Case Study 1: Hurricane Andrew (1992) — When the Severity Tail Came Home

A real, public event, told with public facts only. No specific dollar figure, market share, or company financial below is fabricated; where exact numbers vary by source or are disputed, the discussion stays qualitative, as the chapter's accuracy rules require. The point is the lesson about risk, not a ledger.

Background

In late August 1992, Hurricane Andrew made landfall in South Florida as one of the most intense hurricanes to strike the United States in the twentieth century, then crossed the Gulf and struck Louisiana. It leveled neighborhoods south of Miami, destroyed or damaged a vast number of homes and businesses, and became — at the time — the costliest natural disaster in U.S. history by insured loss. For the insurance industry, Andrew was not merely a large loss. It was a revelation, and a brutal one: it exposed the fact that much of the property-insurance business had been writing catastrophe-exposed risk without truly measuring the catastrophe.

To understand why Andrew was a turning point, you have to understand the state of catastrophe underwriting before it. For decades, property insurers had priced hurricane-exposed coastal property largely on the basis of historical experience — what losses had actually occurred in the recent past — adjusted by judgment. In a run of mild hurricane seasons, that approach produces comfortable results: the average year is benign, the premiums look more than adequate, the book appears profitable, and competition pushes rates down. Coastal property was widely regarded as good business. The catastrophe models that today simulate hundreds of thousands of synthetic storms and estimate the full distribution of possible losses were, in 1992, in their infancy and not yet widely used. The industry was, in this chapter's vocabulary, pricing the average severity and largely ignoring the maximum severity.

The insurance and underwriting issue

This case is, at its heart, about the two dimensions of loss from §6.3 — and specifically about the dimension the industry had under-weighted. Frequency, for hurricanes, is relatively legible: you can count how often storms of various intensities have historically struck a coast. Severity — and especially maximum severity, the worst plausible single event — is the treacherous dimension, because the historical record may simply not contain the worst case that the physics permits. A coast can go decades without a direct hit from a top-intensity storm on a densely developed area, and during those decades the loss experience understates the true risk. The expected loss in any ordinary year is modest; the loss in the rare bad year is enormous and correlated across the entire book at once. Andrew was that rare bad year arriving with full force.

Three features of catastrophe risk that this chapter has named all converged in Andrew, and it is worth seeing each one in the event:

• The independence assumption failed completely (§6.3, §6.7; Chapter 1). A property book is supposed to be a pool of independent risks, where one loss tells you nothing about the next. A hurricane is the opposite: a single event that damages tens of thousands of insured properties simultaneously. The "diversification" of a large coastal book was an illusion — it was, in the chapter's phrase, a single bet wearing thousands of costumes. Insurers that thought they held a diversified portfolio discovered they held one enormous, correlated exposure.
• Exposure had been measured but accumulation had not (§6.4). Carriers knew, more or less, how much total value they insured in Florida. What many had not adequately grasped was how much of it sat in the same peril zone, exposed to the same storm — the accumulation problem this book develops in Chapter 30. Total exposure measures the size of the bet; concentration measures how much of the company rides on one roll, and the concentration was far higher than the comfortable average results had suggested.
• The maximum severity dwarfed the priced-for severity (§6.3). The price had been built for the everyday loss. The event delivered a loss from deep in the tail of the distribution — the kind of loss that expected severity never sees but that determines whether the company survives.

What it shows

Andrew demonstrates, with unusual clarity, the single most important refinement in this chapter's treatment of severity: for catastrophe-exposed property, the number that prices the everyday risk and the number that threatens the company's survival are two different numbers, and pricing for the first while ignoring the second is how insurers fail. The ordinary-year loss is what competition drives the price toward. The tail loss is what actually arrives, eventually, and decides who is still standing afterward.

It also demonstrates that insurability is a function of the machinery, not a fixed property of the risk (§6.7). Coastal Florida property did not become more hazardous on the day after Andrew — the same homes faced the same storms. What changed was the industry's understanding of the risk and the machinery it brought to bear. The event forced a reckoning that ultimately improved the apparatus: the rapid maturation and widespread adoption of catastrophe models that simulate the full distribution of possible losses rather than extrapolating from recent history; a fundamental rethinking of how much catastrophe risk a single carrier should retain versus cede to reinsurance (Chapter 27); a rethinking of how much capital must stand behind coastal exposure (Chapter 28); and changes in policy structure, including the wider use of percentage named-windstorm deductibles that share catastrophe severity with the insured (Chapter 12). The risk was the same; the machinery for managing it was rebuilt.

The connection to your own desk is direct. Harbor Steel sits on a hurricane-exposed Gulf Coast, in a single county, with no internal geographic diversification of its property exposure — the same structural situation, in miniature, that Andrew exposed at industry scale. Everything you will do to make Harbor Steel writable — pricing the catastrophe exposure adequately rather than for the comfortable average year, attaching a percentage named-windstorm deductible, running the account through a catastrophe model (Chapter 30), and ceding the catastrophe tail to reinsurance (Chapter 27) — is the discipline that the industry learned, expensively, from Andrew.

Outcome

The consequences rippled for years and reshaped the business. A number of insurers were severely weakened by Andrew, and some did not survive it — carriers that had concentrated catastrophe exposure without the capital or reinsurance to absorb a tail event. The reinsurance market repriced sharply as the true cost of catastrophe capacity became visible (a thread Chapter 27 picks up). Regulators and the state of Florida grappled with the availability and affordability of coastal coverage, leading over time to public mechanisms and residual-market entities for risks the private market would no longer freely write — an early chapter in a coastal-property story that continues to this day (and that Chapter 15 develops through the modern Florida and California crises). And catastrophe modeling moved from a niche curiosity to an indispensable tool of property underwriting, accumulation management, and capital allocation.

Lesson

The transferable lesson is the chapter's central quantitative point, paid for in one of the industry's hardest losses: a risk is two numbers, frequency and severity, and for catastrophe the severity that matters is the maximum, not the average. An underwriter who reads only the recent loss experience of a coastal book is reading a record that systematically understates the risk during quiet years — and quiet years are exactly when the price gets competed down and the exposure piled on. The disciplines that protect against this are not optional refinements; they are the difference between an insurer that survives the big one and an insurer that discovers, too late, that its diversified portfolio was a single correlated bet. Price for the tail. Measure the accumulation, not just the total. Cede what you cannot survive. And never mistake a run of quiet years for evidence that the risk is small.

Discussion questions

1. Using the frequency/severity framework (§6.3), explain precisely why a run of mild hurricane seasons makes coastal property look like good business while the underlying risk is unchanged. Which dimension does the recent record measure well, and which does it measure badly?
2. The chapter says the "diversification" of a large coastal book was an illusion. Connect this to the independence assumption from Chapter 1 and the accumulation idea from §6.4. What would genuine diversification of a property book require?
3. Andrew did not make Florida homes more hazardous, yet coastal property became harder to insure afterward. Explain this using the §6.7 claim that insurability is "a function of the risk plus the machinery." What specifically changed?
4. Apply the lesson to Harbor Steel. Name three concrete things you would do — in pricing, in terms, and in reinsurance — because the account sits in a single hurricane-exposed county, that you would not bother with for an identical plant in a non-catastrophe inland location.
5. (Pricing discipline.) A competitor is winning coastal accounts by pricing for the recent quiet years. In the short run their combined ratio looks better than yours. Explain, using this case, why their discipline — not yours — is the dangerous one, and over what time horizon the truth tends to arrive.