. 8
( 18)


managers. This theory, originally founded on the property-rights theory and on the
agency relationship concept borrowed from the principal-agent approach, is aimed
towards a theory of coordination and control applied to organization management and
centered on managers. It applies, in particular, to organizational architecture and
corporate governance.

As Jensen and Meckling specified (1998, p. 8), their goal was to build a theory of
organizations: "Our objective is to develop a theory of organizations that provides a clear
understanding of how organizational rules of the game affect a manager's ability to
resolve problems, increase productivity, and achieve his or her objective."
Since their first writings (in particular, Jensen and Meckling 1976; Jensen 1983), the
founders of this theoretical current had clearly taken care to mark their difference in
comparison to the principal-agent theory, as much from the point of view of their
objectives as of the methodological approach used. However, it appears that the
specificity of PAT often remains ill-perceived. It is either regarded as a non-formalized
alternative of the principal-agent theory, or it appears as a component of the contractual
theories, which is less generally applied compared to that of the transaction-costs theory
(hereafter TCT). These two interpretations are both, if not erroneous, at least simplifying;
they are explained, in particular, by the methodological differences separating these
various currents as well as by the fact that a major part of the work, based on PAT, was
published in accounting and financial reviews, relatively marginal in the field of
organizational economics, such as the Journal of Financial Economics or the Journal of
Accounting and Economics. This origin, which was deeply marked by finance because of
the relative isolation of the paradigms and disciplinary fields, led to the support of certain
misunderstandings, even a certain ignorance of the central PAT characteristics or its
many contributions, that are, however, important for organization and management

This chapter has three objectives. First of all, it points out the central components of PAT.
Second, it aims at rectifying the reading of this theory, comparative to TCT, proposed
by Williamson (1988a). Third, it tries to show the variety of questions tackled by PAT in
fields as diverse as finance, accounting, management control, human resources
management, and corporate governance. The first concern of the founders of PAT was
to offer an analysis framework to managers enabling them to understand the impact of
organizational structure on performance and to guide their actions and decisions. The
reading[2] thus suggested particularly attempts to show the continuity and the originality
of the project of the Business School of the University of Rochester,[3] as much in its
operational concerns, and its sources of inspiration, as in the methodology that it

The differences between PAT and the principal-agent theory will not be specifically analyzed.
These theories differ in particular in their underlying rationality models, in the basic variables
of modeling, and in the methods used.

A more thorough analysis of PAT can be found in Charreaux (1999).
PAT was initially developed by Jensen and Meckling at the Business School of the
University of Rochester where Meckling was the Dean. This theory could also in future be
referred to as being of the Rochester-Harvard School, Jensen having left for Harvard
2 The principal components of PAT
To highlight the original place of PAT,[4] it is necessary to quickly reconsider its main
ingredients and its principal theoretical message, in other words, the modeling of
organizational architecture and the distribution of the economic activities that it proposes.
Let us clarify that if PAT has evolved, its essential components are already present, to
various degrees, in the seminal articles of Jensen and Meckling (1976), of Jensen (1983),
and of Fama and Jensen (1983a, 1983b).

2.1 The building blocks of the theory

Jensen positions PAT as an "integrated" theory of organizations, directed at joining
together two distinct research currents: the research into economic tradition centered on
market operations, and those associated with the fields of psychology, sociology,
organizational behavior, anthropology, and biology, directed at explaining human
behavior, as much on the individual level as on the social one. Thus, Jensen's research
group includes a personality like Argyris, well known for his work in organizational
learning. PAT is thus conceived to be purposely integrative: it must allow for
simultaneous embracing of organizational and market phenomena. In this sense, as well
as in its multifield basis, it is close to the TCT, which moreover originally constituted one
of its sources of inspiration.
PAT, according to the presentation made by Jensen (1998), includes four fundamental
building blocks: a model of human behavior, the costs related to the transfer of
knowledge, the agency costs, and the organizational rules of the game (figure 15.1).

Figure 15.1: The building blocks of PAT source: (Jensen, 1998, p. 3)

Block 1: The model of human behavior
The article that Jensen and Meckling (1994) devoted to "the nature of man" includes an
accurate presentation of the Resourceful, Evaluative, Maximizing Model (REMM). This
model falls under the paradigm of rational approaches of organizations. It is based on
four assumptions:
Individuals are concerned with all that is a source of utility or disutility
and are "evaluators." They are in a position to trade off between the
various sources of utility and their preferences are transitive.
Individuals are insatiable.
Individuals are maximizers. They maximize a utility function, the
arguments of which are not exclusively financial, while under
constraints. The constraints may be cognitive and the choices made
take into account the acquisition costs of knowledge and information.
Individuals are creative and know how to adapt; they are in a position
to foresee the changes of their environment, evaluate the
consequences, and respond by creating new opportunities that they
are able to evaluate.

The design of rationality within PAT is close to that of Williamson, in other words, of
bounded type, while remaining "calculative" under cognitive constraints. The rational
expectations hypothesis used to build the model for optimal financial structure contained
in Jensen and Meckling (1976), which implies a less bounded rationality, seems to have
been adopted only by a concern for simplifying the analysis. It cannot be regarded as
representative of REMM.

This rationality is socially "located." The social norms represent constraints and govern
the actions; according to Jensen and Meckling (1994) "They serve as an external device
that aids in the storage of knowledge about optimal behavior. In addition, they represent
a major force for teaching, learning, disciplining, and rewarding members of a group,
organization or society." They are supposed to evolve according to the changes in the
environment and to those of knowledge, transforming individuals' theories, and
influencing their actions by modifying all opportunities, costs, and gains associated with
the actions. If the aspects related to social embedding are reflected in the individual's
actions, they, however do not dictate them. The status of the norms and institutions is
similar to that they hold in TCT: it is a matter of parameters. However, like TCT, the
theory does not allow an explanation (contrary to North's, 1990, institutional theory) of
the institutional changes. The norms are supposed to evolve when they impose costs
that are too high in the new environment, but the mechanism controlling their
development process is not studied.

The possibility for the individual to be creative and to adapt gives the theory an implicitly
dynamic character. The recognition of the adaptive character of behaviors allows us to
take into account the active neutralizing behavior of certain mechanisms, central, for
example, in the manager's entrenchment strategies, as well as, for that matter, the
positive role that the latter may play in building all of the opportunities.

More recently, Jensen (1994) proposed an extension to REMM by adding the Pain
Avoidance Model (PAM). Its goal is to explain, on the one hand, that in certain cases
individuals acted, with defensive concern, in an irrational manner (from a
consequentialist point of view) by making decisions that are apparently (for a neutral
observer) contrary to their welfare and on the other hand, the limited character of the
individual's learning capacity, in other words, of the adaptive behavior in view of the
mistakes made. The individual would avoid changing their mental (or perceptual) model
because of the resulting psychological costs ("the pain"). This dualistic model of human
behavior would find its origin in the lessons from cognitive sciences and behavior. Let us
specify that it may be possible, according to us, to avoid this problematic dualism from
the methodological point of view. All that is necessary is the interpretation of the PAM
model as an extension of the REMM model, (in a "calculative" sense), by invoking the
high costs (in psychological terms) for individuals, related to the change of their
perceptive models by learning.

Block 2: The costs of transferring knowledge between actors
Although Jensen uses the terms information and knowledge indifferently, knowledge is at
the center of PAT that attributes it a determinant role in the constitution of the
organizational performance. Efficiency basically depends on the capacity of the
organization members to use the "relevant" knowledge, valuable in decision-making. The
distinction, established by Hayek, between "general knowledge" and "specific
knowledge," plays a central role. The solution to the organizational problem consists of
finding the least expensive means to put the relevant knowledge at the disposal of the
decision-makers. This availability, consisting of colocalizing decision rights and specific
knowledge, can be achieved in two ways: (1) either, in a centralized manner, by
transferring knowledge to those holding the decision rights; (2) or, in a decentralized
manner, by transferring the decision rights to those having the knowledge. The choice
depends on the respective costs of transferring knowledge and the decision rights.
However, because of the importance of the non-transferable specific knowledge, the
centralized solutions most often fail. The alignment of decision rights and the localization
of knowledge goes beyond the traditional centralization/ decentralization debate, insofar
as specific knowledge is distributed on the whole hierarchy. The true question relates to
the nature of decision rights to be centralized or decentralized.

Block 3: Agency costs
The decentralized allocation of decision rights creates agency relationships, sources of
conflicts of interest, and of agency costs. The organizational must be conceived so as to
reduce these costs creative of inefficiency, by implementing incentive and control
systems intended to align the interests of the agents with those of the principal(s).

In PAT, the representation of the agency relationship - that would qualify more precisely
as a "cooperation relationship" - evolves according to the analyzed problem. Beyond the
traditional principal-agent asymmetrical relationship, for example between shareholders
and managers, the relationship in certain models becomes bilateral or "dyadic" where
the two parties can alternatively be regarded as principal or agent. Overall, the general
formulation of the organizational problem in terms of efficiency and the representation of
the organization as "a nexus of contracts," or rather a "contracting nexus," leads to the
ability of going beyond these restrictive representations to locate the problem of the
reduction of agency costs at the level of the simultaneous management of all the
relationships between actors that can overlap and be interdependent. The same
organizational mechanism, for example the board of directors, can be used to
simultaneously manage conflicts of interests between shareholders and managers, but
also between creditors and shareholders or between shareholders and employees.

