0% found this document useful (0 votes)

158 views54 pages

Corporate Culture and Economic Theory

Uploaded by

youssefmnaili98

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

158 views54 pages

Corporate Culture and Economic Theory

Uploaded by

youssefmnaili98

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 54

Corporate culture and economic theory

DAVID M. KREPS

INTRODUCTION
In this chapter, I explore how an economic theorist might explain or
model a concept such as corporate culture. While the theoretical con-
struction that is given is far from inclusive (which is to say that many
aspects of corporate culture are not covered), I conclude that economic
theory is moving in the direction of what seems a reasonable story. But
before that story can be considered told, we must employ tools that are
currently missing from the economist's tool kit. In particular, we re-
quire a framework for dealing with the unforeseen.
I can give two explanations for why I present this topic. The first
concerns how economists (and those weaned on the economic para-
digm) deal with the topic of business strategy. If we take Porter (or
Caves or Spence) as the prototype, business strategy could roughly be
called applied industrial organization. The firm and its capabilities are

This work was prepared in 1984 for presentation to the Second Mitsubishi Bank Foun-
dation Conference on Technology and Business Strategy. It subsequently appeared in Jap-
anese in Technological Innovation and Business Strategy, M. Tsuchiya (ed.), Nippon
Keizai Shimbunsha Press, Tokyo, 1986, and appears here in English with the kind per-
mission of the previous publishers. The chapter surveys the current state of research and
is, of course, quite dated now. But, with the kind permission of the current editors, it
appears now much as it was written in 1984, except that references have been updated
where appropriate and punctuation and English have been made more correct. In a very
brief postscript, following the appendix, I engage in a bit of updating and revisionist
thinking. And, in one place in the text, where it is too painful to reread what I wrote, I
alert the reader that something more on this issue will be said in the postscript.
I have benefited from discussion with too many colleagues to give a comprehensive
list, but two individuals must be cited for particularly helpful ideas: Jose Scheinkman,
concerning the overlapping generations model, and Bengt Holmstrom for stressing to me
the important distinction between observability and verifiability. The financial assistance
of the National Science Foundation (Grants SES-8006407 and SES-8408468), the Sloan
Foundation, and the Mitsubishi Bank Foundation are all gratefully acknowledged.

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
more or less taken as givens, and one looks at the tangible characteris-
tics of an industry to explain profitability. It sometimes seems, in this
approach, that there are good industries (or segments of industries) and
bad: Find yourself in a bad industry (low entry barriers, many substi-
tutes, powerful customers and suppliers, many and surly competitors),
and you can do nothing except get out at the first opportunity. Now,
this is assuredly a caricature of the Porter approach. The size of entry
barriers, relations with suppliers/customers, and, especially, competitive
discipline within an industry are all at least partially endogenous. Bad
industries can sometimes be made good, and (perhaps a more accurate
rendering of Porter) good niches can be found or formed even in bad
industries.
This approach carries with it a powerful legacy from textbook mi-
croeconomics: The firm is an exogenously specified cost function or
production possibilities set, and market structures (also exogenous) de-
termine how it will fare. The actual purpose of the firm qua organiza-
tion is not considered. This is rather strange, for if one has an
economic mind-set, one must believe that the firm itself performs some
economic (efficiency-promoting) function. From there it is a short step
to consider as part, perhaps the largest part, of successful strategy
those actions designed to increase the firm's organizational efficiency.1
But since textbook economics doesn't explain firms qua organizations,
it comes up empty as a discipline for analyzing this part of strategy.
Of course, disciplines other than economics deal with organizational
efficiency or effectiveness. One could simply assert that Porter's ap-
proach is incomplete, to be supplemented or, better, taken concurrently
with other disciplines and approaches. The dangers here are that pre-
scriptions from the economic approach may interact negatively with
organizational efficiency in specific cases and that individuals trained
in one approach may ignore the other. In order to reduce these dangers,
it seems a good idea to develop a theory that addresses issues of orga-
nizational efficiency in the language of economics and then to integrate
it with Porter-style analysis.
When I say that economics has not come to grips with issues of im-
plementation, I mean standard textbook economics. I believe that the
foundations for such analysis have been and are being laid. There are
the obvious and very visible contributors: Williamson and his col-
leagues. But I believe that "higher" theorists (which, in economics,
means more mathematical) dealing in such topics as agency theory, the
theory of repeated games and reputation, and (less formally) the theory
of focal points in noncooperative games should also be counted.
This, then, gives rise to my second and primary reason for writing
this chapter. I think economists are moving in a profitable direction,
91

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
and I want to present the outlines of the theory that is developing. I
hope in so doing to interest readers in further developments.
With my objectives stated, I can describe the nature of this chapter.
With significant exceptions, no new theory is being presented here.
What little is new is undeveloped — it constitutes conjecture and little
else. This chapter is meant to be expositional and exploratory and, per-
haps, just a bit synthetic: I want to sketch out the pieces of the theory
that have been developed, to connect them (as they are connected in my
mind), and to conjecture as to what is missing. (Needless to say, what is
missing largely coincides with my current research agenda.) I have tried
as much as possible to stay away from technical details; sometimes,
however, this has been impossible to avoid, and I apologize.
This chapter sprawls somewhat, but I have in mind a very definite
plot that ties things together. At the risk of completely losing the
reader, let me give here an outline of the plot. It has three fundamental
building blocks. The first is that in many transactions, in particular
ongoing ones, contingencies typically arise that were unforeseen at the
time of the transaction itself. Many transactions will potentially be too
costly to undertake if the participants cannot rely on efficient and eq-
uitable adaptation to those unforeseen contingencies. Note that such
reliance will necessarily involve blind faith; if we cannot foresee a con-
tingency, we cannot know in advance that we can efficiently and equi-
tably meet it. (For those who find the notion of an unforeseen
contingency unpalatable, we could equally well imagine how costly it
is to specify how every contingency will be met.)
Transactions can be characterized by the adjudication processes that
meet unforeseen contingencies. In particular, some transactions will be
hierarchical in that one party will have much more authority in saying
what adaptation will take place. The firm (or other organization) is
meant in this theory to play the canonical role of the authoritative
party: When I am employed by a firm, I accept within broad limits the
firm's right, as expressed by my superior, to specify how my time will
be spent as contingencies arise. Or, to take another example, when stu-
dents attend a university, they accept the university's right, through its
administrators, to spell out the terms of the commodity students have
bought.2
If employees or students are to grant such authority to a firm or
university, they must believe that it will be used fairly. What is the
source of this faith? It is that the firm and university are characterized
by their reputations. The way an organization adapts to an unforeseen
contingency can add to or detract from that reputation, with conse-
quences for the amount of faith future employees or students will have.

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
This faith is the glue that permits mutually beneficial transactions to
take place, transactions that would otherwise not be made because
of their costs. The organization, or, more precisely, those in the orga-
nization who have decision-making authority, will have an interest
in preserving or even promoting a good reputation to allow for future
beneficial transactions. Thus, workers or students can trust the orga-
nization to act equitably in its own interest to protect its valuable
reputation. Note that the organization must be an ongoing entity here:
If ever it loses its incentive to protect its reputation, an incentive de-
rived from the incentive to undertake future beneficial transactions,
then it can no longer be trusted, and the hierarchical transaction will
fall apart.3
With these three blocks in place, we come to corporate culture. In
order for a reputation to have an effect, both sides involved in a trans-
action must ex ante have some idea of the meaning of appropriate or
equitable fulfillment of the contract. Potential future trading partners
must be able to observe fulfillment (or lack of) by the hierarchically
superior party. These things are necessary; otherwise the hierarchically
superior party's reputation turns on nothing. When we speak of adap-
tation to unforeseen contingencies, however, we cannot specify ex ante
how those contingencies will be met. We can at best give some sort of
principle or rule that has wide (preferably universal) applicability and
that is simple enough to be interpreted by all concerned. In the lan-
guage of game theory, unforeseen contingencies are best met by the
sort of principle that underlies what Schelling (1960) calls a focal
point. The organization will be characterized by the principle it selects.
It will (optimally) try to promote understanding of that principle in the
minds of its hierarchical inferiors. In order to protect its reputation for
applying the principle in all cases, it will apply the principle even when
its application might not be optimal in the short run. It will apply the
principle even in areas where it serves no direct organizational objec-
tive, if doing so helps preserve or clarify the principle. Because
decision-making authority in a firm is diffuse, those who make deci-
sions in the firm's name will be judged by their diligence in applying
and embracing the principle. In this light, I interpret corporate culture
as partly the principle itself (or, more realistically, the interrelated prin-
ciples that the organization employs) and partly the means by which
the principle is communicated to hierarchical inferiors (so they can
monitor its application) and hierarchical superiors (so they can apply it
faithfully). It says how things are done, and how they are meant to be
done in the organization. Because it will be designed through time to
meet unforeseen contingencies as they arise, it will be the product of

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
evolution inside the organization and will be influenced by the organi-
zation's history.
This, very roughly, is the economic theory of corporate culture that I
wish to lay out. As noted earlier, this theory captures at most one facet
of corporate culture. The economic paradigm also contains explana-
tions that rely on the screening function of internal cultures. Outside of
the economic paradigm, at least so far as I can see, are explanations
that rely on concepts such as need for affiliation and other things that
1 know nothing about. I don't mean to advance the theory as all-
inclusive; rather, it fits in well with recent advances in the economic
theory.
Rather than proceeding with the theory as already outlined, I will
first exposit those pieces of the story that are in the extant literature
and then go on to pieces that are missing or underdeveloped. This will
make the basic plot harder to understand, but if readers can keep this
plot in mind it will make it easier to see how several strands in eco-
nomic theory interrelate. Since underlying this plot is the basic need to
render efficient otherwise inefficient transactions, I begin with a sec-
tion on the basic transactions theory of Coase and Williamson. I give
here as well the standard criticism of Williamson: He analyzes why
(and when) market-based transactions are costly but insufficiently jus-
tifies his assumption that when market-based transactions are costly,
hierarchy-based transactions are not equally (or more) costly.
Next, in the section on Grossman and Hart, I discuss a recent paper
by those authors (1986) that gives an example of an adjudication pro-
cess for dealing with unforeseen (or, more precisely, uncontracted-for)
contingencies. In their model the authority to decide to employ capital
rests with the legal owners of the capital. The authors use this, together
with an inability to contract for certain contingencies, to explain par-
ticular patterns of ownership for particular transactions. This is not
quite the theory I will later employ - authority in hierarchical transac-
tions is typically much less tangibly based than in something like own-
ership — but it provides a reference point for the type of theory I am
suggesting.
In the next section, the parts of game theory that are needed to dis-
cuss reputation are put in place and are used to make a first pass at a
theory of the firm. The section begins by reviewing repeated games, the
folk theorem, and implicit contracts, and then recasts the folk theorem
into a story of reputation.
Finally, I give a simple parable that shows how something as intan-
gible as a reputation could become an economic good - one that eco-
nomic actors would invest in and, when the time comes, sell. This gives
us a rather pat explanation for what a firm is: an intangible asset car-
94

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
rying a reputation that is beneficial for efficient transactions, confer-
ring that reputation upon whoever currently owns the asset.
The theory developed in the section on reputation will seem rather
disconnected from notions such as corporate culture, or even from the
concept of a hierarchical transaction. But in later sections, I will use
the reputation construction to move in the direction of these ideas. The
first step is to make the basic reputation construction encompass trans-
actions that are hierarchical in the sense already given. This is the sub-
ject of the section titled "Hierarchical Transactions." Also in this
section, we refine a point made in the previous section: Reputations for
behaving in a particular way work more efficiently the more deviations
from that behavior are observable.
Through the section on hierarchical transactions, we will work
within the standard framework of neoclassical microeconomics. The
theory developed in the sections on reputation and hierarchical trans-
actions plays entirely within the usual rules of economic theory. As a
consequence of this, however, it does not provide a very good case for
its own importance. A stronger case emerges if one considers the pos-
sibility of contingencies arising that parties to a particular transaction
have not ex ante thought through, either because they were ex ante
unimaginable or because it is simply too costly to think through all
possible contingencies. The section on unforeseen contingencies that
follows also speculates as to how a (useful) formal theory of this sort
of unforeseen contingencies might develop. Then, the section on focal
points takes a brief excursion into Schelling's (1960) very underdevel-
oped area of game theory. This concept of focal points will play an
important role in the theory finally constructed in the section titled
"Corporate Culture."
In this last-mentioned section, the various pieces are assembled into
an economic theory of the role of corporate culture. More precisely,
the outlines of such a theory, together with some conjectures as to
where that theory might lead, are given. It is my hope that the theory
will be in line with the well-developed theories reported in earlier sec-
tions and that readers will see that the final steps, while not yet accom-
plished, are not excessively difficult to traverse. Concluding remarks
and questions are given in the final section.

WILLIAMSON, TRANSACTION COSTS, AND

THE THEORY OF THE FIRM
Following in the footsteps of Coase (1937), Simon (1965), and Arrow
(1974), among others, Williamson and his students have been de-
veloping a theory of what hierarchical organizations such as firms
95

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
accomplish. Williamson states this theory most fully in his 1975 book;
his more recent work and a very good summary statement can be
found in his 1981 article.
The heart of the theory is the concept of transaction costs. For par-
ties to consummate a transaction or an exchange they must expend re-
sources other than those contained in the terms of the transaction.
Among these transaction costs are resources expended to spell out in
advance the terms of the transaction, so that each side knows what it is
getting, and resources expended to enforce the terms of the transaction.
Textbook economic theory, which calls forth images of the exchange of
one physical good for a second or of one physical good for money,
tends to treat transaction costs as being near enough to zero to be ig-
nored. Costs can, however, be substantial in more complex transac-
tions, such as those in which one party sells labor to the other,, in
which the good sold has hidden qualities or in which one side must
sink resources in preparing for the transaction before the other side
fulfills its part of the bargain. In deciding whether to undertake a
transaction, both parties must weigh the benefits they will accrue, net
of the cost of transacting. Transactions that give the parties positive
benefits gross of transaction costs (which, according to textbook eco-
nomics, would therefore take place) may not give benefits sufficient to
cover the transaction costs and so will not take place.
The organizational structure the transaction takes place within can
affect transaction costs. An exchange in the marketplace may be more
or less costly than the same exchange in a hierarchical organization.
Holding benefits constant, a transaction will tend to occur within
whatever infrastructure minimizes its cost. When transactions take
place in firms, the presumption (and direction for analysis) should be
that the transactions are less costly within the firm than they would be
in the marketplace.
That, very briefly, is the basic theory that Coase advanced (1937)
and that Williamson extended and elaborated on. The study x>( markets
and other organizations that transactions take place in becomes a study
of the relative transaction costs within those organizations. Note well
the type of firm in this analysis: The firm is like individual agents in
textbook economics, which finds its highest expression in general equi-
librium theory (see Debreu 1959, Arrow and Hahn 1971). The firm
transacts with other firms and with individuals in the market. Agents
have utility functions, firms have a profit motive; agents have consump-
tion sets, firms have production possibility sets. But in transaction-cost
economics, firms are more like markets — both are arenas within which
individuals can transact. Indeed, we might think of firms as market-

