384
The Nobel Prizes
Allocate bargaining power to the power plant through the contract but assign
ownership of the coal mine to the mine manager. Ownership of the coal mine
may provide the manager with a good outside option—perhaps selling the coal
elsewhere—and since this outside option will increase if the manager becomes
more efficient, some of the gains from increased efficiency will now accrue to
the manager.
Unfortunately, this is not the end of the story. Suppose that the manager’s
outside option if she owns the mine = ̄p – λ
c, where c is the cost of supplying coal
to the power plant and 0 < λ < 1 reflects the fact an increase in efficiency from
supplying the power plant translates partially but not completely into gains from
supplying coal to others. Obviously, asset ownership would be irrelevant if the
coal mine manager could be offered an incentive scheme directly of the form ̄p
+ (1 – λ)c. (Since the manager incurs the cost of supplying coal this makes her
net payment ̄p – λc.) This would indeed be possible if c were verifiable, but at
first sight seems impossible if c is merely observable.
But there is a way to do it, as Eric Maskin and Jean Tirole (1999) have pointed
out, drawing on work of Maskin (1999) and Moore and Repullo (1988). The
following game, to be played ex post, is written into the contract. The coal mine
manager announces her cost of supplying (low-ash-content) coal to the power
plant, c
∗
, say. The power plant can accept this and pay the manager ̄p + (1 – λ)
c
∗
or he can challenge, claiming the cost is c
∗∗
(presumably lower than c
∗
, so that
the manager is paid less). If the power plant challenges, the coal mine manager
pays a large fine F to a third party. At this point the challenge is tested: the mine
manager is asked whether she wants to supply coal at a price = 1/2c
∗
+ 1/2c
∗∗
. If
she does supply, this establishes that the coal mine manager lied since she would
be losing money by supplying if her true cost were c
∗
. In this case the power plant
receives F from the third party. If the mine manager does not supply, that is, the
challenge is proved incorrect, the power plant pays F to the third party.
The unique subgame perfect equilibrium of this mechanism is for the coal
mine manager to tell the truth about her cost and for the manager to receive a
net payment of ̄p – λc. The need for asset ownership has been avoided.
Are there ways of ruling out Maskin-Tirole mechanisms? One objection is
that collusion could take place between the buyer or seller and the third party.
20
However, it is not clear how such a collusive agreement would be enforced since
the parties could make clear in the contract that it is illicit. Also Maskin and
Tirole have shown that if at least one of the parties is risk-averse, a cleverly-
designed lottery can substitute for the third party. Nonetheless, if one is prepared
to assume that (a) the buyer and the seller are both risk neutral, (b) third par-
ties cannot be used, (c) the parties can always renegotiate the contract after any
Incomplete Contracts and Control
385
procedures for revising
it have been completed, e.g., because there is no clear
deadline; then the incomplete contracting story, and the role for asset ownership,
can be resurrected. This is shown in Segal (1999) and Hart and Moore (1999).
But these are strong assumptions and I, for one, do not feel very comfortable
with them. If the model accurately captures reality, one would expect to see some
attempts to use Maskin-Tirole mechanisms, as well as to allocate bargaining
power contractually. I know of no cases of the first and not many of the second.
Also the model as it stands cannot explain ex post inefficiency (except if parties
are wealth-constrained). This seems a significant limitation since the earlier work
of Coase and Williamson argues convincingly that reducing ex post inefficiency
is at least one of the rationales for the existence of firms.
Of course, one could try to incorporate ex post inefficiency by supposing
that the parties are asymmetrically informed.
21
However, as long as there is sym-
metric information at the contracting date, a further set of mechanisms (also not
seen in practice) can be used to overcome this.
22
For these reasons in recent work
I have turned to a different approach.
6. DROPPING RATIONALITY
If parties are fully rational I do not see why they would not include in their con-
tracts mechanisms of the type suggested by Maskin and Tirole. As I have said,
as far as I know, there are no examples of this in practice. One can, of course,
always put the blame on judges: they would not understand and/or would not
enforce such mechanisms.
23
But this just pushes the question one step further:
there are many smart judges and if mechanism design
is the solution to incom-
plete contracting problems one would expect judges eventually to understand,
embrace, and enforce contracts based on the mechanism design approach. It has
been eighteen years since the publication of Maskin and Tirole’s paper and I do
not see any move in this direction.
My view is that the reason we do not see these mechanisms is that parties are
not fully rational. This is in many ways an unfortunate conclusion since while
there is one way to model rationality there are many ways, perhaps infinitely
many, to model irrationality. The lack of discipline makes many economists
uncomfortable, and they are willing to hold on to the rationality approach at all
cost. However, I think that there is no alternative but to abandon it.
I began to do this in a paper with John Moore (Hart and Moore (2008)).
Looking back, one can say that our motivation was three-fold. In no particu-
lar order of importance: First, we wanted to develop a theory immune to the
Maskin-Tirole critique. Second, we wanted to explain why parties do not allocate