WIXLIB

Insurers’ PerspectiveInsurers continually seek ways to more accurately measure the risk of loss to which they areexposed. Before 1992, insurers generally believed that their probable maximum loss, in theaggregate, was much less than the actual insured losses caused by Hurricane Andrew. Insurersbegan to realize that the application of standard catastrophe ratemaking methodologies based onhistorical losses of 25 to 30 years of data were inadequate for low frequency, high severity perilssuch as hurricanes and earthquakes. Hurricane Andrew cost insurers 16.5 billion, but that wasnot the “Big One.” Had Hurricane Andrew struck the coast 25 miles to the north, the insured losspotential would have been in the 50 billion range. Andrew caused major market deteriorationfor the citizens and insurance regulators in Florida. Insurers came to the stark realization thattheir companies’ assets were much more exposed than previously believed. The potential for losscontinues to cause availability problems for coastal states particularly states that have notexperienced hurricanes in recent years.As insurers seek to measure their potential loss exposure, they are motivated to raise prices towhat they consider to be adequate levels or to reduce their exposure. If insurers believe that theirproducts are inadequately priced, they will attempt to raise the prices to levels they believe to beadequate. If regulators resist these price changes, insurers may withdraw from the market anddivert capital to other lines of business. From the insurers’ viewpoint, they are not attempting topunish consumers or make life difficult for insurance regulators. They are simply trying toprotect capital and surplus.In the past, property insurance rates have typically included a factor to reflect exposure fromcatastrophe loss. The typical ratemaking procedure called for capping catastrophe losses at aselected figure and then averaging the excess losses over a long period of time. The time periodranged from 10 to 30 years in duration. History has proven to be a good predictor of the future aslong as there are no significant changes in the environment in which the coverage is to beprovided and a time period that provides a sufficient number of catastrophes is used. Manyinsurers believe that past property insurance catastrophe loadings are no longer needed in areasthat are subject to significant catastrophe losses. In addition, it is possible that weather patternsare changing. Insurers argue that using old ratemaking methods underestimate the cost ofinsuring catastrophe prone areas.The problem with using relatively short periods of historical catastrophe losses, such as 25-30years of insurers’ data, is as follows: A very long history is needed in order for it to be reflective of all the expected futurehurricanes or earthquakes that are likely to occur. On the geologic, climatic and seismictimeframe, 100 years is not a lot of time, however 100 years of recorded scientificinformation is significant. When smaller geographic areas are considered, the availablehistory of such events is even more problematic. Also, the traditional method does not takeinto account the full spectrum of possible catastrophes that could befall any particulargeographic location. Some events have a “return period” of 50, 100 or even 250 years ormore and would not be taken into account using the traditional method; Historical losses are reflective of the exposed structures (both as to numbers and their value)that were there when the historical events took place and not of what are currently exposed—what insurers believe is needed is a means to measure the loss impact on the exposures that 2011 National Association of Insurance Commissioners2

will be exposed at the time of the expected event, reflecting both the growth in the numbersof insureds living in “harm’s way” and the higher values of the structures built in theselocations; People have been and still are building more dwellings and businesses in areas that aresubject to significant catastrophe losses, casting doubt on the validity of historical lossfigures in projecting future loses since fewer buildings were then subject to loss; Building codes and building practices have been more focused on life safety than onproperty preservation and have not kept pace with the changes in construction methods. Inaddition, there are conflicting interests for the construction and real estate developmentcommunities versus those of insurers, reinsurers and emergency management agencies oftencreating situations where owners and occupants are not adequately aware of the risk they arefacing until it is too late; and When only the most recent historical period is used, losses can be over- or underestimated.What is needed is a methodology that creates stability in ratemaking to prevent unnecessaryfluctuations in the premiums charged to policyholders. Many insurers believe thatcatastrophe modeling provides that stability.Advances in technology in recent years have enabled insurers and reinsurers to apply computermodeling to estimate the effect catastrophe losses could have on their books of business. Theresults of these models are used to assist with decisions about where an insurer will choose towrite business and how much to charge for business written. Some insurers are replacing thetraditional catastrophe loading in rate filings with numbers supported by computer models.Catastrophe Modelers’ PerspectiveApplication of catastrophe modeling is relatively new to property insurance. Catastrophemodelers have significant investments in these models and are, as a result, somewhat guarded inthe information they publicly disclose about their models. They are doing what any prudentbusinessperson would—protecting a business asset from disclosure to competitors. As insurersare requested to provide information so that regulators can determine if the processes applied arelogical and result in rates that reasonably reflect expected future loss costs, pressure is placed oncatastrophe modelers to demonstrate that their models produce results that are acceptable to theregulators. Catastrophe modelers are being called upon to disclose information about elementsused in their models and how data provided by an insurer is used. It is now incumbent upon themodeling firms to share enough information with regulators so that they can perform theirregulatory duties to assure that rates are not excessive, inadequate, or unfairly discriminatory.Consumers’ PerspectiveConsumer representatives are skeptical about the application of computer models to propertyinsurance rates and underwriting decisions. There has been a perception that the insurers arehiding behind a cloak of mystery that is held by the catastrophe modelers. The consumerrepresentatives know that a great deal of public information goes into the models, yet themodelers’ claim the models are confidential and will not show consumer representatives howthey work. This cloak of mystery and the varying results produced by catastrophe models tend tomake consumer representatives wary. They often oppose including a catastrophe rate based on 2011 National Association of Insurance Commissioners3

