Risk Pooling Reduces Uncertainty, Not Cost
Adding members to an insurance pool does not make coverage cheaper per member — each brings expected claims in alongside their premium, so both sides of the ledger grow together. What pooling buys is predictability: relative variability of the average falls as 1/√n, so almost all the benefit is captured by regional or national scale. And the averaging only works for independent risks.
The most common misunderstanding about insurance is that a bigger pool makes coverage cheaper per member. It does not. **Pooling reduces the uncertainty of the average outcome; it does not reduce the underlying cost.** ## Why the cost doesn't fall Each house joining a pool brings its own expected claims *in* alongside its premium. If the average house generates roughly $1,000 a year in expected claims, then adding 10,000 houses adds about $10 million of expected annual payouts and about $10 million of premium. Both sides of the ledger grow together. Your house is not less likely to flood because your neighbour signed up. The potluck analogy is close: adding guests does not make the food cheaper, because everyone still eats. What it changes is the *variance* — who pays in any given month, and how predictable the total is. ## What pooling actually buys Independent risks average out. The law of large numbers makes the realised loss ratio converge on the expected one, so the insurer can charge close to expected loss plus a modest margin rather than pricing for a wide range of outcomes. The crucial part is the *rate* of convergence. The relative variability of the average falls with the **square root** of the number of policies — as 1/√n. That has a sharp practical consequence: - 100 → 10,000 policies cuts relative volatility by a factor of 10. Transformative. - 10,000 → 1,000,000 cuts it by another factor of 10, but from an already-small base. - 1 million → 100 million achieves almost nothing measurable. **Nearly all the statistical benefit is captured at regional or national scale.** There is no mathematical engine pushing insurers toward global size — the returns to scale in pooling are exhausted long before that. See Why There Is No Single Global Insurance Company. ## The condition that matters more than size The averaging only works for **independent** risks. When losses are correlated — everyone in the same storm path, the same flood plain, the same financial crisis — the square-root rule fails entirely and a bigger pool concentrates rather than diversifies. See Correlated Risk: The Failure Mode That Actually Kills Insurers. Which is why "how many policies" is the wrong question and "how independent are they" is the right one. A hundred thousand houses in one neighbourhood is a far worse pool than ten thousand spread across a continent. ## What follows Since pooling cannot reduce the expected loss, the premium is chained to it. A well-run insurer can shave the margin, invest the float, and fund loss prevention — worth perhaps 10–30% — but the physical cost of houses needing repair is not something a financial structure can remove. See Combined Ratio: Where an Insurance Premium Actually Goes.