Insurance has exactly one legitimate function — hedging catastrophic, low-probability events. Everything else it does makes things worse.
Links: Economics, Risk and Entrepreneurship, Value and Profit, Scope Confusion, Civilizational Cycles, Words of Wisdom
Insurance is only legitimate where moral hazard approaches zero, which is only where individual control approaches zero, which is only at the catastrophic tail.
This is the single-sentence version. Everything below unpacks it.
Moral hazard is the free rider problem wearing a different hat. In socialism, the free rider consumes without producing. In insurance, the moral hazard actor consumes risk without bearing it. Structurally identical — you’re externalizing cost onto the pool.
The moral hazard gradient maps onto the insurability gradient:
| Event | Individual Control | Moral Hazard | Insurable? |
|---|---|---|---|
| Lightning strike | Zero | Zero | Yes |
| Random aggressive cancer in a healthy 25-year-old | Near-zero | Near-zero | Yes |
| House fire from lightning | Zero | Zero | Yes |
| House fire from candles | High | High | No — “insurance will cover it” changes behavior |
| Car accident (reckless driving) | High | High | No — insured drivers take more risks (empirically documented) |
| Unemployment | High | High | No — you can always accept a lower wage |
| Business losses | Full | Full | No — you can make deliberate losses |
| Suicide | Full | Full | No — obvious |
| Not getting out of bed | Full | Full | No — “if I could insure myself against it, I would always feel ill” (Hoppe) |
The pattern: As individual control increases, moral hazard increases, and the pooling mechanism degrades. At the catastrophic tail — events so rare and so random that individual information is useless — the free rider problem vanishes because there’s nothing to free-ride on. Nobody can game a meteor.
Hoppe (following Knight and Mises) distinguishes:
Hoppe’s framework says insurance works when you can define stable classes. But his own examples keep demonstrating that the class breaks down under refinement. Football players vs. desk workers — different class. But within football players: linemen vs. kickers. Within linemen: prior injuries vs. none. Drinking vs. not. The discrimination ratchet keeps subgrouping because there’s always more information that distinguishes individuals.
This subgrouping process doesn’t converge on stable classes — it converges on individuals. And once you reach the individual, you’re back to case probability, which is uninsurable by Hoppe’s own definition.
The catastrophic framing solves this. When probability is sufficiently low and the event is sufficiently outside anyone’s control, class distinctions genuinely don’t matter. A marathon runner and a couch potato have approximately the same probability of being hit by a meteor. The discrimination ratchet stops naturally because there’s no information that meaningfully distinguishes individuals for truly random catastrophic events.
Insurability is inversely proportional to the degree to which individual information predicts the outcome.
Hoppe (drawing on Mises) traces the trajectory of government insurance intervention:
Each step follows logically from the previous one. Each intervention creates the problem that justifies the next intervention. This is the Misesian critique of interventionism applied to a specific domain.
Universal healthcare makes the government an insurance company. The argument for it — monopsony (single buyer) reduces costs through bargaining power — ignores that monopsony is as bad as monopoly for the same structural reasons:
This is the opposing forces principle: clamping down on markets generates counter-forces. The government creates the shortage, then has to manage the escape valves, each of which invites another intervention.
Canada’s MAID (Medical Assistance in Dying) illustrates the spiral taken to completion. MAID didn’t emerge directly from insurance rationing — it emerged from a separate rights-based legal argument (autonomy, right to die). But once the universal system was rationing by default (18-month wait times for surgery), offering death to people who can’t get treatment becomes functionally equivalent to rationing — even if nobody admits that’s what’s happening.
The government didn’t design MAID as cost management. But the incentive structure doesn’t care about intentions. The system corrupts even well-intentioned policy through structural pressure. MAID started as morality-scope (individual autonomy over one’s own death — perfectly legitimate). It got codified into law-scope (legislation). And now it functions as economics-scope (managing healthcare costs). Three different scopes, one policy, and the scope confusion is invisible to most people defending it.
Mises in 1922: social insurance “weakens the will to health” and “produces illness by subsidizing it.” He was talking about people staying home from work. Scale that mechanism up: subsidize illness → people stay sick longer. Subsidize unemployment → people stay unemployed longer. Subsidize dying → more dying. The incentive structure doesn’t care about magnitude.
“Compassion is always good.” — “So you would go ahead and pet those baby bear cubs over there?”
Compassion without information is indistinguishable from recklessness. The person petting the cubs feels compassionate. The mama bear doesn’t care about feelings. Reality has structure that doesn’t bend to emotional states.
