. 2
( 18)


incentive that maximizes her profit is the sure contract B > 0 and = 0, such that B =
Under asymmetric information, the principal cannot observe the agent's effort. Her
program is then to maximize her expected profit subject to the incentive constraint (IC,
given by (1)) and to the participation constraint (IR) so that the agent receives a non-
negative utility

Substituted into the objective function, this gives

The first-order condition with respect to effort is
Using (1), we find the following optimal share

This result sheds some light on the trade-off between the incentives and insurance
dilemma in a moral-hazard situation. If = 0, there is no insurance. The optimal
incentive scheme (w = B + y) depends only on : the more productive the agent
(increasing ), the greater the payment. If > 0, < so that there is a risk sharing. And
the greater the risk (increasing ), the more the agent risk shares, the smaller .

1.2 Adverse selection

Now, we will consider two agents of different types, which differ only with respect to the
disutility of effort function, which is

for type 1, and

with 2 > 1. Hence the disutility of any particular effort is greater for an agent of type 2.
We shall refer to the first as a "good" type and the second as a "bad" type, since for the
same effort, the principal will have to pay more to the second type than to the first. The
principal will propose to the agents a compensation wi = wi (ei), i = 1, 2, relative to the
effort level observed ei in order to maximize her profit = e1 + e2 (w1 + w2). The
choice of optimal contract (w1, w2) by the principal depends on the information that she
holds on types before the contract design.
If there is no adverse selection problem, the principal can perfectly discriminate between
the two types. The program is then to maximize her profit subject to the participation
constraint (IRi) that each agent receives a non-negative utility

Substituting into the objective function and differentiating, we obtain

The optimum contract is then . Because 1 < 2, w*1 >w*. Agent 1
with the lower disutility of effort ("good" agent) is offered the higher payment and invests
more effort than agent 2 ("bad" agent).
In the case of the adverse selection problem, the principal does not know which agent
belongs to which type. As a result, if the principal offers the two contracts {(e*1, w*1), (e*2,
w*2)} to any agent allowing him to freely select the contract that he most likes, agent 2
will choose the contract that is designed for him, but agent 1 prefers (e*2, w*2)to(e*1, w*1)
in order to receive a surplus . This result can be
avoided if the principal restructures her payment so that the agent's i utility from
choosing (e*i, w*i) is higher than his utility from choosing (e* i, w* i). These are self-
selection constraints or incentive compatibility conditions (ICi)
In order to calculate the best contracts that the principal can offer in this situation, let us
assume that the principal considers the probability of an agent being type i is qi. The
principal's program is then

Only one equation from of each pair has to be used in the optimization procedure. The
other inequality is automatically fulfilled.[2] The optimization problem of the principal

The first-order conditions give

We verify easily that . The optimal wage offers are

with . We can point out that if the "bad" type (agent 2) receives a
smaller wage than under symmetric information , the good type (agent
1) receives a higher wage . The surplus ( ) that he obtains is just big
enough to make it of no interest to him to pretend to be the bad agent (agent 2).

[1] 2
If w N(w, w), the expected utility of the agent is

Because expected utility is increasing in
we can take a monotonic transformation. Then we obtain the utility function given, which
is equivalent to using the mean-variance criterion for choice under uncertainty rather
than the expected utility criterion.

From (IR2) and < 2, we obtain

we conclude that when (IC1) holds, (IR1) is also verified. Moreover (IC1)is a binding
constraint because the principal tries to keep his offer wi as small as possible. Then
substituting (IC1) in (IC2) we get . As < 2, this inequality is
2 2
always strict when e 1 > e 2.
2 Incomplete contract theory
Let us assume a vertical relationship between a buyer (B) and a seller (S) that runs over
two periods of time. During the first period (ex ante period), the parties are supposed to
be able to sign only an incomplete contract at date 0. At date 1, they invest in specific
assets, respectively and . These levels of investment are non-contractible because
these are unverifiable by a court. During the second period (ex post period), the two
parties set up the efficient quantity of exchange (q) (date 3) after the realization of a state
of nature, which was unknown when they signed the initial contract (date 2). We denote
( , q, ) as the buyer valuation and c( , q, ) as the seller cost of production. is
supposed to be increasing and concave in ( , q) and c decreasing in and convex in ( ,
q). We distinguish two kinds of situation according to the degree of incompleteness of
the initial contract: the null contract (sub-section 2.1) and the simple contract (sub-
section 2.2).

2.1 Null contract and property-rights allocation
A null contract is a contract that does not specify a quantity provision (q). This can be
explained by a difficulty describing the quantity variable and/or difficulty making this
variable verifiable by a court (Grossman and Hart 1986, Hart and Moore 1990, Hart
1995). This has two implications. First, the only way to complete the incomplete contract
is to define a property-rights allocation on a set of assets K ={k1, k2}, because ownership
gives formal control over the asset for uses that have not been pre-assigned. It defines
"residual rights of control" that give bargaining power during the renegotiation. Second,
because there is a null contract ex ante, the parties have to negotiate about the
possibility of trade taking place at date 3. There are two possible outcomes at this date:
either the parties agree to trade or they go their own ways:
If they agree to trade, a bilateral negotiation under perfect information
defines an efficient quantity q *( , , ), after , , and have been
observed. Then a total surplus S( , , q *(·), ) = [ ( , q *(·), ) c( , q
*(·), )] emerges. If the parties can commit themselves ex ante to
agreeing to trade ex post, the maximum social surplus at date 1 from
choosing efficient levels of investment is then given by

We denote by * and * the efficient levels of investment solution of the first-
order conditions

If the parties fails to agree, the buyer receives her outside option wB(
KB) and the seller his outside option wS( KS), where KB(KS) is the set
of assets that the buyer (seller) has control over at date 3.
Assume that S wB + wS. Then it is optimal to agree to trade and divide the total surplus
such that the buyer obtains at least wB( KS) and the seller at least wS ( KS). If the
surplus S wB wS is split following Nash's solution (50 : 50), utilities are

Foreseeing these date 3 payoffs, the buyer and the seller take their investment decisions
at date 1. Let us assume that these decisions are made non-cooperatively and that a
0 0
Nash equilibrium results. Let and be the solutions to the following first-order
The only endogenous variable influencing the parties' choice of investment is the
allocation of assets KB and KS (through outside options). In order to analyze how assets
allocation affects investment decisions, it is necessary to introduce further assumptions:

is increasing as the buyer (the seller) controls more assets; the cross-partial is positive

and the marginal returns of investment are supposed to be higher when the parties

The first implication is that the equilibrium level of investment is at or below the efficient
level ( 0 * and 0 *).[3] Therefore, no propertyrights allocation can replicate the
first-best level of investment. The second implication is the definition of a trade-off
principle: when B controls more assets (integration by the buyer), her outside option wB
increases which raises her incentives to invest (from (3)). But at the same time, S
controls fewer assets which reduces his incentives to invest (from (4)). Analyzing
symmetrically the situation where S controls more assets (integration by the seller) gives
us the following comparison of efficiency under different property-rights allocations
(Table 1A.1).

Table 1A.1: Efficiency under different property-rights allocations
Property-rights allocation Investment

no integration (KB ={k1}; KS ={k2}) 0
0 B
Buyer integration (KB = K; KS = )
B 0
< <*
S 0
Seller integration (KB = ; KS = K)
0 S
< <

But who must integrate? Grossman and Hart (1986) define the following criterion: the
property-right allocation which minimizes incentives distortions is the one which gives all
the rights (integration) to the party whose investment has the prominent effect on social

2.2 Simple contract and first-best solution
A simple contract is a contract which specifies a quantity provision in the contract. When
the court can verify only that trade has occurred (q = 1) or not (q = 0), Hart and Moore
(1988) show that a contract (at will), stipulating a trading price (p1) and a penalty (p0)
when there is non-exchange, leads to surplus-sharing which depends on the state of
nature ( ), whereby incentives to invest are not higher than under a null contract
completed by a property right allocation. Nöldeke and Schmidt (1995) show, however,
that if the parties can define a price contingent for the delivery of the good (option
contract), a first-best solution can be obtained. But this option contract solution to the
hold-up problem requires a higher degree of verifiability: a court is supposed to observe
the party which is at fault in the exchange. Chung (1991) and Aghion, Dewatripont and
Rey (1994) show that this additional verifiability assumption is not necessary if an initial
contract (specific performance contract) can design a renegotiation framework that
avoids this hold-up problem. This simple contract is such that :
It allocates all the bargaining power to the buyer, such that she has the
right to make a take-or-leave-it offer (q, p) in the renegotiation subgame
it defines a default option (q0, p0) that generates a status quo outcome to
the seller in case of renegotiation failing (specific performance).
Given this framework, at the sub-game perfect equilibrium the buyer will always offer to
the seller to deliver the efficient quantity q *( , , ) ataprice p which makes the seller
indifferent between accepting and rejecting the offer

the seller's expected utility is then

Let the initial quantity q0 given by

By maximizing his expected utility (6), the seller chooses a level of investment
investment such that (7) is verified. The assumptions on the function cost ensure that
= *.

