The insurance industry over the years has been the source of many of the statistical methods we have come to know and love (if you are in the hate category, you might want to just skip on by this). It is the ultimate exercise in probability, regardless of what is being insured, and indeed, many of the common statistics that you learned as a first year undergraduate were invented as the result of asking a simple question – if we insure this person or thing against damage, what is the risk that we will have to pay out?

For many years now, insurance agencies and underwriters have relied on data to make decisions on who to insure, and for what levels. This is known as actuarial science which is really the original commercial application of data science.  Insurance exists to cover risk, whether that is for the property itself, or the people inside those properties. You know that a certain percentage of your underwritten policies will result in claims, and the question then becomes how much you should charge to underwrite each policy in order to ensure that your total claims are less than your total premiums.

With AGS data, insurance companies have access to a large variety of geo-referenced risk attributes that can aid in the risk assessment for homes or businesses based on location.

The AGS CrimeRisk data provides a highly accurate view of the relative risk of specific crime types for any geographic area down to the census block. For commercial properties in high-risk areas, an underwriter may require the property owner to provide enhanced security measures before agreeing to insure the property, may charge a higher premium, or may forego insuring it altogether.

Weather hazards are also an important aspect of risk assessment, as annual claims for wind, hail, tornadoes, and hurricanes can be enormous. Each of these risk elements is available for every neighborhood nationwide and have served as an important component of the underwriting process for many firms over the past two decades.

With the 2020B release of AGS data, we have introduced another valuable risk element – wildfire.  Wildfires are an annual event in much of the western and southern United States, and as population densities have increased over the years, so has the damage caused to human structures by them. The wildfire risk assessment is finely tuned to that critical “urban-wildland” interface where the city meets the forest or grasslands. It is here that the combination of heat, drought, and winds do the most damage. Local variations in this risk are substantial, often going from extreme to minimal within a few miles – so incorporating this risk into an underwriting model requires a fine geographic resolution.

Actuary science and its practitioners really were the first commercial data scientists and most of us use the fruits of their labors in sample statistics without even knowing it. We are happy to return the favor with data.