Chapter 6 — Key Takeaways

A one-page card. Risk: what it is, how to measure it, and how to think about it like an underwriter.

The core claims

  • Risk is the uncertainty of loss — and the thing you price is the uncertainty, not the loss. A loss you could predict with certainty would be a bill, not a risk. Underwriting converts uncertainty into a defensible price and terms.
  • Insurance addresses pure risk, not speculative risk. Pure risk = loss or no loss; speculative risk carries a chance of gain. The retained upside is exactly what would turn a pool into a subsidized gamble, so the market — not the insurer — bears speculative risk.
  • Peril ≠ hazard. A peril causes a loss (fire, wind, collision, a lawsuit). A hazard makes a loss more likely or more severe (old wiring, an aging roof, a bad venue). You insure against perils; you evaluate hazards, because hazards are what you can see, measure, and often change.
  • There are four families of hazard, and each answers to a different tool: physical (the photographable condition → inspection + loss control + terms), moral (dishonesty/incentive → investigation + disclosure duties), morale (carelessness → deductibles + aligned incentives), legal (the venue/statute → venue-aware pricing + limits).
  • Every loss has two dimensions: frequency × severity = expected loss. Frequency is how often (stable, predictable, trustworthy from short history). Severity is how big (fat-tailed, dragged by rare large losses). Two risks with the same average can be completely different animals.
  • Exposure is what's actually at risk; the exposure unit is how it's counted. A good base is proportional to loss, practical to measure, and hard to manipulate. Choose the thing that grows when the risk grows (value, payroll, sales, vehicles). Interrogate values — undervaluation hides until a loss arrives.
  • Risk classification is the underwriter's primary tool and the place where the stakes are highest. Sort risks into groups of similar expected loss; this is the cure for adverse selection. The line it must not cross: unfair discrimination by protected class, including via a proxy.
  • Insurability is a function of the risk PLUS the terms, price, and machinery — not a yes/no fact. Few risks are flatly uninsurable; most are writable at some price, with some terms, behind some reinsurance, for some carrier.

The key rule of thumb

The underwriter's mental move: exposure → hazard → controls → frequency × severity → a grade. For each thing of value, ask: What's at risk? What makes a loss likely or severe (which family)? What controls exist or can I require? Then, for the controlled risk, how often × how big? The grade is rarely "good" or "bad" — it is usually "controllable, at the right price and with the right requirements."

$$\text{Expected Loss} = \text{Frequency} \times \text{Severity}$$

The numbers that matter (and their limits)

  • Frequency carries signal even from a few years of a risk's own history — counts obey the law of large numbers reasonably well.
  • Severity is treacherous: most claims are modest, a few are enormous, and the maximum (not the average) severity is what decides whether the company survives a catastrophe. Expected severity prices the everyday risk; maximum severity sizes the capital and reinsurance.
  • The blind spot is time. Frequency and severity are backward-looking; they are silent about novel and latent hazards. "No losses yet" is a question, not a verdict (see the asbestos case).

Key terms

pure risk · speculative risk · peril · hazard · physical / moral / morale / legal hazard · frequency · severity · exposure unit · risk classification

What you could defend to your manager

"Harbor Steel is a coastal metal-fabrication risk with a complete inventory across all lines. The controllable physical hazards — aging roof, aging wiring, hot-work near combustibles, aging sprinklers — point to fire frequency and wind/water severity, and each has a requirable control. The strained insurability criterion is the only one I'm truly worried about: 'non-catastrophic to the insurer,' because the coastal accumulation breaks the independence assumption for the windstorm peril and only for it. That single fact tells me what to do — price the catastrophe exposure adequately, attach a percentage named-windstorm deductible, and cede the tail to reinsurance — and turns what a careless reader would decline into a writable, controllable account. I haven't priced or decided it yet; I've seen it correctly, and the verdict is: controllable, at the right price and with the right requirements."