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
places, contrasting them with other marketplaces, such as the stock ex-
changes, within all of which transactions take place.
Williamson goes on to study five factors that make transactions rel-
atively more costly. He divides these factors into two categories - those
that pertain to the transaction itself and those that pertain to the par-
ties to the transaction. The transaction itself can be described according
to its complexity, which includes the amount of uncertainty that the
transaction bears, especially uncertainty about future contingencies;
according to the thinness of the transaction, or the number of alterna-
tive trading partners involved in it; and according to the extent of im-
pacted information in it, information some but not all parties to the
transaction possess. The transacting parties may be more or less oppor-
tunistic, in that they pursue selfish interests in a guileful manner. They
may also be limitedly rational, in that it is costly and sometimes impos-
sible for them to carry out all the computations required to find a
truly optimal course of action or to elaborate and think through all
contingencies that might bear on the transaction. Relatively greater
complexity and/or thinness and/or impacted information, joined with
relatively greater guile and/or relatively more limited rationality, will
raise transaction costs.
Williamson goes on to analyze particular scenarios in which trans-
action costs are high, suggesting that in these cases there is a clear case
for organizing the transaction in a hierarchy rather than in a tradi-
tional marketplace. Examples drawn from Williamson (1981) include
vertical integration, when costs need to be sunk in transaction-specific
capital before the transaction is actually executed, and franchising,
when the quality of the good sold depends in part on services a sales-
person delivers.
Williamson builds quite a substantial case in these and other in-
stances for large transaction costs in market-mediated transactions.
He is less convincing in arguing that transacting through a hierarchical
organization lessens transaction costs. This is a frequent criticism of
his work: He explicitly recognizes that transacting through a hierar-
chical framework incurs costs (what he sometimes refers to as the costs
of bureaucracy), but he doesn't say enough about how they would
differ from market-mediated transaction costs. Increasing the five
factors cited will mainly increase market-mediated transaction costs.
There is little reason, though, to think that these factors will not simul-
taneously increase (and perhaps by more) hierarchy-mediated transac-
tion costs. Therefore we cannot, without a leap of faith, expect to see
more hierarchical mediation and less market mediation in transactions
with large levels of Williamson's five factors. This is not to say that

GROSSMAN AND HART AND THE RESIDUAL RIGHTS

CONFERRED BY OWNERSHIP
One hinge on which the argument could turn is the legally mandated
default clauses in contracts. Grossman and Hart make such an argu-
ment in a recent article (1986). They give a story for Williamson's ver-
tical integration due to transaction-specific capital. Roughly put, they
argue that contracts that would be optimal under vertical disinte-
gration and a market transaction cannot be written, while contracts
that would be more efficient in an integrated setting can be. The key to
their argument is the notion of residual rights; owners of capital equip-
ment own the right to use that capital as they see fit, subject to the
specific contractual arrangements that have been made. If certain con-
tingent contracts cannot be made (because they are costly to make
or enforce), then the second-best arrangement between an upstream
and a downstream entity might well be one where the residual rights
associated with capital ownership are concentrated in one hand. Gross-
man and Hart suppose that detailed contingent contracts cannot
be made and thus that ownership (and the residual rights thereby en-
tailed) changes the space of feasible contracts (a "contract" here means
all clauses, including unwritable ones that are nonetheless created
by a pattern of ownership). In such a case, they show that certain
ownership patterns for physical capital might be more efficient than
others.
Besides providing an interesting analysis, Grossman and Hart point
us in what I believe is the right direction to pursue. Williamson, fol-
lowing Coase, wishes to make the transaction the unit of analysis;
Grossman and Hart do so with a vengeance. Indeed, taking Grossman
and Hart at their word, one might expect a much more active market
in physical capital: As the particular transaction for which that capital
is employed changes, ownership of the capital (optimally) changes as
well. That is, the authors explain ownership patterns for a particular
transaction but not the permanence or stability of that pattern, which
marks the modern corporation. (Of course, Grossman and Hart can
easily defend themselves on two grounds: Insofar as other identical
or even similar transactions follow the one being analyzed, the opti-
mal ownership arrangement might be quite stable. And markets for
physical capital will have severe moral hazard and adverse-selection
problems.)
98

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
My point is that Grossman and Hart study the requirements for a
particular (ideal) transaction and the way various institutional arrange-
ments approximate those requirements. Theirs is not a theory of the
firm per se; rather, they entitle their work "The Costs and Benefits of
Ownership." At the level of their analysis, capital ownership by single
entrepreneurs is as likely a consequence as is ownership by an entity
with a firm's legal status. One can begin with their analysis and, using
other pieces, build a theory of why firms exist.4 But their specific con-
cern is with transaction efficiency. The tie to the theory of the firm
comes from the observation that conditions conducive for efficiency in
the sorts of transactions the authors examine (concentrated ownership
of capital) correlate with conditions in which one finds "firms" (effi-
cient sharing of risk).
This leads to a second (niggling) criticism of Williamson: He tries
too hard to dichotomize directly the market and the firm. It will prove
more fruitful, I think, to characterize particular sorts of transactions
and to then correlate the characteristics that lead to efficient trans-
acting with firms/markets. Drawing a clean line between firms and
markets will not prove possible; cleaner lines can be drawn if the trans-
action is the unit of analysis.
As outlined in the introduction, I develop in this chapter a dichotomy
in transactions that correlates well with the distinction between firms
and markets. I will also attempt to explain the source of the correla-
tion. The dichotomy is between hierarchical transactions and, for lack
of a better name, specified transactions. Roughly, in a specified trans-
action all terms are spelled out in advance. In a hierarchical trans-
action, certain terms are left unspecified; what is specified is that one
of the two parties has, within broad limits, the contractual right to
specify how the contract will be fulfilled. There is, to be sure, less than
a perfect dichotomy here, for one can think of many other variations —
for example, transactions with ex ante unspecified clauses, with ex post
fulfillment determined by negotiation and requiring unanimous con-
sent; or where the authority to determine ex post fulfillment is split
among the parties. Here I will concentrate on arrangements where one
party has the authority to determine ex post fulfillment, comparing
this with cases where there is no need for such authority. (This notion
is far from original to me; see, for example, Simon 1951.)
Note that Grossman and Hart's residual rights from ownership are
very much of this flavor. They assume that the contract cannot provide
for the use of the capital in certain contingencies and that the owner of
the capital has the right to decide on how those contingencies will be
met. The difference between their analysis and the one developed here
is that they assume that ownership confers this authority and that thus
99

PAYOFFS
A B
honor A's trust
» $10 $10

trust B
*~B

abuse A's trust

*- -$5 $15

do not trust B
*- $0 $0
Figure 4.1. The trust game

the efficient placement of this authority determines the pattern of own-

ership. In what follows, ownership will not determine who retains
hierarchical authority. Instead, we will attempt to use the notion of
reputation to endogenize the determination of (efficient) authority.
(This particular endogenization will provide the raison d'etre for the
firm - the source of correlation between firms and hierarchical trans-
actions, and, eventually, the role of corporate culture.) But still, the
similarity is more important than the difference; they, as we, concen-
trate on explicit contracts that are incomplete descriptions of the trans-
actional relationship and on how those contracts might be completed
as circumstances arise.

REPUTATION AND THE ROLE OF FIRMS

We leave for the time being transaction costs and hierarchical trans-
actions in order to supply another of the three pieces from which the
theory will be constructed - the theory of reputations that emerges
from noncooperative game theory.

Repeated games, the folk theorem, and implicit contracts

Consider an individual, A, who is playing the following game against
some second party, B. First A must choose whether or not to trust B. If
A elects not to trust B, then both A and B get nothing. If A elects to
trust B, B is made aware of this and has the option either to honor that
100

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
trust or to abuse it. If A trusts B, and B chooses to honor that trust,
both get $10. But if A trusts B and B chooses to abuse that trust, B gets
$15 and A loses $5. A diagram of this game is given in Figure 4.1.
This is a one-sided version of the well-known prisoners' dilemma
game. The salient feature of this game is that, played once and with no
considerations other than those in the previous paragraph, A would
not willingly trust B. For if A does so, then B must choose between
honor, which nets $10, and abuse, which nets $15. Absent other con-
siderations, B will choose $15. So trust will lead A to pay $5, and A is
better off without trust. Of course, this makes both worse off than they
would be if A had chosen trust and B honor - in the language of eco-
nomics, it is an inefficient outcome. But it is the unique equilibrium
outcome of this game, played once and played noncooperatively (that
is, if we assume that individuals are motivated only by the monetary
payoffs involved - part of our qualification "absent other considera-
tions" — and if we assume that they have no opportunity to sign a bind-
ing and enforceable contract - more of the qualification - then this
outcome is the unique self-enforcing outcome of this game).
This is meant to represent the archetypal transaction with some ele-
ment of moral hazard. A, say, must sink some resources into preparing
for a transaction with B, who can (at personal gain) take advantage
of A's position to an extent that makes the entire thing unworthwhile
for A.
"Absent other considerations" can now be examined. One thing that
the two transacting parties might do is to sign at the outset a contract
that binds B to honor. Note that ex ante each will willingly sign such a
contract as long as it is enforceable, because without it each will net
nothing. But such a contract's execution may be costly. B will have an
incentive to violate the contract once A has sunk resources. So some
enforcement mechanism must be provided; this could also be costly.
These are typical examples of Coase/Williamsonian transaction costs. If
they are sufficiently great - if, for example, enforcement costs are
greater than $20, or if the contract cannot be enforced because courts
cannot distinguish between honor and abuse - then this otherwise mu-
tually beneficial transaction will be foregone, the victim of transaction
costs that are too high.
Now suppose that A and B are involved in this situation not once
but repeatedly. Specifically, suppose that after each round of play, there
is a 90 percent chance that they will play at least once more and a 10
percent chance that this (current) round will be the last. Suppose, for
simplicity, that both want to maximize their winnings (less losses) from
the sequence of plays. (If you wish, you can discount those winnings,
but then the product of this continuation probability times the discount
101

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
factor plays the role of the 90 percent in what follows.) Now the anal-
ysis changes dramatically. A could, for example, say to B, "I will begin
by trusting you, hoping that you will honor that trust. Indeed, I will
continue to trust you as long as you do not abuse that trust. But if ever
you abuse that trust, I will never again trust you." If B hears and be-
lieves this statement, B will indeed honor the trust. The following
becomes the relevant calculation: Abuse in any round will increase
the payoff in that round by $5. But weighed against that is the fact that
the payoff will be nothing in all subsequent rounds (if any). There is a
90 percent chance of at least one more round, and if honor is chosen
in that round, then in the next round at least $10 will be obtained, so
the expected profits in the future from honor in this round outweigh
the immediate gain of $5 from abuse.
Note that B must always have some substantial stake in the future if
this is to work. If, say, there is only a 10 percent chance of a continu-
ation of the game or if, say, A only chooses to trust B 10 percent of the
time, then the calculation will come out the other way: B should (op-
timally) take the money and run.
This sort of result is the subject of the so-called folk theorem of non-
cooperative game theory. It is called the folk theorem because it has
been well-known for so long, and no one has the presumption to claim
to have originated it. (Actually, there are many versions of the folk the-
orem. The earliest ones concern games that are infinitely repeated with
probability one and for which an average-payoff-per-round objective
function is used. For a version appropriate to the discounted game we
have posed, see Fudenberg and Maskin 1986.) Roughly, the folk theo-
rem states that we can sustain feasible expected payoffs as noncooper-
ative equilibrium payoffs for players that are sufficiently above the
worst that others can inflict on them. The term "feasible" means that
there must be some way to play the game and get those expected pay-
offs; for example, since the most in present value either player can get
is $150 ($15/. 1), we can't sustain an equilibrium for the game in which
each side nets an expected $10,000. The term "sufficiently" has to do
with the discount factor and with the most players can get in the short
run by defecting from the arrangement. The bigger the discount factor
(the smaller the probability of continuing for another round at least),
the greater must be the payoffs to be sustained as an equilibrium; and
the greater the possible short-run gains, the greater must be the payoffs
to be sustained. One mechanism for sustaining such payoffs as equilib-
ria is in our example. Each player says to the other, "As long as
you stick to the arrangement that gets us the payoffs we are aiming for,
we will continue to cooperate. But if you try to take short-run advan-
tage, you will be punished." As long as the punishment (including at
102

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
least the loss of subsequent cooperation) looms large enough relative to
the gains from cooperation (the source of the qualifier "sufficiently"),
neither will want to defect from the arrangement, which will become
self-enforcing.
Note well what has been accomplished here. We began with a trans-
action that, on the face of it, looked beneficial to both sides. To assure
that one side does not take advantage of the other, we argued that
transaction costs would need to be expended on writing and enforcing
a contract. Should that contract prove too expensive or even impossible
to write or to enforce, then the transaction might not take place at all
(it must be beneficial net of transaction costs). But by repeating the sit-
uation (with sufficiently high probability), we are able to avoid trans-
action costs entirely; the trust-honor arrangement is self-enforcing.
There are three problems with this result:
1. It says how trust-honor might emerge as an equilibrium outcome,
but it allows for many other equilibria as well and doesn't offer any
guidance as to which we will see. For example, in the game we began
with, B might say to A, "I intend to honor your trust two out of three
times and to abuse it once every three, as long as you continue to trust
me. But if ever you choose not to trust me, then I will abuse your trust
every time 1 get the opportunity." If A accepts such a declaration, A's
best response is to trust B and take lumps every third round. The two
$10 prizes outweigh (even with the 10 percent chance of the game end-
ing at any moment) the $5 loss A sustains every third time. The point
is that each player has many feasible expected payoffs sufficiently
above the maximin point of zero that can sustain an equilibrium. The
theory doesn't say which will emerge; it just says that the repeated
character of the situation makes them possible outcomes.
Indeed, the folk theorem requires that a person repeat this encounter
over and over against the same opponent. Williamson might argue that
this transaction thinness actually increases transaction costs; he cer-
tainly lists thinness as one of the cost-increasing qualities of a transac-
tion. We see something of this arising from the richness of the
equilibrium set: We can get many efficient arrangements once we re-
peat the encounter, but the two participants might then expend time
and resources bargaining over which arrangement they will in fact fol-
low. Having arrived at an agreement, they might expend further re-
sources to guarantee somehow that neither side subsequently tries to
renegotiate. Perhaps the cleanest way to think of the power of repeti-
tions is to imagine that many parties of each type fall into two-party
arrangements; a market-determined agreement is reached through sup-
ply and demand for partners. That is, the equilibrium reached between
A and 6 is determined by their opportunities with other trading part-
103