Block 4: Alienability and the organizational rules of the game
The decision rights refer to the use of assets, of resources. They are, in fact, property
rights, that Jensen and Meckling (1992) break up into two components: the actual
decision right (the right to use the firm's assets) and that of alienating this decision right
and appropriating the product of the transfer ("alienability" of the right).

Alienability is the basis for the existence of the market system, interpreted as a system of
transferable rights. On a market, the colocalization of knowledge and decision is carried
out by way of a transaction directed by prices, by the alienation of the decision right
associated with a voluntary exchange. The decision rights are acquired by those who
attribute the greatest value to them; they are agents who are in a position to use them for
the best, because of the specific knowledge they hold. In the absence of externalities to
the exchange, the colocalization by the market is efficient; it is not necessary to introduce
a control mechanism. Only alienability solves the control problem by way of price,
simultaneously representing a measurement of performance and an incentive system.
On the other hand, the absence of alienability leads to the reappearance of the control

For the intra-firm transactions, the transfer of decision rights does not accompany that of
alienability. This leads to two consequences: (1) effective colocalization is no longer
carried out in an automatic and decentralized manner; (2) an automatic system no longer
exists for performance and incentive measurement, leading agents to use their decision
rights in the interest of the organization. In this case, it is necessary to turn to a
hierarchical authority to solve this problem as well as to various organizational
mechanisms. Organization is explained only when the handicaps related to the absence
of alienability are compensated by certain amount of advantages: for example,
economies of scope and scale, reductions in transaction costs that cannot be obtained
by independent agents, but also the "cognitive" argument put forth by Demsetz (1988):
the firms allow us "to economize knowledge," in particular, because of the long-term
character of the employment relationship.

2.2 A theory of the organizational structure and distribution of the organizational

The construction of PAT, based on optimal use of specific knowledge, results in the
suggestion of two complementary application fields: (1) the internal field of organizational
architecture and (2) the external field relating to the distribution of the organizational
2.2.1 Organizational architecture
The argument presented results in the proposition of a theory of organizational
architecture founded on the allocation of decision rights within the organizations. This
allocation, which does not rest on the voluntary exchange of rights between actors, is
carried out by way of an organizational policy. The rights are distributed by the managers
and respect is assured by the incentive and control systems put in place taking into
account the institutional environment. The distribution results from the arbitration
between the costs related to the misuse of specific knowledge (insufficient
decentralization of decisions) and those associated with the conflicts of interest (owing to
decentralization). The distribution, incentive, and control systems constitute the
"organizational rules of the game."

The theory of organizational structure is thus articulated around two dimensions
constituting the base of the "taxonomy" central to PAT:
Allocation of decision rights within the organization; this allocation may
involve a distribution of decision rights between the decision
management rights, which include the rights to initiate and implement
the allocation of resources, and decision control rights that concern
ratification and the monitoring of decisions.
This distribution of rights corresponds to the decision-making process in
organizations as modeled by Fama and Jensen (1983b, p. 303), in four steps:
initiative, ratification, implementation, monitoring (performance measurement of
the agents, rewards, and punishments). The initiative and implementation
functions, most often entrusted to the same category of agents, are regrouped to
form the function of "decision management." Those of ratification and monitoring
are associated in the definition of the "function of control" (decision control).
The design of the control systems, while distinguishing:
o the evaluation and performance measurement system
o the incentive system, which allows us to specify the
relationship between the performance measurement
and its consequences in terms of rewards and
punishments; it is the coherence and the
complementarity between these two (or three)
dimensions that is supposed to determine the level of
organizational efficiency.
2.2.2 Distribution of organizational forms
PAT also allows us to understand the distribution of organizational forms. Fama and
Jensen (1983a, 1983b) propose an explanation of the various organizational forms that
relies on the central role of specific knowledge and the minimization of agency costs.
Their argument comes from an analysis of the contracts considered as central in any
organization, which are the contracts that specify, on the one hand, the nature of the
"residual claims" and, on the other hand, the allocation of the steps of the
decisionmaking process between agents. This results in establishing predictions for the
distribution of the economic activity according to ownership structure, characterized by
the distribution of residual claims that govern the bearing of risk.

By associating the concept of organizational complexity with that of dispersion of specific
knowledge and by studying the efficiency of the various functional configurations
(decision, control, and risk-bearing), Fama and Jensen (1983a, p. 304) construct two
fundamental hypotheses:
The separation of residual risk-bearing functions from decision
management leads to decision systems characterized by a separation
of the management and decision control functions
The combination of the functions of management and decision control
in the hands of a limited number of agents results in a concentration of
residual claims with these same agents.

Fama and Jensen find a confirmation of their theory in the fact that almost all the
organizations, characterized by a separation of the functions of decision management
and risk-bearing, present the same structures of decision and control.

The texts which allow us to best account for the evolution and for the current state of PAT
are, in addition to Jensen and Meckling (1998), Jensen (1998), and Brickley, Smith and
Zimmerman (1997).

The first version of this article (Meckling 1976), written at the beginning of the 1970s, is
contemporary with the 1976 article.
3 PAT compared with TCT: a second reading of Williamson
While there is frequent opposition to contractual theories, of which the central argument
is efficiency and the evolutionary theories in which cognitive aspects play an essential
role, PAT, that bases efficiency on optimal use of knowledge, occupies in a certain way
an intermediate position. This specificity of PAT implies, in particular, that it should not
be confused with the TCT.

Many features attributed to PAT, as well as the distinctions contrasting it with the TCT,
are often the outcome, because of Williamson's notoriety, of the comparison he carried
out, in "Corporate Finance and Corporate Governance" (1988a). This comparison
constitutes a useful starting point in understanding the origin of certain
misunderstandings concerning PAT as well as its true nature. Even if PAT and TCT have
evolved since then, the main foundations of the two theories had already been explicitly
stated in 1988 “ in particular, the role of specific knowledge and the presentation of the
main arguments making PAT a theory of organizational architecture, and distribution of
the organizational forms occupying a central place in the articles of Jensen (1983) and of
Fama and Jensen (1983a, 1983b) quoted in the references of Williamson's article. The
developments borrowed from subsequent articles on PAT only allow for reinforcement of
the argument presented. The final objective of this second reading of Williamson's article,
however, is not to make an exhaustive comparative assessment of the two theories, but
to propose an analysis of PAT different from that of Williamson.

3.1 The common points: a critical discussion
Williamson identifies two principal common points: the managerial discretion framework
and the contractual efficiency.
3.1.1 The first common point: the managerial discretion framework
The framework of managerial discretion regroups the conception of the nature of the firm
and that of the man (bounded rationality, opportunism and risk neutrality).

Concerning the nature of the firm, in both theoretical bodies, according to Williamson,
there is a rejection of the neoclassical conception of the firm as a function of production.
In TCT, the firm appears as being of a particular governance structure, including a
coordination directed by hierarchy as opposed to spontaneous coordination associated
with the market. As for PAT, it would rest on the conception of a nexus of contracts that
would also constitute an acceptable representation in the TCT.

These conclusions need to be moderated. There is not necessarily a contradiction
between the representations of the firm as a nexus of contracts and as a function of
production. Thus, Jensen and Meckling (1979) explicitly represent the firm as an
enlarged function of production taking into account the organizational choices (the rules
of the internal game) among the factors of production and conditioned by the institutional
framework, when, simultaneously, they resort to the nexus of contracts metaphor.
Should we see an inconsistency? The nexus of contracts concept means nothing other
than that manager contracts in a centralized manner in the name of the firm with all the
stakeholders, the partners who supply production factors or customers. Therefore, it is
not contradictory to claim on the one hand, that the manager optimizes the production for
others by taking into account the internal rules of the game as a production factor and in
an institutional context and, on the other hand, that the nexus of contracts management
is carried out so as to minimize agency costs by choosing an adapted organizational
architecture. The first approach applies to "external" analyses directed towards the
comparison of organizational forms, the second to "internal" analyses focused on
organizational architecture.

The TCT approach, based on the firm as a mode of governance founded only on
hierarchy, appears different from the representation that implies contract and production
management within PAT. In the latter, the nexus of contracts does not necessarily imply
an exclusive recourse to a mode of directed coordination. The only requirement is that
the coordination be carried out so as to ensure the best use of the specific knowledge:
therefore the firm can certainly use directed coordination, but it can also rely on
spontaneous coordination (that does not compare to the simplistic outline of the price
mechanism) or on concerted coordination. This plurality of coordination modes within the
firm is, in particular, retained by Demsetz (1988, 1995), an author who greatly inspired
Jensen and Meckling (1992). Moreover, it is useless to insist on the fact that the
representation of the firm compared to the directed coordination is not easily compatible
with the argument based on the optimal use of specific knowledge.

