Availability Bias: Why Insurers Overreact to the Last Catastrophe

After the 2020 wildfire season, which burned over 4 million acres in California, insurance carriers dramatically tightened wildfire underwriting. After a relatively quiet 2019, many had loosened it. This pendulum behavior isn't strategic. It's a textbook manifestation of availability bias.

Availability bias, identified by Tversky and Kahneman in 1973, describes our tendency to overestimate the probability of events that are easy to recall, typically because they are recent, vivid, or emotionally charged. A devastating wildfire season is all three.

Consider the pattern: in the 12 months following a major catastrophe, carriers may tighten underwriting restrictions by an average of 35%. In the 12 months following a quiet year, they may loosen restrictions by 20%. This creates an asymmetric ratchet that progressively shrinks the addressable market without necessarily improving the risk profile.

Here's the paradox: the statistical probability of a wildfire at any given address doesn't change based on what happened last year (with the exception of recently burned areas, which are actually less likely to burn again). The underlying risk factors, including climate, vegetation, ignition sources, and topography, evolve over decades, not seasons.

Yet carriers often behave as if the probability of wildfire has increased in the year following a major fire and decreased in the year following a quiet one. This is pure availability bias: the salience of the recent event distorts the perceived base rate.

What if carriers with the most volatile underwriting criteria (swinging between loose and tight) end up with loss ratios 4-7 points higher than carriers with consistent, data-driven criteria? Industry evidence suggests this is exactly what happens. Reactive carriers pay a 'whiplash premium,' the cost of constantly over-correcting.

The solution is to anchor underwriting decisions to long-term statistical models rather than recent events. This doesn't mean ignoring new information. Models should be updated annually with the latest data. But the update should be systematic, not emotional. A model that incorporates 30 years of wildfire data won't swing dramatically because of a single bad season.

Carriers who resist the siren call of availability bias and maintain consistent, evidence-based underwriting criteria could outperform their reactive peers by 5-8 points on combined ratio over a 10-year period. In an industry where 2 points is considered a significant improvement, this would be transformational.