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
ners. We have the thickness needed to avoid Williamson's problems
with thin markets. At the same time, repetition allows partners to
trade efficiently without fixing on paper (or needing to enforce) the
market-determined terms of trade. (The astute reader will see a prob-
lem in this argument: If B has many alternative trading partners, why
not abuse a current trading partner and then find another? Somehow,
abuse must have deleterious future consequences or the entire construc-
tion falls apart. This anticipates what will happen in the next subsec-
tion, so we leave the reader to guess the answer that will be supplied.)
2. Suppose that we modify things ever so slightly as follows: We will
play the game over and over, just as before, with a 90 percent chance
each time of proceeding with another round. But should it ever be
reached, the game will definitely terminate on round one hundred mil-
lion. It seems unlikely that this will have much effect on the way the
game is played (in experimental situations, it doesn't), but the theory
suggests otherwise: If we do reach round one hundred million, then
trust, if given, will surely be abused. There is at this point no future to
be traded off against the current benefit from abusing trust. Thus, in
round one hundred million, if reached, A should offer no trust. But
then if A offers trust in round 99,999,999, it is sure to be abused:
There is no point in B honoring it, because it will not be offered in the
last round. Thus, trust will not be offered in round 99,999,999. And so
on - the whole thing unravels from the back.
This problem, noted first by Selten (1978) and called (in a slightly
different context) the chain-store paradox, can be resolved theoreti-
cally, at the cost of complicating the analysis. This is not an appropri-
ate forum to discuss what ensues, but the basic idea runs as follows:
Suppose that, at the outset, there is a small (say, one in a thousand)
chance that B is the type of person who on moral grounds would never
abuse trust. B knows, of course, whether he or she is of this type, but A
is unsure. This small change is enough to restore for most of the game
the trust-honor outcome.5 Indeed, we get the same result if, say, A is
sure that an opponent is not this sort of person and B knows that A
knows this and A knows that B knows this. But B is a little unsure that
A knows that B knows that A knows that B is not of this type. This can
be modeled formally, and it is enough to get us back to the trust-honor
outcome, as long as a large but finite number of rounds are left to go.
(For the basics of the approach, see Kreps, Milgrom, Roberts, and Wil-
son 1982 and Kreps and Wilson 1982. This approach can help with the
problem (1) of too many equilibrium outcomes, but it does not solve
the problem entirely; see Fudenberg and Maskin 1986.)
3. Suppose that A cannot observe directly whether B chooses to
honor or abuse A's trust. Instead, what A sees is simply a payoff from
104

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
this round. Suppose that A's payoffs are not quite what we have dis-
cussed: If A trusts B and B honors that trust, then A's payoff is nor-
mally distributed with mean $10 and variance $1, whereas trust
followed by abuse nets A a payoff that is normally distributed with
mean $—5 and variance $1. Suppose A makes the speech that we began
this section with: A will trust B as long as that trust is met with honor.
A trusts B in the first round, and then A receives a payoff of $-4. If A
complains that this trust has been abused, B could reply (indignantly)
that this is not the case; that A was simply unlucky. And, after all, this
is a possible (but unlikely) outcome. What does A do? Carry out the
threat and close off all possibilities of future cooperation? Or modify
the threat to punish B (by choosing not to trust) for a long but finite
length of time? And, if the second, for how long? And what should
trigger this punishment?
The point is that when one player cannot observe directly that the
agreement is being carried out, and when this player can only rely on
noisy, indirect observations, the problem of finding self-enforcing ar-
rangements is vastly more complicated. Some loss will necessarily re-
sult from the efficient arrangements, because some punishment will be
required in any arrangement: If A never punishes B, then B will opti-
mally respond by always abusing A's trust. To be a viable mechanism
when there is noise, punishment must be used at least occasionally. In
deciding when and how much to punish B, A must consider that the
quicker or more severe the punishment, the more will be lost during
punishment. But the slower and less severe the punishment, the more
incentive B will have to take advantage of A by abusing trust. (For a
formal analysis of this type of situation, see Green and Porter 1984.)
Repetition allows for the possibility of self-enforcing implicit con-
tracts. We needn't write down the terms of the contract, nor need we
provide an enforcement mechanism for it. But because enforcement is
by punishment when the contract is broken, we must be able to observe
compliance. As we become less and less able to observe compliance, we
become less and less able to use this device at all.
Let me throw in a remark at this point that foreshadows later devel-
opments: When we say that compliance must be observable, we natu-
rally suppose that we understand what "compliance" means. In simple
toy problems like the one under discussion, this is not a very strong
supposition, except for the problem of knowing which of the many
equilibria constitutes the implicit contract. But when we turn to the
real world, in which circumstances arise that no one initially foresaw
or in which so many circumstances arise that it would be too costly to
think through what compliance would mean in all of them, knowing
what the contract calls for is problematic at best. The contract may be
105

Reputation
In the game in the preceding section two participants engaged in a
transaction repeatedly. This would seem to limit the game's applicabil-
ity, because many transactions between individuals do not recur much
or even occur only once. So it seems sensible to ask, to what extent
must the participants to the transaction endure as trading partners?
Suppose that, instead of having one individual offer trust and a
second honor or abuse that trust, we had a sequence of individuals A
who must choose whether or not to trust a single trading partner B.
For the sake of exposition, let us call the sequence of individuals
Au A2, • • • Let us make the following formal assumptions: Aj must
decide whether or not to trust B, and then (if trusted) B must choose
to honor or abuse that trust, with payoffs to the two as before.
Then, with 90 percent probability, B faces the same situation with A2.
And so on.
A moment's reflection should convince you that the following ar-
rangement is self-enforcing. Party B carries a reputation from past be-
havior. For simplicity, we will suppose that B begins with an unsullied
reputation, and B's reputation is irrevocably sullied if ever 6 abuses
trust. Any A will trust B if B has an unsullied reputation, and A will
refuse to trust B if B's reputation is sullied. Then B will always honor
trust, and all As will (in sequence) put their trust in B. This is just like
the self-enforcing agreement of the previous section, except here we see
that only B must be enduring, as long as B's opportunities in later
rounds can be tied to behavior in earlier rounds.
There are two parts to this statement. First, B's behavior, not A's
posed the original problem in getting to the efficient transaction, so in
this case only B must endure. Compare this with the situation in which
A and B are both at risk in the transaction. For example, suppose A
and B were engaged in the well-known prisoners' dilemma. In this
game, A and B must simultaneously and without binding contracts
each choose whether to cooperate or take advantage of the other. If
they both cooperate, both get, say, $5. If one cooperates and the other
takes advantage, the one cooperating loses $1 and the one taking ad-
vantage makes $7. If both take advantage, neither wins nor loses any-
106

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
thing. Played noncooperatively and only once, both sides will choose to
take advantage: Each does better doing so, regardless of what the other
does. This leads to an inefficient outcome where neither makes any
money, compared with the $5 that each can make if each cooperates. If
we repeat this, however, with there always being a, say, 90 percent
chance of playing at least one more time, then cooperation can emerge
as an equilibrium outcome. (Each side will cooperate as long as the
other does and threatens, say, never to again cooperate if the opponent
takes advantage.) The point is that in this case both sides must endure.
If one side played a sequence of opponents, each of whom played only
once, then the opponents would always take advantage, and the one
enduring side would have no reason to do otherwise. Whether it is
enough to have only one side enduring (as it is in our model transac-
tion) depends on the nature of the specific transaction.
Second, there must be a mechanism that ties B's opportunities in fu-
ture rounds to past behavior. It is critical to our story that the As are
able to observe B's past actions and that they condition their behavior
on B's actions. If either condition is not met, then B's incentive to
honor trust in any particular round disappears and, therefore, so does
the incentive of the As to give trust.
Note our use of "reputation" to describe what transpires in this sit-
uation. B has a reputation built up from past encounters, and the As
consider that reputation when deciding whether to trust B. The nature
of the reputation is quite circular — it works because it works: B
guards a reputation because it influences future trading opportunities;
it has this influence because B guards it.
In this game, some caveats must be observed.
1. Just as before, many reputations would work. B, for example,
might have the reputation of abusing trust randomly, with, say, a prob-
ability of one-third each time. If any particular A feels there is a one-
third chance that trust will be abused, that A is still willing to trust B.
Of course, such a reputation will allow B to make greater profits. In
general, we would expect that, in a situation where there are many
long-lived Bs, competition among them would limit the extent to which
they can take advantage of the As they deal with. This fleshes out the
point made earlier concerning competition between Bs. Whether or not
the As are long- or short-lived, we can imagine that they can, in any
round, select from among a number of alternative B trading partners,
selecting the B that has the best general reputation for trust. The Bs
will then compete for As through their reputation. When a single B is
the only possible trading partner for the As, then the economics of the
situation suggests that this B will take maximal advantage of his or her
power, just as any monopolist would.
107

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
2. This story about reputations depends critically on there being
no last round. As before, more complex models can get around this
problem.
3. Reputations must be based on observables in order to work. Am-
biguity and uncertainty cause problems. This point is best made by
considering the reputation previously described above in point 1: 6
abuses trust with probability one-third in any particular round. Before,
with only one A and one B, B could abuse trust every third round,
threatening to abuse trust every time if A didn't trust every time (suf-
fering every third). It would be easy for A to monitor compliance with
this arrangement, because there is no uncertainty about what B will do
in any single round. But this won't work at all if one B faces a se-
quence of As. Threatening to abuse trust every third round would only
result in every third A deciding not to trust. Each A must have a suf-
ficiently large probability that trust will be honored. Hence, B can only
do something like abuse trust each round with probability one-third.
But if B tries to build such a reputation, how do the As know
whether 8 is living up to the deal? Suppose that, as is possible but
unlikely, B abuses trust ten times in a row. Should the As conclude that
B is no longer living up to a reputation? At some point the i4s must
punish B for too much abuse, otherwise B will (optimally) abuse trust
every time. But how severe should that punishment be? When should it
be triggered, and how long should it last? These are difficult questions,
and when real-world ambiguity is added to the game, they become
questions that might never get sorted out. This is getting us closer to a
theory of the type of reputation that might be expected — better to use
a reputation that is easier to monitor. (Note that B could abuse trust
based on some ex post observable random number. For example, B
could abuse trust if the closing stock price of AT&T, on a day after
trust is offered or not but before B must choose to honor or abuse
trust, is an even multiple of one dollar, or a dollar plus one-eighth, or
a dollar plus two-eights. Because it is ex post observable whether B
follows this rule, we have perfect ability to monitor, and no loss in
efficiency from punishment periods is required.)

Firms
In the previous section we made the As short-lived, as long as B en-
dured. But what if the Bs were also short-lived? For concreteness, sup-
pose that at discrete dates t — 1, 2, . . . we have an A, and a Bt who
might engage in the transaction described. Absent binding contracts or
some other costly contrivance, the transaction seems unlikely to take
place. A, cannot trust B, because B, has no incentive to do other than
108

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
abuse trust. We are almost out of business, but with one change we
can resurrect our constructions. We suppose that Bt lives for two dates,
at t and at t+1. At date t, Bt engages in the potential transaction with
At; at date t+1, Bt retires to Florida and lives off savings. Also, we
suppose that B, comes endowed in period t with some resources, per-
haps the fruits of labor early in period t.
Now consider the following arrangement: In this society there is a
partnership called B Associates. At date t (prior to the transaction be-
tween At and Bt), B Associates is owned by Bf_j, who is about to move
to Florida. For a price, B,^ will sell a position in B Associates to Bt,
who would purchase it out of preexisting resources.
Why would Bt pay anything for this place? Suppose that B Associates
has a reputation for never abusing trust. At claims to be ready and
willing to trust Bt, if Bt is a member of 6 Associates but not otherwise.
Then B, will certainly be willing to pay something to B M , in order to
have the opportunity to undertake the transaction with At.
Why should A, trust Bt if Bt purchases a place in B Associates and
not otherwise? Suppose that the As have a somewhat more complex
decision rule: They will trust a member of B Associates as long as no
previous member has ever abused the trust of some previous A. In
other words, a member of B Associates will be trusted as long as the
company's reputation is unsullied. Then Bt, having purchased a place
in B Associates and having received the trust of At, must make the
following calculations: The trust of A, can be abused, which will net
$15 today. But then the reputation of B Associates will be sullied, and
B t + 1 will pay nothing tomorrow for a place in it. On the other hand,
honoring the trust of A, will net only $10 today, but it will preserve
the association's reputation, an asset that can be sold to B, + 1 . The pro-
ceeds from that sale can be used to finance retirement in Florida. As
long as the value of that asset is enough greater than five dollars so
that its discounted value to Bt at date t exceeds $5, B, will, having
purchased a place in the association, optimally honor the trust of At.
And At, realizing this, will happily trust B, if B Associates is unsullied.
There is perhaps less to this argument than meets the eye initially.
Suppose, for example, that the market price of a place in B Associates
is $9. (We'll justify this as the approximate value in a second.) Then
buying a place and protecting its reputation will probably be a good
deal for the Bs. Not to buy will net nothing, because the corresponding
A will not grant trust. To buy and then abuse trust nets $6 today ($15
from dealings with A, less the $9 purchase price) and nothing tomor-
row. To buy and then honor trust nets $1 today and $9 tomorrow.
Assuming a discount rate of 0.9 between the two dates, the last course
is optimal, with net present value $9.10. (Hence the $9 price for a
109

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
place; assuming competition for places, we would have an equilibrium
with a slightly higher price or a slightly greater discount rate.)
Now suppose that instead of this elaborate construction we allowed
Bs to make the following contract with As: B, posts a bond of nine
dollars at date t if A, gives trust, a bond that is forfeit if B, abuses that
trust. But if B, honors the trust given, then B, gets back the posted
bond (perhaps with interest?) in period t + 1. It is hardly surprising
that such contracts, if they can be written and enforced, will lead to the
trust-honor outcome. From a mathematical point of view, this is all
that the invention of B Associates has done for us. Although they are
mathematically similar, the two arrangements work differently. With
the bonding arrangement, even in this simple setting, we must have
some agency that enforces the forfeiture of the bond. Courts would be
natural candidates. That is, the two parties could write up a contract
with the bonding provision, giving them recourse to the courts if the
contract provisions are not fulfilled. But the drafting and (if necessary)
subsequent enforcement of the contract might be costly. Such costs
need to be weighed against the benefits accrued from the transaction.
Indeed, because enforcement must occur after the transaction is com-
pleted, it may not be in the interests of A, ex post to go to court if B,
fails to fulfill the contract. A, certainly would want B, to believe that
he or she will in fact go to the courts, but it isn't clear that this is
altogether believable. What does At gain ex post for expending time
and money? If the threat to go to court is not credible, then the bond-
ing arrangement won't work at all. (A simple cure, it would seem, is to
stipulate in the contract that the bond is forfeited to B.)
Another problem, more nettlesome and more fundamental to what
will follow, arises if honor is observable but not verifiable. The distinc-
tion here is important. A and B and others as well may be able to
observe whether B honors A's trust, but to substantiate this in a court
of law is quite another thing. Verification or substantiation will be re-
quired if we are to have the courts as an enforcement agency, and this
verification may be very costly or even impossible. If the costs of veri-
fication rise above the value to either party of any effective bond that
can be posted, or if verification is simply impossible, a bonding scheme
would not work.
With B Associates, however, the arrangement is self-enforcing as
long as the As can observe whether trust has been abused or not. It
needn't be verified in court; no third party need be employed at all to
enforce the agreement. If future As simply refuse to transact with a
sullied B Associates, then any B's honoring of trust is enforced by self-
interest - namely interest in recovering the full value of the asset orig-
inally purchased.
110