The conception of the nature of the firm seems rather different in the two theories.
Reduced to the status of a particular mode of governance (directed coordination) in the
TCT, the firm constitutes a complex system in PAT, the arrangement of which must allow
for the best use of specific knowledge. The frequent references made to Demsetz may
even lead to a conclusion that Jensen and Meckling (1992) would not reject the definition
that he gives of the firms: "repositories of specialized knowledge and of the specialized
inputs required to put this knowledge to work" (Demsetz 1988, p. 171).

Let us consider now the conception of the nature of the man retained in the two theories.
The conception of rationality that the TCT implies is a "calculative" bounded rationality,
allowing for a long-term view. It is of a consequentialist nature and not procedural. The
presentation of the rationality models associated with PAT (a dualistic model: REMM and
PAM) shows, at least for the REMM model, a very similar conception. The individuals
evaluate, maximize (under cognitive and institutional constraints), are creative, and
adapt; in particular they create new sets of opportunities to respond to environmental
evolution. The insistence on the adaptive character, and more recently on the learning
phenomena (by way of PAM, in particular), leads, however, to the conclusion that PAT
will now retain a "broader" and more organic conception of rationality.

The opportunism hypothesis is often quoted as being central to the TCT. In fact, it does
not imply that individuals are systematically opportunistic, but only that they may be. In
PAT, opportunism does not have a particular role; its presence induces only an increase
in agency costs. On the other hand, its absence does not induce the elimination of
conflicts of interest. The fact that individuals have unequal access to information or have
different cognitive models is enough to justify the existence of obstacles to cooperation
and conflicts of interests.

Finally, according to Williamson, the risk neutrality hypothesis is not common to TCT and
PAT, but to TCT and the principal“agent theory. In TCT, the justification of this
hypothesis is of an instrumental nature; it allows us to focus the analysis on the most
essential aspects of efficiency. PAT does not retain this hypothesis for it would lead to
the inability to explain diversification behaviors. The risk aversion hypothesis, on the
contrary, is retained; it allows for the explanation of the distribution of activities between
the various organizational forms. In particular, it justifies the important role assigned to
the risk-bearing function.
3.1.2 The second common point: contractual efficiency
Contractual efficiency ("efficient contracting") constitutes the second point common to
PAT and TCT. The sources of efficiency would, according to Williamson, be within the
capacity of organizational forms "to economize rationality" and to protect transactions
against the risks from opportunism. He adds that PAT is concerned mainly about the
second aspect and pays almost exclusive attention to the contractual aspects ex ante. A
reading, faithful to the spirit of PAT, focused on the central argument for the use of
knowledge, leads to different conclusions. Concerning the sources of efficiency, the
place attributed to the cognitive component in PAT leads, on the contrary, to the
conclusion that both sources of efficiency are being considered. As for the exclusive
attention paid to these aspects ex ante, it concerns only particular models, such as the
financial structure model proposed by Jensen and Meckling (1976), which rests on the
rational expectations hypothesis and remains very close to the traditional models of the
principal“agent theory. In the majority of PAT developments, this hypothesis is
disappearing and the aspects ex post are being taken into account. Moreover, it is
important to specify that in PAT, contrary to TCT, the conflict analysis is not carried out
transaction by transaction, which is not very compatible with the representation of the
firm in terms of a nexus of contracts. The management of conflicts is conceived globally,
on the level of the considered organizational system, "the joint production team" (that is
to say, all the cocontracting production factors), to employ the term used by Alchian and
Demsetz (1972); it is this representation that constitutes the very foundation of PAT as a
theory of organizational architecture.

3.2 The differences: a critical debate
If the common points identified by Williamson need to be moderated, his analysis of the
differences also seems questionable. The main differences relate to the analysis unit, the
distinction between agency and transaction costs, and the favored organizational
dimension. To these three differences, he adds two others considered as secondary: the
differences in selection processes and the neutrality of the nexus of contracts.
3.2.1 The main differences: the analysis unit, the nature of costs,
and the central organizational dimension
Within TCT, the transaction is the basic unit of analysis. The study of organizational
forms is done according to the transaction features. The explanation is based on the
alignment of transactions/modes of governance in order to minimize transaction costs.
The extent of the assets' specificity plays a central part. On the other hand, in PAT,
according to Williamson, the central unit of analysis would be the individual, that would
result in neglect of the transaction dimensions.

If the transaction is really the basic unit of TCT, the fact that it concentrates the attention
should not mask the important role that actors preserve in this theory, based on
methodological individualism. In fine, it is the actors who bear the costs and their
analysis is done obviously from the point of view of the transaction participants. Let us
recall furthermore that TCT gave rise to a certain number of criticisms, in particular, the
imprecise character of the transaction concept and the central role attributed to the
causal link between asset specificity and the choice of governance mode.

The status of PAT is rather similar to that of TCT. Although the theory also rests on
methodological individualism, that does not imply that the individual is the basic unit of
analysis. As in any modeling of organizational phenomena of this kind, the analysis on
the individuals' level is of a relational nature and is located structurally within a system.
Thus, in PAT, the basic unit is the agency relationship. If it sometimes takes the status of
an asymmetrical relationship (shareholders with managers, for example), conversely in
the most complex versions, the analysis relates to the whole of the nexus of contracts
and organizational architecture. The representation of this agency relationship is
contingent on the organizational phenomenon studied, which can be, according to the
case, the board of directors, the total quality management (TQM)[7] or the financial
structure (regarded as a particular organizational mechanism). For example, for the
board of directors, the unit of analysis may be the agency relationship between
shareholders and managers, but can be enlarged to become the nexus of relationships,
shareholders/managers/employees/financial creditors/other stakeholders: the board of
directors is therefore explained according to its capacity to minimize the agency costs on
the whole nexus of relationships. In this approach, relationships between actors are
overlapping; there is no single causality link, for example, between specificity and the
mechanism of governance to be implemented. Thus, if the specificity of the manager's
human capital implies that the board of directors may be interpreted as a mode of
governance allowing for its preservation, PAT extends this kind of reasoning while
claiming that the manager decides on his investment effort in specific human capital
according to the nature of controls he submits to and, in particular, of his own capacity to
control the board. The optimal solution, finally, depends in this context on the capacity to
best use the manager's specific (present and future) knowledge.
The second difference referred to relates to the nature of costs: agency costs against
transaction costs. Relying on the categories of agency costs (monitoring costs, bonding
costs, residual losses) identified by Jensen and Meckling (1976), Williamson concludes,
regarding the only model of financial structure, that the agency costs are exclusively ex
ante, and therefore restrictive. He then emphasizes that transaction costs also cover the
costs ex post that, for him, play the most important role. This reading is also debatable
because, as already mentioned, the hypothesis of rational expectation set out by Jensen
and Meckling in their financial model has only one instrumental character specific to this

In addition, in the case of its restrictive character it takes the side of the concept of
transaction cost rather than that of agency cost, because of the generality of the residual
loss concept. From its very definition, the transaction cost is associated with a particular
transaction (the basic unit). The same does not apply for the residual loss concept that
constitutes the loss of value in regard to an ideal hypothetical situation associated with
the absence of conflicts and an optimal use of specific knowledge. Thus, this concept is
not dependent on one single transaction; it takes on significance only with regard to the
organizational phenomenon, the subject of the research.
Finally, the third principal difference would relate to the favored organizational dimension.
According to Williamson, because of the exclusive attention that PAT allegedly pays to
the mechanisms ex ante, this theory would neglect the explanation of the organizational
mechanisms as modes of conflict resolution ex post. Thus, PAT would not be concerned
with internal organization, focusing only on the residual claims. Such a conclusion is
wrong. On the one hand, agency costs ex post are not ignored by PAT and on the other
hand, since its first developments, this theory was conceived to cover the internal
organizational architecture and to simultaneously call on the external and internal
governance mechanisms, relating to the market and organization.
3.2.2 The secondary differences: the nature of the processes and
the neutrality of the nexus of contracts
According to Williamson, the modes of the natural selection process would differ
between the two theoretical bodies. While the TCT, based on the remediability criterion,
would be based on a "weak" form of the natural selection principle “ only the more
comparatively adapted, "the fitter," and not the most adapted (in the absolute sense of
the fittest) would survive “ it would be different in PAT. An attentive reading shows,
however, that this difference seems to be illusory. Jensen and Meckling (1976, 1992),
following the example of Williamson, are inspired by Coase and Demsetz. When Jensen
(1983) evokes the concept of "the fittest," he does so in a context of constrained
efficiency, in other words, relative to the existing organizational forms. He does not
exclude the individual being creative, that more efficient organizational forms may

As for the nature of the adjustment processes, even if the treatment offered by TCT may
appear more thorough, one cannot conclude, as Williamson does, that the adjustments
within PAT are all comparable to the market mechanism directed by prices. Since 1976,
Jensen and Meckling invoked compensation, audit, and management control systems.
More recently (Jensen and Meckling, 1992, p. 261), they wrote, in a very explicit manner,
that the allocation and the implementation of decision rights in organizations depend on
organizational policy, not on voluntary exchange[9] between actors. Within PAT, the
adjustment mechanisms are based as much on hierarchy as on markets, or rather on the
various forms of coordination (directed, concerted, spontaneous). Moreover, they must
not be analyzed with regard to a single transaction, but relatively with the whole nexus of
contracts, conceived as a complex equilibrium system.

