Oliver Hart Prize Lecture: Incomplete Contracts and Control

386 

The Nobel Prizes

bargaining power contractually to solve the hold-up problem. Third, we wanted 

to develop a model that allows for ex post inefficiency. It is this multiplicity of 

motives that perhaps explains why we did not try to introduce cognitive limi-

tations for the parties but instead focused on ideas of fairness and reasonable 

behavior. In this respect we were much influenced by the large behavioral litera-

ture on the latter topic.

24

To understand our approach, consider a simple situation of a buyer B and a 

seller S, who meet at date 0. At that time there is a competitive market for buyers 

and sellers, but after date 0 B and S will pair off and will be isolated from the 

market. At date 1 there are gains from trade. S can supply one widget at cost c

and B obtains value v > c from it. All returns are measured in money (but these 

returns are not verifiable).

For simplicity, suppose that the seller’s reservation utility determined in the 

date 0 market for buyers and sellers is zero. One contract that B could offer to S 

that will give B all the gains from trade is the following: The contract states that at 

date 1, B will make an offer to S that S can accept or reject; S cannot make any offers 

to B. As we have seen, under standard rationality assumptions, B will offer S just 

above c at date 1, S will supply the widget, and B will receive the full surplus v – c.

Now the evidence from the famous ultimatum game experiments suggests 

that things may not work this way in practice. These experiments find that in 

situations like this B will end up offering considerably more than c, and that 

moreover if B does not do so S will reject the offer.

25

 Note, however, that the 

parallel between the ultimatum game and our case is not exact since there is no 

prior contract in the ultimatum game.

Moore and I could have constructed a model based directly on the idea that 

S will turn down ungenerous offers. We did not do so for two reasons. First, we 

were concerned about the possible criticism that the ultimatum game evidence 

concerns relatively small payoffs.

26

 Second, we wanted our model to apply to 

more general situations than just ones where a seller can choose not to trade. 

For example, what is the analogy of ultimatum game behavior when parties are 

asked to play a Maskin-Tirole mechanism?

We therefore proceeded as follows. We assumed that even ex post perfect 

contracts cannot be written and so it is possible for both the buyer and seller 

to provide less than ideal performance while staying within the terms of the 

contract: we refer to less than ideal performance as “shading.” In the buyer-seller 

example, the seller might shade by supplying a widget of deficient quality, while 

the buyer might shade by not providing information that would make the seller’s 

task easier. A critical assumption is that a party will shade if and only if he does 

not feel well treated. So in the case of the seller who receives a low ball, but 

---

Incomplete Contracts and Control 

387

nonetheless profitable, offer, the seller will accept the offer but then punish the 

buyer by shading.

We made the further crucial assumption that the initial contract circum-

scribes what parties feel is fair. The date 0 competitive market for buyers and 

sellers is an important element here. The idea is that the broad terms of the 

contract are regarded by both parties as reasonable since they are negotiated 

at arms-length and neither party blames the other for the equilibrium terms 

of trade. As a result, neither B nor S feels entitled to an outcome outside the 

contract. In contrast any discretionary decision made by one of the parties at 

date 1—when the competitive market is no longer there to provide an objective 

benchmark—may be found unreasonable by the other party and may lead to 

shading. To make things as stark as possible, Hart and Moore (2008) suppose that 

each party is subject to an extreme self-serving bias and feels that a reasonable 

outcome is one that maximizes that party’s payoff over all outcomes permitted 

by the contract that are individually rational for the other party.

Return to the widget example and the contract that specifies that B will make 

a take-it-or-leave-it offer to S. Let B offer a price just above c. S will consider 

this unreasonable given that B could have been more generous. Indeed, the best 

outcome for S under the contract would be for B to offer v (anything more than 

v would involve B’s making a loss and so would not be individually rational). 

How much does S shade given the actual offer c? Hart and Moore (2008) assume 

that shading is a fraction of how much S is shortchanged or aggrieved, where 

the latter is the difference between the payoff S feels entitled to—here v – c—and 

what she gets—zero. Specifically, S reduces B’s payoff by θ(v – c), where 0 < θ < 

1. Shading does not affect the payoff of the party doing the shading.

In sum, under the contract that gives B the right to make a take-it-or-leave-it 

offer to S, there will be a deadweight loss of θ(v – c). Note that there is no way 

of negotiating around this. Coasian bargaining fails because shading is noncon-

tractible. B could, of course, offer more than c to reduce S’s aggrievement, but it 

is not in his interest to do this: offering a dollar more increases B’s cost by a dollar 

but reduces shading by only θ.

There is, however, a solution to this problem in this simple example. B and S 

could fix the price in advance: they could write a contract at date 0 that specifies 

the date 1 price of the widget to be c. In this case neither party has any discretion 

at date 1. B and S both regard the price c as fair since it is negotiated at arms-

length in a competitive market at date 0. There will be no shading or deadweight 

losses at date 1 and the full surplus v – c will be earned. The first-best is achieved.

This simple framework suggests an alternative rationale for the existence of 

a contract. A contract negotiated before parties are locked into each other gets 

---