Why insurance has enemies at all
By now you have met the engine room of insurance: a crowd of people each facing a small chance of a large loss hand their risk to a pool, and because the law of large numbers tames the average, everyone trades a scary maybe-loss for a calm certain premium. That story quietly assumed something tidy — that everyone in the pool looks much the same, and that nobody acts differently just because they are covered. Real people break both assumptions, and the breaks are not random noise. They lean in one direction, always against the insurer.
The two breaks have names, and they are the great twin headaches of the whole industry. Moral hazard is the problem of behaviour changing after you buy: a covered driver parks a little less carefully. Adverse selection is the problem of who buys in the first place: the people most certain they will claim are the people most eager to sign up. One is about deeds, the other about identities, but both share a root cause — the customer knows things about their own risk that the insurer cannot fully see. Economists call that gap *asymmetric information*, and these two enemies are simply its two faces.
Moral hazard: the cover changes the bet
Recall what insurance does at its heart: it removes the financial sting of a loss. That is the whole point — and it is also exactly the trouble. Once the sting is gone, so is part of your reason to avoid the loss in the first place. A driver who knows a dented bumper costs them nothing will, on average, brake a fraction later. A homeowner whose flooded basement is fully covered will postpone fixing the leaky pipe. Nobody is plotting; the incentive has simply, invisibly, shifted. The act of insuring a risk nudges that risk slightly larger.
Actuaries split this into two flavours, because they need different defences. *Ex-ante* moral hazard happens before a loss: you take less care to prevent it. *Ex-post* moral hazard happens after a loss has occurred: you claim more lavishly than you otherwise would — the cheap repair becomes the expensive one, the two-night hospital stay becomes three. Both inflate the cost the pool must carry. And here is the honest twist: a tiny bit of moral hazard is not even a flaw. Some of it is the insurance doing its job — letting a sick person finally see the doctor they were avoiding for cost. The art is telling wasteful behaviour-change from the valuable kind.
Adverse selection: the sick rush the door
The second enemy strikes before anyone has even behaved at all — it is about who chooses to buy. Suppose an insurer offers one flat health premium to everybody, set at the average cost of the whole population. To a healthy person that price looks like a bad deal, so many of them decline. To a chronically ill person it looks like a bargain, so they sign up eagerly. The pool that actually forms is therefore sicker than the population the price was based on. Claims run higher than the premium can cover.
Now watch the trap spring. To stay solvent the insurer raises the premium. But the new, higher price drives away the next-healthiest layer of customers — the people who were just barely willing to stay. The pool gets sicker still, so the price must rise again, which chases out yet more healthy people. In its worst form this spiral feeds on itself until only the very riskiest remain and the price becomes absurd. Insurers call it the death spiral, and it is adverse selection's signature catastrophe: a market that, left unguarded, can eat itself alive.
Notice how cleanly this differs from moral hazard. Adverse selection needs nobody to change their behaviour at all — the chronically ill person is being entirely honest and reasonable in buying a bargain. The damage comes purely from the *mix* of who shows up, driven by what they privately know about themselves. In health insurance the same idea wears a sharper name, antiselection; it is the reason questions about pre-existing conditions exist at all.
The defences, before the policy is sold
The first line of defence is to close the information gap before signing anyone up. This is underwriting: gathering enough about an applicant — age, health, driving record, the building's wiring — to judge their true risk and price accordingly. Closely paired with it is risk classification, the act of sorting applicants into groups that genuinely behave alike, so each group can be charged its own fair rate. A 19-year-old with three speeding tickets and a careful 45-year-old simply do not belong in the same priced bucket.
Classification is exactly the antidote to the death spiral. If healthy applicants are offered a price that reflects *their* low risk, they no longer flee, so the pool does not curdle. But the same tool raises a permanent tension that actuaries must own honestly. Refine your classes too finely — charging people for risks they cannot control, such as a genetic condition or where they were born — and you have technically efficient pricing that society may judge deeply unfair. The line between fair discrimination by risk and unfair discrimination against people is not a maths question; it is set by regulation, ethics, and public values, and it genuinely moves over time.
Two blunter structural defences also fight selection. Mandates — laws that make everyone buy in — refill the pool with healthy people who would otherwise opt out, which is why most countries require everyone to carry motor liability cover. And waiting periods plus a contestability period (a window in which the insurer can void a policy for misstated facts) discourage people from buying only once they already smell trouble coming.
The defences, written into the policy
The second family of defences fights mostly moral hazard, by leaving the policyholder with some of the loss so they keep a reason to care. An deductible makes you pay the first slice of any claim yourself; a coinsurance arrangement makes you pay a fixed percentage of every dollar; a policy limit caps what the insurer will pay at the top end. Each one hands a portion of the risk back to you — and that retained sliver quietly rebuilds your incentive to avoid losses and not over-claim.
Loss = 5,000 Deductible = 500 Coinsurance: you pay 20% above it You pay : 500 + 20% x (5,000 - 500) = 500 + 900 = 1,400 Insurer : 80% x (5,000 - 500) = 3,600 ---------------------------------------------------------- Total : 5,000 You carry 1,400 of every 5,000 loss -> a real reason to be careful.
These features also feed back into price. Because a deductible strips out the swarm of small, cheap, easily-triggered claims, it lowers the insurer's expected payout — so a higher deductible buys a lower premium. That trade-off is itself a quiet selection tool: a careful, low-risk customer is happy to accept a big deductible for a cheaper price, while someone expecting to claim constantly is not. The policy design lets the right people sort themselves into the right price, an idea actuaries lean on heavily when setting the full gross premium.
Why this is the actuary's quiet obsession
Step back and the two enemies explain a striking amount of how insurance actually looks. Deductibles, medical questionnaires, waiting periods, no-claims discounts, the very existence of underwriters — almost every feature you find irritating about buying insurance is there because, without it, the insurance arrangement would quietly poison itself. They are not bureaucratic spite; they are the immune system that keeps a fragile, beautiful idea alive. Strip them away and you do not get cheaper, friendlier insurance — you get no insurance, because the pool collapses.
It also reframes what makes a risk worth covering at all. Back when you met the checklist for an insurable risk, one quiet requirement was that the insurer must be able to assess and monitor it. That requirement is precisely a demand that moral hazard and adverse selection stay manageable. A risk the buyer can secretly trigger at will, or that only the doomed ever buy, is uninsurable not because the maths fails but because these two enemies overwhelm it. Watching for them, and designing the defences, is a thread that runs through every job an actuary ever does.