The buyer's expected utility is

where is the net surplus from renegotiation that she captures

After simplification, her expected utility can be written

As the buyer captures the social surplus minus a constant US, she has the appropriate
incentives to invest at the first best level ( = *). So the investment game equilibrium is
such that the first-best level ( *, *)is achieved.
Now let us show that the extreme bargaining power allocation to the buyer can be
sustained by a financial hostage provision. Let us assume that the seller rejects any offer
(q, p) made by the buyer in the sub-game and makes a counter-offer such that
. Then it is possible to design in the initial contract a
hostage such that . That is to say, the buyer
prefers to capture the hostage and makes the offer q = 0 rather than accepting ,
which does not maximize her utility. At the sub-game equilibrium the seller never rejects
the buyer's "take-it-or-leave-it" offer (q, p), and the buyer effectively obtains all the
bargaining power. Then a simple contract (q0, p0, t*) enables the parties to renegotiate
the default quantities according to a bargaining rule that cannot be modified during this
process. This ensures the credibility of the initial commitments and, therefore, the
optimal levels of specific investment by each party.

But the Aghion, Dewatripont and Rey solution requires quite a strong constraint of
verifiability (and actually a much stronger verifiability constraint than in the Hart and
Moore model) because the judge needs to know the delivery and the payment date in
order to be sure that he would be able to impose the performance of the contract.

The seller's investment incentives, determined by (5) are such that
then they will push him to under-invest. The buyer's return of investment will be then
lowered owing to the complementarity of the investments. So she will reduce her
investment, which lowers the seller's incentives to invest, and so on until a (sub-
optimal) Nash equilibrium is achieved.
3 Transaction-cost theory
The transaction-cost approach holds that the institutions of capitalism are to be
understood in transaction-cost economizing terms. Such economies are realized in a
discriminating way by aligning governance structures (market, hybrid forms, and firm)
with the attributes of transaction, of which the condition of asset specificity is the most
important (Williamson 1985, 1991). Unlike Incentives Theory, transaction-cost theory
(TCT) analyzes only discrete choices because it assumes that economic agents choose
between alternative governance structures and not a continuum of contracts. Moreover,
as compared to ICT, incompleteness in the transaction cost approach is not due to
verifiability problems but to the limited rationality of economic agents (contracting parties
and courts) and the uncertainty of the environment.
We will extend the Riordan and Williamson (1985) model in order to formalize the trade-
off between governance structures. Let r (q) be the revenue from producing a quantity q,

with , and c(q, A) the production costs of governance structures
procurement, with and . Asset specificity A is available at the
constant per unit cost of . The profit is given by

In a world without transaction costs, a first-best level of quantity (q*) and asset specificity
(A*), solutions of the first-order conditions

is achievable.
In world with transaction costs, the transaction costs of governance structure choice are
defined by the function TC = + z(A), where is the fixed cost of the chosen
governance structure, and z(A) an increasing function of asset specificity. z(A) takes the
form (A) when the governance structure is the market, w(A) when it is an hybrid form,
and x(A) when it is a firm. Let the subscripts M denote market, Hy hybrid forms and F the
firm. The transaction costs of these governance structures are given by

where > > 0 and 0 <
1 0

The corresponding profit functions for governance structures in a world with transaction
costs are

First-order conditions are

In each case, optimal output is defined in order to minimize production costs .
Optimal asset specificity is however chosen in order to minimize the sum of production
costs and transaction costs . As the first-order
condition for the output is identical for the three governance structures, then
. But the first-order condition for asset specificity is different. Indeed, as
is everywhere below , which is everywhere
below . Then, the A solutions of the optimization problem are such that A* > AF >
AHy > AM (see figure 1A.1). As , then the q solutions are such that q * > qF > qHy
> qM.

Figure 1A.1: omparative efficiency of the three governance structures

So, the optimal choice of governance structure depends only on asset specificity: market
procurement supports transactions with slight asset specificity, whereas the hybrid form
is more efficient as the condition of asset specificity deepens and internal procurement
(firm) as asset specificity is high.
1. If w N(w, w), the expected utility of the agent is

Because expected utility is increasing in

we can take a monotonic transformation. Then we obtain the utility function
given, which is equivalent to using the mean“variance criterion for choice
under uncertainty rather than the expected utility criterion.
2. From (IR2) and < 2, we obtain

we conclude that when (IC1) holds, (IR1) is also verified. Moreover (IC1)is a
binding constraint because the principal tries to keep his offer wi as small as
possible. Then substituting (IC1) in (IC2) we get . As 1
2 2
< 2, this inequality is always strict when e 1 > e 2.
3. The seller's investment incentives, determined by (5) are such that

then they will push him to under-invest. The buyer's return of investment will
be then lowered owing to the complementarity of the investments. So she will
reduce her investment, which lowers the seller's incentives to invest, and so
on until a (sub-optimal) Nash equilibrium is achieved.
Contracts, Organizations, and Institutions
Part II:
Chapter 2: The New Institutional Economics
Chapter 3: Contract and Economic Organization
Chapter 4: The Role of Incomplete Contracts in Self-Enforcing Relationships
Chapter 5: Entrepreneurship, Transaction-Cost Economics, and The Design of
The New Institutional Economics
Chapter 2:
Ronald Coase

It is commonly said, and it may be true, that the New Institutional Economics started with
my article, "The Nature of the Firm" (1937) with its explicit introduction of transaction
costs into economic analysis. But it needs to be remembered that the source of a mighty
river is a puny little stream and that it derives its strength from the tributaries that
contribute to its bulk. So it is in this case. I am not thinking only of the contributions of
other economists such as Oliver Williamson, Harold Demsetz, and Steven Cheung,
important though they have been, but also of the work of our colleagues in law,
anthropology, sociology, political science, sociobiology, and other disciplines.

The phrase, "the New Institutional Economics," was coined by Oliver Williamson. It was
intended to differentiate the subject from the "old institutional economics." John R.
Commons, Wesley Mitchell, and those associated with them were men of great
intellectual stature, but they were anti-theoretical, and without a theory to bind together
their collection of facts, they had very little that they were able to pass on. Certain it is
that mainstream economics proceeded on its way without any significant change. And it
continues to do so. I should explain that, when I speak of "mainstream economics," I am
referring to microeconomics. Whether my strictures apply also to macroeconomics I
leave to others.
Mainstream economics, as one sees it in the journals and the textbooks and in the
courses taught in economics departments has become more and more abstract over
time, and although it purports otherwise, it is in fact little concerned with what happens in
the real world. Demsetz has given an explanation of why this has happened: economists
since Adam Smith have devoted themselves to formalizing his doctrine of the invisible
hand, the coordination of the economic system by the pricing system. It has been an
impressive achievement. But, as Demsetz has explained it is the analysis of a system of
extreme decentralization. However, it has other flaws. Adam Smith also pointed out that
we should be concerned with the flow of real goods and services over time “ and with
what determines their variety and magnitude. As it is, economists study how supply and
demand determine prices but not with the factors that determine what goods and
services are traded on markets and therefore are priced. It is a view disdainful of what
happens in the real world, but it is one to which economists have become accustomed,
and they live in their world without discomfort. The success of mainstream economics in
spite of its defects is a tribute to the staying power of a theoretical underpinning, since
mainstream economics is certainly strong on theory if weak on facts. Thus, for example,
in the Handbook of Industrial Organization, Bengt Holmström and Jean Tirole (1989, p.
126), writing on "The Theory of the Firm," remark that "the evidence/theory ratio is
currently very low in this field."
This disregard for what happens concretely in the real world is strengthened by the way
economists think of their subject. In my youth, a very popular definition of economics was
that provided by Lionel Robbins (1935, p. 15) in his book An Essay on the Nature and
Significance of Economic Science: "Economics is the science which studies human
behavior as a relationship between ends and scarce means that have alternative uses."
It is the study of human behavior as a relationship. These days economists are more
likely to refer to their subject as "the science of human choice" or they talk about "an
economic approach." This is not a recent development. John Maynard Keynes said that
the "Theory of Economics is a method rather than a doctrine, an apparatus of the mind,
a technique of thinking, which helps the possessor to draw correct conclusions"
(introduction in H. D. Henderson 1922, p. v). Joan Robinson (1933, p. 1) says in the
introduction to her book The Economics of Imperfect Competition that it "is presented to
the analytical economist as a box of tools." What this comes down to is that economists
think of themselves as having a box of tools but no subject matter. It reminds me of two
lines from a modern poet (I forget the poem and the poet but the lines are indeed
I see the bridle and the bit all right
But where's the bloody horse?