HIERARCHICAL TRANSACTIONS
Consider the following elaboration on the simple game between A and
B. For concreteness, we will adopt a version of the game with a se-
quence of As (labeled, where necessary, Au A2,.. . ), and a single, en-
during B. But what follows can be adapted to the other two scenarios
as well.
In this elaboration, At may hire B to perform some task that requires
B's expertise. It is unclear at the outset how hard B will have to work
to accomplish this task. We suppose that the task is either easy or dif-
ficult, each with probability one-half. The difficulty of the task is irrel-
evant to At in determining his or her value for it. A, simply values it at
$9 if it is adequately accomplished and A, can tell whether or not it has
been adequately accomplished ex post. The difficulty is relevant to B: If
it is easy, a $3 compensation is adequate to B; if it is hard, a $13 com-
pensation is required. A bargaining problem between the two parties
must be solved here, but that bargaining problem is not germane to my
point. So I will assume that for some reason the equilibrium arrange-
ment gives B a $3 compensation if the job is easy and $13 if it is hard,
111

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
as long as this arrangement can be enforced. Also, I assume that this
equilibrium generates positive surplus for B, for reasons that will be-
come apparent later.
We assume that ex ante A doesn't know how hard the job is. Imag-
ine that the two could sign and enforce the following contract: A pays
B $3 if the job is easy and $13 if it is hard, with no payment paid to 6
if the job is inadequately done. Then, assuming that the no-payment
clause gives B the incentive to do an adequate job, A nets $9 in benefit
against an expected payment of $8, making this a worthwhile transac-
tion for A.
Now suppose that this contract cannot be enforced. Specifically, we
assume that its provisions are observable but not verifiable. The dis-
tinction here is just as before: ex post both sides can observe the diffi-
culty of the task and whether the task was adequately performed. But
offering adequate proof of either thing in the courts is impossible.
If both A and 8 deal once only, this will kill the precise transaction
already described. Indeed, the transaction will be killed even if we can
enforce adequacy of performance. This is so because we cannot enforce
payment — A will always want to pay the lesser amount, and B will
want to collect the greater amount. How is the payment amount to be
determined and enforced if we cannot enforce clauses contingent on the
job's difficulty? (If it is possible to enforce such a contingent contract,
we would still need to reckon in the cost of enforcement.) Seemingly,
we must have a payment that is not contingent on the job's difficulty.
Now if B is risk-neutral this might allow another transaction to take
place: The two agree that B will be paid $8 regardless of the difficulty
of the task, which is the same to B as the contingent payment. But the
entire transaction may be rendered infeasible if B is risk-averse. The
certain payment required to get B to undertake the task may then ex-
ceed the $9 value that A places on it. (We could, by elaborating the
situation still further, get other reasons that lack of contingent fees
would make the transaction impossible. For example, we could sup-
pose that A ascribes a higher value to a more difficult task, give A the
ability at some cost to ascertain the difficulty of the task ex ante, and
then put the usual adverse selection argument to work.)
But we are concerned here not with one A and one B but rather with
a B who might perform this service for a sequence of i4s. Then we can
get the contingent fee arrangement, even supposing that adequacy of
performance is observable but not verifiable. That observability is suf-
ficient to ensure adequate performance is our old story: Future As may
refuse to deal with 6 if ever B fails to perform adequately. (Of course,
B must derive enough positive surplus from this transaction so that
benefits from future transactions outweigh any present benefits from
112

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
inadequate performance.) But a similar construction will yield the con-
tingent fee structure. Imagine that A, and B sign a contract that leaves
payment unspecified and at the discretion of B, subject to a limit of
$13. That is, A, agrees ex ante to pay any bill B submits up to that
limit. As adopt the following rule in deciding whether to transact with
B: They will enter into transactions only if B has never gouged a
former client; that is, if B has always charged $3 when the task was
easy and $13 when it was hard. Because we are assuming that diffi-
culty is observable, the As have sufficient data to adopt such a rule. If
the value to B of future transactions exceeds the immediate $10 extra
to be had by gouging a customer, this arrangement is self-enforcing.
In a mathematical sense, there is absolutely nothing new here. If we
simply reinterpret honor to mean to perform adequately and to bill
appropriately and abuse (now multifaceted) to mean to do anything
else, then we are noting that sufficient potential gains from future
trade, combined with As who observe and react to what 6 does in the
right way, suffice to get the trust-honor outcome. The new twist is that
we see how, in certain circumstances, trust could involve agreeing ex
ante to obey dictates from the other party that are ex ante unspecified,
awaiting the resolution of some uncertainty. In this sense, trust can en-
compass transactions that are hierarchical, where one party grants to
the other the right to specify ex post just what the contract in fact calls
for (always within limitations).
Even though mathematically there is nothing novel here, the reinter-
pretation in terms of hierarchical transactions and the connection be-
tween such transactions and reputation is key. We often see contracts
that are ex ante quite vague about what will happen as contingencies
arise. These contracts often include adjudication procedures. There
may be specified recourse to independent arbitration of some form or
another. Further negotiation may be called for (without specification as
to what will happen if those negotiations fail to come to an agree-
ment). In a large number of cases, discretion is left to one party or the
other (always within broad limits). I contend that such transactions are
characteristic of hierarchies; at some point, the hierarchically superior
party, either explicitly or implicitly, has the right to direct the inferior
party. Why would anyone ever willingly enter into such a contract in
the inferior position? It might be because the worst that could happen
is good enough so that the transaction, even on those terms, is worth-
while. But when the superior party has a reputation to protect or en-
hance, a reputation that turns on how that party exercises authority,
then the inferior party need not presume the worst. That party can
count on the superior party to live up to an implicit contract in his or
her own interests.
113

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
The notion of a hierarchical transaction and the connection with rep-
utation goes back to Simon (1951). He makes clear the distinction be-
tween this sort of transaction and the usual exchange of goods for
money that is the basis of standard economic theory. He argues that
hierarchical transactions are particularly prevalent in employment rela-
tionships. Beginning with a transaction in which A presumes the worst
(or, rather, that B will act solely for short-run interests), Simon shows
that in such cases the authority that A willingly grants to B can be
quite small. Then he argues, just as we have done, that this authority
can be enlarged (or the compensation demanded by A decreased) in
cases where A and B may wish to repeat the transaction over time,
with B implicitly threatened by suspension of the transaction for abuses
of hierarchical authority.
Of course, in this particular explanation for hierarchical transaction,
authority must rest with a long-lived party. Suppose that A is long-
lived, participating in this transaction with a sequence of Bs. If there is
no problem in enforcing adequate performance, then the proper hierar-
chical form could have A paying whatever is appropriate, with each 8
agreeing ex ante to let A determine what is appropriate (no less than
$3). (What if adequate performance is not enforceable? With a small
enough discount rate, we could even have A's payment of whatever is
appropriate be zero in cases involving inadequate performance.)
In the story just told, it seems crucial that the task's difficulty be
observable. What would happen if we changed the story so that this is
not so? Note well that this unobservability need only be on the part of
future As; they and not the current A must be able to monitor compli-
ance with the implicit contract. Again, we know the answer from the
previous section: Some inefficiencies will enter, but all is not necessar-
ily lost. If, say, the difficulty of the task cannot be observed, then B
could claim to follow the billing rule already explained but will have
to be watched carefully and punished (by withholding transactions)
occasionally. The As can ascertain whether B is indeed charging $3
around half the time, according to the billing rule described. If there
is a history of too many $13 charges, then the As can (for a while)
refuse to transact with B. Note that it is crucial that B be punished
occasionally. We cannot arrange things so that punishment is avoided
altogether, for that would mean accepting without question any bill
B submits. B would then charge $13 every time. There is no way to
avoid some loss of efficiency when the As are unable to observe the
task's difficulty. But lack of observability needn't completely prevent a
transaction.
Considerations of observability will play a large role in what is to
follow, so let me make this point one more time in a different way.
114

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
Imagine that the task's difficulty is indeed unobservable, but concrete
signs are observable that imperfectly indicate whether the job was dif-
ficult. To be precise, suppose that each job either does or does not re-
quire calculus. Of the difficult jobs, 97.5 percent require calculus, and
2.5 percent don't. Of the easy jobs, 20 percent require calculus, while
the other 80 percent don't. If it is observable whether calculus is re-
quired for the job but not whether the job is difficult, then it might
well be better to base payment on whether the job required calculus
than on whether the job was really difficult. That is, we can, using the
reputation construction, have a self-enforcing arrangement in which
the bill is $13 if the job requires calculus and $3 otherwise, even if the
requirement to use calculus is observable but not verifiable. This may
not be the first-best arrangement; that will be the one for which pay-
ment is based directly on the job's difficulty. But to try to base the bill
directly on difficulty runs us into inefficiencies that exist because the
job's difficulty is unobservable. It may be better to contract implicitly
on some contingency that is clearly observable but that is not the ideal
contingency to base the arrangement on, because what is lost in mov-
ing away from the ideal contingent contract may be regained from the
greater efficiency of the reputation arrangement. (In the spirit of results
by Holmstrom [1979], it is natural to conjecture that the most efficient
implicit contract will not be based solely on the observable contin-
gency. In the simple example we have described, one can show that this
is so. But I am unaware of any general result in this direction.)
A concrete example of this phenomenon may be helpful in under-
standing it. In the United States, it is a commonplace observation that
doctors run too many tests on patients in order to protect themselves
from malpractice suits. The phrase "too many" presumably means that
more tests are given than are necessary, given the doctor's information.
But it is hard to observe (ex post) whether a doctor's subjective ex ante
judgments were professional. It is far easier to observe whether the
doctor followed some pattern of generally accepted practice that allows
for few subjective options. That such a clearly laid-out pattern of prac-
tice is suboptimal relative to the (first-best) application of subjective
judgment is obvious. But it is also irrelevant to any reasonable analysis
of the problem: We have to consider the full equilibrium implications
of having doctors rely more on subjective judgments. (So that no one
gets upset, I am not suggesting that what we see in the United States is
the most efficient feasible arrangement; I haven't thought about it
nearly enough to have an informed opinion. But the simplistic argu-
ment that one often hears is incomplete, ignoring as it does the costs
of monitoring, enforcement, and so on.) Another, quite similar exam-
ple, comes from the accounting profession, where accounts are kept
115

UNFORESEEN CONTINGENCIES
We have now reached the point where orthodox economic theory will
be abandoned because it lacks the two final ingredients of the stew
being concocted. Before we head off into less orthodox waters, let me
offer the following summary of where we've been.
In transactions where one side must trust the other, the reputation of
the trusted party can be a powerful tool for avoiding the transaction
costs of specifying and enforcing the terms of the transaction. Indeed,
when the contingencies upon which the terms are based are observable
but not verifiable, reputation may be the only way to effect the trans-
action. Reputation works as follows: The trusted party will honor that
trust because to abuse it would preclude or substantially limit oppor-
tunities to engage in future valuable transactions. Such a reputation ar-
rangement can work even when the reputation rests in a wholly
intangible entity (the firm), as long as those who make decisions or
take actions in the entity's name have a stake in preserving its reputa-
tion. Among the types of contracts to which this pertains are those that
are hierarchical, where the contract calls ex ante for one party to de-
cide ex post how the contract will be fulfilled, with the second party
agreeing ex ante to abide by the first party's dictates, within broad lim-
itations. Finally, this arrangement works best when the actions of the
trusted party are based on contingencies observable to all concerned;
reputation based on unobservable contingencies is not impossible, but
it will always involve some degree of inefficiency. Put another way, the
best reputation, from the point of view of effecting the type of arrange-
ment we are talking about, is one that is clear-cut and easy to monitor.
All of these conclusions have been derived from orthodox economic
theory. But the examples so far discussed present a fairly weak case for
the importance of this theory. Too much seems to rest on the distinc-
tion between observable and verifiable contingencies. How many of
those are there, and how important are they in real-life transactions?
I contend that there are many such contingencies and that they are
very important, if we stretch this theory to include unforeseen contin-
gencies. An unforeseen contingency is a set of circumstances that ex
ante the parties to the transaction had not considered. Unforeseen con-
tingencies need not be unimaginable: Individuals may simply be unwill-
ing ex ante to spend time thinking through all possibilities, on the
grounds that it is too time-consuming and expensive to do so. Or it
116

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
could be that the circumstances really are ex ante unimaginable. From
the point of view of our development, either interpretation is fine.
Before adding unforeseen contingencies to our analysis of transac-
tions, consider briefly how individuals act when faced with unforeseen
contingencies. Introspection suggests that while a particular contin-
gency may be unforeseen, provision for it is not completely impossible.
While the exact circumstances of future contingencies may be unimag-
inable (or too costly to think through), aspects of those contingencies
can be anticipated. I contend that unforeseen contingencies follow pat-
terns. At least, I, and, I suspect, others, act as if this is so. Accordingly,
my provisions for the unforeseen are somewhat evolutionary. I examine
what has happened that was surprising in the recent and sometimes
distant past, and I provide for roughly similar contingencies.
Formal models of behavior such as this are easy to produce. Imagine,
for example, that I have at my disposal several possible remedies in
varying amounts, remedies that may or may not be applicable to a
wide range of circumstances. Simple examples are cash and fire extin-
guishers. Holding these remedies at the ready is expensive: Cash on
hand doesn't earn interest, and fire extinguishers are costly. The
amount of each remedy that I keep on hand is related to how useful I
suspect it will be. Insofar as I can anticipate relatively greater need for
a remedy in a particular circumstance, I will adjust my holdings up-
ward. (For example, having just bought a house, 1 am holding rela-
tively more cash in my checking account to meet the numerous small
expenses that I discover.) But I will also be guided by the relative use-
fulness of the various remedies in the recent past. (In the month that
has elapsed since I bought the house, I have discovered that I need to
keep more cash in my checking account than I had originally antici-
pated was necessary.) If there is indeed a stationary pattern to unfore-
seen contingencies, then in the long run I will wind up holding correct
levels of the remedies. If, as seems more likely, there is a pattern that is
not stationary but that has secular trends, then although I may not
achieve the best levels of remedy holdings, I will still be better off by
paying attention to recent events than not.
I meander concerning how individuals (or at least how I) act in the
face of the unforeseen because economic theory does not provide an
accepted model of behavior. One knows the appropriate model for, say,
decision making under uncertainty, from the axiomatic development by
von Neumann and Morgenstern, Savage, and others. But there is no
corresponding standard model of decision making through time when
there are unforeseen contingencies.
Although there is nothing to my knowledge written on the subject, it
might be instructive to speculate on what a standard model of choice
117