The last difference would relate to the "neutrality" of the nexus of contracts, in the sense
that all the contractual relationships are simultaneously determined, according to a
complex balanced process. From Williamson's point of view, even if this hypothesis of
neutrality is also shared by TCT, the structure of the latter would allow us, contrary to
PAT, to take into account the strategic behaviors which can break up this neutrality, at
least temporarily. For example, the manager, because of his central position, can
broadcast information in a selective manner, profitable for him. Such a conclusion seems
equally invalidated by the numerous developments of PAT, that explicitly takes into
account the managers' entrenchment strategies to analyze the organizational
mechanisms that adapt to assure a restoration of equilibrium.
3.3 A revised comparative assessment

The comparative assessment drawn up by Williamson thus seems to be a rectification.
PAT is often presented in a restrictive manner, apparently because of a hasty
comparison with certain models of the principal“ agent theory or with the 1976 model of
financial structure. In particular, the central role played by specific knowledge was not
perceived by Williamson, the latter having placed, in his comparative table (1988, p. 575),
a question mark in the heading "focal dimension" (translated as a fundamental variable
of efficiency).
The revised comparative assessment may take the shape of table 15.1. One may
conclude from table 15.1 that PAT constitutes a more general and flexible analytical
framework, in particular because of the following aspects:
no central role is given to opportunism and asset specificity, which are
only some dimensions among others in PAT
adjustments are made at the level of the organizational system, the
nexus of contracts, and not on the transaction level. PAT does not retain
the direct causality link that is retained by TCT between the
characteristics of the transaction and the mode of governance.
Adjustments are made by displacement of the complex organizational
equilibrium that aims for an assurance of optimal use of knowledge while
minimizing agency costs.

Table 15.1: PAT versus TCT: a revised comparative assessment

Common Dimensions PAT TCT

Unit of
"Calcluative" Agency Transaction
bounded relationship
rationlity (more (with
organic in multiple
PAT) representati
Principle of Speciffic Asset
efficiency knowledge specificity
Focal cost Residual Maladaptation
concern loss (ex ante cost (more
and ex post focused on ex
dimensions) post)
Contractual Ex ante and Mainly ex
focus ex post post
governance governance
(internal and

In fact, the majority of the interpretation errors seem attributable to an assimilation of TPA to
this particular model that constitutes but one aspect of the 1976 article. This article
furthermore contains more general developments of PAT which shows that this theory has a
much more ambitious vocation.

For example, Wruck and Jensen (1994) developed a very innovative analysis of TQM.
Bonding costs are those associated with the mechanisms allowing the agent to reassure the
principal on the credibility of his commitments, for example, the costs associated with a
voluntary audit.

"The assignment and enforcement of decision rights in organizations are a matter of
organizational policy and practice, not voluntary exchange among agents."
4 Contributions and influence of positive agency theory
The central place that the manager occupies in PAT predestined this theory to play a
determinant role in the development of management sciences. Reading the scientific
journals, often considered as being among "the best" in their respective fields “ for
example, the Journal of Financial Economics, the Journal of Accounting and Economics,
or the Journal of Strategic Management “ allows us to grasp the extent of the influence
of this theory on management sciences. If PAT initially appeared as a theory of finance,
it quickly extended beyond the financial field to propose new analyses in accounting,
management control, human resources management, manufacturing, or marketing
management. Some works, for example, that of Watts and Zimmerman (1986) in
accounting, revolutionized research in their field. In addition, PAT is at the origin of new
theoretical fields such as "corporate governance." The corporate governance theory[10]
permitted, in particular, a renewal or prolongation of the analyses regarding the
comparison of the performance of economic systems (for example, the traditional debate
opposing public to private companies) and the various organizational forms (companies,
mutuals, cooperatives ).

Accompanied at times by TCT in which some features can be easily integrated, PAT
became one of the main "grammars" used in management sciences. In particular, it
allowed the establishment, if not restoring of, links between disciplinary fields, which had
often evolved independently.

Rather than trying to do a survey of the multiple contributions of PAT to the various fields
of management sciences, it seems preferable to examine the presentation made by
Jensen and Meckling (1998, p. 17) of the four main axes constituting the PAT research
program. This presentation will be complemented by that of another contribution of PAT,
the enrichment of traditional methodologies through the use of clinical studies.

4.1 The four main axes of the PAT research program
The modeling of the nature of human behavior. The objective is to build,
in order to go beyond the existing models of REMM and PAM, a model of
human behaviour integrating the results of work from economists,
psychologists, and neuroscience specialists in order to understand both
rational (calculated) and non-rational behaviors. The goal is to under-
take, on the level of the individual, the same effort as the one undertaken
on the level of the organization, in other words, to burst the individual
"black box" in order to better theorize human behavior.
The study of the promotion, compensation, and performance
measurement systems. In extension of the work of Baker, Jensen and
Murphy (1988) and of Jensen and Murphy (1990), the goal of this axis is
to provide a theory for the management of human resources that
managers can rely on to design and implement systems allowing the
best use of human resources within organizations, according to the
argument that is implied by PAT, considered as a theory of
organizational architecture. The theorization effort is also integrative,
trying to take into account not only the teachings of labor economics and
human resources management but also those of the human resources
school in the field of behavioral sciences or the research in management
The study of the links between task structure, organizational boundaries,
and nature of technology. The goal is understanding how the overlapping
systems, connecting organizations and markets, can allow us to benefit
best from the opportunities offered by various technologies.[11] In other
words, it is a matter of analyzing the relationships between the nature of
technology and organizational architecture. This somewhat new axis
within the contractual perspective allows us to compensate for the gap
frequently underlined in these theories, the neglect of manufacturing
The study of the links between corporate governance systems, corporate
finance, and organizational performance. The objective of this is to
understand how the relationships of the firm with its financial suppliers or,
more generally, with all the stakeholders, influence the strategy, the
processes of decision-making, and the creation and distribution of
value. This axis includes, in particular, work on corporate governance,
focused on the organizational rules of the game, which constrain the top
managers' decisions “ for example, the board of directors. Work on
corporate governance is in direct connection with corporate finance
research, in particular that relating to financial and ownership structures.

Initially focused on an external vision of the organization, work of this last axis, in
particular that realized in corporate finance, centered on valuation, tended to neglect
internal aspects, but is nevertheless the most important to understand the creation of
value. The current developments, by incorporating the lessons from research in
organizational behavior, try to integrate the internal aspects of governance such as the
formulation of strategy, the role of managers, or the distribution of decision rights within
the hierarchy. These questions should permit an understanding as to how to make
effective investment choices, a subject paradoxically neglected by financial research, as
emphasized by Jensen (1993). Beyond investment decisions, the research topics also
relate to restructurings or to the new forms of ownership structures, for example,
leveraged buy-outs (LBOs) or employee stock ownership.

4.2 Methodological contribution of positive agency theory

Beyond the content of the research questions tackled, PAT built an original research
methodology in the field of organizational economics, by developing the use of clinical
studies as a legitimate research method. It is one of the aspects that separates PAT
most significantly, on the one hand, from the principal“agent theory exclusively based on
quantitative modeling and, on the other hand, from the traditional econometric

Jensen emphasized the limits of formal modeling approaches as much as econometric
studies in order to understand organization behavior:
many important predictions of the research on positive organization theory and positive
accounting theory will be characterizations of the contracting relations, and much of the
best evidence on these propositions will be qualitative and institutional evidence By its
nature, much of this institutional evidence cannot be summarized by measures using real
numbers. (1983, p. 332)

Jensen also insisted on the nature of modeling in PAT (analytical but not mathematical)
and on the central variables, very different from those of the principal“agent theory. Thus,
the variables judged as important in the latter (preferences and information structures)
are regarded as secondary in PAT, that favors the aspects relating to informational costs,
institutional environment, or control systems.
This critical remark about the traditional tools of the competing theories resulted in the
proposition, as an addition to the traditional approaches, of turning to clinical studies. As
the editors of the Journal of Financial Economics (Jensen et al. 1989, p. 4) emphasized,
clinical studies can direct the work of mathematical economists and econometricians
towards more relevant theories by providing them with thorough analyses of the most
important dimensions of real organizational phenomena. Many articles based on clinical
studies were thus published in this journal. In July 1999, a conference was organized by
Harvard Business School on the theme of the complementarity of the various methods of
research in finance, insisting again on the contribution of clinical studies. This
complementarity might, of course, be extended to other fields of management.

Detailed references, including the work realized in France, on the developments of PAT
and, more particularly, on its applications to corporate governance, may be found in
Charreaux (1997).
See, for example, Baldwin and Clark (1992).