I have expressed the same thought by saying that we study the circulation of the blood
without a body.

In saying this I should not be thought to imply that these analytical tools are not
extremely valuable. I am delighted when our colleagues in law use them to study the
working of the legal system or when those in political science use them to study the
working of the political system. My point is different. I think we should use these
analytical tools to study the economic system. I think economists do have a subject
matter: the study of the working of the economic system, a system in which we earn and
spend our incomes. The welfare of a human society depends on the flow of goods and
services, and this in turn depends on the productivity of the economic system. Adam
Smith explained that the productivity of the economic system depends on specialization
(he says the division of labor), but specialization is possible only if there is exchange “
and the lower the costs of exchange (transaction costs if you will), the more
specialization there will be and the greater the productivity of the system. But the costs of
exchange depend on the institutions of a country: its legal system, its political system, its
social system, its educational system, its culture, and so on. In effect it is the institutions
that govern the performance of an economy, and it is this that gives the "New
Institutional Economics" its importance for economists.
That such work is needed is made clear by another feature of economics. Apart from the
formalization of the theory, the way we look at the working of the economic system has
been extraordinarily static over the years. Economists often take pride in the fact that
Charles Darwin came to his theory of evolution as a result of reading Thomas Malthus
and Adam Smith. But contrast the developments in biology since Darwin with what has
happened in economics since Adam Smith. Biology has been transformed. Biologists
now have a detailed understanding of the complicated structures that govern the
functioning of living organisms. I believe that one day we will have similar triumphs in
economics. But it will not be easy. Even if we start with the relatively simple analysis of
"The Nature of the Firm," discovering the factors that determine the relative costs of
coordination by management within the firm or by transactions on the market is no
simple task. However, this is not by any means the whole story. We cannot confine our
analysis to what happens within a single firm. This is what I said in a lecture published in
Lives of the Laureates (Coase 1995, p. 245): "The costs of coordination within a firm and
the level of transaction costs that it faces are affected by its ability to purchase inputs
from other firms, and their ability to supply these inputs depends in part on their costs of
coordination and the level of transaction costs that they face which are similarly affected
by what these are in still other firms. What we are dealing with is a complex interrelated
structure." Add to this the influence of the laws, of the social system, and of the culture,
as well as the effects of technological changes such as the digital revolution with its
dramatic fall in information costs (a major component of transaction costs), and you have
a complicated set of interrelationships the nature of which will take much dedicated work
over a long period to discover. But when this is done, all of economics will have become
what we now call "the New Institutional Economics."

This change will not come about, in my view, as a result of a frontal assault on
mainstream economics. It will come as a result of economists in branches or sub-
sections of economics adopting a different approach, as indeed is already happening.
When the majority of economists have changed, mainstream economists will
acknowledge the importance of examining the economic system in this way and will
claim that they knew it all along.
Chapter 2 was originally published in American Economic Review, 88(2), May 1998. It is
reprinted with the permission of Ronald Coase and The American Economic Association.
Contract and Economic Organization
Chapter 3:
Oliver E. Williamson
1 Introduction
As discussed elsewhere, the New Institutional Economics works predominantly at two
levels: the institutional environment, which includes both the formal (laws, polity, judiciary)
and informal (customs, mores, norms) rules of the game, and the institutions of
governance (markets, firms, bureaus) or play of the game (Williamson 1998). The
transaction-cost economics approach to economic organization is concerned principally
with the latter, with special emphasis on the governance of contractual relations. As it
turns out, this approach to economic organization has wide application, generates a
large number of refutable implications to which the data are broadly corroborative, and
has many public policy ramifications “ especially to anti-trust and regulation but to
include labor, corporate governance, corporate finance, privatization, and the list goes on.
That the study of governance has such broad application is because any issue that
arises as or can be reformulated as a contracting problem can be examined to
advantage in transaction-cost economizing terms. Many issues present themselves
naturally in this form “ the mundane make-or-buy decision being an example. The
comparative contractual choice to be made here is whether a firm should contract out for
the provision of a good or service or take the transaction out of the market and manage it
internally. The contractual nature of other transactions is more subtle “ as with the
corporate finance decision, where the choice needs to be made between debt and equity.
Ordinarily debt and equity are treated as strictly financial instruments, but they are also
usefully viewed as alternative modes of governance “ where debt is the more market-like
mode of contracting for project finance and equity is the more administrative form and is
akin to hierarchy. Still other transactions need to be reformulated to bring out their
contractual nature, the oligopoly problem being an example. The contractual issues
surface here not when the problem is posed in Cournot or structure“conduct“
performance (SCP) terms but as a cartel problem. When does the unenforceable and
often illegal "contract" among members of a cartel work well or poorly, and why?

But there is a puzzle. If the comparative contractual approach to economic organization
has wide application and generates new and testable propositions, why did it take so
long to take hold? Also, where does it go from here?
2 Obstacles
Major obstacles to the comparative contractual approach to economic organization were
that (1) orthodoxy was uncritical in its treatment of the firm in technological terms, partly
because it was committed to full formalization, (2) contract had come to be viewed as
unproblematic because of the presumed efficacy of contract law and its enforcement,
and (3) organization was ignored, dismissed, or suppressed. Consider each in turn.

The theory of the firm-as-production function (or as production possibility set) was both a
major conceptual achievement and a great analytical convenience for the progressive
mathematization of economics in the immediate post-war era. To be sure, other social
scientists were unpersuaded by some of the more arid abstractions of economics. The
Graduate School of Industrial Administration at Carnegie Tech aside, however, the
business schools lacked the academic credentials to dispute economic orthodoxy
(Gordon and Howell 1959). And the gulf between economics and sociology was vast
(witness the quip by James Duesenberry that "economics is all about how people make
choices; sociology is all about how they don't have any choices to make," 1960, p. 233).

The ideas that contracts were complete and that the laws on contract (regarding offer
and acceptance, breach, etc.) were well conceived and were enforced by well-informed
courts in a legalistic way effectively removed contract from the research agenda. Upon
treating contracts as unproblematic and fully within the purview of the law, the self-
contained nature of the economics enterprise was reinforced.

The propensity of economists to delimit microeconomics to price and output served
further to limit the scope. As Harold Demsetz put it, "It is a mistake to confuse the firm of
economic theory with its real-world namesake. The chief mission of neoclassical
economics is to understand how the price system coordinates the use of resources, not
to understand the inner workings of real firms" (1983, p. 377). The contributions of
organization theory to the study of economic organization and contract could thus also
be set aside.
3 Growing discontent
In addition to the price and output purposes described by Demsetz, economists were
also expected to advise on public policy. This very same theory of the firm was also used
by Industrial Organization specialists to inform anti-trust and regulation. That was an
embarrassment, in that the interpretation of non-standard and unfamiliar contracting and
organizational practices in strictly technological terms invited convoluted and even
preposterous public policy “ although that was not evident until someone observed that
the emperor was scantily dressed (Coase 1972). Concurrently, the legal centralism
approach to contract law and its enforcement was also coming under criticism from
lawyers, whence the readiness of economists to be dismissive of contract was being
questioned. The growing importance of the modern corporation was also bringing issues
of organization and governance more forcefully to the fore. The upshot is the economic,
legal, and organizational foundations for the orthodox theory of the firm were all under
assault. Consider each in turn.

So long as the firm was viewed in strictly technological terms, students of public policy
were prone to condemn structures and practices that did not have obvious technological
origins or serve technological purposes. For example, vertical integration that lacks a
"physical or technical aspect," such as integrating the production of assorted
components or forward integration into distribution, was believed to be lacking in
economizing purpose and effect and, therefore, to be deeply problematic “ whereupon
excesses of vertical integration and firm size were projected (Bain 1968, p. 381). More
generally, non-standard and unfamiliar contracting and organizational practices were
believed to have anticompetitive purpose and effect, there being no legitimate
economizing purpose that could accrue thereto. The then head of the Antitrust Division
of the US Department of Justice thus treated "customer and territorial restrictions not
hospitably in the common law tradition, but inhospitably in the tradition of antitrust."

