Climate Risks to US Insurance
The US experienced $140 billion in insured losses in 2024
Weather catastrophes were responsible for 97% of insured losses
That’s a lot of money to lose on a rainy day.
The insurance industry is based around probabilities. There is a huge amount of complexity that goes into calculating these probabilities. In simple terms, to insure a product one must calculate the expected range of losses, and then charge enough to cover for it. For some of these bets, the low probability, high expense scenario will happen and the company will pay out. But this scenario will be rare, and on average insurance companies will pay out less than they gain from insurance premiums.
Of course this rests on one keystone: It requires accurate modelling of the possible losses, and that problem is complicated because the different losses are not independent. If one premises burns down, neighbouring properties are at higher risk.
Of course one way insurance companies work on this problem is by hiring actuaries: mathematicians and statisticians who specialise in building these complex models. However, even the best model in the world can be completely undermined if the data it’s built upon is wrong, incomplete, or the analysis is flawed.
Further, it is precisely the quality of these models that determines how good the insurance company can compete and grow market share. Companies behind the curve will get stuck with “lemons” their nimbler counterparts avoided.
Oh look, another opportunity for alternative data.
How do we solve this problem?
Let’s take a use-case for an example. Imagine you are supplying insurance to a manufacturing company based in Houston, WidgetMaker Industries. WidgetMaker makes many things across multiple manufacturing facilities spread across Houston.
Since 1980, Houston has experienced twelve rain, thunderstorm and/or flood events and when insuring any company with a physical footprint in Houston, this needs to be taken into account, especially for a company that leans so strongly on physical manufacturing. So in order to provide insurance to WidgetMaker Industries, this climate risk needs to be understood.
On simplistic level, we can calculate the probability of flooding in Houston and combine that with the potential loss and give a general estimation for flooding risk in Houston. However, flood risk in Houston is not uniform; some areas are lower risk whereas others are much higher. By taking a blanket approach to modelling flood risk we could end up on the wrong side of this transaction.
Drilling down
In order to accurately assess the risk, we need to know two things:
The location of WidgetMaker Industries’ physical buildings
The flood risk in these locations
In particular, we’re interested in the locations of WidgetMaker Industries’ manufacturing facilities. If a flood hits head office, there will be damage to equipment which will cause a loss, however it is likely that staff can work-from-home temporarily, creating minimal loss of productivity and likely no loss of product. However, as a manufacturing company the potential losses of factories are much more impactful.
So in order to get point one, the insurance company can just ask WidgetMaker Industries for its manufacturing locations, floor plans, etc which they will be happy to provide if they want insurance.
Calculating Flood Risk
One of the ways flood risk can be calculated is to use data from previous flooding events. Areas that have been flooded multiple times in the past are more likely to be flooded again in the future. This methodology was used by The University of Texas at Austin School of Architecture as part of their Center For Sustainable Development to create a map of flooding risk over the Texas metropolitan area.
Here is the map they produced, with darker blues indicating areas with a higher risk of flooding:
Combining this information with the locations of WidgetMaker Industries premises, each can be given a flood risk score able to accurately estimate the flood risk.
We can also go further and track local precipitation to give early warning signs that could indicate a flood is coming. For example, while the odd spike of rainfall can cause a problem, it’s actually a build up of multiple days of bad weather that exacerbates flooding risk significantly. By keeping track of local weather patterns and trends, we can assess the likelihood of flooding at an even more granular level.
A chart of 18 months of daily precipitation at a local weather station in Houston, Texas.
Working at scale
While this methodology works great for our narrow use case (just ask the company for its locations and overlay them with this map of Houston), in reality this problem becomes far more difficult to solve at scale. A small insurer may be able to get by just by asking for manufacturing facilities and overlaying this information with the local flood risk.
Smaller insurers will find this work inefficient and expensive because they lack the depth in their data science departments to do this analysis. Once the academic study becomes out-of-date, they will need to find a replacement. They would find an off-the-shelf data product, delivered quarterly, much more economical.
For big insurers, the challenge comes rather from working with big companies whose manufacturing facilities are spread across the country or world. This data is required at scale, and the problem becomes piecing it togther from fragmented and inconsistent data sources. Just locating data providers to cover all required geographies is a major undertaking.
Fortunately, finding hard-to-find data for hard-to-solve problems is 411 Data’s speciality.
We can provide climate risk products covering US, UK and Europe. All are data quality checked and standardised, so that only one onboarding is required. Many additional forms of data can be overlaid in order to incrementally improve the model, keeping the customer competitive versus their peers and avoid “holding the can” on the most high-risk items.
Keen to find out more? Book a demo and we’ll show you.