CAS Annual Meeting: Actuarial Models’ Effects on the Underwriting Cycle
CHICAGO—CAS Annual Meeting attendees learned about the varying effects of actuarial models on the underwriting cycle at a general session on the topic held November 7, 2011. Property/casualty insurance has long been noted for the sharp rise and fall of rates through the underwriting cycle. At the conference, Stephen Mildenhall, chief executive officer for Aon Benfield Analytics, John Beckman, senior vice president and chief underwriting officer at CNA Commercial Insurance, and David Bassi, chief underwriting officer with The Plymouth Rock Company, discussed how model improvements have influenced the cycle and how they may affect it in the future.
As actuaries have developed more sophisticated statistical and computer models to help understand insurance problems, the models’ effects on the underwriting cycle have become increasingly complex. The answer as to whether models flatten the cycle or make it more volatile is not entirely clear. Good models can help companies better understand and manage their risks, which would reduce the cycle. However, if models miss a major event, it creates uncertainty that could exacerbate the cycle.
Actuaries routinely use several newer types of models:
- Catastrophe models, which estimate the likelihood and the size of natural disasters like hurricanes and earthquakes.
- Predictive models, which use factors like credit scores to help determine an accurate rate to charge customers.
- Economic capital models, which try to show how changes in the business and economic environment could affect an insurer’s health.
Mr. Mildenhall noted that a model that projects off of the previous year’s losses and the premium collected over the past two years explains 90 percent of the cycle. Mr. Bassi added that local factors, like state regulatory changes and reinsurance, also drive price adequacy. Mr. Beckman noted that flexible capital management and better underwriting tools make a difference, too.
All of the newer models have elements that could either flatten or accentuate the underwriting cycle. Catastrophe models, for example, have been getting better over the years, which lend stability to the market. However, if those models come up short, as they did in 2005 with Hurricane Katrina, the ensuing uncertainty can create a spike in prices. Even changing the model can roil the market, as we observed after a leading catastrophe modeling firm updated its model.
Still, catastrophe models and their predecessors have done an excellent job of containing catastrophe risk. Catastrophes cause fewer than 10 percent of insurance company insolvencies, Mr. Mildenhall said. Underpricing and underreserving cause five times that amount.
Predictive models have not been around for as long. They have become well-established in personal lines, like private passenger auto, but are only now emerging in commercial lines pricing. Panelists agreed that predictive modeling will continue to grow, starting with pricing small commercial risks, then later in pricing larger risks.
As the models become more widespread, actuaries will have to help underwriters understand what the models do and don't do well. “Underwriters need to understand how well the models capture key risk factors and the role of judgment,” Bassi said.
Underwriter understanding is also crucial to a company’s relationship with agents and policyholders. The underwriter needs to know why a predictive model is behaving as it does. “We can’t say [to a policyholder], ‘You get a rate increase because we have this new predictive model,’” Mr. Beckman said. “That won’t create a satisfied customer.” The underwriter needs to know and be able to communicate which characteristics drive the model indication for each insured.
Economic capital models recreate an insurance company’s risk portfolio—risks it underwrites and the investments it makes—and show how the portfolio would react to a wide variety of economic and industry scenarios. The models let management know where the company is weak; managers can then shore up the weakness or be prepared in case a threat emerges.
The panelists remarked that when models reflect what management already believes, the interpretive limitations can be viewed as a strength. Mr. Bassi posited that when the model reflects management’s point of view, it is easier for managers to “buy into” its output. Consequently, they will find it difficult to discount any discouraging news coming from the model in the future.
Sometimes, however, management relies on the model without a full understanding of the inherent uncertainties. When it fails, they overreact, and overreacting by raising or lowering prices too much drives the underwriting cycle, according to Mr. Bassi. As a result, “It has the potential to create less frequent but more severe cycles,” he said.
Capital models help regulators and rating agencies understand a company’s strengths and weaknesses. Capital models will play an important part in Solvency II, a more rigorous system of monitoring insurers than the EU currently has under Solvency I. Solvency II is scheduled to come into effect on January 1, 2013. Mr. Beckman noted that as regulators better understand a company’s capital model, their understanding of the company itself improves. That, too, is valuable.
Though models will become more important to the business, they will not be entirely self-sufficient. “We’re still a people business,” Mr. Beckman said. The models will play important roles, but the ability to explain the models to managers, underwriters, investors, regulators and rating agencies will be critical. “The people who are the best at communicating [the model] will be the ones who use the model best. And the ones who use the model best are the ones who will succeed.”