See, for example, Jensen (1993) and Charreaux (1997).
5 Conclusions
PAT occupies an original position, sometimes unrecognized, within organization theories;
in particular, it should not be compared to the principal“agent theory or the TCT. If it
remains positioned within the contractual paradigm, the central role that it attributes to
specific knowledge, as well as the evolution of the rationality model on which it is based,
tends to bring it closer, in particular in its most recent developments, to the evolutionary
theory or strategic theories whose central focus is on resources or capabilities. It can
thus, in some respects, be regarded as a first attempt at a compromise between the
theories based on knowledge and those based on opportunism (Conner and Prahalad,
1996; Hodgson, 1998).

The current contributions are numerous and important. They strongly influenced the
various fields of management sciences by renewing the analytical frameworks. The
current developments of PAT lead, moreover, to a connection with the other traditional
research paradigms in management, inspired in particular by the organizational learning,
behaviorist, and managerial theories or certain streams of organizational sociology.
Finally, the specific methodological orientations that it proposed, by recommending an
important recourse to clinical studies, contributes to an emphasis of the influence on
management sciences.
Chapter 15 was originally published as "La th©orie positive de l'agence: positionnement
et apports," in Revue d'Economie Industrielle (92, 2000).

The author thanks the two referees for their comments and suggestions.
1. The differences between PAT and the principal-agent theory will not be
specifically analyzed. These theories differ in particular in their underlying
rationality models, in the basic variables of modeling, and in the methods
2. A more thorough analysis of PAT can be found in Charreaux (1999).
3. PAT was initially developed by Jensen and Meckling at the Business
School of the University of Rochester where Meckling was the Dean.
This theory could also in future be referred to as being of the Rochester-
Harvard School, Jensen having left for Harvard University.
4. The texts which allow us to best account for the evolution and for the
current state of PAT are, in addition to Jensen and Meckling (1998),
Jensen (1998), and Brickley, Smith and Zimmerman (1997).
5. The first version of this article (Meckling 1976), written at the beginning of
the 1970s, is contemporary with the 1976 article.
6. In fact, the majority of the interpretation errors seem attributable to an
assimilation of TPA to this particular model that constitutes but one
aspect of the 1976 article. This article furthermore contains more general
developments of PAT which shows that this theory has a much more
ambitious vocation.
7. For example, Wruck and Jensen (1994) developed a very innovative
analysis of TQM.
8. Bonding costs are those associated with the mechanisms allowing the
agent to reassure the principal on the credibility of his commitments, for
example, the costs associated with a voluntary audit.
9. "The assignment and enforcement of decision rights in organizations are
a matter of organizational policy and practice, not voluntary exchange
among agents."
10. Detailed references, including the work realized in France, on the
developments of PAT and, more particularly, on its applications to
corporate governance, may be found in Charreaux (1997).
11. See, for example, Baldwin and Clark (1992).
12. See, for example, Jensen (1993) and Charreaux (1997).
Testing Contract Theories
Part V:
Chapter 16: Econometrics of Contracts”An Assessment of Developments in the
Empirical Literature on Contracting
Chapter 17: Experiments on Moral Hazard and Incentives”Reciprocity and Surplus-
Econometrics of Contracts-An
Chapter 16:

Assessment of Developments in the
Empirical Literature on Contracting
Scott E. Masten, St©phane Saussier
1 Introduction
The growth in the analysis of inter-firm contractual relationships that has occurred in
recent years is an indication of the importance economists associate with the issue of
contracting and contract design. On the theoretical side, understanding how and why
economic agents use contracts to coordinate their activities is crucial to understanding
the organization and efficiency of economic exchange. For policy-makers, understanding
the functions and implications of various contract terms is a prerequisite to distinguishing
between efficient and anti-competitive practices and to developing appropriate policies
with respect thereto.

Over time, two approaches have come to dominate the analysis of contracting: agency
theory and transaction cost economics (TCE). Of the two, agency theory is widely
regarded as having had the greater success developing formal models of contracting
behavior. But on the empirical side, the assessment is generally reversed. Compared
to agency theory, TCE is seen as having been far more successful both at generating
testable hypotheses and in explaining actual contracting practices.
In this chapter, we review the empirical research on contracting, with special emphasis
on the relative contributions of agency and transaction-cost theories, first, in providing
structural guidance to empirical researchers and, second, in identifying observable
determinants of both the decision to contract and the design of contractual agreements.
We begin in section 2 with a description of the underlying structure and specification of
contracting and contract duration models, followed by assessments of, first, the
contributions of agency and transaction-cost theories to the formulation of hypotheses
about contracting decisions and, second, the evidence pertaining to those hypotheses.
Section 3 extends the analysis from the decision to contract to the issue of contract
design or, more precisely, to the relative success of agency theory and TCE in explaining
the structure and content of contractual agreements. Section 4 comments briefly on the
implications of empirical research on vertical integration for our understanding of
contracting. Finally, in section 5, we provide an overall assessment of the literature's
progress to date and discuss some remaining issues.

Reviews of the agency literature can be found in Hart and Holmström (1987) and Furubotn
and Richter (1997), among other sources. For purposes of this chapter, we include under the
heading "agency theory" complete contract theory (in the tradition of Myerson 1982),
incomplete contract theory (such as Grossman and Hart 1986), and linear contract theory, the
latter consisting of the set of models that restrict consideration to linear sharing rules (see,
e.g., Allen and Lueck 1999 and Lafontaine and Slade 2000). See Masten 2000 for a
discussion of the relation among these models.
2 Why contract?
Agency and transaction-cost theories of contracting differ on the first and most basic
question: why contract? Whereas the primary motives for contracting in the agency
literature are risk transfer (insurance) and incentive alignment (see, generally, Hart and
Holmström 1987), transaction-cost economists tend to view contracts more as devices
for structuring ex post adjustments and for constraining wasteful (rent-dissipating) efforts
to influence the distribution of gains from trade, including, especially, ex post bargaining
and "hold-up" activities in transactions supported by relationship-specific investments
(Williamson 1975, 1979; Klein, Crawford, and Alchian 1978) and ex ante sorting and
search in contexts where additional information serves merely to redistribute rather than
expand the available surplus (Barzel 1982; Kenney and Klein 1983; Goldberg 1985). Yet,
despite these differences, the theories possess a common underlying structure. Before
turning to the implications for empirical research of the differences in the theories, it will
be useful to outline the basic decision structure that unites them.

2.1 Structure and estimation
2.1.1 The contracting decision
In its most general form, the decision to contract represents a standard discrete choice
problem: Transactors will choose to contract if the expected gains (net of transaction
costs) from doing so are greater than those of organizing the transaction in some other
way, or formally,

where GC represents contracting, Ga an alternative to contracting, VC and Va (the
transactors' beliefs about) the corresponding values of the transaction under contracting
and the alternative, and G * represents the governance form actually chosen.

Because the returns transactors expect from governing their transactions in different
ways are difficult, if not impossible, to observe, a testable theory of contracting requires
that the theory relate the benefits and costs of alternative governance arrangements to
observable features of the transaction. To the previous arguments must thus be added
relations of the form


where X represents a vector of observable attributes affecting the gains from trading
under the relevant governance arrangements, and ec and ea are error terms that may
reflect either variables omitted by the investigator or errors or misperceptions on the part
C a [3]
of decision-makers about the true values of V and V . If we assume, for practical
reasons, that the preceding relations can be represented linearly as


we can represent the probability that contracting will be chosen over the alternative
C C a [4]
governance form as Pr(G * = G ) = Pr(V > V ) = Pr(ea ec < ( )X). In words, an
element of X whose effect on the expected gains from trade under contracting, , is
greater than its effect under the alternative arrangement, , will increase the likelihood
that contracting will be the observed form of governance. Theories of contracting inform
the analysis by identifying which attributes empirical researchers should focus on and
predicting the differential effects (i.e., ) of those attributes on the value of
transacting and, potentially, by providing guidance on the functional form of the V(X, e)s.
2.1.2 Contract duration
An alternative to the categorical formulation presented above is to treat the contracting
decision as a question of contract duration: instead of choosing between contracting and
not contracting, transactors could be viewed as choosing how many periods (if any) their
contract should cover. The absence of a contract, under this formulation, would then
correspond to the limiting case of contract duration equal to zero. Conversely, one could
view the contract duration decision as a series of discrete choices, in which transactors
decide, for each future period, whether or not to govern exchange by contract. Drawing
on this correspondence, we could represent the continuous analog to the discrete choice
decision represented by (1) as

where represents contract duration, T the potential (possibly infinite) duration of the
relationship between the parties, V ( ) the cumulative value of contractual exchange
periods covered by the contract, and Va(T
over the ) the value of trade in the
periods following expiration of the contract. Optimal contract duration, *, is the value of
that satisfies the first-order condition

In words, the parties would continue to increase contract duration until the value of
transacting under a contract for an additional period was just equal to the (forgone) value
of transacting without a contract in that period.
As in the discrete choice case, our inability to observe the contracting parties' subjective
expectations of VC and Va necessitates development of hypotheses that relate these
values to observable attributes of transactions. Letting X and e again represent
observable and unobservable factors, we can rewrite (5) as