Reversing such a policy was not easy. It takes a theory to beat a theory (Kuhn 1970),
and a rival theory needed to be fashioned. Ongoing developments in law and
organization contributed to this purpose.
The legalistic approach to contract law had come under criticism from Karl Llewellyn in
1931, but that took time to register. Llewellyn's early distinction between the prevailing
contract as legal rules approach and his proposed contract as framework approach is
basic. The contract as framework approach recognizes that all complex contracts are
unavoidably incomplete and holds that a contract between two parties "almost never
accurately indicates real working relations, but affords a rough indication around
which such relations vary, an occasional guide in cases of doubt, and a norm of ultimate
appeal when the relations cease in fact to work" (Llewellyn 1931, p. 737). The main
contractual action thus takes place between the parties in the context of private ordering,
to which court ordering appears late for purposes of ultimate appeal, if at all.
That reverses the "legal centralism" tradition, which holds that "disputes require ˜access™
to a forum external to the original social setting of the dispute [and that] remedies will be
provided as prescribed in some body of authoritative learning and dispensed by experts
who operate under the auspices of the state" (Galanter 1981, p. 1). The facts, however,
reveal otherwise. Most disputes, including many that under current rules could be
brought to a court, are resolved by avoidance, self-help, and the like (Galanter 1981, p.
2). That is because in "many instances the participants can devise more satisfactory
solutions to their disputes than can professionals constrained to apply general rules on
the basis of limited knowledge of the dispute" (Galanter 1981, p. 4). Private ordering
through ex post governance is therefore where the main action resides.

A growing appreciation for the importance of organization and, more generally, of
governance was also taking shape. Alfred Chandler's study of the modern corporation in
the first half of the twentieth century revealed that significant organization form changes
had taken place with the result that the managerial discretion problem with which Adolf
Berle and Gardiner Means (1932) were concerned was being brought under more
effective control (Chandler 1962). In that event, the firm was more than a production
function. The structure of the corporation, especially as between centralized (U-form)
and divisionalized (M-form), had governance/economizing consequences as well.

The quotation is attributed to Donald Turner by Stanley Robinson, New York State Bar
Association, Antitrust Symposium, 1968, p. 29.

Recourse to the literal language of the contract and access to the courts for purposes of
ultimate appeal are important so as to delimit threat positions.
4 Fashioning a response
The comparative contractual approach to economic organization is responsive to all
three of these critiques. Rather than hold law, economics, and organization apart, a
combined law, economics, and organizations approach began to take shape. The firm is
described as a governance structure in which (1) economizing transcends technology to
include contract and organization, (2) comparison with alternative modes of managing
contracts is featured, and (3) organization form matters.

Describing the human actors whose behavior we are studying turns out to be important
to this project. So does naming the unit of analysis.

4.1 Human actors
According to Herbert Simon, "Nothing is more fundamental in setting our research
agenda and informing our research methods than our view of the human beings whose
behavior we are studying" (Simon 1985, p. 303, emphasis added). That challenges the
propensity of economists to describe human actors in a fashion that served their
analytical convenience “ as illustrated by the triple of omniscience, omnipotence, and
benevolence to which Avinash Dixit refers (1996, p. 6) in his description of old-style
public policy analysis.

The transaction-cost treatment of human actors emphasizes three features: the cognitive
ability of human actors, their self-interestedness, and their capacity for foresight.
Describing human actors as boundedly rational “ that is, intendedly rational, but only
limitedly so (Simon 1961, p. xxiv) “ undermines the idea of complete contracting. Instead,
all complex contracts are unavoidably incomplete “ hence contain errors, gaps,
omissions, and the like. Such incompleteness is of special concern where human actors
are given to opportunism, hence will not reliably self-enforce all promises. Instead, they
will sometimes behave strategically “ by sending false or misleading signals, by
interpreting the data to their advantage, by costly repositioning, and by otherwise
withholding best efforts to realize mutual gains. Mere promise, unsupported by credible
commitments, is not self-enforcing by reason of opportunism.
A redeeming feature, however, is that human actors possess the capacity for conscious
foresight. As Richard Dawkins puts it, the "capacity to simulate the future in
imagination [saves] us from the worst consequences of the blind replicators" (1976, p.
20). Parties to a complex contract who look ahead, recognize potential hazards, work out
the contractual ramifications, and fold these into the ex ante contractual agreement
obviously enjoy advantages over those who are myopic or take their chances and knock
on wood. Human actors with conscious foresight will take steps to mitigate contractual
hazards by crafting responsive governance structures.

4.2 Unit of analysis
But wherein do the potential hazards reside? What does working out the contractual
ramifications entail? How does the ex ante contractual agreement get reshaped? John R.
Commons' prescient insights apply. It was his position that "the ultimate unit of activity
must contain in itself the three principles of conflict, mutuality, and order. This unit is a
transaction" (Commons 1932, p. 4).

Taking the transaction to be the basic unit of analysis has turned out to be an instructive
way of uncovering contractual hazards. If some transactions pose few hazards and
others pose many, then presumably there are systematic differences between them.
Identifying the key attributes of transactions that give rise to differential hazards has
been instructive both for the theory of contract and economic organization and for
empirical investigations that appertain thereto (which sometimes take the form of
focused case studies, as with John Stuckey's study of vertical integration and joint
ventures in the aluminum industry, 1983, but more often involve cross-section studies
that employ conventional econometric techniques, as with Paul Joskow's study of coal
contracting for electric power generation, 1987).
4.3 Operationalization
The idea that the transaction is the basic unit of analysis needs to be harnessed to an
economic purpose. The Commons' triple invites the concept of governance “ where
governance is the means by which order is accomplished in a relation in which potential
conflict threatens to upset or undo opportunities to realize mutual gains. Economizing
purposes that transcend technology are thereby realized.

Combining the idea that economizing is the main purpose served by economic
organization with the proposition that mitigating contractual hazards (in cost effective
degree) is among the chief economizing purposes to be served leads to the following
hypothesis: transactions, which differ in their attributes, are aligned with governance
structures, which differ in their cost and competence, so as to effect an economizing
result. Transaction-cost economics realizes much of its predictive content from this
discriminating alignment hypothesis.
Implementation of this hypothesis requires that alternative modes of governance be
identified and their defining attributes described. There being no single, all-purpose
superior form of organization, all evidently have strengths and weaknesses. That is
because each generic mode of organization is defined by an internally consistent
syndrome of attributes to which differential performance competencies accrue. As
discussed elsewhere (Williamson 1991, 1999), key attributes of governance include (1)
incentive intensity, (2) administrative controls, and (3) the applicable law of contract.
Both different types of markets (spot markets and various forms of long-term contracting)
and different types of hierarchies (firms, regulation, public bureaus) are distinguished. In
general, incentive intensity decreases and administrative controls build up in moving
across the succession shown in figure 3.1 (where h denotes hazards and s denotes

Figure 3.1: ncentive intensity and administrative controls

What is furthermore noteworthy is that each generic mode of governance is supported by
a distinctive form of contract law. The contract law of spot markets is that of legal rules,
which is the ideal transaction in both law and economics: "sharp in by clear agreement;
sharp out by clear performance" (Macneil 1974, p. 738). This legal rules approach gives
way to Llewellyn's concept of contract-as-framework as the importance of continuity
builds up and incomplete long-term contracting is adopted. That in turn undergoes
change when transactions are taken out of the market and organized internally. The
implicit law of contract now becomes that of forbearance. Thus whereas courts routinely
grant standing to firms engaged in inter-firm contracting should there be disputes over
prices, the damages to be ascribed to delays, failures of quality, and the like, courts will
refuse to hear disputes between one internal division and another over identical technical
issues. Access to the courts being denied, the parties must resolve their differences
internally. Accordingly, hierarchy is its own court of ultimate appeal. That firms and
markets differ in their access to fiat is partly explained by these contract law differences.
The concept of contract thus has a pervasive influence on the study of economic
organization. Consider the following five features: (1) the transaction (trade, exchange,
contract) is the basic unit of analysis; (2) all complex contracts are incomplete (by reason
of bounded rationality); (3) many contracts pose hazards (because mere promise,
unsupported by credible commitments, is not self-enforcing “ by reason of opportunism);
(4) governance structures, which are the institutional frameworks within which the
integrity of contract is decided, are hazard mitigating responses; and (5) each generic
mode of governance is supported by a distinctive form of contract law.

