Comm. and Reneg. SSRN-Id581721_2

8/8/2019 Comm. and Reneg. SSRN-Id581721_2

1/27

Communication and Renegotiation

Birger Wernerfelt*

August 24, 2004

*J. C. Penney Professor of Management Science, MIT Sloan School of Management,

Cambridge, MA 02142, 617-253-7192,[email protected]. The paper has benefitedgreatly from discussions with Bob Gibbons, Oliver Hart, Tracy Lewis, Rohan Pitchford,

Duncan Simester, Steven Tadelis, Miguel Villas-Boas, and numerous seminarparticipants. Of course, the usual disclaimer applies.

JEL Codes: D2, L2

Key Words: Contract Theory, Communication, Renegotiation.
mailto:[email protected]:[email protected]


2/27

Communication and Renegotiation

Abstract

In a bilateral monopoly, we are looking at the sellers incentives to propose an

improved widget design under two different negotiation rules. With Renegotiation, the

players negotiate a price after a design has been agreed upon, and with Commitment, they

negotiate the price beforehand. The comparison depends on two effects. (1) Bargaining

power: Since communication reveals information about his preferences, a seller with little

bargaining power may prefer to remain quiet in anticipation of ex post negotiation. (2)

Incentive transfer: If the gains from adjustment are large, ex post negotiation gives the

seller a chance to share the gains. On balance, Renegotiation is better when adjustments

are rare but large, while Commitment is better when adjustments are frequent but small.

The comparison might help explain why some contracts have more features left

incomplete and throw some light on the nature of the employment relationship.

2


3/27

I. INTRODUCTION

One of the central lessons of the incomplete contracting paradigm (Grossman

and Hart, 1986) is that the possibility of renegotiation may keep players from

implementing specific investments. In the present paper we show that another cost of

expected renegotiation is that players will withhold certain types of information.On the

other hand, we also show that other types of information will be withheld if contracts can

not be renegotiated.More generally, when identification of efficient trades require one

player to communicate information to the other, should the players negotiate the price

before or after communication?

We identify three effects of renegotiation: It makes prior communication

- (1) unattractive, because senders prefer opponents with bargaining power to know as

little about their preferences as possible (the bargaining power effect),

- (2) attractive, because the ability to negotiate for a share of the gains from a trade can

give a player incentives to help implement it (the incentive transfer effect), and

- (3) attractive, because the revealed information contributes to the efficiency of the

negotiation process (the bargaining efficiency effect).

Put differently, early negotiation is good because players can communicate without

worrying about revealing their preferences, but bad because payments can not be

contingent on what is communicated. The relative magnitudes of these effects depend on

the extent to which communication reveals preferences, the gains from communication,

and the players relative bargaining power.

3


4/27

The analysis compares a sellers incentives to propose an improved widget design

to a buyer in two different game forms. In the Commitment game form, the players first

negotiate price and then agree on a design, while the sequence is reversed in the

Renegotiation game form. We imagine that the new design may or may not be more

costly to produce. So if the players are committed to a fixed price, the seller will only

propose it (communicate about it) if it in fact is cheaper for him. Since the new design is

better (more valuable for the buyer), this is inefficient. However, if the seller has

bargaining power, renegotiation may help by allowing him to appropriate part of the

incremental value. This then gives him incentives to communicate even if the new design

is unattractive on a strict cost-basis. However, the argument is complicated by the fact

that the buyer can use the sellers communication-decision to make inferences about his

costs of the two designs. If he proposes the new design, it must be cheap to make, and if

he does not, the old design must be cheap. We rig the model such that the buyer, ceteris

paribus, learns more if the seller communicates than if he stays quiet. In particular, we

assume that the seller only sometimes knows about the improved design. So if he

communicates, the new design is known to be relatively cheap, but if he does not

communicate, the old design could be expensive. To the extent that the buyer has

bargaining power, this reduces the sellers incentives to communicate.

