How Insurance Pools Risk

Imagine a thousand families on a block. Every year, by bad luck, about three of those houses catch fire. Nobody knows which three. If your house is one of them, the loss is ruinous — maybe half a million dollars gone in a night. If your house is not, you pay nothing and feel lucky. The problem is that nobody can afford to be one of the unlucky three, and nobody can predict in advance who they will be.

Insurance is the trick of turning that situation into a different one. Suppose each of the thousand families agrees to pay $2,000 a year into a shared pot. That pot now holds $2 million. When three houses burn, the pot pays out roughly $1.5 million to rebuild them. The families who didn't have a fire have spent $2,000 each — annoying, but survivable. The families who did have a fire are made whole. Nobody is ruined.

The deep idea here is risk pooling. A loss that is catastrophic for one person becomes a manageable expense for a thousand people sharing it. The unpredictability of which house burns hasn't gone away — it just doesn't matter as much, because the group as a whole faces a much steadier number. Across a thousand houses, the count of fires per year wiggles a little, but it stays close to three. This steadying effect, where the average of many independent events is far more predictable than any single event, is called the law of large numbers.

For pooling to work, two conditions have to hold. First, the losses have to be roughly independent — meaning one family's fire doesn't make the next family's fire more likely. Independence is what keeps the yearly total steady. Second, the people in the pool have to be similar in their actual risk, or else priced accordingly. If you secretly let in a family whose house is already on fire, the math collapses.

This is where the mechanism gets fragile. Two specific failures come up over and over. The first is correlated losses. A wildfire or a hurricane doesn't burn three houses out of a thousand — it burns three hundred. The pot was never built for that, because the assumption of independence was wrong. The second is adverse selection. If the insurance company charges everyone the same $2,000, the safest families (brick house, no woodstove, sprinklers) start to think the deal is bad and drop out. The remaining pool is riskier than the price assumed, so the company has to raise prices, which pushes out the next-safest layer, and so on. The pool unravels from the inside.

Real insurance companies spend enormous effort fighting these two failures. They diversify across regions so one hurricane can't sink them. They use detailed pricing — your premium depends on your roof, your zip code, your claims history — so each person pays roughly what their own risk costs the pool. They require everyone in a group plan to join, so the safe people can't quietly leave.

So insurance is not really about predicting your future. It's about joining a group large enough, and similar enough, that the group's future is predictable even when yours is not. The price you pay is the cost of trading a small chance of ruin for a steady, survivable bill.