4.4 Additional features

Omitted from the discussion but important to an understanding of contract and
organization are (1) the institutional environment “ constitution, laws, polity, judiciary “
which define the rules of the game, (2) the central importance of adaptation, of both
autonomous (Hayek 1945) and cooperative (Barnard 1938) kinds, to economic
performance, and (3) the distinctive process attributes of organization, in auditing,
accounting, informal organization, bureaucratization, and politicking, to include the
ramifications of each on comparative economic organization. Suffice it to observe here
that the study of contract and economic organization is an ambitious interdisciplinary
undertaking. (For a discussion, see Williamson 1991.)

Such profound insights failed to impress critics of older-style institutional economics, who
held that "Without a theory [American institutionalists] had nothing to pass on except a mass
of descriptive material waiting for a theory, or a fire" (Coase 1984, p. 230).

Surveys of empirical transaction cost economics are reported in Howard Shelanski and
Peter Klein (1995), Keith Crocker and Scott Masten (1996), Bruce Lyons (1996), and Aric
Rindfleisch and Jan Heide (1997).
5 Looking ahead
The transaction-cost approach to economic organization has progressed through a
series of stages. Beginning with informal (Coase 1937) and preformal (Williamson 1975;
Klein, Crawford and Alchian 1978) stages, transaction-cost economics has moved into
semi-formal (Klein and Leffler 1981; Williamson 1983, 1991; Riordan and Williamson
1985), and fully formal (Grossman and Hart 1986; Hart 1995) work.
Although full formalization is vital to a progressive research agenda, it can also be
problematic. Here, as elsewhere, there are trade-offs. Thus although Simon once argued
that "mathematical translation is itself a substantive contribution to theory because it
permits clear and rigorous reasoning about phenomena too complex to be handled in
words" (1957, p. 89) and subsequently asserted that the "poverty of mathematics is an
honest poverty that does not parade imaginary riches before the world" (1957, p. 90),
provision also needs to be made for the possibility that core features of the theory are left
out or obscured by the translation. There is, after all, such a thing as prematurely formal
theory. David Kreps speaks to the issues as follows (1999, p. 122):
If Markets and Hierarchies has been translated into game theory using notions of
information economics, it is a very poor translation In particular, mathematics-based
theory still lacks the language needed to capture essential ideas of bounded rationality,
which are central to transaction costs and contractual form. Anyone who relies on the
translations alone misses large and valuable chunks of the original.

Kreps has reference especially to the "property rights theory of the firm," which is the
fully formal theory to which I refer above. My reservations about this theory have been
discussed elsewhere (Williamson 2000) and will not be repeated here. More to the point
is that a series of promising full formalization efforts are taking shape "even as I write."
These include the unpublished paper by Oliver Hart and John Moore (1999b), the
unpublished paper by Patrick Bajari and Steven Tadelis (1999), and the unpublished
paper by Gene Grossman and Elhanan Helpman (1999). I am confident that these are
harbingers of more to come.

Such theoretical developments in combination with the vast and growing empirical
literature in transaction-cost economics lead me to project that the comparative
contractual approach to the study of economic organization will remain an active area for
research well into the new millennium. Public policy has been and will continue to be a

See n. 3.
6 Concluding remarks
Whereas once the subject of contract was relegated to an obscure closet in the house of
economics, that has changed as greater appreciation for more veridical attributes (as
against analytically convenient attributes) of human actors has set in, the limits of legal
centralism have been conceded, and the apparatus for doing comparative contractual
analysis has been progressively built up. One of the most important developments with
respect to this last has been to go beyond the "black box" theory of the firm (according to
which the firm is a production function) to view the firm in comparative contractual terms
“ as a governance structure.[6] As Kreps observes (1990, p. 96):
The [neoclassical] firm is like individual agents in textbook economics Agents have
utility functions, firms have a profit motive; agents have consumption sets, firms have
production possibility sets. But in transaction-cost economics, firms are more like
markets “ both are arenas within which the individual can transact.
This reconceptualization of firms and markets as alternative modes of governance with
discrete structural differences has had ramifications for anti-trust and regulation and has
promise for helping to reshape public policy analysis more generally. Avinash Dixit's
monograph on The Making of Economic Policy has precisely that ambition (1996, p. 9):

Economists studying business and industrial organization have long recognized the
inadequacy of the neoclassical view of the firm and have developed richer paradigms
and models based on various kinds of transactions costs. Policy analysis also stands to
benefit from such an approach, opening the black box and examining the actual
workings of the mechanism inside. This is the starting point, and a recurrent theme, of
this monograph.

I conclude that the examination of alternative modes of organization through the lens of
contract and transaction cost economizing has been and will continue to be a productive
research enterprise.

This is responsive to Kenneth Arrow's advisory that "Any standard economic theory, not just
neoclassical, starts from the existence of firms. Usually, the firm is a point or at any rate a
black box Butfirmsare palpably not points. They have internal structure. This internal
structure must arise for a reason" (1999, p. vii).
Chapter 3 was originally published as "Contract and Economic Organization," in Revue
d'Economie Industrielle (92, 2000).
1. The quotation is attributed to Donald Turner by Stanley Robinson, New
York State Bar Association, Antitrust Symposium, 1968, p. 29.
2. Recourse to the literal language of the contract and access to the courts
for purposes of ultimate appeal are important so as to delimit threat
3. Such profound insights failed to impress critics of older-style institutional
economics, who held that "Without a theory [American institutionalists]
had nothing to pass on except a mass of descriptive material waiting for
a theory, or a fire" (Coase 1984, p. 230).
4. Surveys of empirical transaction cost economics are reported in Howard
Shelanski and Peter Klein (1995), Keith Crocker and Scott Masten
(1996), Bruce Lyons (1996), and Aric Rindfleisch and Jan Heide (1997).
5. See n. 3.
6. This is responsive to Kenneth Arrow's advisory that "Any standard
economic theory, not just neoclassical, starts from the existence of firms.
Usually, the firm is a point or at any rate a black box Butfirmsare
palpably not points. They have internal structure. This internal structure
must arise for a reason" (1999, p. vii).
The Role of Incomplete Contracts in
Chapter 4:

Self-Enforcing Relationships
Benjamin Klein
1 Introduction
A major advance in economics involves the recognition that contracts adopted by
transactors are incomplete. This fundamental insight has produced two main strands of
economic research. One strand of research emphasizes the importance of self-
enforcement in assuring contractual performance. Building upon Stuart Macaulay's
pioneering study documenting that performance is secured in most business
relationships not by the threat of court enforcement but by the threat of termination of the
relationship, this work develops models of self-enforcement where a termination sanction
is sufficient to assure transactor performance.

The other, more extensive, unrelated strand of economic research flowing from
incomplete contracts is the principal“agent contract design literature. This work
examines the role of contract terms in minimizing transactor malincentives given that
performance can only imperfectly be contracted on. The major point of this chapter is
that the incomplete contract terms actually used by transactors in the marketplace can
be understood only by combining these two strands of research. What follows is a
summary of my research on contracts from this perspective of integrating our research
on incomplete contracts.
One way to integrate the two lines of research on incomplete contracts is to add self-
enforcement considerations to the principal“agent model. The alternative way I have
attempted this integration is by extending the simple model of self-enforcement to take
account of the role of contract terms in facilitating self-enforcement. Contract terms are
used as an aid to self-enforcement because the transactors' reputational capital through
which the self-enforcement mechanism operates is limited (in the sense that transactors
can credibly promise to pay only a finite maximum future amount to their transacting
partners in return for current performance). Therefore, although Macaulay and others are
correct in noting that many business relationships are self-enforced, transactors are not
indifferent regarding the contract terms they choose to govern their self-enforcing

Rather than explaining the incomplete contract terms chosen by transactors in terms of
the minimization of direct transactor malincentives, contract terms are considered here
as devices that economize on transactors' limited reputational capital to facilitate self-
enforcement. Transactors use contract terms to get close to desired performance without
creating too much rigidity and to shift future rents between transacting parties so as to
coincide more closely with each transactor's potential non-performance gain. In these
ways contract terms assure that the transactors' business relationship remains self-
enforcing over the broadest range of likely future market conditions. Within this
framework where contract terms are used to efficiently define the self-enforcing range of
the transactors' contractual relationship, self-enforcement and courtenforcement are not
alternative enforcement mechanisms, but are complementary instruments used by
transactors in combination to guarantee transactor performance.

Macaulay (1963).

See Klein and Leffler (1981) for an early example of such a model.