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
with unforeseen contingencies might look like. The usual models of
choice under uncertainty, of which Savage's theory (1954) is the exem-
plar, presume that individuals can foresee all conceivable future contin-
gencies; they are uncertain only over which contingencies will develop.
Or rather, the standard model posits that individuals act as if that were
so at any point in time. This could continue to be so for choice at any
single point in time, but with one major change: One of the fundamen-
tal pieces of the standard model is the state space — the set of all con-
ceivable future contingencies. This is a given in the model; in the usual
interpretation this piece of the model is objectively fixed, not a part of
an individual's behavior. But if there are unforeseen contingencies and
if individuals attempt at any point to make provision for them, then
one might better regard the state space in an individual's model as part
of the subjective inputs the individual provides. That is, just as in the
standard model the individual's choice behavior reveals probability
assessments and utility function, so in a model with unforeseen con-
tingencies might subjective choice behavior reveal the individual's con-
ception of what the future might hold. In such a model, it would then
become important to speak of how the individual's state space (and
probability assessments and utility function) evolves through time: I
have a strong bias toward models in which past surprises are taken as
guides to what the future might hold. More precisely, such a model
would generate behavior in which provision for future contingencies is
positively influenced by what would have been useful in the past; this
evolutionary behavior will be modeled by corresponding evolution in
the individual's state space and probability assessments. These are no
more than speculations, but I do wish to indicate the importance that
will be attached to the evolution of behavior: The past, and especially
past surprises, will guide how individuals prepare for future surprises.
A model of dynamic choice behavior when there are unforeseen con-
tingencies is an essential precursor to a solid theory of transactions in
the face of unforeseen contingencies. As I do not have and do not know
of the former, it is somewhat dangerous to speculate on the latter. But I
will try, nonetheless. Unforeseen contingencies make explicit and com-
plete contracting impossible. How can we provide ex ante for contin-
gencies that ex ante we cannot anticipate? Yet many transactions live so
long that unforeseen contingencies must be met. An especially good
example is that of individuals seeking employment. In many cases, in-
dividuals commit themselves to more than a day's or a week's work.
Individuals will develop human capital peculiar to a particular firm,
and the employment relationship will be relatively more efficient if it
continues for many years in the future. Insofar as this is so, it is impos-

118

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
sible for workers and the firm to specify everything the workers will
do during their employment. Future contingencies are extraordinarily
complex, involved, and even unimaginable. Consider also the contract
between students and a university. Ex ante, students have little idea
what courses they will want to take, when they will want to take them,
and so forth. There is no possibility of enumerating and providing for
the myriad contingencies that could arise.
In such transactions, the workers and the firm, and students and the
university, agree ex ante not so much on what will be done in each
particular contingency as they do on the procedure by which future
contingencies will be met. The workers and students have certain
rights that cannot be violated by the firm and the university. They have
the right to terminate the relationship at will, but workers and students
usually agree at the outset that, in the face of unforeseen contingencies,
adaptation to those contingencies will be at the discretion of a boss or
dean. That is, the adaptation process is hierarchical, in just the sense of
the last section. (This can be qualified in important ways when work-
ers or students have a body representing their interests — a labor union
or student union, for example.)
Why will workers and students enter into such arrangements? What
protection do they get? It is that the firm or university develops and
maintains a reputation for how it meets unforeseen contingencies by
the way in which it actually meets those contingencies. How the con-
tingencies will be met is not verifiable, at least not in the sense that
workers or students could take a firm or university to court and en-
force a violated implicit contract. But the meeting of the contingencies
is observable ex post; others can see what the firm or university did
and decide whether to enter into similar transactions with either.
Of course, our previous qualifications to the reputation construction
continue to hold. The firm or university does not make decisions: Boss
B or Dean B does so in the organization's name. These people must
have some real stake in maintaining their organization's good repu-
tation. Most crucially, the meeting of unforeseen contingencies must
conform to some pattern or rule that is observable - that is, the orga-
nization's reputation must be for something. This is especially prob-
lematic in the case of unforeseen contingencies. Once they arise, we
may know what they are, but how do we know that they have been
met as they are supposed to be met? Because the contingencies are
unforeseen, we cannot specify in advance how to meet all possible con-
tingencies and then observe that advance specifications have been ful-
filled. We are in a situation somewhat more analogous to that where
the meeting of future contingencies is not perfectly observable, with

119

FOCAL POINTS
Consider the following relatively simple game. Here are eleven letters,
A, B, C, D, H, K, L, M, N, P, S. Assigned to each letter is a number of
points, between 1 and 100. I won't tell you what assignments the let-
ters have, except to tell you that N is assigned the highest number of
points (100) and K the least (1). I ask you and another person, un-
known to you, to pick simultaneously and independently a subset of
these letters. Your list must have the letter S on it, and your opponent's
must have the letter B. Each of you is aware of the requirement im-
posed on the other. Otherwise, you are free to pick as many or as few
letters as you want. We will compare lists and make payments to each
of you as follows: For any letter on one list and not on the other, the
person listing that letter wins as many dollars as that letter has points.
For any letter appearing on both lists, both players must forfeit twice
the number of dollars as that letter has points. If the two lists precisely
partition the set of letters so that each letter appears in only one list,
prizes will be tripled.
Before going further, what list of letters would you submit? If you
were assigned the letter B instead of S, what list would you submit?
Notice that this game has a vast number of equilibria - two lists of
letters, one for each player, such that neither player would wish to
change his or her list unless the other person did so. Namely any of the
512 partitions of the nine other letters constitutes an equilibrium. Yet I
am fairly confident that many readers came up with the list L,_M, N, P,
S. And, putting yourselves in the role of the player assigned letter B,
you thought of the list A, B, C, D, H, and perhaps K. The rule is
simple: alphabetical order. One player takes the first five letters, the
other player takes the last five. The letter K is problematic, because
there are eleven letters. But many of you may have replicated the fol-
lowing argument: The player with the end of the alphabet is getting N,
which is the best letter to have, and K is worth only one dollar, so why
not let the other person have it?
Note that the rule applied here is wholly dependent on context. I
have played a very similar game with Stanford MBA students and Har-
vard undergraduates that comes to a very different solution. I tell them
120

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
I have a list of eleven U.S. cities, namely Atlanta, Boston, Chicago, Dal-
las, Houston, Kansas City, Los Angeles, Miami, New York, Philadel-
phia, and San Francisco. Each city has been assigned a number of
points reflecting its relative importance to commerce, trade, and the
arts in the United States, with New York the highest at 100 points and
Kansas City the lowest at 1. Two students unknown to each other, one
from Harvard and the other from Stanford, are to list simultaneously
and independently some subset of the cities, with the Harvard partici-
pant required to list Boston and the Stanford participant required to
list San Francisco. The game continues as before.
Before reading on, how would you proceed if you had the Stanford
role? If you had the Harvard role?
In a surprisingly large number of cases (my rough estimate is 75 per-
cent), Harvard people select Atlanta, Boston, Chicago, Miami, New
York, and Philadelphia, while Stanford people take the complement
Dallas, Houston, Kansas City, Los Angeles, and San Francisco. When
asked why, the usual response referred to which side of the Mississippi
River the city is on — Harvard gets everything east of the Mississippi,
Stanford everything west. (Kansas City causes problems for individuals
unschooled in geography. Miami also sometimes causes problems,
when people use Sunbelt/Snowbelt division as the principle for their
selections, although the joint presence of Atlanta, which belongs to the
Sunbelt, and Miami, usually causes students to reject this principle be-
cause of the unequal numbers of Snowbelt and Sunbelt cities. Substitut-
ing Detroit for either Atlanta or Miami is a good way to make the
Sunbelt principle live in at least some minds, and substituting Detroit
for Dallas definitely favors it. Putting either Minneapolis/St. Paul or
New Orleans into the list causes great confusion. Foreign students re-
sent this game ex ante - they seem to know what the nature of the
principle applied will be without knowing just what the principle will
be or how to apply it, and they are quick to point out how unfair this
is to them.)
These are examples of focal points. This concept, derived from
Schelling (1960), refers (roughly) to some principle or rule individuals
use naturally to select a mode of behavior in a situation with many
possible equilibrium behaviors. (More precisely, the focal point is the
equilibrium suggested by some focal principle.) Schelling's discussion is
informal but persuasive; he is able to cite many games like these two in
which, somehow, most participants know what to do. He finds that
most focal points can be characterized by simple qualitative principles.
Symmetry is a powerful focal principle, when it can be applied. When
one principle singles out a unique equilibrium and other principles do
not give a clear-cut answer, the first tends to be applied; uniqueness is
121

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
a fairly powerful principle. Often the focal point seems a product of
culture. For example, the use of the Mississippi River to divide the
eleven cities occurs with astonishing frequency to Americans; non-
Americans rarely if ever come up with that means of division.
The notion of a focal point is well outside orthodox economic the-
ory. No formal work I am aware of addresses the concept. (I suspect
that psychologists have something quite substantial to tell economists
on this score, and I am looking forward to getting references from
readers of this book.) So, again, reliance on casual introspection seems
in order.
One point the game examples just described make is that in any par-
ticular situation many focal points may be applied. The individual, try-
ing to decide between them, will look for which one fits best. Which is
most suggested by context? (A geographical rule seems more appropri-
ate to cities than does an alphabetical rule. One wonders if an Ameri-
can student who knows that his opponent is from overseas and is
relatively unschooled in U.S. geography then focuses on the alphabeti-
cal rule?) Which does least violence to other principles? (For example,
the presence of both Atlanta and Miami in the list of cities makes ap-
plication of the Sunbelt principle problematic, since it violates badly
the equal division principle. But if 1 took away Dallas and added De-
troit, then the Mississippi River principle would give a seven-to-four
division, and, I expect, the Sunbelt principle would get somewhat more
play, especially because it lumps New York into a group of five and
Kansas City into a group of six.) A choice of a principle to apply must
be made, and the choice is usually far from capricious.
Second, it seems to me that focal points arise in part because of ev-
olutionary fitness. A good, useful focal principle in a particular situa-
tion will tend to have had successful wide applicability in similar past
situations for the individual using it. This will tend to favor principles
that are more universal or broader and, of course, that are clearer in a
particular context.
A related point, supported by experimental evidence, is that focal
points can be learned - and quite quickly. Roth and Schoumaker
(1983) have conducted the following experiment. Two individuals play
a game in which they bargain over 100 chips. The bargaining proce-
dure is simple: Each simultaneously and independently proposes a
number of chips that he or she would like to have. Let the two initial
bids be xA and xB. If these bids are compatible in the sense that xA +
xB < 100, then each gets the number of chips asked for. If the sum of
the bids exceeds 100, then each is asked to concede or stick to the ini-
tial bid. Those who concede get 100 less the number of chips originally

122

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
requested by their opponents. If one sticks and the other concedes, the
one who sticks gets his or her original bid. If both stick, then both get
no chips. These chips are then redeemed: If a player has won n chips,
that player is given an «/100 chance at winning a monetary prize. The
first player's prize is $10, while the second's is $40. Both players are
told all this, and both are told that the other is being told. Insofar as is
practical, all the above is made common knowledge between the play-
ers. Bidding and conceding/sticking are done via computer; communi-
cation between the two players is strictly limited. Players participate in
this game not once but repeatedly; they are told before each round how
well their opponent has done (in terms of chips) in previous rounds.
As a formal game, the bidding situation described has many equilib-
ria. For any number between 0 and 100, if one player is going to ask
for and then stick to that number of chips, the second's best response is
to ask for and stick to 100 minus that number. That is, there are 101
pure strategy equilibria here. Moreover, there are two somewhat natu-
ral focal points: Split the chips 50—50, or split the chips so that each
player has the same expected value, that is, 80 chips to the first player
and 20 chips to the second. (Another focal point, perhaps to econo-
mists only, is the efficient outcome where one player gets all the chips.)
In previous experiments, Roth found that the second (80—20) focal
point seems to predominate in the population he tested (students at the
University of Illinois).
Roth and Schoumaker add a new wrinkle. Unbeknownst to the par-
ticipants, in their first ten or so rounds of play, they are matched
against a computer. In some cases the computer is programmed to in-
sist on the 50-50 split of chips (or, when the computer is the $10
player, to accede to this split); in other cases the computer is pro-
grammed to insist on or accede to the 80—20 split. After this training
period, players are matched against each other with predictable results:
Those whose training has equipped them to come to exact agreement
stay at that agreement. Those whose training leads to inconsistent de-
mands (one demanding 50 and the other 80) tend not to come to any
agreement. Those whose training leads them to leave chips on the table
(one asks for 20 and the other for 50) tend to head for one of the two
focal points after a while. (Interested readers should consult the paper,
which I am abridging with abandon here.)
The point is clear: Focal points are in part the product of experience.
They can be taught through repeated application. It is a heroic leap
from this experiment to the evolutionary process we have already con-
jectured, but, having made the conjecture, I can now finally throw it
into the stew and emerge with corporate culture.

123

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
CORPORATE CULTURE
Recall where we were in the main development before detouring into
the subject of focal points: Unforeseen contingencies provide both a
golden opportunity for the reputation/hierarchical transaction con-
struction to take life and provide a problem for that construction. The
problem is that the reputation argument turned on the ability of future
potential trading partners to observe and monitor their degree of
compliance with the implicit contract they have with current trading
partners. Current trading partners could enter into a hierarchical trans-
action in the inferior position with equanimity because each could
trust the other to carry out the implicit contract in her or his own
interests, to protect their reputations and safeguard future beneficial
transactions. But the implicit contract is more than implicit in the face
of unforeseen contingencies. Practically by definition, it cannot be clear
ex ante precisely what is called for in a contingency that ex ante has
not been foreseen. So on what might a reputation turn?
It is not logically impossible for the reputation construction to work
without flaw. What is needed is (i) the ability, after observing a partic-
ular contingency, to know what should be done and (ii) a belief ex ante
by the hierarchical inferiors that application of what should be done
will be good enough to warrant undertaking the transaction. Truly un-
ambiguous and universal rules could be imagined that could be applied
ex post to any contingency that arises, and as long as test (ii) above is
passed we meet all of the requirements for the reputation construction
with such rules. (More elaborate, real-life adjudication procedures have
this flavor. For example, the (legal) hierarchical superior may seek
third-party mediation in the case of any disputes and have a reputation
for always abiding by the recommendations of the mediator.)
But it seems unlikely that unambiguous and universal rules exist, at
least in most situations. Or rather, it is unclear that such rules will
exist that at the same time will pass test (ii). (An unambiguous and
universal rule is: "I decide according to my whims." But this will prob-
ably inspire trust insufficient to pass test [ii].) Because every contin-
gency that arises must be dealt with somehow (which is to say that the
rule actually applied must be universal), one should expect ambiguity.
The hierarchical inferiors, and, more importantly, the population of
potential future hierarchical inferiors, will sometimes be uncertain that
the rule is being followed. By analogy with formal work on reputation
based on unobservable or partially observable behavior, we expect
some loss of efficiency.
Yet, as in the formal literature, some loss need not mean complete
loss. One might doubt whether, in the face of unforeseen contingencies,
124

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
there can ever be a rule the application of which is sufficiently unam-
biguous and sufficiently advantageous to the hierarchical inferior to
permit the reputation construction to live at all. But, again arguing by
analogy, I contend that the literature on focal points indicates that such
rules or principles can in fact exist. Faced with a competitive situation
that one has never imagined before, a situation with many equilibrium
actions that could be taken, one is often able to see how to proceed,
applying a general principle. Indeed, this is so even if one is aware of
only the slightest details of one's fellow players and if one must identify
on one's own what principle is appropriate. It would seem that this is
more likely to be true in cases where the fellow player has a long track
record of applying a particular principle in similar but not quite iden-
tical situations.
Moreover, features that make for a good focal principle ought to
make for a good rule to base the implicit contract/reputation on. Ex
post unambiguity, as long as test (ii) is passed, is the sine qua non?
Note in this regard that a quite effective scheme might be to exclude
purposely some contingencies from the rule. That is, the rule is to act
in some specified manner (or according to some principle) in most
cases, but it is not meant to be applied in a few others. Here it will
only be necessary that we can agree, ex post, that a particular contin-
gency was one in which the rule was meant to be applied. "We do
XYZ as long as it makes sense" can be quite effective as long as it
makes sense most of the time and everyone can agree on cases where it
doesn't.
Enter corporate culture. Let us consider first the organization as a
single decision-making entity or as a sequence of decision-making enti-
ties who, when they have decision-making authority, have that author-
ity all to themselves. This entity's problem is to identify a rule that
permits relatively efficient transactions to take place and on which a
viable reputation can be based and then to communicate that rule to
current and potential future trading partners. In the preceding formal
analysis, we supposed a particular equilibrium, giving passing mention
of the problem of multiplicity. In real life, communication becomes
much more of a problem, especially where the rule is to some extent
abstract. The communication of this principle is crucial: Potential trad-
ing partners are judging our decision-making entity on its faithful ap-
plication of the principle; it is clearly important that the entity let
others know just what is the principle. Especially in a world with pri-
vate information, the entity will want to have a principle tailored to it.
Perhaps more importantly, the principle, if communicated well, can be
used effectively as a screening device, warning off potential trading
partners for whom the arrangement will not be good.
125