Linearizing those functions as


and substituting into (5 ) yields an expression for optimal contract duration, *, of the


and e =(ec ea)/( 1 1). For values of strictly between 0 and T, elements of X that
increase the value of contracting for another period more than the forgone benefits of
transacting without a contract in that period ( 2 2 > 0) will result in contracts of longer
predicted duration. As in the discrete choice version of the model, the contribution of
contract theories lies in identifying the attributes likely to affect the efficiency of
contracting and its alternatives and in predicting the direction of their net effects.
Econometrically, (8) would seem to fit neatly the standard regression model. Two
aspects of contract duration, however, necessitate departures from the standard model.
One is duration's natural lower bound of zero, which affects how the distribution of the
error term is parameterized. The second consideration is more peculiar to contracting.
Because only contracts whose durations are at least as long as the difference between
the sampling date and the contracting date "survive" long enough to appear in the
sample “ contracts written x years before the sampling date with durations greater than
x years will appear in the sample but contracts with durations of less than x years will
have expired and will not be represented “ samples drawn from populations of contracts
in existence at a point in time will tend to be over-represented by longer-term
agreements. If the unobserved determinants of contract duration are correlated with the
observed variables, ordinary least squares (OLS) estimates of the coefficients in (8) will
be biased (see, e.g., Maddala 1983, pp. 166“7). Empirical research on contract duration
has generally recognized this data censoring problem and has sought to account for the
potential bias using maximum likelihood estimation techniques (see Crocker and Masten
1988 and Joskow 1987). A third issue, not addressed in the literature but relevant to
studies of contract duration, is heteroskedasticity arising from the decreased precision
with which contracting parties are likely to be able to assess the trade-offs from altering
contract duration at more distant dates. In the data, this phenomenon is manifested in
the tendency for contract duration to be more "finely tuned" for shorter-term agreements,
which vary by intervals of days or months while the durations of longer-term contracts
tend to cluster at discrete intervals of one, five, or ten years.

2.2 Predictions
2.2.1 Agency theory
Surprisingly, agency theory contains little explicit discussion either of the decision to
contract or the choice of contract duration. Though seemingly a serious omission for a
theory of contracting, the agency literature's inattention to those questions is consistent
with the theory's inclusive use of the term contract to encompass any transaction (cf., for
instance, Lyons 1996, p. 27). Under such a broad definition of contracting, it makes little
sense to inquire whether contracting is desirable; the only question is what form the
contract will take. Even where agency theorists nominally distinguish between explicit
and implicit contracts, agency theory provides no reason for transactors not to make their
agreements explicit: contracts deduced from agency axioms are complete, and therefore
efficient, in the sense that (1) they specify each party's obligations for every possible
contingency and (2) they yield the best possible outcome given the information available
at the time the agreement is carried out and thus "never need to be revised or
complemented" (see Holmström and Tirole, 1989, p. 68). Combined with the assumption
that courts enforce verifiable provisions costlessly, issues of contracting and contract
duration are effectively removed from consideration.

To generate testable implications for contracting or contract duration from agency theory,
it is thus necessary to step outside the deductive agency framework by invoking
contracting (transaction) costs (see Hart and Holmström 1987, pp. 131“3). Given some
non-trivial impediment to contracting, we can extrapolate that factors that increase the
benefits of contracting will increase the likelihood and duration of contractual agreements.
Since in agency theory those benefits derive from the sharing of risk and alignment of
incentives, the theory would predict contracting and contract duration to be positively
related to the level of risk (or uncertainty) and to the importance of effort and information
to payoffs.
2.2.2 Transaction-cost economics
In contrast with agency theory, the decision to contract and determinants of contract
duration have been central concerns of the transaction-cost literature. First, transaction-
cost economists, unlike agency theorists, tend to draw a clear distinction between
contractual and non-contractual exchange, reserving the term contract for formal, legal
commitments to which transactors give express approval and to which a particular body
of law applies. By contracting, transactors expose themselves to potential third-party
(judicial) sanctions for failing to honor their commitments and alter the procedures for
resolving disputes and adapting to changing circumstances. In particular, whereas
parties transacting without a contract are generally free to haggle, stall, or walk away as
they please if dissatisfied with the terms of trade currently tendered, the law restricts the
ability of contracting parties to extort concessions from their counterparts by unilaterally
refusing to deal or threatening not to perform.
In terms of the models of section 2, the principal benefits of contracting, V , or, in the
contract duration context, of contracting for an additional period, V C, in the transaction-
cost framework are (1) the greater willingness of transactors to take actions whose value
is conditional on the other party's performance, and (2) a reduction in the costs of
(repeated) bargaining, while the costs of contracting (i.e. forgone benefits of not
contracting), Va, include (1) the costs of anticipating, devising optimal responses to, and
specifying future contingencies (formation costs), and (2) the losses associated with
efforts to enforce, evade, or force a renegotiation of the contract's terms and the
"maladaptation" costs of failing to adjust to changing circumstances (execution costs).
The benefits of contracting, according to the theory, increase with the value of
relationship-specific investments, increasing the likelihood of contracting and the
duration of contractual agreements. More complex or uncertain transactions, meanwhile,
make performance specification and verification more difficult and increase the risk that
the contract will impede desirable adjustments or induce costly evasion or renegotiation
efforts, thereby discouraging transactors from entering formal, long-term agreements.

2.3 Evidence
2.3.1 Agency theory
We are unaware of any empirical studies of the decision to contract or of contract
duration from an agency perspective. As we discuss below, however, where variables of
interest in the theories overlap, evidence from transaction-cost studies offers little
support for agency theory concerns.
2.3.2 Transaction-cost economics
At least two large-scale empirical studies of the choice between formal contracting and
informal agreements using a transaction-cost perspective have been published. In the
first, Allen and Lueck (1992b) examined the use of written versus oral leases for
farmland. Farmers and landowners, they found, were more likely to adopt formal, written
contracts for land requiring investment in and maintenance of irrigation systems (which
are location-specific) while familial and other ongoing relations favored reliance on
informal, oral agreements. In the second, a study of the contracting practices of UK
engineering sub-contractors and their customers, Lyons (1994) found that the probability
that firms adopted a formal contract was significantly higher the greater the share of the
sub-contractor's output accounted for by the customer, the greater the percentage of that
output specifically designed to that customer's requirements, and where production
required significant investments in specific capital. The likelihood of a formal contract fell,
on the other hand, where the subcontractor employed expensive, but flexible, equipment
and where the firm produced an advanced technology product: the greater complexity
and uncertainty likely to be associated with advanced technology products would tend to
make contract specification and enforcement more difficult. The evidence is thus
consistent with transaction-cost predictions regarding the benefits of contracting in the
presence of relationship-specific investments and the liabilities of contracting in complex
and uncertain environments.
The transaction-cost determinants of contract duration have also been the subject of
several studies. An early and well-known example is Joskow's (1987) econometric
analysis of the duration of approximately 300 contracts between coal mines and coal-
fired electricity generators. Joskow's analysis exploited (1) regional differences in the
characteristics of coal and transportation alternatives across the United States, (2)
differences in the proximity of mines and power plants, and (3) variations in contract
quantity to create proxies for the degrees of physical-asset specificity, site specificity,
and dedicated assets. Joskow found the duration of coal contracts to be approximately
eleven years greater in Western states, where coal is more heterogeneous, mines are
larger, distances greater, and transportation alternatives fewer, than in the eastern
United States, where coal tends to be more homogeneous, mines are smaller and more
numerous, distances are shorter, and transportation alternatives abundant, with
Midwestern coal contracts intermediate both in duration and characteristics. Longer still,
by approximately twelve years, were contracts for coal supplied to power generators
located at the mouth of a mine. Finally, contract duration increased by an additional
thirteen years for each additional million tons of coal contracted for.
While Joskow's study provided evidence that contract duration varies with the benefits of
contracting, Crocker and Masten's (1988) study of 245 natural gas contracts sought to
assess variations in the costs as well as the benefits of contracting on the duration of
contractual agreements. Like Joskow, Crocker and Masten found evidence of a positive
relation between contract duration and appropriation hazards; contracts tended to be of
shorter duration for wells in gas fields (1) served by larger numbers of producers and
pipelines (reducing appropriation hazards) and (2) in which only a single producer
operated, eliminating the risk of pipelines exploiting the common-pool drainage problem
to extract price concessions. Crocker and Masten also found, however, (1) that natural
gas contracts written during the period of greater uncertainty following the 1973 Arab oil
embargo tended to be shorter (by an average of three years) than contracts written
before the embargo, and (2) that misaligned incentive provisions (a byproduct of price
regulation) reduced contract duration by an average of fourteen years. Finally, in addition
to the study's substantive findings, Crocker and Masten developed a model of the
contract duration decision that, though rudimentary, nevertheless yielded specific
functional relationships for their estimations.