An important paper by George Baker, Robert Gibbons and Kevin Murphy (1999) does this
by adding a self-enforcement mechanism to the standard Grossman and Hart (1986)
principal“agent model of the firm. Although Baker, Gibbons and Murphy provide a number of
valuable insights regarding the operation of the self-enforcement mechanism in this context,
they do not identify what I consider to be the key advantage of vertical integration that
facilitates self-enforcement discussed below, post-contract flexibility.
2 Incomplete contracts
Contracts are incomplete because there are significant information and measurement
costs surrounding most business transactions. When a large number of possible
contingencies exist regarding future events, the use of the fully contingent complete
contract of economic theory is too costly. Transactors use incomplete contracts in these
circumstances not only to avoid the significant "ink costs" of writing fully contingent
contracts, but, more importantly, because incomplete contracts avoid the wasteful search
and negotiation costs that otherwise would be borne by transactors. The attempt to
specify desired performance completely for a very large number of unlikely possibilities
primarily involves the costly search by transactors for an informational and negotiating
advantage over their transacting partner. Contractual specification of performance for
such extremely low-probability contingencies creates potential wealth distribution effects,
where one transactor will receive a transfer in the event some unlikely contingency
occurs, with little or no allocative benefits in terms of creating proper ex ante incentives.
Therefore, while these real resource costs associated with complete contractual
negotiation will lead individual profit maximizing transactors to stop short of complete
contract specification, transactors may jointly decide to reduce the wasteful rent
dissipating activity of increased contractual specification even further. Transactors enter
relationships knowing they have left some unlikely contingencies unspecified,
recognizing that if such a contingency develops, it will have to be handled after the fact.

In addition to avoiding the rent dissipating search and negotiation costs involved in
complete contractual specification, contracts are incomplete because of measurement
costs. Some aspects of performance, such as the taste of a hamburger or the energy an
employee devotes to a task, may be prohibitively costly to contractually specify in a way
that breach can be demonstrated to a third-party enforcer. Therefore, performance along
these not easily measured dimensions will not be fully specified in the contract.

Because the contract terms used by transactors are necessarily incomplete, transactors
are cautious regarding what they write in their contracts. Incomplete contract terms may
create opportunities for transactors to engage in a hold-up by using the court to enforce
the literal imperfect contract term in a manner that is contrary to the intent of the
contractual understanding.[4] This is one of the primary economic lessons of the General
Motors“Fisher Body case. In that case Fisher took advantage of the long-term, cost-plus
exclusive dealing contract designed by the parties to encourage Fisher to make GM-
specific investments to hold up General Motors. The long-term contract used to protect
Fisher's GM-specific investments which locked Fisher into GM-created contractual
specificity that locked General Motors into Fisher. Fisher then took advantage of this
long-term GM contractual commitment by refusing to locate an important body plant next
to the GM assembly plant. As a consequence, Fisher produced very costly (but highly
profitable) automobile bodies that General Motors was compelled to buy.[5]

The General Motors“Fisher example illustrates that, contrary to most models, increased
contractual specification can make things worse.[6] Increased contractual specification
not only produces benefits, but also creates costs. In particular, rigidity is created when
an agreement is formalized in a long-term explicit contract. Only by declaring bankruptcy
could General Motors have unilaterally opted out of not performing to the literal imperfect
terms of the long-term Fisher Body contract. Unless a side payment was made to Fisher
Body (and vertical integration was the form in which such a side payment was ultimately
made), General Motors was forced to continue buying bodies at cost-plus from
improperly located plants until the contract expired. If, on the other hand, the Fisher“
GMunderstanding had not been formalized in a long-term written contract, the parties
would have been able to flexibly alter their supply arrangements without being forced by
the court to adhere to the conditions of the imperfect written agreement.

The extent of contractual specification chosen by transactors involves trading-off the
obvious benefit of being able to use the court to enforce elements of performance with
these less obvious costs of contractual specification. Increased contractual specification
involves rent dissipating search and negotiation costs that results in an imperfect, rigid
agreement which can then be used by transactors to hold up one another. The existence
of these costs, not the narrow transaction costs associated with contractual specification,
is why transactors often decide to intentionally leave some elements of performance

I am assuming for analytical and expositional simplicity that the court enforces written terms
and does not enforce unwritten, understood terms. While courts in practice interpret both
written and unwritten terms when enforcing contractual agreements, under English common
law the amount of discretion exercised by courts with regard to unambiguous written terms is
generally limited. In any event, as transactors cover additional contingencies with explicit
imperfect contract terms, it is reasonable to assume that after some point there is an
increased likelihood the court will effectuate a hold-up by enforcing the contract in a manner
that is contrary to the parties' contractual understanding.

Klein, Crawford and Alchian (1978) and Klein (2000). These transitional hold-up costs
conflict with the costless ex post renegotiation assumption generally made in the incomplete
contracting/property-rights literature that has developed from the pioneering work of
Grossman and Hart (n. 3). These models assume, contrary to what occurred in the Fisher“
GM case, that in cases where a potential hold-up exists, ex post renegotiation of the contract
instantaneously and costlessly takes place, so that, after a lump sum is paid to the transactor
that can engage in the hold-up, price and cost quickly move to the efficient level. Therefore,
instead of designing contractual arrangements to minimize the ex ante expected hold-up
potential and, hence, the real resource costs incurred during the hold-up process (as the
transactor engaging in a hold-up attempts to convince its transacting partner of the extent and
magnitude of the hold-up), these models focus on ex ante investment inefficiencies as the
economic motivation for contractual organization. Although the reduced willingness to make
specific investments (as well as the wasteful expenditure of resources during the initial
contracting process to protect against future hold-ups) are costs of potential hold-ups in this
framework, the costless renegotiation formulation of the problem makes it difficult to justify the
post-contract flexibility advantages of vertical integration discussed below.

Bernheim and Whinston (1998) present a model where increased contractual specification
may make things worse by creating asymmetric non-performance gains for one party.
3 Self-enforcing arrangements
Transactors can freely avoid the costs associated with complete contractual specification
because they have available a self-enforcement mechanism to assure performance.
Rather than court enforcement of written contract terms, a self-enforcement mechanism
operates by threatening termination of the business relationship for non-performance of
the unwritten contractual understanding. Transactors compare the short-term gains they
can achieve by not performing consistent with the contractual understanding, W1, with
the discounted expected future profit stream they will lose if the relationship is terminated
for such non-performance, W2. Performance is assured when

W2, the capital cost of the lost expected future profit stream that is imposed upon a non-
performing transactor when the relationship is terminated,

is called the transactor's reputational capital. The magnitude of each transactor's
reputational capital determines, according to (1), the efficacy of the self-enforcement

When sufficient reputational capital exists, transactors will rely on selfenforcement rather
than court-enforcement. Self-enforcement avoids the costs associated with contractual
specification described above and reduces the time lag and noise involved in court
detection and sanction of non-performance. Court-enforcement entails an imperfect
timeintensive process of contract interpretation to determine whether a contractual
understanding has been violated or not, followed by a further period to determine an
appropriate penalty. Rather than relying on necessarily imperfect contract terms to
communicate the elements of agreed-upon performance to the court, such third-party
contract interpretation and enforcement problems are avoided entirely with a
selfenforcing mechanism. With self-enforcement, once transactors learn that their
transacting partner has not performed, a termination sanction is imposed. Therefore, if
sufficient reputational capital exists, transactors always will prefer to handle contract
performance with self-enforcement.

If General Motors had possessed sufficient reputational capital, an explicit long-term
contract would not have been used to induce Fisher Body to make its GM-specific
investments and the subsequent costs associated with the contract would have been
avoided. A long-term Fisher“GM contract would not have been necessary because
General Motors would have had more to lose in the long run than it could gain in the
short run from holding up Fisher Body for its GM-specific investments. Therefore, Fisher
would have been assured that General Motors would not engage in a hold-up and would
not have required the long-term exclusive dealing contract that later led to the Fisher
hold-up of General Motors. It has been extensively documented that Japanese
automobile manufacturers avoid these costs of court-enforcement in their dealings with
parts suppliers in exactly this way.[7] By relying primarily on the threat of non-renewal of
the relationship Japanese manufacturers induce their suppliers to make the required
specific investments and to charge reasonable prices that are adjusted downward at
regular intervals as sales increase and supplier costs fall.[8]
However, although self-enforcement is preferable to courtenforcement, transactors
cannot always rely entirely on a selfenforcement mechanism because the magnitude of
the private sanction that can be imposed for non-performance, W2, is limited.
Presumably, this is the reason why General Motors could not use a Japanese-type
supply arrangement in its dealings with Fisher Body. General Motors' lack of sufficient
reputational capital (W2) compared to its hold-up potential given the magnitude of
Fisher's required specific investments (W1) made it impossible for Fisher and General
Motors to use the superior, largely self-enforcing alternative. Instead, they were forced to
rely to a large extent on court-enforcement.

