, where we can model out many different variants of each type of event to capture a broad range of the potential losses that could occur in the future. This is like using a historic catalog of natural catastrophe events to create a stochastic catalog of events that haven’t happened yet but could happen, given the nature of the parameters governing the loss process. Even more importantly, it lets us capture and begin accounting for the risk posed by (previously uncontemplated) emerging risks.
This concludes Part 1 of our article. In Part 2, we’ll demonstrate how this framework can be used to classify different types of emerging risks.