What if the disaster models are wrong?

For nearly a quarter century, the Office of the Superintendent of Financial Institutions (OSFI) has required Canadian P&C carriers to demonstrate capacity for handling major earthquakes, including so-called one-in-500 year events.

Canada’s P&C industry is well capitalized and backed by high levels of reinsurance from global firms. Nothing can go wrong, right?

Wrong, said Alister Campbell, president and CEO of the Property and Casualty Insurance Compensation Corporation (PACICC) in a recent column in Solvency Matters.

“Nature is not fair,” he wrote. “And in due course, Canada will be tested by a major earthquake. It could happen in British Columbia. It could happen in the Montreal/Ottawa corridor…And when it does, our models and our industry will be tested.”

While odds favour the P&C industry’s prudence and capitalization levels withstanding a major quake, Campbell asked: “What if the models are wrong?”

“PACICC research has consistently highlighted a gap in the public infrastructure of our country ‒ the lack of a federal backstop mechanism of some form to protect Canadians and the industry in the case of a tail-risk event,” he said. “An example of this would be a mega-quake above and beyond the one-in-500 year threshold.”

He noted PACICC’s modelling put the industry threshold at an insured loss of $35 billion – but emphasized that modelling is based on insured losses, not probabilities. “So, we do not speak to how remote this tail risk is. Quake modelling does that,” he wrote. “And here is where the model risk lies.”

At something beyond a one-in-500-year scale event, companies can fail and possibly spark a cycle that causes the whole industry to follow.

“All the modelers are drawing upon the same limited event sample set, the most recent seismological maps and their own proprietary scientific analysis,” he said. “The list of vendors is short, and all are highly professional and deeply serious about their work. But, there remains a significant margin of error. And, with any model, there is always that risk that it is flat-out wrong.”

The same bulletin contained guest commentary from Donna Sirmons, manager of the modeling program, Florida Hurricane Catastrophe Fund. She addressed the state’s efforts to improve hurricane models following Hurricane Andrew in 1992, which included creation of the Florida Commission on Hurricane Loss Projection Methodology.

Sirmons noted the Florida Legislature saw a need for expert evaluation of computer models and other new or improved actuarial methodologies for projecting hurricane losses, “in order to resolve conflicts among actuarial professionals, and in order to provide for both immediate and continuing improvements in the sophistication of actuarial methods used for determining catastrophe loads in residential rate filings.”

Hurricane and flood models, she said, are often developed by private companies and are usually proprietary. So, the review process helps “address the proprietary nature of the loss-projection models, while still allowing the commission to have full and complete access to the models and their underlying assumptions.”

Overall, she said, this can bring transparency to the model-building process and promote and establish “the validity of catastrophe models in general.”

Campbell said Canada may not need to go that far and added OSFI’s traditionally been model agnostic.

Feature image by iStock.com/hepatus