Universal healthcare feels compassionate. The incentive structure doesn’t care. You subsidize illness, you get more illness. The compassion was real. The results are structural. And the people who point at the results get accused of lacking compassion — because critics are evaluating intentions while the system is producing outcomes.
This connects to the empathy debate finding: empathy WITH information reduces to rational analysis; empathy WITHOUT information is empty projection. Compassion is the same — with information it becomes wise action; without information it becomes the bear cubs.
The Monopoly AI project produced a concrete demonstration of the catastrophic-tail principle. When analyzing optimal cash reserves (the “insurance premium” against landing on an opponent’s developed property):
The liquidity curve is sharply convex:
| Shortfall Size | Effective Cost | What Covers It |
|---|---|---|
| $50–$200 | 0.10–0.16x | Mortgage buffer (cheap) |
| $300–$500 | 0.17–0.53x | Mortgage + some house selling |
| $600 | ~1.0x | Crossover point |
| $750+ | 2.3–6.2x | House selling, bankruptcy risk |
For small, routine shortfalls: the cost is trivial (0.1x). Insurance adds no value — you can self-insure for essentially nothing.
For catastrophic shortfalls ($750+): the cost explodes to 2–6x. This is where “insurance” (holding cash reserves) earns its keep. Without it, a single bad roll causes liquidation spirals.
The optimal strategy uses a flat 1.0x risk loading across all shortfall sizes — overcharging on routine events to fund catastrophic protection. This is exactly how insurance premiums work. And it’s the only version that produced statistically significant improvement (Z=2.28 vs static reserves).
Insurance pools face a fundamental tension in how they classify risk:
If risk classes and premiums become perfectly specific to the individual:
If risk classes are broad and premiums are pooled:
There is no middle ground that resolves both problems simultaneously. The more actuarially accurate you get, the less “insurance” you’re providing. The more you pool risk, the more you distort incentives.
The catastrophic tail is the only zone where this paradox dissolves — because at the tail, individual risk classes are genuinely indistinguishable, so pooling doesn’t create systematic redistribution, and the event is outside individual control, so moral hazard doesn’t exist.
Everything above describes what insurance should be. What insurance companies actually do is systematically worse — because the profit motive points in the opposite direction from the legitimate function.
An insurance company is a profit-seeking firm. Its revenue is premiums; its cost is payouts. To maximize profit: collect as much premium as possible, pay out as little as possible. This creates a structural inversion:
The result: the industry covers what doesn’t need covering and finds ways to exclude what does. The one scenario where pooled risk is theoretically justified is the one scenario where the insurer fights hardest not to pay.
Insurance doesn’t just fail to solve the cost problem — it creates the cost problem. When a third party (insurer) pays the bill, neither the provider nor the patient has incentive to minimize cost:
Remove insurance from the equation and market competition on price reasserts itself — because the person paying is the person choosing. The third-party payer problem is why healthcare costs are opaque, inflated, and rising faster than any other sector.
If insurance is a bad deal in practice, what’s the alternative? Put the premiums in your own interest-bearing account:
Lived example: Chris was diagnosed with rectal cancer at 54. Radiation, chemo, surgeries, exploratory procedures — the paper bill was ~$500k, the real cost ~$100k. Out-of-pocket with insurance: <$8k. But decades of premium payments into a savings account would have covered the real cost several times over. The insurance “worked,” but it was his own money all along — minus the insurer’s profit margin and the price inflation the insurer’s existence caused.
The steel-man for insurance: what if catastrophe hits at 25 before the premiums accumulate? This is real — and it’s the fear that drives people to buy insurance. But the fear itself is a moral hazard: it’s the emotional mechanism that keeps people paying into a system that structurally works against them. The rational response — catastrophic-only policies with high deductibles at a fraction of current premiums — exists but is actively discouraged by both insurers (not profitable enough) and regulators (not “comprehensive” enough).
The fear of early catastrophe is legitimate. But the system that exploits that fear doesn’t solve it — it extracts rent from it.
From Risk and Entrepreneurship: “Risk cannot be eliminated — only managed.” Insurance is one tool for risk management, but it’s a narrow tool with a narrow domain:
The entrepreneur’s relationship with risk is instructive. The entrepreneur doesn’t insure against business failure — they accept the risk, manage it through judgment and structure, and earn profit as the reward for bearing it correctly. Insurance for routine events eliminates this feedback loop: you never learn to manage risk if someone else always absorbs the cost.