Catastrophe Loss Development Analysis

By Sanders Cathcart and Joseph Izzo

In the wake of a large catastrophe — a hurricane, tornado, or earthquake, for example — insurers and reinsurers have a big question to grapple with: Management, regulators, and investors all want to know, What will the cost to the company be? But hard information is not readily available on primary companies' actual claims and may not be available for months or longer. For reinsurers, the claim information that ultimately determines their costs is even less accessible.

Actuaries are pressed to provide estimates, but such estimates always include a range of uncertainty, making reserve levels problematic. Over time, as actual catastrophe claims are addressed and more information becomes available, the range of uncertainty will naturally narrow. But depending on several circumstances, the uncertainty may not diminish as quickly as stakeholders would like.

Property losses generally tend to settle more quickly than casualty claims. But starting with some of the outsized catastrophes in the 1990s, such as Hurricane Andrew, insurers found that aggregate claim costs sometimes continued to emerge over a longer period than had been historically typical. (Actuaries refer to this emergence of claim costs as the loss development process.) Hence, estimates that relied on early claims compilations were inadequate — even with large teams of claim adjusters feverishly picking through the rubble to help policyholders and provide cost information to carriers.

Over the past few years, we have received several inquiries as to whether our statistical data shows different loss development patterns between catastrophe and noncatastrophe losses. The answer is yes, though the differences vary according to the size and timing of the catastrophe. We have developed analytics that can help insurers and reinsurers reduce the range of uncertainty by illustrating how historical development patterns can vary across different catastrophes — and between catastrophe and noncatastrophe ­circumstances.

Figure 1 shows the first 36 months of development arising from homeowners losses attributable to Hurricane Katrina and compares that to the development pattern for two categories of noncatastrophe-incurred losses — those 2005 noncatastrophe losses that occurred in the same month as Katrina and those that occurred in a nearby month without catastrophes (in this case, June 2005).

Figure 1
Incurred Loss Development (to 36 months)

market value vs. reconstruction cost


Hypothetically, an insurer with $45 million in emerged losses (paid plus case reserves) after six months might infer, based on the noncatastrophe development pattern, that it needs to establish additional bulk and incurred but not reported (IBNR) reserves of about $5 million to handle the ultimate costs. But the catastrophe development pattern suggests the insurer would need quite a bit more than that — almost $20 million more — to handle the costs that have not yet emerged.

The data underlying this exhibit was derived from our database of homeowners data reported under the Personal Lines Statistical Plan – Other than Auto (PLSP-OTA), as of first-quarter 2010. Losses were attributed to this catastrophe based on the loss date, the cause of loss, and the location of the loss, as reported to ISO. The event date, states affected, and estimated insured property damage information was taken from material produced by ISO's Property Claim Services® (PCS®) unit.

Similar analyses are available for certain other catastrophes and property lines of business, and we can update these analyses periodically, as appropriate.

Sanders B. Cathcart is assistant vice president and actuary of personal lines information and analytics at ISO. Joseph M. Izzo is assistant vice president and associate actuary of commercial lines information and analytics at ISO.