After discussing some related literature in Section II, we present the core

argument in Section III, looking at a model in which the buyer has bargaining power. The

analysis shows that Renegotiation is better when adjustments are rare but large, while

Commitment is better when adjustments are frequent but small. We show in Section IV

that more communication also can be implemented by reallocation of bargaining power

4


5/27

to the buyer. However, we argue that this alternative instrument is likely to be quite blunt.

More sophisticated contracts, an application to the theory of the firm, and empirical

evidence is discussed in Section V. We end with a brief conclusion in Section VI.

II. RELATED LITERATURE

Our effects are known from other contexts, but we here interpret them differently

and in many cases reverse the direction of the argument. The bargaining power effect is a

cost of renegotiation and as such fits into a large literature on the appropriate allocation

of bargaining power. Most work in the area is concerned with hold up problems, but

Farrell and Gibbons (1995) look at the incentives to communicate. They invoke intuition

similar to ours and show that intermediate allocations of bargaining power are necessary

to give both parties incentives. The present analysis is different because we are varying

the game form - comparing ex ante and ex post bargaining while holding bargaining

power fixed. (In fact, our version of the bargaining power effect is more closely related to

the idea that offering a contract reveals information about the offerer as in Aghion and

Bolton (1987) and Hermalin and Katz (1993)). Another way to limit the harm done by

renegotiation is to commit against doing it and a number of other papers have shown how

such a commitment can be sustained by a commitment not to acquire certain pieces of

information (Cremer, 1995, Dewatripont and Maskin, 1995; Spier, 1992). The bargaining

power effect studied in this paper makes the causality go the other way: A commitment

not to renegotiate influences the amount of information revealed.

The incentive transfer effect plays a role in models in which incentive systems

may cause agents not to reveal information that is of value to their principals

5


6/27

(Dewatripont and Tirole, 1999; Rotemberg and Saloner, 1993). The arguments are

variants of the hold-up problem. If only one player has search costs, renegotiation can be

bad, because it may prevent a player from getting enough expected surplus to warrant the

specific investment. In the present paper we highlight the opposite case in which it may

be efficient to spread the surplus in ex post negotiation, rather than concentrate it in ex

ante negotiation. This then provides a possible rationale for contractual incompleteness.

Finally, the bargaining efficiency effect is exploited in a lot of signaling and screening

games in which revelation of a players type allows implementation of previously

impossible trades (Akerlof, 1970; Rothschild and Stiglitz, 1976; Spence, 1973).

The most closely related paper is by Gertner (1999), who also is concerned with

withholding of information. He compares a game form with a neutral arbitrator (a firm)

and an asymmetric game form in which one party has control (a market). So

information is sent to a neutral party in one, but an adversary in the other. Ergo, less

information is sent in the latter case. In the present paper, information is always sent to an

adversary, but either after or before negotiation. Depending on parameter values, we find

that there may be more communication with or without renegotiation. On the other hand

both papers predict that concerns about bargaining power ceteris paribus will hamper

communication.

III. BASIC MODEL

A seller and a buyer may trade a widget. The initial plan is to trade the old

design, but with probability < 1, the seller can also produce a new and improved

design. It is common knowledge that the buyer values the old design at 1 and the new

6


7/27

design at rn > 1. The seller incurs costs co for the old design and cn for the new design.

These costs are privately known, but i.i.d. draws from a uniform distribution over [0,1].

Only the seller knows if he can produce the new design and if he can, he may propose it

to the buyer who then may select either of the two designs. Because she can invert the

sellers equilibrium strategy, the buyer may get information about the relative magnitudes

ofcoand cn from the sellers decision to propose the new design or not. However, since it

is possible that the seller can not produce the new design (< 1), a failure to propose is

generally less informative than a proposal because it leaves open the possibility that co is

high. These inferences will be important because we do not allow contracts to depend

on whether a design is old or new, nor on whether a proposal was made. (We imagine

that outsiders understand very little about the products and that neither the truthfulness,

nor the occurrence, of communication is verifiable.)