A pair of more recent studies by Saussier (1998, 1999) analyzing contracts for coal
transportation in France also examined both the costs and benefits of contract duration.[5]
Saussier found that the duration of these contracts was positively related to the value of
investments in relationship-specific assets (as measured by the value of start-up
investments and guaranteed contract quantities) and negatively related to the level of
demand uncertainty over time. In addition, Saussier used two-stage least squares (2SLS)
and a set of exogenous instruments to endogenize the level of specific investments,
addressing a potential limitation of the earlier literature. He found his results to be largely
robust to this refinement.

Finally, Bercovitz (1999) applied transaction-cost reasoning to analyze the duration of
franchise contracts, an area of study otherwise dominated by agency-oriented research
(see, generally, Lafontaine and Slade 2000). Consistent with the evidence from other
contract duration studies, Bercovitz found that the duration of franchise agreements are
significantly longer the larger franchisees' initial investments, which, she argues, are
likely to be correlated with the franchisees' specific investment. In addition, Bercovitz
argues that the threat of non-renewal under shorter-term contracts allows franchisors
greater ability to discipline opportunistic franchisees. Consistent with this, she found that
franchise agreements tend to be of shorter duration in systems having the greatest
potential for franchisee free-riding (as measured by the value of the system's brand
name and the locational density of franchise outlets).

Although not specifically designed to test agency hypotheses, the results of several of
these studies bear indirectly on the validity of predictions derived from agency concerns.
Thus inasmuch as high-technology projects tend to be riskier than simpler procurements,
Lyons' (1994) finding that engineering sub-contractors adopt formal contracts less
frequently for projects characterized as high-tech conflicts with what would be expected if
risk transfer was a primary motive for contracting. Crocker and Masten's (1988) and
Saussier's (1998, 1999) findings that contracts tend to be of shorter duration in periods
of higher uncertainty appear also to be inconsistent with the use of contracting as a
mechanism to allocate risk.[7]

For a more detailed discussion of the problems of identifying the efficiency of alternative
governance arrangements, see Masten, Meehan, and Snyder (1991).

Potential differences in the set of attributes that affect efficiency under alternative
governance arrangements are captured in the model by the possibility that the estimated
marginal effects of particular attributes equal zero.

This correspondence between the discrete choice framework and transaction-cost
hypotheses was first outlined in Masten (1982, 1984).

Saussier had access to the full population of the contracts written over the period covered
by his study and therefore did not face the censoring problem present in Joskow's and
Crocker and Masten's studies. Saussier's study, however, involved a smaller number of
contracts (twenty-nine or seventy, depending on the specification).

Acknowledging that much of the equipment franchisees use is redeployable, Bercovitz
includes only 10 percent of equipment expenditures in this figure.

Compare Goldberg and Erickson (1987, p. 398).
3 Contract design
3.1 Specification and estimation

The variety of possible contract designs is virtually unlimited: the structure of contractual
agreements may vary with, among other things, the objectives of the contracting parties,
underlying production relations, and the nature and size of informational and strategic
impediments to contract formation and enforcement. As a consequence, theory provides
no unifying structure for the specification and testing of contract design hypotheses.
At a practical level, contract provisions and their analysis, like the contracting decision
itself, take both discrete and continuous forms. Contract terms such as price (Joskow
1988b); royalty rates and franchise fees (Lafontaine 1992; Bercovitz 1999); and take-or-
pay provisions (Masten and Crocker 1985; Mulherin 1986) vary continuously, while other
provisions such as price adjustment methods (Crocker and Masten 1991; Crocker and
Reynolds 1993) and the assignment of authority or discretion often have a discrete, on-
or-off character. In still other cases, researchers have chosen to treat contract terms as
discrete choices empirically even though conceptually the "discrete" alternatives are
actually the limit values of some continuous contract parameter. An example is the
treatment of fixed and variable payment schedules as discrete alternatives (see Leffler
and Rucker 1991, Allen and Lueck 1992a; and Masten and Snyder 1993) even though
the "choices" represent corner solutions within a more general contract containing
continuously varying fixed and variable components.

The most common econometric issues to arise in the testing of contract design
hypotheses consist of reasonably familiar simultaneity and endogeneity concerns and,
for some continuously varying provisions, accommodations for limits on the range of the
dependent variable. The only problem to arise so far that is even remotely peculiar to
contracting concerns the systematic over-sampling of longer-term contracts discussed
earlier, which could bias estimates of the coefficients in contract design regressions to
the extent that the errors in contract duration and design equations are correlated.
Empirical studies of contract terms that have recognized and made efforts to control for
this potential bias include Joskow (1988b); Crocker and Masten (1991).

3.2 Predictions
3.2.1 Agency theory
Despite the profusion of agency theoretic models, neither complete nor incomplete
contract theory (ICT) has produced much in the way of testable hypotheses. In the case
of complete contract theory, the potential complexity of optimal incentive schemes and
their "extreme sensitivity" to changes in information assumptions have prevented
formulation of general hypotheses about contract form (Hart and Holmström 1987, pp.
80-1, 105). On the other hand, ICT, despite its name, is actually a theory of ownership
rather than contracting that, by imposing severe restrictions on feasible contract forms,
assumes in essence what a theory of contracting seeks to explain.
Partly because of its relative tractability, the literature on linear sharing contracts has
been more successful than complete and incomplete contract theories at generating
predictions. The main prediction of that literature is that efficient sharing rules will
balance incentives for one party against inefficient risk-bearing by that party or the
incentives of trading partners; larger shares should tend to be assigned to the party with
(1) lower aversion to risk and (2) higher marginal productivity of effort. More recently, the
generalization of the linear agency model to multitask settings has augmented the list of
agency predictions, most notably with the prediction that contracts should provide agents
who perform multiple or multidimensional tasks, some aspects of which are difficult to
measure, with low-powered incentives (Holmström and Milgrom 1991).
3.2.2 Transaction-cost economics
Transaction-cost economists acknowledge the role of contract terms in aligning marginal
incentives but see an additional function of contract design in preventing wasteful efforts
to redistribute existing surpluses. In contrast to moral hazard, which represents a
deviation from joint surplus maximizing behavior within the terms of an extant contract,
this second form of opportunism includes efforts to evade performance or to force a
renegotiation of a contract's terms by imposing costs on one's trading partner. Because
the incentive to engage in such efforts is likely to be related to the ex post distribution of
contractual surpluses - parties greatly disadvantaged by the terms of a contract are more
likely to want to evade or renegotiate a previous deal - contracting parties will seek to
design contracts to divide ex post rents "equitably" (Masten 1988), keep the relationship
within the agreement's "self-enforcing range" (Klein 1992), or, equivalently, achieve what
Oliver Williamson has called "hazard equilibration" (1985, p. 34) (see, in addition,
Goldberg 1985 and Goldberg and Erickson 1987). The more uncertain the environment
and the harder it is to accommodate changing circumstances within the contract, the
more likely it will be that parties will sacrifice the precision and ease of implementation of
definite contract terms for more cumbersome but flexible "relational" contract terms that
define performance obligations less precisely or establish procedures for negotiating
adjustments in the terms of trade within the contract.

3.3 Evidence
3.3.1 Agency theory
In addition to its failure to generate testable hypotheses, complete contract theory has
been criticized, by agency theorists themselves, for failing to account for even the most
basic features of real-world contracts. Thus, whereas the theory predicts detailed and
complex-payment rules specifying each party's performance obligations for every
possible contingency (in the case of discrete contingencies) and elaborate non-linear
pricing rules (in the continuous case), actual contracts incorporate few if any explicit
contingencies and generally use simple, typically linear, pricing schemes (Holmström
and Hart 1987; Bhattacharyya and Lafontaine 1995).

Agency models that impose linearity at the outset thus start out with an obvious
advantage in explaining observed contracts. Among the settings that have been
analyzed in linear agency terms are franchising (Mathewson and Winter 1985; Lal 1990),
agricultural share-cropping (Stiglitz 1974; Eswaran and Kotwal 1985), and product
warranties (Priest 1981; Cooper and Ross 1985). Yet despite the variety of settings to
which agency models potentially apply, empirical studies of the determinants of contract
terms from an agency perspective have been limited (see Lafontaine and Slade 2000).
One reason for this is the difficulty of finding workable proxies for contracting parties'
relative risk aversion and the marginal contributions of their efforts to joint surplus. To the
extent that these factors are difficult or impossible to measure, acceptance of the
theory's predictions often turns on accepting non-falsifiable risk preference and marginal
productivity assumptions (Stigler and Becker 1977; Allen and Lueck 1995).
Where risk arguments have been subjected to formal tests, they have not done well. For
example, in franchising, where much of the recent empirical work has been directed,
observed correlations between uncertainty and royalty rates are inconsistent with the
standard assumption of franchisee risk aversion (Lafontaine 1992). Risk-sharing as a
motive for contracting has fared poorly in other settings as well. Allen and Lueck (1992a,
1999), for instance, conclude that the incidence of crop-share versus fixed-rent contracts
between farmer-tenants and land owners is inconsistent with the maintained assumption
that farmers are more risk averse than land owners. Similarly, Leffler and Rucker (1991)
reject risk-sharing as inconsistent with the incidence of lump-sum versus royalty
payments in contracts between timber harvesters and land owners.

