The insurance industry has been using statistical models to provide guidance towardsdata-driven decision-making for many years.

Life insurance companies have mortality tables going back acentury or more, allowing actuaries to determine rates for today'slife insurance customers that will maintain future profitabilitywhile fulfilling the promises made in the policies.

The explosion of computer technology in the last 50 years hascreated opportunities for data gathering and analysis that could only bedreamed of even a generation ago. Insurers now have the ability tosift through mountains of data to help derive useful insights andestimate the potential impact of a variety of possible futureevents.

Earthquake activity

The invention of the seismograph in the early 20th centuryprovided scientists and later insurers with hard data aboutearthquakes as they happened. The U.S. Geological Survey (USGS) was establishedin 1879, the bill authorizing its creation signed by PresidentRutherford B. Hayes. Their initial mandate was to improvetopological mapping of the United States. In the 138 years sincetheir creation, the USGS has developed high-resolution maps ofnearly the entire country, and provided detailed information aboutlandslide potential, liquefaction susceptibility and soilconditions (along with numerous other types of data).

This data, along with probabilistic estimates of the frequencyand magnitude of future earthquake events, is combined intosophisticated algorithms that provide insurers and reinsurers withestimates of the losses they can expect from future earthquakeevents. Gone are the days when underwriters could keep track of thebuildings that they insured by sticking push-pins into a map.Today's fast-paced business climate requires a much moresophisticated approach to accumulation management and riskanalysis.

Predicting the flu

Catastrophe models don't stop with earthquakes. The National Oceanographic andAtmospheric Administration (NOAA) uses computer models to helppredict weather patterns — movement of the jet stream,hurricanes, tornadoes, severe thunderstorms and the like.Statistical models are also employed in analysis of sea surfacetemperatures, sea level rise andocean currents. The Centers for Disease Controland Prevention (CDC) uses complex epidemiological models tohelp predict the spread of diseases such as influenza, viralhemorrhagic fevers, HIV and many more. These models also helpepidemiologists determine what strains of influenza to include inthe influenza vaccines each year.

(Photo: Shutterstock)

CAT model uses

Insurance underwriters and actuaries use catastrophe (CAT)models to help them convert immense amounts of individual datapoints into actionable information. The models provide a frameworkfor ratemaking decisions, capital allocation decisions, lines ofbusiness to target or avoid, even how much reinsurance a companyshould purchase. It's important to bear in mind, however, thatthese models make no guarantee of accuracy.

There is no way to precisely and accurately predict when thenext major earthquake will strike the San Francisco Bay Area, forexample. What the models offer is a range of potential outcomes,based on the best science available that can be used in conjunctionwith the risk appetite of each organization to make data-drivendecisions about business practices.

Using these statistical CAT models is relatively straightforward(exactly how the models work is a topic for another article). Foran earthquake model, the first step is geo-coding and hazardretrieval. The street addresses for each risk are entered into themodel, which then converts each location address tolatitude/longitutde coordinates. The model then overlays the USGSsoil, landslide and liquefaction maps on top of the geo-codedlocations, and takes note of the known conditions at each location.Distance to known faults is calculated and weighted based on theexpected return period for each fault as well as the expectedmagnitude of earthquake that each fault segment is expected toproduce.

How the magic happens

All of this data is coupled with expectations about groundmotion attenuation (how ground motion is dampened or increased byvarious soil conditions and distance from the epicenter). Then themagic happens — the model developer's "secret sauce"involves the algorithms that convert all of these disparate piecesof data into expected loss information. Each model uses a differentalgorithm, the exact details of which are highly protectedproprietary information.

For the end user, whether underwriters, actuaries or seniorexecutives, the real trick is determining how to interpret the lossestimates, and how to craft guidelines and business practices tomaximize profit while keeping overall risk within establishedtolerances. The models can't tell whether or not a majorearthquake, hurricane, tornado or pandemic will occur in a giventime period. All the models can offer are estimates of thefinancial impact of these events should they transpire, along witheducated guesses about the likelihood of such events occurringwithin the given time horizon.

Output from an earthquake analysis would include an estimatesuch as, "There is a 0.4 percent chance of an earthquake occurringwithin the policy year of such magnitude as to cause $X (orgreater) losses to the company after deductibles and reinsuranceare applied." This can also be thought of as the 250-year returnperiod, or the one in 250-year event. With floods, we often hear reference to the 100-year flood,which really means flooding that has a 1 percent chance ofoccurring each year. (It does not mean that such a flood happensonly once in a century.)

Better understanding loss exposures

It's the "or greater" bit of the model output that causes thesleepless nights for risk managers. Perhaps the most useful benefitderived from the use of a CAT model is being pushed out of one'scomfort zone, and forced to consider the potential outcomes fromvery low probability, high severity events. Failure to consider thesmall possibility of such an event might expose an organization togreat risk of ruin in the event that actual losses aresignificantly higher than expected.

The ever-growing quantity of data available in the world todaysuggests that dependence on statistical analysis will increase inthe future. As the insurance industry continues down this path ofbetter understanding loss exposures and enterprise-wide riskmanagement, expect the continuous evolution of CAT modeling anddata-analytics to help plot the course.

Michael Brown ([email protected]) isvice president and property department manager at Golden BearInsurance Company.