Rate Hike for Quake Insurance Stymied : Government: At issue is the accuracy of ‘modeling,’ a technique the industry uses to predict losses.

An attempt by insurers to double earthquake insurance rates in California is being held up by an angry dispute over widely different assessments of quake frequencies and future quake losses, and the state may step in to certify the firms that do such studies.

Opponents of the rate increase are saying that the firm making the cost predictions--based on earthquake “models”--has vastly overestimated the frequency and damage in order to justify the rate increases. The firm strongly disagrees and defends its work.

The dispute features one of the state’s leading earthquake “modelers,” Risk Management Solutions of Menlo Park, against the consumer organization led by insurance activist Harvey Rosenfield, backed up by a prominent UCLA seismologist.

In the middle is the state Department of Insurance, impatient with the months-long legal test and frustrated with the dueling cost estimates. As a result, Milo Pearson, a deputy commissioner who heads rate regulation for the department, says a commission may be named to set standards for the firms and their cost estimates.

The earthquake modeling is so imprecise that “that’s certainly a concern to the Department. . . . We have recently come to the conclusion that we should provide some level of certification to these companies,” Pearson said. “We’re grappling now with what that level should be. It’s a $64,000 question.”

In the present dispute, Risk Management is trying to help justify a 100% increase in commercial earthquake insurance rates for State Farm. Protracted hearings were held in August, and briefs are still being submitted to an administrative law judge while the rate increase is delayed.

The issue is regarded as an important test case. If Insurance Commissioner Chuck Quackenbush accepts the Risk Management estimates of future quake frequency and loss, and grants the increase of rates in this instance, dozens of homeowners insurance sellers can be expected to apply for similar increases.

Risk Management is being challenged by Rosenfield’s group and by a seismologist it has hired--David D. Jackson, a leader of a scientific panel that in recent years has drafted earthquake probabilities for Southern California. Jackson says that the firm’s model “is inconsistent with the seismological history of California, predicting three times the observed rate of large earthquakes.”

The Risk Management model is also “inconsistent with geological information on fault movement, predicting 1.3 times the observed fault slip rate,” the seismologist adds. “The earthquake magnitude and recurrence data used in the . . . model are incompatible, and their combination leads to a biased result.”

In rebuttal, Heresh Shah, a director of Risk Management and a Stanford professor of structural engineering, says: “I’m disappointed in my colleague, David Jackson. He’s acting as an adversary, giving numbers that are not true.”

In a lengthy Times interview, Shah and Risk Management President and Chief Executive Officer G. Thompson Hutton defended their estimates as well within accepted scientific parameters, and Hutton said: “We will always understate risk. Of course, we understate it.”

Hutton and Shah said that Risk Management, which works for more than 100 insurance companies, reinsurers, brokers and others, has been hired by the Federal Emergency Management Agency to perform an analysis of earthquake risk throughout the United States, showing the high regard in which the firm is held.

There are seven firms that do modeling, providing estimates of quake frequency and loss in California. They are Risk Management, EQE/Guy Carpenter, Applied Insurance Research, Dames & Moore, Risk Engineering, J.H. Wiggins Co., and Tillinghast.

Although Risk Management is on the hot seat now, most of the firms have, in the wake of the extremely damaging Northridge and Kobe earthquakes, upped their estimates of quake damage potentials.

The Insurance Department’s Pearson said, “We might give them all our own quake scenario, and see what losses they come up with, who’s high, who’s low, and so forth. That could occur rather quickly, but that doesn’t say much about how these models work.”

So, he said, the department may create a “peer review panel” made up of seismologists, actuaries and other experts “to review these models and reach a judgment as to whether a company like RMS ought to be certified or not.”

“It’s clear we don’t have the expertise in the Department of Insurance to determine this,” Pearson said.

Such a California panel could follow procedures being pioneered by a new commission in Florida that will examine and set standards for hurricane frequency and loss modeling by several firms that perform those functions.

Hurricanes, like earthquakes, are essentially unpredictable, given the periods of years that are necessary for setting insurance rates.

Normally, insurers rely on actuarial studies for determining how much to charge for insurance against auto accidents or thefts, or fires in dwellings, because they can tell within a relatively narrow statistical range how many of these events will occur. But with infrequent, sporadic and very possibly catastrophic events such as hurricanes or earthquakes, modeling has come to be the favored method.

The trouble, according to Deputy Insurance Commissioner Richard Wiebe, is, “We don’t know that one estimate by a modeler is any more valid than another.”

Quite aside from the sharply varying estimates as to how much damage may occur in a future quake--with Risk Management ranging in the past year from $125 billion to $220 billion for a magnitude 7.0 quake on the Newport-Inglewood fault in Los Angeles and Orange counties--the quake modelers have not even been very accurate in assessing how much damage there has been in quakes that have already occurred.

After the Northridge earthquake, for example, several firms were asked to estimate for the insurance industry how much would be paid to quake victims. Initially, the estimates totaled about $2.5 billion. Instead, the payments turned out to be $12.5 billion and may go higher.

Jack Nicholson, the Florida official who heads that state’s new Commission on Hurricane Loss Projection Methodology, reports that similarly low estimates were initially made on insurer payments to victims of Hurricane Andrew.

“The firms doing it said they would be $8 to $10 billion,” Nicholson recalled. “And they turned out to be $16 billion.”

The assignment of the Florida commission is to consider various methods and models for future hurricanes and “evaluate those for appropriateness, accuracy and reliability for predicting losses,” Nicholson said. “We will make certain findings of acceptability, and we have to adopt standards and guidelines by Dec. 31.” The commission began functioning in July.

The Risk Management officials who were interviewed said they are concerned, if anything, that even sharply elevated earthquake insurance rates may prove to be too low to allow payment of all insured losses in a big quake without bankrupting the companies.

“There isn’t enough data historically to define what the rates ought to be, because today’s buildings weren’t here in [the great San Francisco quake of] 1906,” Hutton said.

“If all we did was to simulate what would be the cost given historic events in today’s exposures, the result would be a rate in excess of what State Farm is asking for,” he said.

Meanwhile, Gina Calabrese, lead attorney for the Rosenfield organization in the August hearings, said a fundamental concern of the Rosenfield group, the Proposition 103 Enforcement Project, is that there is a powerful incentive for the quake modeling firms to increase their estimates to please their insurer clients.

In a brief dated Nov. 30, Calabrese contended that throughout the August hearings “RMS exhibited a disregard for accuracy and reliability in defending its model. If the testimony reflects what happens internally, [their model] cannot be relied upon as a tool for formulating earthquake insurance rates.”

Hutton and Shah, however, said that Risk Management has other clients, such as reinsurers and brokers, whose interests conflict with the insurers. To the extent that estimates are high, the reinsurers may get nervous and increase their prices to the insurers, they said, so the quake modelers have no real incentive to artificially increase their estimates.