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
As in most communication problems of this kind, the simpler the
message being sent and the more internally consistent it is, the easier it
will be to communicate. By analogy with focal principles, simplicity
and consistency will be virtues in application. As readers no doubt
guessed long ago, I wish to identify corporate culture with the prin-
ciple and with the means by which the principle is communicated. My
(limited) understanding of corporate culture is that it accomplishes just
what the principle should — it gives hierarchical inferiors an idea ex
ante how the organization will react to circumstances as they arise; in
a strong sense, it gives identity to the organization.
More than this, corporate culture communicates an organization's
identity to hierarchical superiors. Firms and other large organizations
do not have a single decision maker but instead have many individuals
who make decisions in the organization's name. Even if we suppose
that these individuals all have internalized as their objective functions
the common good of the organization, it will be hard for them, in the
face of circumstances they hadn't foreseen, to know how to proceed.
Costly communication will render infeasible completely centralized de-
cision making, yet it will be advantageous to have some consistency
and coordination in the decentralized decisions of the many 6s in the
organization. Corporate culture plays a role here by establishing gen-
eral principles that should be applied (in the hope that application of
that principle will lead to relatively high levels of coordination). But it
is more than merely a coordinating device: It is especially useful in
coordinating the exercise of the organization's hierarchical authority. If
the organization is to have a reputation in its hierarchical transactions,
it must be consistent in exercising hierarchical authority. Thus, the or-
ganization has a crucial task: to communicate the general decision rule
it applies to all those who undertake the actual application. The culture
inside the organization will do this as well - it will communicate the
principle to all concerned.
Corporate culture also provides a means of measuring the perfor-
mance of hierarchical superiors. In many organizations, individuals
have not fully internalized the common good — they are-concerned with
their own welfare. Thus, an organization must monitor and control in-
dividual performance. If individuals within an organization who exer-
cise hierarchical authority are supposed to exercise that authority
according to some clear principle, then it becomes easier ex post to
monitor their performance. (This is simply another variation on our ex
post observability story.) Hence, in the usual fashion, efficiency can be
increased by monitoring adherence to the principle (culture). Violation
of the culture generates direct negative externalities insofar as it weak-
ens the organization's overall reputation. Rewarding good outcomes
126

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
that involve violations of the culture generates negative externalities
indirectly through the chain: This weakens individual incentives to fol-
low the principle and thus increases (potentially) the costs of monitor-
ing and control.8
This, then, is how economic theory (or rather an economic theo-
rist) might try to explain the phenomenon of corporate culture. It
clearly gives corporate culture an economic role to play; indeed, the
role is part and parcel of the organization's role. As such, design
and maintenance of the culture is crucial to efficient organization. If
strategy consists of finding economic opportunities and then maintain-
ing and protecting them, this puts corporate culture in the center of
strategy.
The theory (if something so incomplete and bare-boned can be dig-
nified with that title) given here may indicate that corporate culture
can be rendered in economic terms, but it would be a good deal nicer if
that rendering suggested some consequences. I close by giving a few
suggested consequences, although readers will quickly discern that I
pay more attention to intuition than to any theoretical constructs. (Per-
haps it would be more honest to call what follows my wish list for this
theory, things I would like to derive from a fully developed theory.) I
will also point out conclusions that are not borne out by any of the
formal constructs that I have reported here, but that I think ought to
be obtainable.
1. Consistency and simplicity being virtues, the culture/principle will
reign even when it is not first best. There are three parts to this hypoth-
esis. First, as part of the communication process, the culture will be
taken into areas where it serves no direct purpose except to communi-
cate or reinforce itself. This seems outside the realm of economic the-
ory, which does not have good models of the difficulties encountered in
real-life communication. One possibility suggested by Roth and Schou-
maker, which can be modeled formally, would be to regard application
in irrelevant areas as training for others - that is, to communicate the
principle, we administer it repeatedly so that others learn it. This does,
however, raise the question, why not engage in direct communication
instead? (Or we could spin an allied economics story by discussing the
use of culture as a screen, a device to select from among potential
transaction partners those individuals who are most appropriate to the
transaction desired.)
Second, contingencies will arise in which the principle will not be in
the best interests of the two immediate parties involved, yet it will still
be applied for the benefit of third parties in order to ensure a general
reputation for applying the principle. This follows directly from the
formal constructions discussed in the chapter.
127

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
Third, cases may arise in which everyone concerned understands that
the principle is inefficient, yet still it will be applied. This will come
about in cases where the basis of a reputation is a belief that faithful
application of the principle per se is valuable to a hierarchical superior.
To provide formal justification for this, we would have to use the rep-
utation arguments that turn on incomplete information, such as in
Kreps, Milgrom, Roberts, and Wilson (1982). We have refrained from
giving details on this literature, and this is no time to begin, so simply
note that a key to this argument is that something may be in everyone's
best interests and everyone may know that this is so, but still it may
not be done if, say, not everyone is aware that everyone knows that it is
in everyone's best interests. In formal terms, it may not be common
knowledge.
2. In general, it will be crucially important to align culture with the
sorts of contingencies that are likely to arise. There are planted axioms
here: Unforeseen contingencies in a particular enterprise do follow pat-
terns. Even though one cannot think through the contingencies, one
might be able to predict what principle will be good at meeting them.
Principles are better or worse depending on how they adapt to the con-
tingencies that do arise. At this level, this observation has little sub-
stance. But there is something potentially substantial here: One can
expect difficulties when an organization's mission changes, because
reputations grow and die hard (that is, reputations are assets fairly spe-
cialized and immutable once created). The theory presented earlier is
completely inadequate to provide a reason that this would be so. There
is nothing to prevent reputation from having secular trends or discon-
tinuities, as long as everyone understands what is expected of B in
every situation and 6 fulfills those expectations. In order to derive for-
mally the desired rigidity in reputations, it again seems necessary to
access the incomplete information construction of reputation, with an
appeal to simplicity in the initial conjectures of the As concerning what
motivates B. (This last remark is likely to be incomprehensible to
readers unfamiliar with the reputation and incomplete information
literature.)
(As long as I am putting in remarks incomprehensible to all but the
cognoscenti, let me add another: Note that a story based on personal
quirks of the 6s is not going to fly in the context of the story of B
Associates unless we believe that becoming a member of B Associates
changes B,'s preferences or if we make the characteristics of the various
B/s exchangeable or something similar or if we embellish the story by
supposing that B Associates, as part of its economic function, screens
potential partners. Any of these would get us the probability of a per-
sistent quirk that is necessary to make the incomplete information con-
128

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
struction fly. And the last, which is my favorite, also gets us another
screening role for corporate culture.)
3. What will corporate culture (or the principle) be based on, and
how can we expect it to evolve? I have stressed the evolutionary char-
acter of beliefs about unforeseen contingencies and focal principles. In-
sofar as the principle is meant to provide an adaptation to unforeseen
contingencies and is to be based on something like a focal principle, I
would expect the similar evolutionary adaptation. Of course, I have no
theoretical models to back this up with, and it is a conclusion to which
I would like to add caveats. Suppose one could argue convincingly that
reputation is somewhat rigid and immutable once in place. Then, one
would conclude that the early history of an organization is likely to
play a decisive role in the formation of that organization's culture/prin-
ciple. It is then that the organizational reputation is largely formed,
for better or worse. Because the culture will form at least in part in
response to unforeseen contingencies that do arise, the nature of an
organization will be strongly influenced by early happenstance.
Suppose moreover that interrelated parts of a particular reputation/
principle all live or die together. That is, trust of an organization (and
its adherence to its various implicit contracts) is not perfectly divisible
- violation of one implicit contract raises doubts about the commit-
ment to others. (This could be derived theoretically using the reputa-
tions and incomplete information technology by controlling the
probability assessments held ex ante about managers. The details,
though, seem complex, for the same reasons that make the incomplete
information story harder to tell. Then the evolution of the organiza-
tional culture/principle will be episodic and discontinuous. Events that
cause some portion of the contracts to be violated (or that give greater
opportunity to break those contracts) will tend to be accompanied by
redefinition of the entire culture/set of principles.
4. Since I suggested as subtitle for this paper "The Economics of
the Focus Strategy," let me now make that connection. The theory
sketched in this chapter has a natural extension into considerations of
the optimal size of an organization, when we recast that as the optimal
span of the implicit contract. Of course, insofar as an implicit contract
permits greater transactional efficiency, an expansion in the span of the
contract will be beneficial. But weighed against this is the problem that
as the span of the contract is increased, the range of contingencies
that the contract must cover also must increase. Either it will then be
harder for participants to determine ex post whether the contract was
applied faithfully or the contract will be applied to contingencies for
which it is not suited. Let me take an analogy. If your opponent in the
eleven cities game is an individual, B, who plays games like this often,
129

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
and B has a reputation for using rivers or mountain ranges to divide
the cities, then following play of the eleven cities game, you would have
no problem in checking whether B is living up to this reputation by
using the Mississippi River to divide the cities. But suppose B also plays
other games using the alphabet as the focal principle and still others
using other principles. In playing the eleven cities game, B uses the
principle, among the many applicable ones, that gives the best advan-
tage. Since B did have a number of choices, it would be hard to say
that B did not live up to his or her end of the deal, which is quite
ambiguous. But it is equally true that, in playing with B, you might get
the feeling that his or her selection of rule is not entirely random. Or
imagine that B faces a vast array of games where sets must be divided
and always uses a geographical rule to divide the set. When dividing
sets of fruits, B divides according to where they grow; for books, ac-
cording to where their authors reside; and so on. The problem with
such an opponent is that geographical division will be hard to apply to
a wide range of division games. B's principle is not ambiguous, but it is
appropriate only to some of the possible games.
The point is simple: Wider scope, in the sense of more types of con-
tingencies that must be dealt with, can be dealt with in one of two
ways. One could employ a wider range of principles/contracts, but then
one may increase ambiguity about how any single contingency should
be handled. Increased ambiguity is bad for maintaining reputations. Al-
ternatively, one can keep (in a larger and larger span) the same quite
clear focal principle/implicit contract/corporate culture. But then as the
span or type of contingencies encountered increases, that principle/
contract/culture is being applied in contingencies to which it will be
less and less appropriate. At some point, the benefits from widen-
ing the scope of the organization are outweighed by the inefficien-
cies engendered, and we will have a natural place to break between
organizations.
Of course, transactions will need to take place across the break
points. But these can be transactions undertaken under a different
implicit contract - the contract of the impersonal marketplace, for ex-
ample, where caveat emptor rules. Where we have formal, recognized
breaks, it will be easiest to have a change in the contract. Included here
are formal breaks inside a legal entity - different divisions in a single
corporation (cf. Williamson and the M-form hypothesis), or "plants
within a plant" (see Skinner 1974), and so on.
From this perspective, the focus strategy becomes a strategy of re-
ducing the range of contingencies with which the implicit contract
must deal, in order to deal better — less ambiguously — with those that
are met.
130

CONCLUDING REMARKS AND QUESTIONS

I hope that this chapter has been able to communicate to a wide range
of social scientists answers to the questions: How might an economist
explain the role of corporate culture? How does that explanation tie in
to extant economic theory? What pieces of theory are missing, and
what might they look like when put in place? As I imagine is the case
with most disciplines, economists tend to be intellectual imperialists,
trying to render everything into their own terms of reference. This,
then, has been an exercise in imperialism.
I have little doubt that this is a useful exercise for economic theorists.
Until recently, we have had very little to say of any substance concern-
ing the role of organizations. The theory outlined here is full of lacu-
nae, and it is far from all-encompassing. But, following from a number
of quite active areas of research, it does give a definite role to organi-
zation and culture within the organization. In so doing it helps to think
(in terms of economics) about the role of organization/culture in strat-
egy. (To readers who are economists: Is this so?)
131

APPENDIX
I provide here a simple formal model of some of the ideas discussed in
the chapter. Consider the situation where A must decide whether to
contract with B to perform a task that is either easy or difficult, each
with probability \. A values performance of this task at $9 and is risk-
neutral. B is risk-averse, and the utility B derives from performing the
task and being paid for it depends on the task's difficulty and the
amount paid. We give B's von Neumann-Morgenstern utility as a func-
tion of the amount paid and the task's difficulty (for several values of
the amount paid) in tabular form. (You can check that these numbers
are consistent with utility functions that are concave in dollar
amounts.)

Table 4.1. Utilities for B for certain payment and task difficulty
combinations

Payment to B
$3 $3.40 $8 $9 $12.50 $13
Easy Task 2 2.32 4 4.3 4.9125 5
Difficult -16 -15 -6 -4.4 1.2 2

We will assume that B can obtain utility equal to zero in any period
when not allowed to perform this task for A. Moreover, we assume
that B's marginal utility for money when the task is difficult and pay-
ment is $13 equals the marginal utility for money when the task is
easy and payment is $3. (This is consistent with the numbers given in
the table.)
We will not be explicit for a while as to which version of the story
we are telling - a story with one A and one B or one B and a sequence
of As or a sequence of each type and B Associates — for any of these the
calculations to follow will work. (For the sake of definiteness, we will
speak in terms of the first of these three stories.) Also, we will speak in
terms of an infinite horizon version of the story, with a per-period dis-
count rate, rather than in terms of a probability of continuing. Of
132

The base case: observable difficulty in a repeated setting

If the task difficulty is observable and we have a long-lived B con-
cerned with reputation, then the reputation construction works. Sup-
pose the pair agree that B will submit a bill of no more than $13 after
the task is done (and they know how difficult it was), with the implicit
understanding that the bill will be $3 if the task is easy and $13 if
hard. This gives B a per-period expected utility of 2. Moreover, if B
will be forestalled from any further work of this type by gouging A in
any period (that is, if A will never deal with B again if B does gouge A
once), then by gouging (charging $13 for an easy task) nets B utility of
5 this period and zero ever after. This is as opposed to (at least) utility
of 2 stretching out into the future, which, with a discount factor of 0.9,
has a present utility value of 20. Thus B does better not gouging, and A
can trust B not to do so, giving A an expected gain of $1.
One further point on this case: Our statement earlier on the equality
of B's marginal utilities, combined with A's risk neutrality, makes this
particular contingent (implicit) contract efficient. There is no way to
make B better off without making A worse off, and vice versa. This is
not to say that this is the only efficient arrangement. By lowering pay-
ments, keeping the marginal utilities equal, we will decrease the ex-
pected utility of B and increase the expected payoff to A. Raising both
payoffs will do the reverse. These increases can continue until A's ex-
pected value from this arrangement hits zero - at the point where the
expected payment is $9. On the other side, the decreases in the two
payments must stop before we reach an expected utility level of zero
per period for B, because at zero utility, B will not keep to the equilib-
rium. The constraint for decreases in B's payments comes at the point
where, faced with an easy job, B would prefer to charge the higher
amount and have no further dealings of this sort to continuing with the
implicit contract.
133

v = . 5 ( 2 + .9v) + . 5 ( 2 + . 9 n + l v ) ,

where the first term on the right-hand side is the utility derived if the
task is easy, and the second is the utility derived if the task is hard.
Solving for v, this gives

v = 1 - (.45 + .5 x .9, n + l \

For this v, in order to verify that submitting the correct bill is a best
response by B, we must verify that

2 + . 9 r > 5 + .9n + iv and 2 + .9n + xv > - 16 + .9v,

or 18 >.9(1 - .9")v> 3 .