See Asanuma (1989). Similar descriptions of Japanese auto parts supply contracts are
provided in Cusumano and Takeishi (1991) and Sako and Helper (1998).
A self-enforcement mechanism may work well for Japanese automobile producers because
of (until recently) the high level of expected future demand growth and because of the
increased social cohesiveness and likely communication of non-performance to other
participants in the economy who may also impose a sanction by refusing to deal with the non-
performing transactor. Both of these factors imply a high level of the parties' reputational
4 Contract terms complement self-enforcement
In this framework the fundamental economic motivation for the use of court-enforceable
contract terms is to supplement self-enforcement. Court-enforced explicit contract terms
are a necessary evil that are used by transactors solely because the transactors possess
limited reputational capital. This has broad implications for the economic analysis of
contracts. Looking at contract terms in this way, it makes no sense to analyze the
malincentive effects of contract terms in isolation from self-enforcement. It suggests that
incomplete contract terms are likely to be used by transactors only to get close to desired
performance, with transactors using a self-enforcement mechanism to move behavior
the remainder of the way towards the desired level. As a consequence, the standard
principal“agent view of incomplete contracts, where contract terms are considered solely
as devices that create optimal incentives on imperfect court-enforceable proxies for
performance, provides a biased view of contractual arrangements. Without considering
self-enforcement, the malincentives that remain in most actual contractual arrangements
are likely to be enormous. Incomplete contract terms cannot be understood without
recognizing that their role often is to control W1 so that it remains below W2.
Recognition of the role of contract terms in facilitating selfenforcement explains, for
example, why Fisher and General Motors used such seemingly inappropriate cost-
plus/exclusive dealing contract terms. These contract terms may appear to have created
an incentive for Fisher to increase the costs of auto bodies to the contractually "locked-
in" General Motors. But the terms can be understood only within the context of
selfenforcement, where contract terms, although imperfect, are designed to create
conditions where each transactor has more to lose from termination of the relationship
than it has to gain from not performing. Within this self-enforcement framework, the
Fisher“GM contract terms were efficient when the parties entered into their contractual
arrangement in 1919. In fact, although Fisher always had the ability to exploit the
imperfect Fisher“GM body supply contract, the contract functioned extremely well for
more than five years. Presumably, Fisher had more to lose from GM's non-renewal of
the agreement than it had to gain. It was only in 1925, when GM's demand for Fisher
bodies increased dramatically (along with new large required Fisher-specific investments)
that Fisher began to take advantage of the contract. The next section discusses what
occurred in the Fisher“GM relationship to make it no longer self-enforcing. But the role of
incomplete contract terms in facilitating self-enforcement is first discussed in some more
Equation (1) suggests that transactors can use incomplete contract terms to facilitate
self-enforcement in two fundamental ways, by either reducing W1 or increasing W2.
Reducing W1 is the common motivation for contractual specification in the economic
literature. By defining a particular element of performance, the ability not to perform
along this dimension is directly controlled with court-enforcement. But, contrary to
standard economic literature, the goal of such contractual specification is to make the
residual W1 (that is too costly to reduce further because of the contract specification
costs discussed above) less than W2.
In addition to contract terms operating on the left-hand side of (1) to reduce the expected
gains from non-performance and hence the amount of reputational capital necessary to
make the arrangement self-enforcing, contract terms also can operate on the right-hand
side of (1). In particular, by shifting expected future rents and, therefore, reputational
capital between transactors, contract terms can make each transactor's reputational
capital coincide more closely with the transactors' potential expected gain from non-
performance. This effect provides an economic rationale for many of the contract terms
used in distribution arrangements, such as resale price maintenance (RPM) or exclusive
territories. By limiting the extent of intra-firm competition faced by a manufacturer's
dealers, these contract terms create future rents that dealers operating under such
contractual arrangements can expect to earn. Hence, these contract terms facilitate self-
enforcement of dealer performance by, in effect, shifting some of the manufacturer's
reputational capital to its dealers. The contract thereby increases the limited amount of
dealer reputational capital relative to the dealers' non-performance potential, creating a
situation where dealers have more to lose if they do not perform as desired.

Such a shift in rents can occur only if the manufacturer can credibly make such a
commitment, that is, only if the manufacturer has more to lose if it reneges on the
commitment than if it pays the dealer the promised future rents. This will depend on the
cost to the manufacturer of organizing distribution in some less efficient alternative way.
For example, in franchising arrangements franchisors can credibly commit to pay
franchisees a future premium stream at most equal to the present discounted value of
the cost savings of handling distribution with a franchising system than with the next
most efficient non-franchising system, such as operating its outlets with employees. Any
promised future franchisee premium stream greater than this will lead the franchisor not
to pay the premium and, instead, bear such higher distribution costs. This implies the
paradoxical result that a credible commitment is less likely to be made by a franchisor or
manufacturer as the cost of the next most efficient alternative distribution arrangement
decreases. Franchisees or dealers will believe they will receive the future profit premium
promised by the franchisor or manufacturer only if paying it is cheaper for them than not
paying it.[10]
Of course, both effects of contract terms in facilitating self-enforcement may operate at
the same time. For example, consider exclusive territory arrangements, where a
manufacturer designates a dealer as the exclusive supplier of the manufacturer's goods
or services within a particular area. Such an arrangement increases the dealer's
probability of repeat sales, internalizing dealer actions and thereby decreasing the
dealer's shortrun gain from non-performance, W1. But granting a dealer an exclusive
territory also may increase the dealer's future continuing profit stream, thereby creating a
valuable dealer asset that can be lost by termination for non-performance, W2.

This analysis illustrates a fundamental complementarity between court-enforcement and
self-enforcement. The two enforcement mechanisms are substitutes in demand, in the
sense of a positive cross-elasticity of demand, so that an increase in the price of one
mechanism leads to an increased use of the other mechanism. (For example, an
increase in the cost of using the court, such as in Russia, will lead to the increased use
of self-enforcement by transactors.) But the two enforcement mechanisms are
complements in supply, in the sense of a positive cross-elasticity of supply, so that an
increase, for example, in the quantity of reputational capital leads to an increase in the
marginal productivity of courtenforcement. That is, the two mechanisms work better
together than either of them do separately.

See Klein and Murphy (1988) and Klein (1999).

See Klein (1995), pp. 22“3. In Kenney and Klein (1983), the ability of DeBeers to commit to
promise to pay siteholders a future profit premium stream in return for not rejecting diamonds
that have been only grossly sorted analogously depends upon the cost savings of the
DeBeers marketing arrangement.
5 The self-enforcing range of contractual relationships
Transactors will design their contractual arrangements, i.e., combine court-enforced
written contract terms with self-enforced unwritten terms so as to optimally define the
self-enforcing range of their relationship. In particular, as the Fisher Body“GM case
illustrates, contract terms facilitate self-enforcement at the point of contracting but more
generally how the contract terms minimize expected costs of hold-up possibilities over
time. That is, since the future market conditions and hence the future gains from non-
performance are uncertain at the time individuals enter into their contractual agreements,
W1 and W2 should be thought of probabilistically.
The amount of each transactor's reputational capital, therefore, should be thought of as
defining the self-enforcing range of the contractual relationship, or the extent to which
market conditions can change (thereby altering the value of sunk specific investments
and the gains to one or the other party from non-performance) without precipitating
nonperformance. Within the self-enforcing range, in spite of the change in market
conditions, each transactor's gain from non-performance remains less than the self-
enforcing sanction that can be imposed. Whether the contract terms chosen by
transactors facilitate self-enforcement in either of the two general ways outlined above,
namely by controlling the expected gains from non-performance or by shifting
reputational capital between the parties, the intended result is to widen the extent to
which ex post market conditions may change unanticipatedly yet performance remains

This probabilistic self-enforcing framework explains why hold-ups sometimes occur.[11] In
the Fisher“GM case it does not make sense to assume that Fisher Body took advantage
of General Motors because General Motors was na¨±ve or because Fisher Body was
able to deceive General Motors into entering an imperfect long-term, exclusive dealing,
cost-plus contract.[12] Relying on the ability of transactors to deceive their transacting
partners is a highly unsatisfactory, usually untestable, way to explain why hold-ups occur.
General Motors and Fisher Body were two large, sophisticated business firms that likely
were fully cognizant of the malincentive problems inherent in the imperfect contract they
entered into. General Motors and Fisher adopted the contract in spite of these problems
because they expected it to function satisfactorily in combination with a self-enforcement
mechanism. That is, Fisher and General Motors expected their contractual relationship to
remain within the self-enforcing range defined by each transactor's reputational capital.
As noted above, the contract, in fact, worked well for more than five years and, under
normal circumstances, would have remained self-enforcing.

