From insurance practice, formalized by Kenneth Arrow (Uncertainty and the Welfare Economics of Medical Care, 1963), Mark V. Pauly (The Economics of Moral Hazard, 1968), and Bengt Holmström (Moral Hazard and Observability, 1979).
Identifies where one party is shielded from the cost of its own risky choices, predicts the extra risk it will take, and designs terms that restore the cost.
The more of a fleet's crash losses the insurer covers, the more crashes the fleet has
A delivery company with 50 vans takes care in proportion to the share of each loss it still bears. Across the plot, the insurer covers more of every loss; the line shows crashes per 100 vans a year, against the rate the premium was priced on; an average crash costs $40,000. Add a deductible the company pays on every claim, or telematics with a safe-driving bonus, and watch the line come back down at the 90 percent cover the policy sells. Illustrative.
Crashes per 100 vans, with these termsRate the premium was priced on
An example
A regional insurer wrote a fleet policy for a parcel delivery company with 50 vans, covering 90 percent of every crash loss with no deductible. The premium assumed six crashes per 100 vans a year, the company's record when it carried most of its own losses. In the first policy year it ran at thirteen per 100.
The underwriter ran a moral hazard review before renewal. Who was protected: the delivery company. Which costs it no longer bore: repairs, downtime, and most of the hit to its own budget from a crash. What had become more attractive: tighter delivery windows, newer drivers on longer routes, and skipped van maintenance. Could the insurer see any of it: no.
The renewal changed the terms, not the price alone. The company now paid the first 30 percent of each loss, and every van got telematics, with a quarterly bonus for drivers whose braking and speed scores stayed in range. The insurer could now see driving, and the company again felt each crash. The next year the fleet ran at seven per 100.
Why it works
People respond to the costs they bear. When a contract, a guarantee, or a bonus plan moves the cost of a risky choice onto someone else, the protected party makes more of those choices, usually without bad intent. Nothing about them changed except the price of carelessness, which fell.
The analysis works because it asks two questions in order. First, what behavior does the protection make cheaper? That predicts the extra risk before it shows up in the losses. Second, can the protector observe that behavior? If it can, the fix is monitoring and terms tied to what is seen. If it cannot, the fix is to hand some of the loss back, through a deductible, co-payment, or clawback, so the protected party's own interest does the monitoring.
The thinking behind it
Kenneth Arrow Kenneth Arrow, one of the founders of modern economic theory, wrote a 1963 paper in the American Economic Review on uncertainty in medical care, and its welfare economics, that is the usual starting point of the modern economics of moral hazard. He observed that insurance changes the behavior of the insured, which limits what can be insured at all. The term itself came from underwriters, who used it long before economists formalized it.
Mark V. Pauly Mark V. Pauly, a Wharton health economist, answered Arrow in The Economics of Moral Hazard: Comment (1968). He showed that insurance raises the use of medical care because it lowers the price the patient pays at the point of use. That is a rational response to a lower price, so moral hazard needs no failure of character to explain it, and it can be predicted the way any response to price can.
Bengt Holmström Bengt Holmström's Moral Hazard and Observability (1979) showed that when an agent's effort cannot be observed, a contract should reward any measure that carries additional information about that effort. It is the logic behind telematics on a fleet policy and behind tying bonuses to indicators beyond the final result. He shared the 2016 Nobel Memorial Prize in Economic Sciences with Oliver Hart.
How to run it
- Name who is shielded and who pays. Write the protected party, the party absorbing the loss, and the contract, guarantee, or plan that links them.
- List the costs that moved. Name each cost the protected party no longer carries: repairs, downtime, losses, reputation, or effort.
- Predict the behaviors that got cheaper. Ask what a sensible person would now do more of, and estimate how much more.
- Check what the protector can see. Separate the behaviors that can be observed or measured from those that only show up in the losses.
- Choose the fix to match. Monitor and reward what is visible; for what is not, hand back a share of the loss through deductibles, co-payments, or caps.
- Watch for side effects. A deductible can push people to hide small losses, and a metric can be gamed, so review claims and scores together.
Where it goes wrong
The common failure is reading moral hazard as bad character. The protected party gets blamed, a warning letter goes out, and nothing changes, because the incentive is untouched. It shows up as the same losses recurring after the scolding. The recovery is to change the terms, not the tone.
A second failure is overcorrecting. A deductible so large that it defeats the purpose of the cover drives good customers away or leads them to hide incidents. Size the retained share so it is felt without being ruinous. A third is monitoring the wrong thing: a measure that is easy to game gets gamed. Pick indicators close to the behavior and check them against the actual losses.
When to use something else
- Agency Theory. When the problem is the wider relationship between an owner and the person acting for them, beyond one protected risk.
- Skin in the Game. When the question is whose advice or forecast to trust, judged by what each source loses if wrong.
- Goodhart-Aware Metric Selection. When the fix will tie pay or terms to a measure, and the risk is the measure being gamed.
Where it sits in SOLVE eX
Go to the source
- Uncertainty and the Welfare Economics of Medical Care, American Economic Review, 1963
- The Economics of Moral Hazard: Comment, American Economic Review, 1968
- Moral Hazard and Observability, Bell Journal of Economics, 1979