134

this being the expected monetary value from the immediate period ($1)
plus the discounted average continuation expected monetary value - if
the task is easy, u discounted one period; if the task is difficult, u dis-
counted by twelve periods. This is « = 2.45, as compared to the $10
expected discounted present value that A receives without the observ-
ability problem. That is, both A and B get approximately 25 percent of
their best-expected discounted monetary value/utility when there are
problems of observability and we use this equilibrium.
By giving A the ability to use a slightly more complex strategy, we
can push up these values a bit. If A punishes B for eleven periods,
then B has a strict incentive to send the correct bill. That is, the
punishment above is more strict than need be. We could lessen the pun-
ishment by having A, say, punish B for eleven periods with probability
.99 each time that B submits a bill of $13. Let me be quite precise
about this. I mean to assume that A randomizes, and this randomiza-
tion is observable to B. Suppose, for instance, that there is a month
between B's submission of a bill and the next time A must decide
whether to contract with B, and in this interim period the government
issues, say, a monthly report on the level of Ml. A then announces the
following strategy: Following the submission of a bill of $13, A
will look up the next report on the level of Ml; specifically, A will
look at the third and fourth significant digits of that report. If the third
and fourth digits are, respectively, 1 and 7, A will continue to deal
with B; if the two digits are anything else, A will refrain for eleven
periods. Note well that B can monitor A's compliance with this ran-

135

2 + .9v>S +.9(1 -p)v + .9"+ipv,

where v, as before, is the value of following the correct bill strategy.

This inequality simply ensures that, when the task is easy, submitting
the correct bill does at least as well as submitting the incorrect, in-
flated bill. (There is a second inequality to check — that punishment is
not so severe that B will submit a bill of $3 even when the task is
difficult. We'll refrain from explicitly doing this calculation in what
follows.) And v will be given by the functional equation

v = 2 + (.5 + .5(1 - p)).9v + .5p{.9"+l)v .

If we call 6 = .9(1 - .9")p, the functional equation and the inequality

become

and 3 ov
.56 - '

respectively. Solving for the largest v that satisfies the inequality yields
v - 5 and 6 = .6. (Readers should not be misled by the fact that both
v and the utility of $13 when the task is easy are 5; this is a coinci-
dence. Indeed, as we lower the utility of $13 when the task is easy,
we will raise the value of the game to B, since lesser punishment is
necessary.)
Similar calculations will yield that the expected discounted monetary
value to A of using a strategy with a p and « as above is a simple
function of 6, namely l/(.l + .56), so for any p and n combination
that yields 6 = .6 (just large enough to keep B honest), the value to A
is $2.50.
All this was based on the hypothesis that A could play an observable
randomized strategy (in determining whether to start a punishment pe-
riod). Suppose that A cannot do this. A always has the ability to ran-
136

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
domize, but we suppose now that 6 cannot be given proof that A has
carried out an intended randomization. This will raise problems for the
equilibria above. If we tried to implement the precise 6 = .6 equilib-
rium above, then p cannot be zero or one. And then, once a bill of $13
is submitted, as long as B is submitting correct bills, it is in the inter-
ests of A to go on trusting B. If the randomization cannot be observed
(and penalties cannot be enforced if A fails to follow through), then A
will have an incentive to pretend that the randomization came out in
favor of further trust. This then destroys the equilibrium.
We can get an equilibrium with a p different from one, but it has
an unfortunate property. We do this by (i) supposing that 6 is set so
that, given an easy task, 6 is indifferent between submitting bills of
$3 and of $13, and by (ii) having B randomize between these two bills
in a way that makes A indifferent between triggering punishment
(in the face of a $13 bill) and continuing to trust B. Assuming that u is
the game's value to A, requirement (ii) is .9« = .9 n + 1 «. Of course, this
has a single solution (for « # 1, which must be so to induce correct
billing by B) — namely, u = 0. This simple set of strategies yields an
equilibrium only when the value to A is driven to zero. This does not
raise the value of the game to B, who must be indifferent between
gouging and not, and so the same functional equation defines v (as a
function of 6).
This scenario pertains only to the simple strategies given here. Since
B is having problems with observability, we might consider an A-like
strategy for B - say, gouge with high probability for a while if A ac-
cepts a $13 bill. To compute such an equilibrium gets quite tedious,
however, so we'll leave this notion here.
Observability of randomization is not, of course, a problem when A
is not randomizing. We always have at our disposal the n = 11, p = 1
equilibrium originally given. (A comment is irresistible here, but read-
ers who are getting a bit lost should skip to the next paragraph: Note
the sense in which this pure strategy equilibrium is really the limit of
the sort of equilibrium suggested in the preceding paragraph. A is not
supposed to accept a bill of $13 and proceed blithely on. What keeps A
to this is the supposed out-of-equilibrium behavior by B: Should A ac-
cept a bill of $13 and not begin a punishment period, B responds by
gouging perpetually. This is clearly the limit of a strategy of A accept-
ing a $13 bill, then, with positive probability, gouging for some length
of time.)
We have assumed that A follows a simple type of strategy, and we've
derived from that equilibria that give B up to an expected utility of 5
and A an expected (overall) payment of $2.50. Are there equilibria,
involving more complex sorts of strategies, in which B can do better
137

Using imperfect signals

This analysis presumes that the A's could not observe the task's diffi-
culty and that payment can be based on nothing else of relevance. Now
consider the possibility of there being an imperfect signal that payment
could be based on. Namely, suppose that each task either did or didn't
require calculus. Easy tasks require calculus with probability 0.2, and
hard tasks require calculus with probability 0.975. It is observable (ex
post) whether the task really did require calculus. Even though this is
observable, we assume it is not verifiable. With the usual construction
we can have a scheme in which B is given the authority to submit any
bill up to $13; B bills A, that amount if the task required calculus and
$3 if it didn't. Or, rather, we can make this an equilibrium as long as
both sides have an incentive to participate in this scheme and as long as
B hasn't the incentive to submit a higher bill when the task didn't re-
quire calculus.
To check that this is an equilibrium, note first that it results in the
following round by round: With probability 0.1, the task is easy and
the task required calculus, giving B utility 5. With probability 0.4, the
task is easy and no calculus is required; B's utility is 2. With probabil-
ity 0.4875, the task is difficult and calculus is required; the bill is $13
and utility 2. With probability .0125 the task is difficult and calculus
wasn't required — a bill of $3 and utility —16. In each round, then, the
expected bill is $8,875 — enough to keep A interested in the arrange-
ment. And B's expected utility in each round is 2.075. This gives B a
present value of 20.75 in keeping to the scheme. And so B will be will-
ing to submit the correct bids: The crucial computation is what hap-
pens if the task is difficult and calculus is not required. If B tries to
gouge A by submitting a bill of $13, B gets utility 2 today and (assum-
ing punishment forever) zero thereafter. Submitting the correct bill nets
-16 today and a (discounted) .9 x 20.75 = 18.675 in the future.
Hence B slightly prefers to stick to the arrangement.
Is this an efficient equilibrium among all those that base payments
solely on whether calculus is or isn't needed? No. Recall the assump-
tion about marginal utilities of money to B; this told us that bills of $3
138

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
Corporate culture and economic theory
and $13 gave an efficient arrangement when the task's difficulty was
observable. But now, when the task's difficulty is unobservable, unless
the marginal utilities to B of money at $13 when the task is easy and at
$3 when it is hard are equal (and equal to the other two marginal
utilities), we will not have a second-best scheme. Assuming concavity
of B's utility function, these margins will not equate given the values in
Table 4.1. So we do better for B, holding A's expected value constant,
if we simultaneously increase the smaller payment and decrease the
larger payment in such a way that the expected value of the payment
stays fixed, up until the point where the conditional expected marginal
utilities (conditional on the task requiring calculus or not) are equated.
This, note, will not upset the equilibrium, because B will be more in-
clined to follow the equilibrium in the one case where it matters -
where calculus is not required but the job is hard. (The other con-
straint, that B will not wish to upset the arrangement by charging the
greater amount when the job is easy and calculus is not required, is not
binding.)
How does this compare with the equilibria we computed without the
use of calculus? B is certainly better off: an expected utility of 20.75,
versus 5 before. But the A's are worse off; they make a bare $0,125 per
round in expectation, versus $0.25 on average that they got before.
Happily, we can make both sides better off as follows.
Lower the payment that B receives when calculus is required to
$12.50 and raise the payment when it is not to $3.40. This gives an
expected payment of $8.74625 per period - just better than before. It
gives 8 an expected utility of 18.1675 per period, certainly better than
before. And it is an equilibrium: The constraint to check is that B
doesn't want to break up the deal when a hard job comes along that
doesn't require calculus. By breaking up the deal, B charges $12.50, for
utility level 1.2 this round and zero thereafter. By sticking to the deal,
B must swallow a utility of -15 this round, plus a continuation value
of 18.1675. Discounting the continuation value appropriately, B's net
for keeping to the arrangement is 1.35, and B will keep on.
In the construction just made, we barely made A better off than if
we had ignored the imperfect calculus signal. Note that B is much bet-
ter off than in the equilibrium where we ignored calculus and depended
on punishments — a utility of 18 or so versus one of 5. It would have
seemed that we could borrow more of that surplus from B in order to
make A better off. But this is an illusion. B's utility must be kept fairly
high in the calculus-dependent equilibrium so that, faced with a hard
job that does not require calculus, B will swallow that large dose of
negative utility. It is a very severe constraint to keep B happy enough to
swallow that dose.
139

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
The constraint gets no easier as the odds of a hard job that doesn't
need calculus fall to zero. As long as this has positive probability of
happening and we wish to keep 6 honest in these circumstances, B
must be getting a large continuation level of utility. Note the disconti-
nuity here: When the probability of this is zero, we don't need to
worry about this constraint; then the binding constraint is that B must
be willing to settle for the lower payment when there is an easy job
that doesn't require calculus.
This suggests that, given this imperfect signal, we might well do bet-
ter to stop trying to satisfy this constraint. Instead of breaking cooper-
ation forever if B charges the high amount when calculus is not
required, A will stop dealing with B for a few periods. This punish-
ment will be enough to keep B from charging the higher amount for an
easy job that doesn't require calculus, but it will be insufficient to keep
B from charging that amount for a hard job that doesn't require calcu-
lus. That is, in the equilibrium, B will be expected to charge the higher
amount for hard jobs that don't require calculus, paying for this priv-
ilege with a loss of work for a few periods. And even though this pay-
ment will be worthwhile when the job is hard, it will be too much to
pay when the job is easy (and calculus is not needed).
Specifically, suppose that the two enter into the following arrange-
ment. The bills will be B's choice of $12.50 or $3.00. A bill of $12.50
will be considered warranted if calculus was indeed required, and A
will enter into the agreement again in the next period. A bill of $3.00
is always happily accepted by A — and the agreement will again ensue
in the next period. But a bill of $12.50 when calculus is not required
will be considered unwarranted, and A will refuse to deal with B for
three periods thereafter.
In the equilibrium, B will set the bill at $12.50 whenever calculus is
required or the job is hard. B will bill $3.00 only for easy jobs that
don't require calculus. In this case, one can work out the expected util-
ity for B in any period when the two deal. It is the solution to

v = .1 x (4.9125 + .9v) +.4 x (2 + .9v) + .4875 x (1.2 + .9v)

+ .0125 x (1.2 + .94t/),

which is v = 18.35296. To ensure that this is an equilibrium, we must

ensure that B is willing to charge the lesser amount for an easy job that
doesn't require calculus: Charging the lesser amount nets utility 2 +
.9v = 17.718, while charging the greater amount nets utility 4.9125 +
.94v — 16.95, and we see that B will go along. Note that B is doing
better here than in the equilibrium we computed just earlier.

140

POSTSCRIPT - 1988
In the four or so years that have passed since I wrote "Corporate Cul-
ture," there has been a fair amount of energy devoted to some of the
themes surveyed there. Even given the normal publication lag, it would
be unfortunate not to pay some attention to those themes. At the same
time, it is amusing to reread the last line of the introduction: "readers
will see that the final steps, while not accomplished, are not excessively
difficult to traverse." As some of those final steps are barely closer to
fulfillment now than then, either I misestimated their difficulty or their
inherent interest. I think the former is more likely to be true.
The one point at which, in rereading the chapter, I winced so hard
that I felt compelled to signal that I would say something more is
where I criticized Williamson for having little to say about the relative
advantages and disadvantages of internal organization. (Hereafter, I
will use his term and call it unified governance.) I would have profited
enormously in writing this chapter from a close reading of Williamson
(1985) and, especially in this instance, in reading Chapter 6, on the
inefficiencies of unified governance. To recast slightly what Williamson
has to say there, when one moves from market governance to unified
governance, one trades one set of inefficiences for another, as the na-
ture of what is exchanged changes. In market governance, a decision
maker deals with others only to purchase the outputs derived from
their labors. When we move to unified governance, the single decision
maker can command the use of all the capital, but now that person

141

Downloaded from https:/www.cambridge.org/core. University of Sussex Library, on 03 Mar 2017 at 11:38:30, subject to the Cambridge Core terms
of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9780511571657.006
David M. Kreps
must employ the labor services of those he or she earlier dealt with
only for the output of their labor. The high-powered market-based in-
centives for labor are replaced by lower-powered internal incentives.
Why are market-based incentives more high-powered? Williamson
gives a number of specific reasons, many of which come down to the
fact that it is hard in some cases to preserve ex post in internal orga-
nization the ex ante incentives that one has. To say more will take us
too deeply into a whole new subject, except to say that Williamson
(1985) answers my initial criticism rather substantially. Moreover,
Williamson's more verbal theorizing on this point has been met by a
number of more mathematical contributions, most notably by Holm-
strom (1982) on the general issues of career concerns and by Milgrom
(1986) and Milgrom and Roberts (1987) on quasi-rents and influence
activities within organizations. (See also the treatment of these issues in
the chapter by Milgrom and Roberts in this volume.) Note well that
these papers all speak to the diseconomies of unified governance,
whereas I was criticizing transaction-cost economics for not being ex-
plicit about unified governance's relative economies. (The chapter
by Milgrom and Roberts in this volume looks at both sides of this
picture.) But this is a case where analysis of one side makes clear
the other.
Another point on which I would have benefited from a close reading
of Williamson concerns the dichotomy I draw between traditional and
hierarchical transactions. There is hardly a dichotomy here. Williamson
(Chapter 3) adds to this stark picture bilateral and trilateral relational
transactions and, of course, unified governance. Having concentrated
my own attention on hierarchical transactions, I might accuse him of
not paying enough attention to that form of governance. But it is clear
that the picture is much richer than the naive reader of this chapter
might come to believe.
This chapter is also severely unbalanced in the role it gives to repu-
tation as the sole raison d'etre for the firm. I did wish to stress that
role, because I think it connects the firm and its culture, but I don't
mean to dismiss other reasons for having firmlike organizations. Be-
yond the one paper in the chapter that I do cite (i.e., Grossman and
Hart 1986), there has been a substantial amount of work recently on
those other reasons and especially work on reasons related to the in-
ability to write complete contingent contracts. The very excellent sur-
vey by Holmstrom and Tirole (1987) should be consulted for a more
balanced view of the subject than one gets here.
There have been some substantial advances on the technical fronts
associated with the ideas of the section on reputation. In particular, the
technology appropriate for dealing with noisy observables in the folk
142

