Catastrophe (Cat) Strip Product

Helping reduce the costs of wildfire reinsurance and home insurance in CA

Intro

California has experienced an unenviable increase in the frequency and severity of wildfires over the past 15 years.  This has caused extreme dislocation in pricing comparative to the actual risk.  That said, even in 2020 (the year with the most acres burned in history) only 4% of the state has been burned by wildfire and only roughly 10,500 structures, less than 0.1% of total households in California.The aggregation of large blocks of risk essentially create a certainty of loss because the lower level majority of risks are imprisoned with those severely exposed. Primary carriers in California have an opportunity to re-shape risk transfer through the ‘fractionization’ of huge bundles of risk. This would reintroduce outcome diversity , improve transparency in underwriting and create a more stable market that can attract out of state risk capital even if the impact of climate change expands.


The example below is a map highlighting a pro-forma portfolio that includes 13,000 risks from a traditional carriers’ portfolio that contains risks exposed to wildfire.  If we write this portfolio as a whole, it’s necessary to charge an extremely high rate (30-35%) in order to price effectively for the risk volatility caused by the toughest risks being assumed.

Acres vs Structure Burned in California Over Time

What we have found through our detailed technical analysis is that 12% of the homes in CA actually have a significant chance of wildfire damage. However, those 12% of homes are driving 90% of the rate increase paid by far less exposed risks. If there was a way to surgically remove risk and price it on an individual home by home basis, there is a significant opportunity to spread and balance the risk to reduce the costs of reinsurance for primary insurers, and ultimately home owners.

Pro-forma Portfolio of 13,000 Risks

Product

Using proprietary deep learning neural networks, we are able to select a specific tranche (1-5k) out of a carrier portfolio of 50,000 homes, which we believe are being incorrectly rated.  Due to the changing climate and increase in wildfires in CA the industry has increased prices across the entire state to try and compensate. However, even in the worst wildfire years in history, less than 0.08% of the structures in the state burn down.

Kettle’s reinsurance underwriting includes three steps: the first step involves simulating millions of wildfires using Artificial Intelligence (AI) and advanced computing technologies and applying the resulting risk estimates to assess wildfire risk for each property. The second step involves a pricing technique that accounts for co-location and risk clustering. During the third step, the Kettle algorithm strategically forms cat strips that optimize for risk-and-return appetite.

Comparison of Modeling: Current industry view wildfire risk as hyper conservative, while Kettle is hyper detailed in the true nature of risk.

Solution

So far, the expected loss ratio for 2020 wildfires for most major primary carriers that underwrite for wildfire damage is well above an unsustainable 100%. A simple and direct application of Kettle’s risk estimate of wildfire probabilities improves pricing and reduces the loss ratio by significantly more than half. Further application of Kettle’s pricing algorithm that equates for co-location and tail loss would reduce the loss ratio by more than 89%.

Different strips could be made depending on risk appetite and rate of return requirement. Similar to the way hedge funds operate, Kettle’s portfolio optimization algorithm finds a mix of properties that deliver the lowest risk given an expected rate of return. The entire Kettle risk optimization and portfolio management process is done with state-of-the-art machine learning techniques, mathematical solvers, and hyper-efficient cloud computing machines.

Loss Ratio Comparisons of Baseline Portfolio vs Kettle Pricing

The majority of risk emanates from a limited number of peak risk locations. Being able to price accurately and select intelligently allows us to generate cat strips that optimize for risk and return appetites. As seen in Figure 7, 50% of losses often come from 6% of the highest risks, if a model can identify these highest risks and avoid them, it can dramatically reduce the price for the rest of CA homeowners.

Original Book Composition vs Expected Loss Composition

Cat strip 1, cat strip 2, and cat strip 3 are generated based on different expected returns on collateral, which reflect the tail risk at 1 in 250 events. The cat strip compositions are different in the number of higher risk homes they include. So far in 2020, none of them has experienced a loss. The expected returns on collateral are 49.9%, 61.9%, and 81.8% respectively. The standard deviations are 14.5%, 15.1%, and 18.4%.

Rate of Return of Different Cat Strips

Contact us to Schedule a demo of our fire prediction model

Contact Us