In this short film, Milliman actuary Matt Chamberlain discuss how geographic information systems can be leveraged to create a more granular analysis of hurricane risk.

Video transcript

Matt Chamberlain: Insurance companies constructed their rating territories originally back in the 50s when they didn’t have sophisticated computers, so they constructed things that were simple and that were based off of existing boundaries, usually counties or zip codes or something like that. So to the extent that there’s science that can be used to understand what the drivers of risk are, we should use that.

One thing that we’re doing that works really well is to look at external third-party data sources that have a causal connection with what we’re trying to predict. It’s not just a question of tabulating what the insurance losses are for past events. We can look at what are the causal mechanisms, what is it that’s really creating some areas to be higher risk and lower risk and then use those to create prices.

So the key to my approach to modeling risk for hurricanes is to bring in geographic information systems so that we can look at insured locations at the location level and utilize geographic information systems layers such as distance to coast or surface roughness.

After I complete an analysis, I’ll have maps that are able to distinguish what the expected losses are on a very granular level.

I definitely think that there’s an opportunity for smaller companies to use their nimbleness to adopt more sophisticated technologies faster than larger companies. Larger companies have a large book of business that they are reluctant to dislocate because of the large rate changes either upwards or downwards that occur when you make a major change to your rating algorithm, so a small company is in a better position to make major changes just because they have a smaller book.

What we’re doing here at Milliman is going to have a huge impact on the insurance industry and, in particular, homeowners insurance. I’m very excited about that.