Four Major Lessons We Learned About Wildfires in California

Here at UrbanStat, we've been researching and modeling wildfires for about 2 years now. When we first started thinking about this phenomenon, we were shocked to see that most models only utilize a few variables that are not necessarily the causes of wildfires such as vegetation/fuel, slope/topography, and access.

Knowing this we were sure that there is room for improvement and we started working on this problem. During this process, we've learned a lot about wildfires and we believe sharing these findings will help us all as the industry.

Commonly used variables are not enough and they are not necessarily the best data points to explain wildfires. Although the factors like fuel, slope, and access are important to determine the wildfire severity and probability of a location, our analysis on historical wildfires shows that there are more factors to consider.

When you think about fuel, slope, or access, one should ask: are any of these factors actually the cause of wildfires? The answer is pretty clear, a tree or a steep land cannot start its own fire. Most of the fires start because of power infrastructure failures, lightning strikes, arson, camp fires, cigarettes, fireworks, controlled burning debris, or glass acting as magnifier in dry and high temperature zones.

Any of these factors can also enhance the probability and severity of wildfires based on climate (e.g. extreme heat and severe drought). There are also factors that might reduce the risk: humidity/precipitation, existence of natural barriers such as lakes and rivers, and fortifications to the property.

Unfortunately the algorithm that insurance companies are primarily relying on doesn't consider most of these factors. Our analysis shows that there are more than 50 factors that explain the probability of wildfires.

Wildfires are more frequent and more severe in last 10 years compared to last 50 years. Historically, about 0.53% of California experiences wildfires every year. However, when we look at the trend, we see that this ratio has almost doubled between 2008-2018 and reached 0.90%.

Most wildfires occur in previously not impacted areas. About 72% of the areas which were hit by a wildfire was never hit by another wildfire again in over 50 years. It takes 17.3 years on average for the second fire to occur in the same location.

Accepting properties in previously impacted areas are actually more profitable. When we run simulations on different portfolios in California between 2007 and 2018, we see that profits are improved by 2 points although the loss ratio is higher.

An underwriting strategy that allows renewal of previously impacted properties combined with more sophisticated wildfire models would definitely improve the profits and the loss ratios.