390
The Nobel Prizes
1. A randomization device determines c and both parties see the outcome. Under
the flexible contract B then chooses p from the range [10,16]; p is required to be
at least c. Trade then occurs. At this point S can choose whether to shade. In the
experiment shading is a discrete action that imposes a small cost on the seller
and a large cost on the buyer.
In contrast, under the rigid contract, trade takes place only if c = 10. Again
after trade the seller can choose whether to shade.
The probability π of the high cost state is chosen to be relatively small. The
experiment is repeated several times with buyers and sellers being re-matched
randomly each time.
If buyers and sellers are fully rational, we would predict a very simple out-
come. Start at the end. Since shading is costly a rational seller will never engage
in it. (Notice that this is why we modified the assumption from the theoretical
model that shading is costless.) Hence a buyer can safely ignore shading and
will choose the flexible contract since this guarantees trade in both states. B will
choose p = 10 when c = 10 and p = 16 when c = 16. The first-best is achieved.
This is not what happens in the experiment. Buyers choose rigid contracts
a significant fraction of the time, and these contracts are more profitable than
flexible contracts. With the flexible contract buyers offer more than 10, and sig-
nificant shading occurs, in the low cost state. Shading is rare in the rigid contract.
These results are broadly consistent with the Hart-Moore model. It is par-
ticularly striking that there is little shading in the rigid contract. This is in spite
of the fact that this contract allocates all of the surplus to the buyer. It appears
that the seller accepts and is not angry about her low (zero) payoff given that this
was determined competitively.
29
Fehr et al. (2015) extend the experiment to permit communication and rene-
gotiation. Allowing the buyer to send a message to the seller at the contract
formation stage, explaining how he plans to choose price in each state, improves
the efficiency of the flexible contract but the trade-off between rigid and flexible
contracts remains.
30
Allowing for renegotiation improves the efficiency of the
rigid contract relative to the flexible contract. Under the rigid contract trade
occurs if c = 10. On the other hand, if c = 16, the buyer can offer a new (renegoti-
ated) contract. Some shading occurs in this case, since the seller is aggrieved that
she does not receive 20. Still trade now takes place in both states. Interestingly,
we find that the seller does not expect renegotiation to occur when c = 10, and is
not aggrieved and does not shade when this fails to happen.
The Hart-Moore model can be used to re-examine some of the issues dis-
cussed in property rights theory. First, given that there may be some ex post
inefficiency, asset ownership will matter, but for different reasons from before.
Incomplete Contracts and Control
391
Since an inefficient allocation may occur (as in the rigid contract without rene-
gotiation), outside options, determined by asset ownership, may be exercised
even if they are first-best inefficient. Hence ex ante asset allocation will affect
ex post surplus. Also asset ownership will determine the size of the gains from
trade, and therefore entitlements and shading. Hence, even if the final outcome
is independent of asset ownership, the deadweight losses from shading will gen-
erally depend on the asset allocation. Hart (2009) uses these ideas to develop a
theory of asset ownership based on payoff uncertainty rather than relationship-
specific investments.
Second, the model can throw light on the employment relationship. Suppose
that, instead of uncertainty about the seller’s cost, there is uncertainty about what
type of good a buyer wants from a seller. Refer to the type of good as the seller’s
task. In Hart and Moore (2008), it is shown that under some assumptions the
optimal contract will take the following form: the price paid to the seller is fixed
and one of the parties is given the right to choose the task. If the buyer is allo-
cated the right, this can be interpreted as an employment contract. If the seller is
allocated the right, it can be interpreted as independent contracting. The reason
for fixing the price is that there will be less shading if ex post disagreement over
entitlements is limited to the choice of task rather than to the choice of the price
and the task.
This version of the model relates to earlier ideas about the employment rela-
tionship (see Coase (1937), Simon (1951), and Alchian and Demsetz (1972)). But
in the earlier work of Coase and Simon it was assumed that the price or wage is
fixed. Here there is an explanation.
This model of employment is used in Hart and Holmström (2010) to develop
a theory of firm scope. Consider two firms that want to coordinate on a techno-
logical platform. They can do this as separate firms and write a contract. Or they
can merge. In the first case since it is hard to specify all the details of coordination
there will be ex post inefficiency, either because renegotiation fails or because of
shading. In the second case, a boss of the merged firm can require coordination
(the boss chooses the task), but in doing so will not fully internalize the work-
ers’ costs of shifting to the new technology. Hart and Holmström (2010) show
that the relative importance of these effects will determine which organizational
form is better.
Finally, in some recent work with Maija Halonen-Akatwijuka, I have explored
dynamic issues that arise when contracts are reference points. If parties write a
sequence of contracts, then the first contract will become a reference point for
the second contract, and so on. This interdependence can make it easy for the
parties to write a new contract if not much changes over time, since the parties