Predictions from agency models based on incentive (as opposed to risk) considerations
have fared somewhat better in general. Lafontaine (1992), for example, found that
royalty rates across franchises tend to vary with the relative importance of franchisor and
franchisee effort. On the other hand, "franchise fees are in general not negatively
correlated with royalty rates, despite the fact that the standard principal-agent model
suggests that they should be" (Lafontaine and Slade 2000). Empirical research on
agency contracting has also been criticized for focusing on such a highly limited range of
contract terms (e.g. Bercovitz 1999).
3.3.2 Transaction-cost economics
Empirical transaction-cost research on contract design has looked primarily at three
types of provisions: incentive provisions, pricing structures, and price adjustment
methods. Like the agency literature, transaction-cost studies of incentive provisions have
sought to determine whether contract terms align the interests of the contracting party
and promote efficient adjustments to change. These studies have sought to explain more
than just sharing rules, however. Studies by Masten and Crocker (1985) and Mulherin
(1986), for instance, analyzed the incidence of take-or-pay provisions in natural gas
contracts. Using data sets covering different periods in the history of gas contracting, the
studies found that take-or-pay obligations varied with the alternative value of gas
reserves, supporting an incentive interpretation over the alternative view that take-or-pay
provisions served distributional or risk-sharing purposes (e.g. Hubbard and Weiner
1986).[10] Though these studies approached the issue of take-or-pay obligations from a
transaction-cost perspective, their hypotheses and results are broadly consistent with an
agency theoretic approach.

The overlap between transaction-cost and agency theory predictions with respect to
incentives can also be seen in two studies on the inclusion of "protective" provisions in
franchise contracts. Although Bercovitz (1999) and Brickley (1999) analyze the use of
various "non-pricerelated" restrictions on behavior in franchise agreements in similar
terms and derive similar predictions, one describes her analysis as "building on the
transaction cost framework" (Bercovitz 1999) while the other "uses agency theory to
develop testable implications" (Brickley 1999).[12] The papers provide evidence that non-
compete clauses (Bercovitz), passive ownership prohibitions, area development plans,
and mandatory advertising requirements (Brickley) are positively related to proxies for
the potential for franchisee free-riding and/or the size of initial investments.

Transaction-cost economists have taken a distinct approach to the analysis of pricing
structures in contracts, however. Whereas agency theory analyzes pricing structures in
moral-hazard and risk-sharing terms, transaction-cost economists have viewed the
choice between fixed and variable payment terms as reflecting efforts to economize on
transaction costs. An example is Leffler and Rucker's (1991) study of lump-sum versus
per-unit pricing structures in timber harvesting contracts. In Leffler and Rucker's analysis,
fixed-payment contracts give purchasers an incentive to engage in extensive pre-sale
measurement of timber quality and quantity, whereas per-unit contracts discourage
harvesters from harvesting timber efficiently and require greater post-agreement
monitoring and enforcement efforts. Using a sample of 188 North Carolina timber
contracts, Leffler and Rucker found the use of per-unit pricing to be more prevalent on
relatively heterogeneous timber tracks for which pre-sale search costs were likely to be
higher. Allen and Lueck's (1999) finding of a positive correlation between the variance in
crop yields and the use of cash rent (fixed-price) contracts for farm land - the exact
opposite of the prediction based on farmer risk aversion - is consistent with a hypothesis
that farmers are better able to misreport crop yields as output variations owing to
exogenous factors (weather, pest infestations) increase (see also Allen and Lueck 1998).
Transaction-cost considerations also figure prominently in Masten and Snyder's (1993)
analysis of pricing arrangements in equipment leases and in Bessy, Brousseau and
Saussier's (2001) analysis of payment schemes in technology licensing agreements.
Empirical research also supports the role of "hazard equilibration" in contract design.
Crocker and Masten (1991), for instance, conclude from their study of price adjustment
in natural gas contracts that circumstances favoring the use of long-term, fixed-quantity
agreements favor the adoption of relatively indefinite price adjustment provisions over
formulaic adjustment mechanisms that, although less costly to implement, are more
likely to induce efforts to evade performance obligations in extreme situations. As
Goldberg and Erickson (1987) have noted, greater reliance on renegotiation provisions in
fixed versus variable quantity contracts is difficult to reconcile with incentive alignment
motives. Crocker and Reynolds' (1993) study of jet engine procurement contracts found
that contracts tended to contain more flexible price adjustment mechanisms as
performance horizons lengthened and technological uncertainty increased, while
contractor litigiousness and the absence of alternative engine suppliers favored more
definite price terms. More generally, Saussier (2000) provides evidence that the
"completeness" of contracts, as measured by the number of dimensions of performance
specified in the contract, varies with the attributes of the transaction: contracts for the
transport of coal in France tend to contain more detail the greater the level of asset
specificity but include fewer provisions as uncertainty increases. Thus, overall, the
available evidence supports the notion that, in designing contracts, transactors are
sensitive to the trade-off between the specification costs and rigidities associated with
detailed performance obligations in uncertain or complex transactions, on the one hand,
and the greater flexibility but higher expected cost of establishing terms of trade ex post
with less definite "relational" contract provisions, on the other.

See Lafontaine and Slade (2000). Empirically, contracts often do contain only fixed or
variable payments, not both, a fact that suggests discontinuities in how the terms operate.

While sympathizing with the view that individuals are not capable of dealing with unlimited
complexity, purists complain that, in the absence of an accepted model of bounded rationality,
restrictions on feasible contract forms are unavoidably arbitrary and ad hoc (e.g. Tirole 1994,
pp. 15-17; Hart and Holmström 1987, pp. 133, 148).

Crocker and Masten's (1988) finding that distortions in the size of take-orpay provisions
significantly reduced the willingness of parties to engage in long-term contracting offered
further support for the incentive interpretation of take-or-pay provisions. Case studies
describing the use of minimum purchase requirements for coal (Carney 1978), petroleum
coke (Goldberg and Erickson 1987), and bauxite (Stuckey 1983), among other products, also
corroborate this finding (see Masten 1988, pp. 91-2, for a discussion).

Compare, for example, Shavell's (1984) theoretical development of efficient breach
analysis with the characterization of the optimal take-or-pay provision in Masten and Crocker

In particular, Brickley (1999) interprets his results as being consistent with the multitask
agency model of Holmström and Milgrom (1991, p. 747).
4 Contracting versus vertical integration
While our discussion of contracting in section 2 presumed that the alternative to
contracting was a simple, arm's-length or "market" transaction, in practice the relevant
alternative to contract is often vertical integration or "internal organization." For space
reasons and because the empirical literature on vertical integration versus contracting
has been discussed at length in several recent and extensive reviews (e.g. Shelanski
and Klein 1995; Crocker and Masten 1996; Coeurderoy and Qu©lin 1997; and Lafontaine
and Slade 1997, 2000), we confine ourselves here to a few observations.

4.1 Agency theory

Complete contract theory is unable to distinguish contracting from other institutional and
organizational forms and thus unable to inform the choice between contracting and
integration. Incomplete contract theory, by contrast, was developed specifically to
explain the existence of the firm and might therefore be expected to inform the contract
versus vertical integration decision. Contracting and integration are not treated as
alternatives in the incomplete contract framework, however. Rather, the theory asks only
which of two (or more) parties to a contract should own a particular asset; the
relationship between the parties themselves remains contractual regardless of who owns
the asset.[14]

Given the limitations of complete and incomplete contract theories, linear contracting
models have again, as in the case of contract design issues, been the primary source of
predictions concerning the choice between contracting and integration within the broad
agency framework. Or, to be more precise, models of the optimal linear share parameter
have been used to extrapolate predictions about integration decisions: Assigning a large
share of the residual to agents corresponds to the high-powered incentives associated
with arm's-length contracts, while a low share generates the low-powered incentives
conventionally attributed to integration (see Lafontaine and Slade 2000, pp. 5, 11).
Agency theory thus predicts that integration is likely to dominate contracting where
conditions favor allocating more of the risk to the principal, namely where the principal is
the lower cost risk-bearer or the value of the principal's (non-contractible) effort is greater
than that of the agent. Although the evidence supports the predicted effect on relative
effort contributions, the empirical literature "strongly rejects" the prediction that higher
risk leads to more integration (Lafontaine and Slade 2000, p. 39). Finally, Holmström and
Milgrom (1991) interpret Anderson and Schmittlein's (1984) and Anderson's (1985)
findings that the importance of non-selling activities and difficulty measuring performance
of sales agents leads to greater integration as support for the predictions of multitask
agency theory; the inability to measure some of multiple dimensions of an agent's effort
favors reliance on lower-powered incentives and the imposition of restrictions on agent
behavior frequently associated with integration (Holmström and Milgrom 1994).

4.2 Transaction-cost economics
The vertical integration, or make-or-buy, decision has been the most extensively studied
question in the empirical transaction cost literature (for overviews, see Joskow 1988b;


. 8
( 18)