Profitably Underwriting Properties with Wildfire Exposure
Using AI to Increase Unit Sales without Adverse Selection
Today, the property insurance industry is contracting. While revenues may be rising due to higher pricing, unit sales are down. Property insurance companies are retreating from risks they cannot quantify, particularly those related to climate change, such as hurricanes and wildfires. This trend is especially pronounced in states with significant wildfire risk.
Without tools to correctly price, underwrite and reserve for the losses, insurers cannot remain profitable. Therefore, they leave the market.
Enter Athena’s Voice of the Acre® with superior wildfire insights. By analyzing the characteristics of the land in past wildfires, Athena can generate probabilities for insured losses, even in areas newly exposed to wildfire risk due to weather volatility. As a data vendor, Athena offers Green/Yellow/Red underwriting decision flags, making it easy for underwriters to integrate this data directly into their existing software.
Discussion of multiple aspects of the model can be found elsewhere on this blog, but we have found that a demonstration based on areas you are familiar with is the most effective way to showcase the power of Athena’s prefire analysis. (Sign up for a demo here.)
Athena’s model is designed to give consistent results about wildfire risk.
Consistency good enough for actuarial analysis and financial service modeling.
Consistency good enough to be useful for mitigation work and disaster planning, in advance of any major fire.
Consistency that is understandable, statistically rigorous enough withstand scrutiny by expert witnesses in legal action by power companies.
Below are two property level examples, the first is from California:
Athena has developed an algorithm that can determine the probability of a structure being inside or outside of a wildfire’s perimeter if a wildfire occurs in the area. We believe this tool can enable profitable underwriting and pricing of wildfire risk.
Athena processes massive amounts of highly disaggregated data from a variety of sources in multiple formats. This data is converted into 30-square-meter pixels, each with a unique alphanumeric code and a time sequence. Everything related to wildfire within several miles of each pixel is incorporated into the conditional profile. These profiles are grouped into cohorts, and within a bioregion, each wildfire from the past nine years is evaluated. The conditional profiles in and around each wildfire are used to create probabilities, reflecting the current conditions for each pixel, regardless of the local history.
Athena’s approach leverages deep learning and artificial intelligence (although we prefer the term, augmented intelligence) to enhance the accuracy and reliability of wildfire risk assessments. By utilizing machine learning algorithms, Athena can analyze vast datasets and multiple variables to identify patterns that are not immediately apparent to human analysts. These advanced information technologies enable the algorithm to create probabilities. This integration of disaggregated, yet best-in-class, data and machine learning positions Athena at the forefront of wildfire risk management, providing unparalleled insights and tools for underwriters.
Here is another example from California:
Raw data is converted to indexes and derivatives and, using deep learning (sometime called Artificial Intelligence), yet traceable back to nine (9) key attributes. These attributes are defined and described in this conditional profile’s alpha numeric code.
Translated into English, the code is:
- WUI Density is Medium
- Proximity is Interface
- Conditional Risk is Low
- Probabilistic Risk is Elevated
- WHP (Wildfire Hazard Potential) is Green
- Burn Probability is Green
- CFL (Conditional Flame Length) is Green
- Building Exposure is Green
- Risk to Structure is Green
In California, over 5 years, this particular conditional profile has presented itself 14,282 times — and burned twice.
If a wildfire comes to Marin County, California, the probability that this property is inside the perimeter of a wildfire is 1 in 861 (or 11.6 basis points — one 10th of one percent)
Athena has this information for every address in the western US, including Alaska and Hawaii. Here is an example for Boulder, Colorado:
The nine key attributes for this property are:
- WUI Density is High
- Proximity is Interface
- Conditional Risk is Low
- Probabilistic Risk is Low
- WHP (Wildfire Hazard Potential) is Yellow
- Burn Probability is Green
- CFL (Conditional Flame Length) is Green
- Building Exposure is Green
- Risk to Structure is Green
In Colorado, over 5 years, this particular conditional profile has presented itself 12,220 times — and none have ever burned.
If a wildfire comes to Boulder County, Colorado, the probability that this property is inside the perimeter of a wildfire is 1 in 666,809 (or basically zero).
Accurate and mathematically precise information about future wildfire risk will make informed decisions by underwriters easier. For more examples of how the information can be used, please visit the Case Studies at AthenaIntel.io.
Athena Intelligence is a data vendor with a geospatial, conditional, profiling tool that pulls together vast amounts of disaggregated wildfire and environmental data to generate spatial intelligence, resulting in a digital fingerprint of wildfire risk. (website: AthenaIntel.io)
Clients include electric utilities, communities, insurance and financial services companies, where Athena’s geospatial intelligence incorporated into wildfire mitigation plans (WMP) and public safety power shutoffs (PSPS), Community Wildfire Protection Plans (CWPP), property insurance underwriting and portfolio risk optimization.
