Case Study 2: Wildfire, the Late-Arriving Model, and the Moving Baseline

Note on figures and framing. This study draws on the real, public arc of wildfire as an insurance catastrophe — the severe Western U.S. and California wildfire seasons of recent years, the maturing of wildfire catastrophe models, and the public regulatory debate over their use. All magnitudes are kept qualitative, and no specific loss, market-share, or rate figure is invented. The structure deliberately mirrors Case Study 1 from a different peril and a forward-looking angle, so you can see the same machinery succeed and strain under different conditions.

Background

For most of the catastrophe-modeling era, wildfire was the poor relation. Hurricane and earthquake models were mature and central to property underwriting; wildfire was often treated as a more localized, secondary peril, modeled less rigorously and priced more crudely. That treatment reflected a real history: for decades, wildfire losses, while serious, did not rival the largest hurricanes and earthquakes in aggregate insured cost, and the peril seemed more containable.

That changed. A series of severe Western United States wildfire seasons in recent years — California especially, but across the West — produced some of the most destructive and costly wildfire losses on record, with whole communities in the wildland-urban interface (WUI) destroyed. Wildfire moved, in the industry's mind, from a secondary peril to a primary catastrophe peril, and the gap between how well the industry modeled it and how much loss it was producing became impossible to ignore. The catastrophe-modeling vendors responded by building far more sophisticated wildfire models — incorporating fuel, terrain, weather, ember spread, and the structural vulnerability of homes at the WUI — and these models began to be used to set PMLs, manage accumulation, and price wildfire risk the way hurricane models had long been used on the coast.

The insurance and underwriting issue

Wildfire concentrates several of this chapter's hardest lessons, and from a complementary angle to Katrina.

A peril that was under-modeled, then repriced sharply. When a previously crude peril is suddenly modeled well, the indicated price can jump — not because the risk changed overnight, but because the measurement of it improved and caught up with a worsening reality. This is §30.2 and §30.6 together: better hazard and vulnerability modeling revealed catastrophe contributions that older, cruder methods had understated. For an underwriter, a model upgrade that raises a zone's PML is not a modeling error to be argued away; it is the discipline doing its job, and it can change which accounts you can write.

The moving baseline, made vivid. Wildfire is the cleanest illustration of §30.6's non-stationarity. Drier conditions, longer fire seasons, and more development pushing homes deeper into the WUI mean a wildfire model trained on the past systematically understates the present and near-future risk for many areas. The direction of the error is known — understatement for an intensifying peril — which is exactly why the disciplined response is to treat the historical-calibration number as a floor and to favor forward-looking, climate-conditioned views where they are credible.

Modeling versus regulation (the §30.4 Compliance Corner, live). Wildfire is where the tension between the actuarially indicated cat load and the politically approvable rate became most public. In California, the long-running debate over whether, and how, insurers may use forward-looking wildfire catastrophe models in rate filings — historically constrained, more recently the subject of reform efforts — is a textbook case of the rate-regulatory friction described in §30.4. When the model says the price must rise and the approved rate cannot, the gap does not vanish; it drives the availability problem in §30.7.

The protection gap and the residual market. As wildfire risk rose and adequate pricing was hard to charge or to get approved, some insurers reduced their wildfire-exposed writings or declined to renew the most exposed properties — and the displaced risk flowed to the state residual mechanism (the California FAIR Plan), whose own exposure and concentration climbed. This is §30.7 enacted: rigorous accumulation discipline by individual carriers, summed across the market, widens the protection gap and shifts catastrophe risk onto a public-facing backstop.

What it shows

Where Katrina showed a mature model surprised by a novel event, wildfire shows a late-arriving model catching up to an intensifying peril — and both point to the same conclusion. The cat model is not a fixed oracle; it is machinery that improves, and whose improvement can sharply change the picture. Wildfire shows three things especially clearly:

  1. Insurability is machinery-dependent (§30.7). Wildfire risk did not become "more insurable" or "less insurable" purely because the peril changed; insurability shifted as the modeling, the pricing freedom, the reinsurance, and the mitigation crediting changed around it. A peril can be uninsurable at a suppressed price and insurable at an adequate one with mitigation credits — the risk is the same; the machinery is different.
  2. Mitigation is part of the model (§30.2 vulnerability). Hardened roofs, ember-resistant vents, defensible space, and community-level fuel management all change the vulnerability function — they genuinely lower the modeled loss. Wildfire made the case that crediting verified mitigation is not a marketing gesture but a real change to the risk the model sees, and a partial answer to the protection gap.
  3. The regulatory gap is a real driver (§30.4, §30.7). When the indicated price cannot be charged, the model's accuracy does not protect the policyholder — capital leaves, and the gap widens. Modeling and regulation are not separate problems.

Outcome

The public arc is ongoing, but the broad strokes are documented. Wildfire is now treated as a primary catastrophe peril; the models are far more sophisticated than a decade ago; the use of forward-looking wildfire models in rate-making has been a sustained regulatory battleground, with movement toward permitting such models under conditions tied to mitigation; residual-market wildfire exposure has grown; and mitigation crediting and community-level resilience have become central to the availability conversation. The protection gap in the most exposed WUI areas remains a live public-policy problem, with private coverage, residual markets, and proposals for public-private structures all part of the response (the future-of-insurability discussion continues in Chapter 36).

Lesson

Wildfire teaches the same humility as Katrina, from the opposite direction — the danger of under-modeling and non-stationarity rather than of a mature model's blind spot — and adds the regulatory dimension:

  • A model upgrade that raises the PML is the discipline working, not a glitch. When better measurement reveals a higher catastrophe contribution, you re-underwrite to the new number; you do not negotiate it back down because it is inconvenient.
  • For intensifying perils, the historical number is a floor. Wildfire's moving baseline is the §30.6 argument in its starkest form. Favor climate-conditioned views, load for non-stationarity, and revisit zone limits more often than a stationary world would require.
  • Credit real mitigation — it changes the vulnerability function. Hardened structures and defensible space lower the modeled loss for real, and crediting them is both good underwriting and part of narrowing the protection gap.
  • The approvable rate and the indicated rate can diverge — and where they do, capital leaves. Your internal pricing must reflect the true modeled risk even where the filed rate cannot; persistent divergence is the engine of the availability crisis, not a footnote to it.

Discussion questions

  1. Contrast this case with Case Study 1: Katrina was a mature model surprised by a novel event; wildfire is a late-arriving model catching up to an intensifying peril. What single lesson about catastrophe models do both cases teach (§30.2, §30.6)?
  2. A wildfire model upgrade raises your WUI zone's 1-in-100 PML substantially, tightening or eliminating its headroom. A senior colleague wants to keep using the old model because "the new one will kill our growth." Using §30.2 and §30.5, explain why that is the wrong response and what the right one is.
  3. Explain how verified mitigation (hardened roofs, defensible space) enters the vulnerability module (§30.2) and genuinely lowers an account's modeled catastrophe loss. Why does this make mitigation crediting both good underwriting and a partial answer to the protection gap (§30.7)?
  4. Walk the §30.4 / §30.7 chain for wildfire: a forward-looking model raises the indicated rate → the approved rate cannot keep up → trace the consequences for private writings, the FAIR Plan, and the gap.
  5. The chapter argues insurability is "a property of the risk plus the available machinery." Identify three distinct pieces of machinery in this wildfire story (a kind of model, a pricing/regulatory freedom, and a mitigation or backstop mechanism) and explain how changing each one moves the line of what is insurable.