The Fisher“GM case vividly illustrates that the use of imperfect contract terms solves
non-performance problems in some states of the world but creates non-performance
problems in other states of the world. If General Motors' demand for Fisher's auto bodies
had not grown so dramatically after 1925 increasing Fisher's short-run gains from
nonperformance, the contract Fisher and General Motors had adopted, although
imperfect, would have remained self-enforcing. The gains to Fisher from taking
advantage of the contract would have remained less than Fisher's reputational capital
and, therefore, the hold-up potential associated with the cost-plus contract terms would
not have mattered. It was only after General Motors' demand for Fisher's bodies and
Fisher's required specific investments increased late in the contract term that the
contract's "inefficiencies" were acted upon by Fisher. It was only then that Fisher found
itself outside the self-enforcing range, where Fisher's reputational capital, or the private
sanction that could be imposed on Fisher by General Motors, became less than Fisher's
short-term gain from not performing. Fisher then found it profitable to violate the intent of
the contractual understanding by taking advantage of the imperfect terms of the
agreement, refusing to make the necessary capital investments required to produce
bodies efficiently for General Motors.[13]

Fisher and General Motors presumably recognized when they entered their contractual
relationship and made their specific investments that their reputational capital was limited,
that the written contract terms they had chosen were imperfect and incomplete and,
therefore, that there was some probability the contract would fail and a hold-up would
occur if changes in market conditions moved either of them outside the "self-enforcing
range," as occurred during 1925 when General Motors' demand for the bodies supplied
by Fisher greatly increased. At that point the pressure placed on the imperfect
contractual agreement used to facilitate self-enforcement became greater than the
contract could withstand and the Fisher Body“GM relationship moved outside the self-
enforcing range.

Klein (1996).

This is the basis of Oliver Williamson's definition of opportunism. He states that "[b]y
opportunism I mean self-interest seeking with guile. This includes but is scarcely limited to
more blatant forms, such as lying, stealing and cheating. Opportunism more often involves
subtle forms of deceit More generally, opportunism refers to the incomplete or distorted
disclosure of information, especially to calculated efforts to mislead, distort, obfuscate, or
otherwise confuse" (Williamson 1985, p. 47).

In particular, Fisher refused to build an important body plant close to a GM production
facility in Flint, Michigan. Fisher would not be expected to make the new, large specific
investments required by General Motors without a renegotiation (e.g. extension) of the
contractual arrangement. But as part of this renegotiation Fisher took advantage of its existing
GM contract to engage in a hold-up. See Klein (2000).
6 Vertical integration
The Fisher Body-GM analysis explains why transactors, when choosing the imperfect
contract terms that govern their self-enforcing relationships, are more likely to use a
vertical integration type of contractual arrangement when they expect future market
conditions to be highly variable. When the uncertainty of future market conditions
increases, the value of the hold-up potential present in every imperfect contract also
increases. Parties entering contractual relationships can be thought of as buying and
selling what amounts to options related to the probability of a hold-up occurring. As in
standard options pricing theory, the values of these options increase as the value of the
ratio of the underlying asset price increases relative to the exercise price (in our case, as
the value of the hold-up potential increases relative to the transactor's reputational
capital), and as the variance per period of the asset price multiplied by the number of
periods increases (in our case, as the variance of underlying market conditions multiplied
by the length of the contract increases).[14] Since transactors wish to avoid the costs
associated with hold-ups even if they are not risk averse, this makes vertical integration,
with its increased ability to make flexible post-contract adjustments, more likely.
The alternative to vertical integration (in cases where the parties have made specific
investments and W1 is greater than W2, i.e. where the relationship cannot solely be self-
enforced) is an explicit long-term contract. The greater the uncertainty of future market
conditions, the more likely it is that the arrangement defined by this imperfect long-term
contract and the transactors' reputational capital will move outside the "self-enforcing
range." In these circumstances the increased flexibility and control transactors gain from
not using a rigid long-term contract to supplement their insufficient reputational capital is
a primary economic advantage of vertical integration. Transactors using vertical
integration avoid the rigidity costs of long-term explicit contracts illustrated by the Fisher-
GM case, at the cost of increased incentive inefficiencies associated with vertical
integration (that presumably cannot be self-enforced because of the difficulty of
detection). That is, vertical integration increases W1, but makes the relationship more
flexible and, therefore, self-enforcing (or decreases W1 relative to W2) in a wider set of ex
post circumstances.

This analysis highlights the shortcomings in the pioneering Grossman and Hart model of
integration.[17] While this model has the advantage of taking the incompleteness of
contracts seriously, it does not consider the key aspect of the contractual arrangement
we identify with the firm, namely that it involves less explicit contractual specification and
more flexibility. Moreover, even within the context of this model, the primary conclusion
that unspecified residual rights (what Grossman and Hart identify with the firm form of
contract) should be allocated to the transactor that will misuse the rights the least makes
sense only if we ignore self-enforcement. Because contract terms are not designed
solely to minimize inefficiencies, how asset ownership is allocated is not determined
independent of the reputational capital of the parties. Transactors must also take account
of the reputational capital of the parties, in addition to their incentives to take advantage
of residual rights not to perform, to determine who will be the owner of a particular asset.
For example, even if ownership by one transactor causes increased gains from non-
performance, this does not imply that the transactor is not the correct owner of the asset
if its reputational capital is higher.

See Klein (1996).

This effect of increased uncertainty on vertical integration when transactors are not risk
averse is distinct from the effect increased uncertainty may have on increased contractual
incompleteness. If the parties are risk neutral, increased incompleteness, in itself, has no
effect on vertical integration in the standard property-rights (Grossman and Hart-type)
approach to the theory of the firm. If the parties are risk neutral, increased uncertainty and
increased contractual incompleteness does not affect organizational form (or which party
owns which assets) in these models because the models ignore self-enforcement.

Klein (1988, 2000).
A summary of the continuing literature in the Grossman and Hart tradition can be found in
Hart (1995).

Klein and Murphy (1997).
7 Conclusion
To increase our economic understanding of contracts, it is necessary to get one's hands
dirty and discover how particular contracts actually work in practice. However, to make
progress in this empirical analysis one must have an appropriate organizing framework.
In particular, one must recognize that the goal of contractual specification often is not to
create optimal incentives on some imperfect court-enforceable proxy for performance.
Rather than focusing solely on these direct incentive effects of contract terms now
emphasized in the incomplete contracting literature, economic analysis of contract terms
must also consider how contract terms may be used to facilitate self-enforcement.
Contractual arrangements can be fully understood only by recognizing that transactors
use court-enforced imperfect contract terms, including vertical integration, as a
complement to their limited reputational capital in order to make a particular relationship
self-enforcing over the broadest range of likely post-contract market conditions.
Chapter 4 was originally published as "The Role of Incomplete Contracts in Self-
Enforcing Relationships," in Revue d'Economie Industrielle (92, 2000).
1. Macaulay (1963).
2. See Klein and Leffler (1981) for an early example of such a model.
3. An important paper by George Baker, Robert Gibbons and Kevin Murphy
(1999) does this by adding a self-enforcement mechanism to the
standard Grossman and Hart (1986) principal-agent model of the firm.
Although Baker, Gibbons and Murphy provide a number of valuable
insights regarding the operation of the self-enforcement mechanism in
this context, they do not identify what I consider to be the key advantage
of vertical integration that facilitates self-enforcement discussed below,
post-contract flexibility.
4. I am assuming for analytical and expositional simplicity that the court
enforces written terms and does not enforce unwritten, understood terms.
While courts in practice interpret both written and unwritten terms when
enforcing contractual agreements, under English common law the
amount of discretion exercised by courts with regard to unambiguous
written terms is generally limited. In any event, as transactors cover
additional contingencies with explicit imperfect contract terms, it is
reasonable to assume that after some point there is an increased
likelihood the court will effectuate a hold-up by enforcing the contract in a
manner that is contrary to the parties' contractual understanding.
5. Klein, Crawford and Alchian (1978) and Klein (2000). These transitional
hold-up costs conflict with the costless ex post renegotiation assumption
generally made in the incomplete contracting/property-rights literature
that has developed from the pioneering work of Grossman and Hart (n.
3). These models assume, contrary to what occurred in the Fisher-GM
case, that in cases where a potential hold-up exists, ex post
renegotiation of the contract instantaneously and costlessly takes place,
so that, after a lump sum is paid to the transactor that can engage in the
hold-up, price and cost quickly move to the efficient level. Therefore,


. 2
( 18)