confidential models. In states where consumer representatives are allowed to intervene in ratecases, they have sometimes hired their own experts to challenge the rates filed by an insurer.Consumers have legitimate questions about the effects of the use of computer models on theirinsurance bills. In Florida, homeowners’ rates have increased more than 100 percent sinceHurricane Andrew. In addition, coverage for hurricane has been reduced with higher deductiblesand lower effective policy limits. Yet, there are still many areas where availability andaffordability remain problematic.Some of the areas where consumers have not received satisfactory explanations are: Why their premiums may differ depending on which model is used; Why the geographic groupings seem arbitrary; Why contiguous counties or territories may have very different costs; Why a method with so much uncertainty is assumed to be correct to the exclusion of othermethods. This can be particularly confusing since previous assumptions about expected costswere also purported by companies to be accurate enough for ratemaking; Why a model’s estimates do not always match closely to actual storms; Why models that seem to overestimate large storms are allowed to be used; Why rates keep increasing even after significant increases have already been incorporated; Why the probability of a large hurricane in any given year is very small, yet premiums paiddo not appear to generate any investment returns that benefit policyholders; and Why premiums are based on expected annual losses, yet many of the assumptions seem to bebased on 250 or 500 years of generated experience for a location at the current time.Consumer representatives call for public disclosure of the input and output from catastrophemodels so independent tests can be run to assure that model results are reasonable.Regulators’ PerspectiveRegulators are faced with an extraordinary challenge as they attempt to reconcile the vagaries ofmodel-based loss costs with the obligation to assure that the resulting rates are appropriate andfacilitate a vibrant industry. Regulators must weigh the relative merits of replacing a traditionalmodel that has evolved very methodically over the years with a methodology that is in its relativeinfancy in terms of producing reliable results.The table that follows highlights the wide range of answers obtained from catastrophe models’estimations of expected loss costs. 2011 National Association of Insurance Commissioners4

CountyDadeDeSotoDixieDuvalEscambiaFlaglerModel 12.200.210.260.20.038.044Model 21.710.530.260.310.490.35Model 31.690.320.130.080.820.19Model 42.860.270.110.071.070.14Model 070.44Lowest1.690.140.110.070.160.14These numbers are expected loss costs per 1,000 of insurance that were provided to the FloridaCommission on Hurricane Loss Projection Methodology. Inputs were the same for every modeland each of the models was subjected to the same standards. Yet the highest modeled loss costswere almost seven times as high as the lowest within a county.Regulators are reluctant to embrace the use of models for ratemaking without demandinganswers to some relevant questions, given such a wide range of results.Initial attempts by regulators to gain information about catastrophe models were met withcaution by insurers and the entities that have developed the models, in part due to claimsregarding the propriety and confidential nature of some aspects of the models. The typicalregulatory response was to disapprove the filing containing the model results, since regulatorswere unable to obtain information necessary to evaluate the appropriateness of model output fora rate filing. Even now, no modeler has requested advisory organization status for use of its losscosts, which could facilitate model output use in ratemaking.Most modelers have agreed to share basic information with regulators who have requested it.Regulators have been allowed to look at model input and output and review some formulas andalgorithms within the model. This happens after an agreement is reached with a modeling firmabout what will and will not be made part of the public record. Regulators recognize that somejurisdictions’ freedom of information laws require public disclosure of all information providedto them. Both regulators and modelers are working to meet the challenge of providing enoughdisclosure to make informed decisions, while preserving the confidentiality of proprietarydetails. This handbook is an important step in that process.Regulators are trying to better understand catastrophe models and their applications to insurers’books of business. These applications include underwriting, ratemaking and solvencymonitoring. Regulators are constantly seeking new sources of information and objectiveguidance in understanding model input, output and methods. 2011 National Association of Insurance Commissioners5

2011 National Association of Insurance Commissioners6

SECTION IISELECTED CATASTROPHE PERILSThe basic physical characteristics of natural phenomenon must be understood to see howmodelers develop assumptions about the characteristics of each catastrophic event. Thishandbook addresses the hurricane and earthquake perils. Models are used for other perils as well,such as non-hurricane wind. However, it is these two types of events that are the mostcontroversial.Much of the uncertainty about potential damage is caused by the low frequency of hurricanesand earthquakes. This adds uncertainty to projections of future insured loss estimates. Otherinsurance, such as private passenger auto, need relatively short periods of time to generatereliable data. Accurate estimates can be calculated with traditional mathematical and statisticaltechniques. When dealing with these, the time span that is meaningful is so long that accuraterecords do not exist of frequency or severity of events. It is this lack of recorded data combinedwith the potential severity of the events that creates the challenges of estimating the impacts ofcatastrophes.EarthquakeAn earthquake is a shaking of the ground caused by the sudden breaking and shifting of largesections of the earth’s rocky outer shell. A severe earthquake may release energy 10,000 times asgreat as that of the first atomic bomb. Rock movements during an earthquake can make riverschange their course. Earthquakes can trigger landslides that cause great damage and loss of life.Large earthquakes beneath the ocean can create a series of huge, destructive waves calledtsunamis that flood c

the Catastrophe Insurance Working Group by the Personal Lines Property and Casualty Insurance (C) Committee. The committee is charged with providing a forum for discussing various issues related to catastrophe modeling and to monitor issues that will result in changes to the Catastrophe Computer Modeling Handbook.