143

Charles Larmore - The Morals of Modernity (Modern European Philosophy) (1996)
No ratings yet
Charles Larmore - The Morals of Modernity (Modern European Philosophy) (1996)
238 pages
Tamás Gyulavári - Emanuele Menegatti (Editors) - Decent Work in The Digital Age - European and Comparative Perspectives-Hart Publishing (2022)
No ratings yet
Tamás Gyulavári - Emanuele Menegatti (Editors) - Decent Work in The Digital Age - European and Comparative Perspectives-Hart Publishing (2022)
383 pages
A Government That Worked Better and Cost Less Evaluating Three Decades of Reform and Change in UK Central Government (Christopher Hood Ruth Dixon)
No ratings yet
A Government That Worked Better and Cost Less Evaluating Three Decades of Reform and Change in UK Central Government (Christopher Hood Ruth Dixon)
257 pages
Alchian - Economics of Property Rights
No ratings yet
Alchian - Economics of Property Rights
11 pages
2018 12 How - To - Think - Like - An - Anthropologist PDF
No ratings yet
2018 12 How - To - Think - Like - An - Anthropologist PDF
2 pages
Game Theory Problem Set 1
No ratings yet
Game Theory Problem Set 1
2 pages
Adler Heckscher - THE FIRM AS A COLLABORATIVE COMMUNITY PDF
No ratings yet
Adler Heckscher - THE FIRM AS A COLLABORATIVE COMMUNITY PDF
603 pages
Strøm, Kaare, Wolfgang C. Müller, and Torbjörn Bergman. 2006. Delegation and Accountability in Parliamentary Democracies. OUP Oxford PDF
100% (1)
Strøm, Kaare, Wolfgang C. Müller, and Torbjörn Bergman. 2006. Delegation and Accountability in Parliamentary Democracies. OUP Oxford PDF
783 pages
Coleman-Foundations of Social Theory-Ch 12
100% (1)
Coleman-Foundations of Social Theory-Ch 12
23 pages
Market Failure Tyler Cowan
100% (1)
Market Failure Tyler Cowan
34 pages
Limits of Public Choice
100% (1)
Limits of Public Choice
460 pages
Zerbe, Richard - Economic Efficiency in Law and Economics PDF
100% (1)
Zerbe, Richard - Economic Efficiency in Law and Economics PDF
335 pages
National Sports Governance Observer 2 Final Report
100% (1)
National Sports Governance Observer 2 Final Report
322 pages
Quezon City University: Bachelor of Science in Accountancy
50% (2)
Quezon City University: Bachelor of Science in Accountancy
13 pages
Hooker & Little 2001 Moral Particularism
No ratings yet
Hooker & Little 2001 Moral Particularism
332 pages
Asian Regionalism Gr6
100% (1)
Asian Regionalism Gr6
31 pages
Readings in Public Choice and Constitutional Political Economy
100% (2)
Readings in Public Choice and Constitutional Political Economy
616 pages
Workplace Democracy
No ratings yet
Workplace Democracy
63 pages
(Nozick) Anarchy, State, and Utopia
No ratings yet
(Nozick) Anarchy, State, and Utopia
191 pages
Epistemic Values in Economics: Economic Methodology As A Normative Object
100% (1)
Epistemic Values in Economics: Economic Methodology As A Normative Object
12 pages
Reconstructing The Cornerstone of
No ratings yet
Reconstructing The Cornerstone of
20 pages
Bayesian Econometrcis in Finance
No ratings yet
Bayesian Econometrcis in Finance
89 pages
About Equality: Irving Kristol
No ratings yet
About Equality: Irving Kristol
7 pages
Ever Bilena Cosmetics Inc in Beauty and Personal Care (Philippines)
100% (1)
Ever Bilena Cosmetics Inc in Beauty and Personal Care (Philippines)
4 pages
Mikko Ketokivi, Joseph T. Mahoney - Efficient Organization - A Governance Approach-Oxford University Press (2023)
No ratings yet
Mikko Ketokivi, Joseph T. Mahoney - Efficient Organization - A Governance Approach-Oxford University Press (2023)
313 pages
Braverman
No ratings yet
Braverman
183 pages
TYBMS SEM VI Retail Management TYBMS Marketing Unit I
No ratings yet
TYBMS SEM VI Retail Management TYBMS Marketing Unit I
26 pages
09 - Venture Capital's Role in Financing - Compressed
No ratings yet
09 - Venture Capital's Role in Financing - Compressed
32 pages
Bayesian Methods in Finance: Eric Jacquier and Nicholas Polson
No ratings yet
Bayesian Methods in Finance: Eric Jacquier and Nicholas Polson
92 pages
Petronas Maritime Services Vetting Services Included
100% (1)
Petronas Maritime Services Vetting Services Included
20 pages
The Republican Playbook - Frank Luntz
No ratings yet
The Republican Playbook - Frank Luntz
135 pages
Reconsidering Evidence Based Policy Key Issues and Challenges
100% (1)
Reconsidering Evidence Based Policy Key Issues and Challenges
19 pages
Bourgeois Virtues?: by Deirdre N. Mccloskey
No ratings yet
Bourgeois Virtues?: by Deirdre N. Mccloskey
6 pages
Project
No ratings yet
Project
99 pages
Vietnam Fuji Electric Vietnam Co., LTD Biling
No ratings yet
Vietnam Fuji Electric Vietnam Co., LTD Biling
2 pages
Strategic Implimentation-Organising For Action
No ratings yet
Strategic Implimentation-Organising For Action
35 pages
Badiou - Absolute (Berlin Presentation)
100% (1)
Badiou - Absolute (Berlin Presentation)
15 pages
Tp. Insurance. 2019 - Golden - Notes - Mercantile - Law
No ratings yet
Tp. Insurance. 2019 - Golden - Notes - Mercantile - Law
22 pages
Cowen - The New Monetary Economcs
No ratings yet
Cowen - The New Monetary Economcs
25 pages
Amy Gutmann, Dennis Thompson - Democracy and Disag - 240421 - 190313
No ratings yet
Amy Gutmann, Dennis Thompson - Democracy and Disag - 240421 - 190313
510 pages
Jual Janji
No ratings yet
Jual Janji
6 pages
6 - LOWI, Theodore J. American Business, Public Policy, Case-Studies, and Political Theory - 1964
No ratings yet
6 - LOWI, Theodore J. American Business, Public Policy, Case-Studies, and Political Theory - 1964
29 pages
Module 1 and 3 Assignment
No ratings yet
Module 1 and 3 Assignment
12 pages
Malaysia Airlines
0% (1)
Malaysia Airlines
28 pages
Hayekian Socialism
No ratings yet
Hayekian Socialism
30 pages
Arrow - Some Ordinalist-Utilitarian Notes On Rawls
No ratings yet
Arrow - Some Ordinalist-Utilitarian Notes On Rawls
20 pages
Public Servants To Public Managers: The European Story: Transformation or Torture?
100% (1)
Public Servants To Public Managers: The European Story: Transformation or Torture?
12 pages
Do Institutions Cause Growth
No ratings yet
Do Institutions Cause Growth
33 pages
Aca Assignment: Case Analysis
No ratings yet
Aca Assignment: Case Analysis
5 pages
Regular Expressions Tutorial
No ratings yet
Regular Expressions Tutorial
2 pages
Public Choice and Constitutional Design
No ratings yet
Public Choice and Constitutional Design
42 pages
Creating Public Value
No ratings yet
Creating Public Value
35 pages
Making Globalization Work: Joseph E. Stiglitz
No ratings yet
Making Globalization Work: Joseph E. Stiglitz
26 pages
How To Sell Golden Road 2.0
No ratings yet
How To Sell Golden Road 2.0
4 pages
Roger E.A. Farmer - Prosperity For All - How To Prevent Financial Crises-Oxford University Press (2017)
No ratings yet
Roger E.A. Farmer - Prosperity For All - How To Prevent Financial Crises-Oxford University Press (2017)
297 pages
Regular Expression Cheat Sheet by Charles Severance
No ratings yet
Regular Expression Cheat Sheet by Charles Severance
1 page
Krueger (2007) Full Book To Course Business Cycles
No ratings yet
Krueger (2007) Full Book To Course Business Cycles
111 pages
Lý thuyết tự luận MNC
No ratings yet
Lý thuyết tự luận MNC
8 pages
The Scharff Technique On How To Effectiv
No ratings yet
The Scharff Technique On How To Effectiv
20 pages
Public Opinion Polling and Political Behavior
No ratings yet
Public Opinion Polling and Political Behavior
12 pages
Advanced Accounting: Akuntansi Keuangan Lanjutan SEMESTER GASAL 2015/2016
No ratings yet
Advanced Accounting: Akuntansi Keuangan Lanjutan SEMESTER GASAL 2015/2016
7 pages
Save Capitalism From Capitalists
No ratings yet
Save Capitalism From Capitalists
20 pages
Blake, Global Distributive Justice-Why Political Philosophy Needs Political Science
No ratings yet
Blake, Global Distributive Justice-Why Political Philosophy Needs Political Science
16 pages
Cambridge International Examinations Cambridge International Advanced Subsidiary and Advanced Level
No ratings yet
Cambridge International Examinations Cambridge International Advanced Subsidiary and Advanced Level
9 pages
DOBUZINSKY George Indictment
No ratings yet
DOBUZINSKY George Indictment
7 pages
Qjae14 1 2
No ratings yet
Qjae14 1 2
28 pages
Daly - 1974-The Economics of The Steady State
No ratings yet
Daly - 1974-The Economics of The Steady State
8 pages
The 6 Laws of Technology Everyone Should Know - WSJ
100% (1)
The 6 Laws of Technology Everyone Should Know - WSJ
5 pages
ME 11 Midterm Exam
No ratings yet
ME 11 Midterm Exam
3 pages
National Power and The Structure of Foreign Trade - by Albert O - Hirschman - Publications of The Bureau of Business and Economic - University of - 1150808252 - Anna's Archiv
No ratings yet
National Power and The Structure of Foreign Trade - by Albert O - Hirschman - Publications of The Bureau of Business and Economic - University of - 1150808252 - Anna's Archiv
196 pages
Economic Anxiety or Cultural Backlash: Which Is Key To Trump's Support? (Essay 4)
No ratings yet
Economic Anxiety or Cultural Backlash: Which Is Key To Trump's Support? (Essay 4)
24 pages
The Philosophy of Fre Will-Paul Russel
No ratings yet
The Philosophy of Fre Will-Paul Russel
56 pages
Bill Mitchell 2011 Debt Deficits and Modern Money Theory
No ratings yet
Bill Mitchell 2011 Debt Deficits and Modern Money Theory
6 pages
PADM MA Assignment
No ratings yet
PADM MA Assignment
3 pages
Final Master Thesis Dwi Sasetyaningtyas
No ratings yet
Final Master Thesis Dwi Sasetyaningtyas
196 pages
Role of Sebi in Investor Protection
No ratings yet
Role of Sebi in Investor Protection
9 pages
Practice Exercise Ch18
No ratings yet
Practice Exercise Ch18
3 pages
Principles of Markting
No ratings yet
Principles of Markting
20 pages
FINA4303 - Topic 01 - Pricing of Bonds
No ratings yet
FINA4303 - Topic 01 - Pricing of Bonds
43 pages
Unit 1-2.hi - en Hi To en 2024-04-25 10-15-17
No ratings yet
Unit 1-2.hi - en Hi To en 2024-04-25 10-15-17
72 pages
Contemporary World Prelims Midterm Lessons
No ratings yet
Contemporary World Prelims Midterm Lessons
12 pages
IDFCFIRSTBankstatement 10135408693
No ratings yet
IDFCFIRSTBankstatement 10135408693
31 pages
26-7 NationalGovernment
No ratings yet
26-7 NationalGovernment
88 pages
Summary of Economic and Financial Data May 2025
No ratings yet
Summary of Economic and Financial Data May 2025
16 pages
The Problems of Philosophy
From Everand
The Problems of Philosophy
Bertrand Russell
No ratings yet
Difference Maker: Enacting Systems Theory in Biology Teaching
From Everand
Difference Maker: Enacting Systems Theory in Biology Teaching
Christian Moore-Anderson
No ratings yet
Demystifying Economic Regulation: A Practitioner's Guide
From Everand
Demystifying Economic Regulation: A Practitioner's Guide
H Sama Nwana
No ratings yet
The Greek and Macedonian Art of War
From Everand
The Greek and Macedonian Art of War
Frank E. Adcock
4/5 (10)
The Quest for a 'National' Nationalism: E.J. Pratt’s Epic Ambition, ‘Race’ Consciousness, and the Contradictions of Canadian Identity
From Everand
The Quest for a 'National' Nationalism: E.J. Pratt’s Epic Ambition, ‘Race’ Consciousness, and the Contradictions of Canadian Identity
George Elliott Clarke
No ratings yet
Resounding Events: Adventures of an Academic from the Working Class
From Everand
Resounding Events: Adventures of an Academic from the Working Class
William E. Connolly
No ratings yet

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.

Corporate Culture and Economic Theory

Uploaded by

Corporate Culture and Economic Theory

Uploaded by

Corporate culture and economic theory

WILLIAMSON, TRANSACTION COSTS, AND

GROSSMAN AND HART AND THE RESIDUAL RIGHTS

abuse A's trust

the efficient placement of this authority determines the pattern of own-

REPUTATION AND THE ROLE OF FIRMS

Repeated games, the folk theorem, and implicit contracts

CONCLUDING REMARKS AND QUESTIONS

The base case: observable difficulty in a repeated setting

2 + . 9 r > 5 + .9n + iv and 2 + .9n + xv > - 16 + .9v,

2 + .9v>S +.9(1 -p)v + .9"+ipv,

where v, as before, is the value of following the correct bill strategy.

v = 2 + (.5 + .5(1 - p)).9v + .5p{.9"+l)v .

If we call 6 = .9(1 - .9")p, the functional equation and the inequality

Using imperfect signals

v = .1 x (4.9125 + .9v) +.4 x (2 + .9v) + .4875 x (1.2 + .9v)

which is v = 18.35296. To ensure that this is an equilibrium, we must

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.