Case Study 2 — The Flood Protection Gap: Adverse Selection, Catastrophe, and the Limits of Insurability

DataField.Dev

Case Study 2 — The Flood Protection Gap: Adverse Selection, Catastrophe, and the Limits of Insurability

This case draws on the well-documented public history of flood insurance in the United States, including the catastrophic flooding of Houston during Hurricane Harvey in August 2017. It is the dark mirror of Case Study 1: where the Philadelphia Contributionship shows insurance working, the flood protection gap shows what happens when the conditions that make a risk insurable (§1.3) and the forces that poison a pool (§1.4) are not managed. Figures here are stated qualitatively where exact statistics are not certain; the pattern is what matters.

Background

In August 2017, Hurricane Harvey stalled over southeastern Texas and dropped an extraordinary amount of rain — by some measures the largest rainfall event in the recorded history of the continental United States — on the Houston metropolitan area. Tens of thousands of homes flooded, many of them far from any coast or river, in neighborhoods that had never flooded before. The human and economic losses were immense.

Then came the insurance reckoning, and it exposed a structural problem that the industry and the federal government had been living with for decades. A very large share of the homes that flooded had no flood insurance at all. In many of the hardest-hit areas, only a minority of flooded homes were covered for the peril that destroyed them. The losses were real, the damage was catastrophic, and yet the mechanism that exists precisely to spread such losses — insurance — simply was not in place for most of the victims. This shortfall between the economic loss and the insured loss is what the chapter (§1.3) calls the protection gap, and flood is its starkest example.

The insurance issue: why flood is the hardest peril to insure privately

Flood is, in the language of §1.3, a risk that strains almost every criterion of insurability at once, and understanding why is a master class in the chapter's concepts.

It is catastrophic to the insurer — losses are not independent. This is the killer. An ordinary house fire is independent: your house burning tells us nothing about your neighbor's. A flood is the opposite — the same storm inundates an entire region simultaneously, so a private insurer concentrated in a flood zone faces thousands of correlated claims at once. The pool the law of large numbers needs (§1.2) does not exist, because the losses arrive together. This is exactly the independence assumption failing, and it is why, for most of the twentieth century, private insurers largely withdrew from residential flood, judging it uninsurable on standard terms.
Adverse selection is severe. Because flood is bought separately (in the U.S., largely through the federally run National Flood Insurance Program, the NFIP, created in 1968 — you will study it in Chapter 15), the people who buy it are disproportionately those who know they are exposed: homeowners in designated high-risk flood zones, many of them required to buy it as a mortgage condition. Homeowners just outside those zones — like many of Harvey's victims — often decline, reasoning that their risk is low. The pool skews toward the worst risks (§1.4), which drives up the price, which drives the better risks further away. The result is a program perpetually strained, and a vast population of uninsured "moderate" risks who discover their exposure only when the water arrives.
The premium is hard to make economically feasible. For the highest-risk properties, an actuarially adequate flood premium can approach the cost of the expected loss itself — economically infeasible for the homeowner (§1.3) and politically explosive. Decades of subsidized NFIP rates kept coverage affordable but also muted the price signal that would have told people how much risk they were really living with, and arguably encouraged building in harm's way.

A note on honesty about the numbers: precise figures for flood-insurance take-up in specific Harvey neighborhoods vary by source and are debated. What is not debated, and what matters for this chapter, is the pattern: a catastrophic, regionally correlated peril, sold separately, bought mainly by those who know they need it, leaving a large protection gap exactly where the chapter's theory predicts one.

What it shows

The flood protection gap is the chapter's framework run in reverse — a catalog of what happens when the conditions for insurability are absent and the forces against the pool go unmanaged:

When losses are correlated, the law of large numbers fails (§1.2, §1.3). No amount of pooling helps if one event hits the whole pool at once; this is why catastrophe needs the special machinery of Part V (catastrophe modeling, reinsurance, public backstops) rather than ordinary pooling.
Voluntary, separately-sold catastrophe coverage breeds adverse selection (§1.4). When only the obviously-exposed buy, the pool cannot be sound at any politically tolerable price.
The "social function" of insurance is most visible where insurance is absent (§1.7, theme 6). The uninsured Harvey homeowner is the human face of the protection gap: a family for whom the financial catastrophe was not spread across a pool, and who bore the full ruin the chapter's opening describes.

Outcome and lesson

Harvey intensified a long-running debate about how to close the flood protection gap: reforming the NFIP's rates to reflect real risk (the program has since moved toward risk-based pricing under an initiative called Risk Rating 2.0), expanding the nascent private flood market (newly possible because catastrophe models and global reinsurance now let some private insurers price and lay off correlated flood risk in ways that were impossible decades ago), and rethinking where society allows building at all. None of these has fully closed the gap, and climate change (Chapters 30 and 36) is widening the underlying hazard.

The lesson for the underwriter is twofold. First, insurability is not a fixed fact about a peril; it is a function of the tools available to manage it. Flood was effectively uninsurable privately in 1960 and is partially insurable privately today — not because flood changed, but because catastrophe modeling, reinsurance, and capital markets changed what underwriters can do with correlated risk. Second, the protection gap is a standing reminder that the chapter's sixth theme is not sentimental decoration: underwriting decisions about who can get coverage, and at what price, determine who is protected from ruin and who is not. When you underwrite Harbor Steel's hurricane exposure later in this book, you will be working at exactly this frontier — making a correlated, catastrophe-exposed risk insurable through price, terms, and reinsurance, rather than simply declaring it uninsurable and walking away.

Discussion questions

Using the six insurability criteria from §1.3, build the case that residential flood is the hardest common peril to insure. Which single criterion does the most damage, and why?
Explain precisely how selling flood coverage separately (rather than bundling it into every homeowners policy) worsens adverse selection. What would bundling do to the pool?
The chapter says insurability "is not a fixed fact about a peril; it is a function of the tools available to manage it." Defend or challenge this claim using the flood example and one other peril.
Subsidized flood rates kept coverage affordable but muted the price signal of risk. Using moral and morale hazard (§1.5), explain how artificially low prices for a hazard can change behavior in ways that increase total losses over time.