The Tenant

The gaps a contract leaves to chance are one problem; the conduct it must actively shape is another. For many contracts are not merely silent about the future but charged with a tension in the present: one party is hired to act for another, his effort matters greatly to the outcome, and his effort cannot be watched. How to write a contract that makes such a hireling exert himself — when shirking is invisible and the temptation to shirk is real — is the problem of incentive design, and its emblem is the oldest contract on the land: the bargain between the owner of a field and the man who works it.

The structure is the one the volume on the asymmetry named moral hazard, here seen from the side of the contract that must answer it. A principal — a landowner, an employer, a body of shareholders — engages an agent — a tenant, a worker, a manager — to act on his behalf; the agent's effort bears on the result the principal cares about; but the principal cannot observe the effort, only the result, which depends on the effort and on chance together. A poor harvest may mean the tenant loafed, or merely that the rains failed; the principal, seeing only the harvest, cannot tell which. And the agent, knowing his effort is hidden, is tempted to give less than he promised, pocketing the ease and blaming the rains. The principal's problem is to design a contract — a rule for paying the agent — that makes the agent want to exert himself, even though his exertion cannot be seen, by tying his pay to something that can. This is the principal-agent problem, and the contract that solves it is an incentive contract: one that aligns the hidden interest of the agent with the visible interest of the principal by the way it pays.¹

The field makes the choices vivid, for there are three classic ways to pay the man who works it, and each is a different contract. The first is the wage: pay the worker a fixed sum, and keep the whole harvest. This loads all the risk on the owner — he bears the bad years and enjoys the good — but it gives the worker no incentive at all, for his pay is the same whether he toils or loafs, and so, unwatched, he loafs. The second is the fixed rent: let the tenant pay the owner a fixed sum and keep the whole harvest himself. This gives the tenant the sharpest possible incentive — every extra bushel is his — but loads all the risk on him, for he bears the bad years alone, and a poor tenant cannot bear a ruinous one. The third, the sharecrop, splits the difference: owner and worker divide the harvest in agreed shares. This gives the worker some incentive — a share of every extra bushel — and leaves him some insurance — the owner shares the bad years too. Between the wage that insures but does not motivate, and the rent that motivates but does not insure, the sharecrop balances the two; and which balance is best depends on how much the worker fears risk and how much his hidden effort matters.²

The three contracts reveal the tension at the heart of all incentive design: the trade-off between incentive and insurance, between the spur and the shield. To motivate the agent, the principal must tie the agent's pay to the result — must make the agent bear the consequences of his own effort, rewarding the good outcome and penalizing the bad. But the result depends not only on effort but on chance; and so, in tying pay to result, the principal also loads the agent with risk, making him bear the bad luck along with the bad effort. The sharper the incentive, the heavier the risk: a contract that rewarded the agent fully for the outcome would motivate him perfectly but expose him utterly to chance, while a contract that shielded him from all risk would insure him perfectly but motivate him not at all. Every incentive contract must therefore strike a balance, trading the spur of reward against the burden of risk — giving the agent enough stake in the outcome to make him exert himself, but not so much as to crush him with risk he cannot bear or would pay dearly to avoid. The optimal contract is the one that best resolves this tension, and where it falls depends on how averse the agent is to risk and how responsive the outcome is to his effort.³

Holmström drew from this tension a principle of surprising reach: pay the agent on every signal that tells you anything about his effort, and on nothing that does not. For the trouble with tying pay to the harvest alone is that the harvest is a noisy signal of effort, muddied by the weather; if the principal could find a cleaner signal — or a second signal that helped filter out the noise — he could reward effort more precisely, and so impose less needless risk. So the principle says: condition the agent's pay on any information bearing on his effort, however indirect. If the neighboring farms all had poor harvests too, that tells the owner the rains, not the tenant, were at fault, and he should not penalize the tenant for the shortfall — so the contract should pay the tenant relative to his neighbors, filtering out the common weather. If a manager's firm did badly while rival firms thrived, that suggests the manager, not the market, was at fault — so his pay should be judged against his rivals, not against a market they all share. The informativeness principle thus tells the designer not merely how hard to spur the agent but what to spur him on: every scrap of information about his hidden effort should enter his pay, and nothing else should, so that he is rewarded for what he did and shielded from what he could not help.⁴

Yet the incentive contract, for all its ingenuity, cannot fully solve the problem it addresses, and its limits point to what lies beyond the written term. Real agents do many things at once, not all measurable, and a contract that sharply rewards the measurable will draw effort away from the unmeasurable — the teacher paid for test scores neglects what tests do not capture, the manager paid for this quarter's profit starves the long term. Tie pay too tightly to the few things you can measure, and you distort the many you cannot. And some of what we want from an agent — judgment, loyalty, care, the willingness to do the right thing when no clause requires it — cannot be bought by any incentive at all, but only trusted to. So the incentive contract, like the default rule before it, is a partial fix: it shapes the conduct it can measure, and must leave the rest — the unmeasurable effort, the unforeseeable case, the good faith no bonus can buy — to something other than the contract. That something is the trust and relationship the next plate takes up, and the foundation the flagship reaches: the unwritten understanding that every written contract, however cleverly its incentives are designed, must finally rest upon.⁵

The tenant in his field is the emblem of the contract that must shape conduct it cannot watch: the bargain that makes a self-interested agent exert himself by tying his pay to what can be seen, when his effort cannot. Its forms — the wage that insures, the rent that motivates, the sharecrop that balances — reveal the tension at the heart of all incentive design, the trade-off of the spur against the shield, the incentive bought always at the cost of risk. Holmström's principle refines it: reward the agent on every signal of his effort and nothing else, judging him against his fellows to filter out the chance they share. But the incentive contract distorts what it cannot measure and cannot buy what must be trusted, and so it too leaves gaps — the unmeasurable, the unforeseeable, the good faith no clause commands. What fills those gaps, in the contracts that work without the law's defaults or the cleverest incentives, is the relationship itself: the trust between parties who deal again tomorrow. The next plate turns to the contract that needs no court — the promise kept not for fear of the sword but for the value of the bond.