Moral Hazard: Why Insurance Can Make You Reckless

Two drivers. Same car, same roads, same skill.

One has no insurance. Dent the bumper backing out of a tight spot, and it comes out of his own pocket.

The other has full comprehensive coverage. Dent the same bumper, and insurance absorbs almost the whole cost.

Ask yourself honestly: which one backs out of that spot a little faster?

Neither driver is being irrational. Both are responding correctly to the risk they actually face — it's just that the risk they face isn't the same anymore. That gap, between the risk a person creates and the risk they personally bear, is what economists call moral hazard: protect someone from the downside of a decision, and you change how carefully they make it.

What moral hazard actually is

Moral hazard isn't about dishonest people. It's about honest people responding rationally to an incentive that's been rewired without anyone announcing it.

The formal version: moral hazard happens after a deal is struck, when one party's protection from a bad outcome changes how much risk they take on — because someone else, not them, absorbs the cost if it goes wrong.

Economists usually file it under a wider umbrella called the principal-agent problem: whenever one party (the agent — the driver, the borrower, the bank) acts on behalf of, or at the expense of, another (the principal — the insurer, the lender, the depositor), and the agent knows more about their own behavior than the principal can watch in real time, the agent's incentives drift away from the principal's. Moral hazard is that drift, specifically the version that only shows up after the contract is already signed.

It's worth separating this from its more famous cousin, adverse selection, because the two get confused constantly and a claims or underwriting system that conflates them will get built wrong. Adverse selection is a before-the-deal problem: the people most likely to buy insurance are disproportionately the people who already know they're higher risk — a smoker buying life insurance, a driver with three past accidents shopping for a new policy. Moral hazard is the after-the-deal problem described above: even an average, honest policyholder starts behaving a little differently once the protection is in place. An underwriting model has to price for the first. A claims-monitoring process has to watch for the second. They call for entirely different defenses, and treating them as one problem usually means solving neither.

When protection changes behavior — not just in finance

This isn't a finance-only phenomenon. It shows up anywhere a decision-maker is shielded from a decision's downside.

In India, the clearest version is the farm loan waiver. In 2008, the central government wrote off roughly ₹60,000 crore of agricultural debt under the Agricultural Debt Waiver and Debt Relief Scheme — a genuine relief measure for farmers in real distress. But economist Martin Kanz, tracking beneficiaries afterward in a study published in the American Economic Journal, found something the scheme's designers weren't hoping for: recipients didn't just get relief on the past loan. They invested less in their own farms in the years that followed, leaned more heavily on expensive informal moneylenders for new credit, and showed no measurable productivity gain — the relief arrived, but the borrowing discipline it was meant to preserve didn't come back with it. State governments have repeated the maneuver many times since — Uttar Pradesh, Punjab, Maharashtra, and others have all run their own waiver schemes — and each round makes the next one a little more expected. A farmer weighing whether to repay a loan he can technically afford to repay is now, rationally, also weighing the odds of a future waiver making that repayment unnecessary.

The international mirror is the United States' National Flood Insurance Program. Because private insurers largely won't underwrite flood risk at an affordable price, the federal government stepped in decades ago to guarantee it — a sensible response to a real market gap. But the program's own data shows a small share of properties, officially labeled "repetitive loss properties," account for a wildly disproportionate share of total claims: homes that flood, get rebuilt with insurance money on the exact same low-lying land, and flood again. The insurance didn't cause the flood. It removed the one signal — the cost of rebuilding somewhere that keeps flooding — that would otherwise have pushed the decision the other way.

Same mechanism, two countries, two completely different domains: a subsidy meant to absorb one bad outcome ends up nudging the odds of that outcome repeating.

The too-big-to-fail version

Move the same logic into banking, and the stakes get systemic instead of individual.

In September 2019, the Reserve Bank of India froze most withdrawals from Punjab and Maharashtra Co-operative Bank after discovering it had concealed a massive concentration of loans to a single struggling real-estate group, Housing Development and Infrastructure Limited (HDIL), for years. Depositors — ordinary account holders, not risk-takers by any definition — found their own savings capped, in some cases for years, because India's deposit insurance at the time covered only the first ₹1 lakh of any account. The public anger that followed pushed the government to raise the Deposit Insurance and Credit Guarantee Corporation's cover to ₹5 lakh per depositor in 2020, a necessary fix for ordinary savers who had done nothing wrong.

But deposit insurance is a textbook moral hazard trade even when it's working exactly as designed. A depositor covered up to ₹5 lakh has far less reason to ask hard questions about a bank's loan book before parking money there — the entire point of the guarantee is to remove that fear, which is precisely what also removes the market discipline that a depositor's scrutiny used to provide. Multiply that by every insured depositor in the country, and a bank faces measurably less pressure from its own customers to behave conservatively, because its customers know they're protected either way, right up to the insured limit.

The American mirror is more dramatic. In the run-up to 2008, AIG's Financial Products division wrote enormous volumes of credit default swap protection on mortgage securities — effectively insurance against those securities defaulting — while holding almost no capital against the possibility that they actually would. That wasn't recklessness in a vacuum. AIG was a AAA-rated insurance conglomerate embedded so deeply in the global financial system that its own executives could reasonably assume the US government would never let it collapse outright — upside kept in-house if the bets paid off, downside quietly assumed to be someone else's problem if they didn't. "Heads I win, tails the taxpayer loses" is the phrase people reached for afterward, and it fits: when the mortgage market did default, the assumption turned out to be correct. The government stepped in with a rescue exceeding $180 billion rather than let AIG's failure cascade through every bank that had bought its protection. The bet on being too systemically important to fail was, in hindsight, not really a bet at all.

Why systems people should care

If you're building or maintaining an underwriting engine, a claims-processing workflow, or a credit risk model, moral hazard isn't an abstract economics term — it's a requirement you have to design against, or your system will get gamed by people who aren't doing anything an auditor would call fraud.

Deductibles and co-pays exist specifically to keep this in check: forcing the insured party to keep some skin in the game, even a small amount, measurably changes claims behavior. Waiting periods before a new policy pays out reduce the incentive to buy insurance only after a loss is already likely. Claims-history tracking and no-claim bonuses are a system's way of remembering behavior over time instead of pricing every renewal as if the past never happened. In lending, covenant monitoring and periodic re-underwriting exist because a borrower's risk profile the day after disbursal isn't guaranteed to match the risk profile a bank priced in on day one.

Miss this in a system's design, and the failure mode is specific and predictable: not a spike in obvious fraud, but a slow, statistically invisible drift toward worse average outcomes among people who were never dishonest — they just responded, rationally, to a system that quietly told them the cost of a bad decision belonged to someone else.

The hidden tradeoff

None of this is an argument against insurance, deposit guarantees, or bailouts. Pooling risk across many people is exactly what lets an individual farmer, driver, or depositor survive a bad year that would otherwise wipe them out — and a banking system with zero deposit insurance is a banking system one rumor away from a panic. The protection is doing real work.

The honest tradeoff is that every mechanism built to absorb a bad outcome also, unavoidably, blunts the signal that would have prevented some of those outcomes from happening in the first place. Make the protection too thin, and people are one piece of bad luck away from ruin. Make it too complete, and you've quietly told everyone covered that caution is now optional.

This is the same tension the financial dominoes essay ends on, seen from the other side. Contagion is what happens when protection is too weak to stop one failure from becoming everyone's failure. Moral hazard is what happens when protection is strong enough that fewer people bother trying to prevent the failure at all. No system gets to eliminate both. It only gets to choose, deliberately, where the discomfort sits.