All the asymmetry we have met so far has been an asymmetry of knowledge: one side knows something about the world — the quality of a car, the state of its own health — that the other cannot see. But there is a second asymmetry, of a different and equally corrosive kind, hidden not in what a party knows beforehand but in what a party does afterward, once the contract is signed and its conduct can no longer be watched. This is moral hazard, and it haunts every arrangement that shields a person from the consequences of his own actions.
The classic case, again, is insurance, and the point was sharpened by Kenneth Arrow. A man who insures his house against fire has, by that act, a little less reason to be careful with candles; a man whose car is fully covered has less reason to lock it; a firm whose losses are underwritten has less reason to avoid the risks that produce them. The insurer cannot observe how careful the insured is being — care is a hidden action — and so cannot price or forbid the carelessness that the policy itself encourages. The very transfer of risk that the insured sought, and that makes him better off, dulls his incentive to prevent the loss, and so raises the chance of the loss occurring. Insurance against a hazard manufactures, in part, the hazard it insures.1
It is worth marking precisely how this differs from the adverse selection of the earlier plates, for the two are the field's twin pillars and are easily confused. Adverse selection is a hidden type, known before the deal: the applicant is already sick, the car is already a lemon, and the hidden fact is a pre-existing quality that selection sorts. Moral hazard is a hidden action, taken after the deal: the insured was no more careless than anyone until the policy changed his incentives, and the hidden fact is a behavior the contract itself induces. The one corrupts the pool of who enters the market; the other corrupts the conduct of those already in it. Both flow from the same root — that one party cannot see what the other knows or does — but they call for different cures, and confusing them has wrecked many a well-meant policy.2
Moral hazard reaches far beyond insurance, into the general problem of one person acting on another's behalf — the principal-agent problem, the deep structure of employment, management, and delegation throughout an economy. The employer cannot fully watch the worker; the shareholders cannot watch the executive; the patient cannot judge the doctor; the citizen cannot monitor the official. In each, the agent has his own interests — to exert less effort, to take the easier course, to favor himself — and the principal cannot see enough to tell diligence from sloth, or good judgment from mere luck. Left alone, the agent will serve himself at the principal's expense, not from villainy but from the ordinary pull of incentive upon unwatched action. The agent's hidden action is the moral hazard at the heart of every organization, and managing it is much of what management is.3
The cure for moral hazard is to restore, by the design of the contract, the incentive that observation cannot supply — to make the agent bear enough of the consequences of his own action that he chooses to act well. The insurer imposes a deductible and a co-payment, so the insured keeps some stake in avoiding the loss; the employer ties pay to output, so the worker shares in the fruit of his effort; the firm grants the executive shares, so his fortune rises and falls with the shareholders'. But here lies a cruel trade-off, the deepest in the economics of incentives. To give the agent a stake in the outcome is to expose him to its risk — and the outcome depends not only on his hidden effort but on luck he cannot control. The more sharply we tie his reward to the result, the keener his incentive but the more risk we load upon him; and if he is risk-averse, as the volume on the premium taught, that risk is costly. The principal must therefore trade incentive against insurance: a contract that perfectly motivated the agent would ruinously expose him, and one that fully insured him would leave him idle. The best attainable contract is a compromise, and it is, of necessity, second-best — worse than what could be had if only the action could be seen.4
How much risk to load, and on what to condition the reward, is the subject of a refined theory whose guiding result is intuitive once stated: the agent's pay should depend on any signal that carries information about his hidden action, and only on such signals. Reward him for what reveals his effort, and filter out what reflects mere noise — judge the salesman against his region's sales, not the whole economy's, so that he is neither blamed for the recession nor credited with the boom. Yet the principle, however elegant, does not abolish the trade-off; it only manages it. Where effort is many-sided, rewarding the measurable part distorts effort toward it and away from the rest — the teacher drilling the test, the surgeon shunning the hard cases that would mar his record — so that the cure for one hidden action breeds a new distortion of its own. Moral hazard, like adverse selection, is not solved but only contained, and the containment leaves its own marks.5
The hazard is the asymmetry of action, as the lemon was the asymmetry of knowledge: the conduct a contract cannot see and therefore cannot govern, the carelessness that insurance invites, the shirking that delegation permits, the self-dealing that unwatched agency allows. Its cure is to write incentive back into the bargain, but only at the price of loading risk onto those who would rather be insured, so that every real contract settles for a second-best that the full sight of action would improve upon. With the hidden type and the hidden action both before us, and the costly devices each demands, the volume can at last turn upon the thing all this machinery exists to repair. The flagship returns to the price — the signal we trust to carry, all by itself, the knowledge a market needs — and asks what becomes of it when knowledge is precisely what one side withholds.