We model the bargaining game by assuming that one of the players has all the

bargaining power and makes a take-it-or-leave-it (TIOLI) offer. In this Section, we

assume that the buyer makes the offer. This negotiation mechanism is clearly arbitrary

and in general less than second best. However, it is very simple to work with, and it has

the property that bargaining is more efficient when players have better information, such

that we can isolate the bargaining efficiency effect.

Given the situation described above, we will compare two game forms. In the

Commitment game form, the players make decisions about the design in light of an ex

ante negotiated price, while the price in the Renegotiation game form is negotiated ex

post, once the design is known. It is clearly possible to look at a number of other game

forms. In particular, one could imagine a game form with some intermediate degree of

7


8/27

commitment (Aghion and Tirole, 1997; Rogoff, 1985) to an ex ante negotiated contract.

However, as a first cut, we will focus on a comparative analysis of the two extreme cases

with full and no commitment.

The Commitmentgame form is defined by the following six stages:

1. The players write a contract, which says that the sellerwill deliver an unspecified

design as chosen by the buyer. In return, hegets w if a trade is made. The players

negotiate w according to the bargaining mechanism described above. If negotiations

fail, the game ends with (seller, buyer) payoffs (0, 0).

2. The seller learns whether or not he can produce the new design.

3. The seller learns his costs for the old design, and if he can produce it, those for the

new design as well.

4. The sellermay propose the new design to the buyer.

5. The buyermay ask the seller to deliver a specific design. If she does not, the game

ends with total (seller, buyer) payoffs (0, 0).

6. The seller may deliver the design in question, and if he does, the (seller, buyer)

payoffs are (w - cn, w + rn) when the new design is delivered, and ( w - co, w + 1)

when the old design is delivered. If the seller does not deliver, or delivers the

wrong design, the game ends with total (seller, buyer) payoffs (0, 0).

TheRenegotiation game form is identical except that negotiation takes place after the

players have agreed on a design. It is formally defined as follows:

8


9/27

1. The seller learns whether or not he can produce the new design.

2. The seller learns his costs for the old design, and if he can produce it, those for the

new design as well.

3. The sellermay propose the new design to the buyer.

4. The buyermay ask the seller to deliver a specific design. If she does not, the game

ends with total (seller, buyer) payoffs (0, 0).

5. The players negotiate w according to the bargaining mechanism described above. If

they succeed, they write a contract specifying that the seller will get w in return for

delivering the chosen design. If negotiations fail, there are no further payoffs and the

game ends with total (seller, buyer) payoffs (0, 0).

6. The seller may deliver the design in question, and if he does, (seller, buyer) payoffs

are (w - cn, -w + rn) when the new design is delivered, and (w - co, -w + 1) when the

old design is delivered. If the seller does not deliver, or delivers the wrong design,

the game ends with total (seller, buyer) payoffs (0, 0).

The first best calls for the new design to be chosen when rn - cn> 1 - coand forall

bargains to succeed. However, both game forms will have trouble inducing the seller to

propose the new design in sufficiently many cases, and both will suffer some failed

bargains. Since the Commitment game form is easier to analyze, we look at that first.

9


10/27

The Commitment Game Form.

With ex ante negotiation, communication and choice can not influence

bargaining, so the equilibrium is relatively simple. Analyzing from the back, we first

assume that the price is less than 1. In this case, the buyer will always want the new

design because she values it more than the old design or the price. The sellerwill propose

the new design wheneverco > cnand he will want to deliver if the price is larger than the

smaller of the two costs. The buyers best TIOLI offer below 1, call wb, will be

wb= Argmax[(rn/2 + - w)Prob(w > Min{co, cn}) + (1 - )(1 w)Prob(w > co)]

= Argmax[(rn /2 + - w)(2w w2

) + (1 - )(1 w)w] =

= [rn + 3 + 2/- ( rn2

+ 9 + 4/2 - 6rn + 4rn/)1/2

]/6. (1)

Since this is independent of the realized costs, the seller communicates iff he can produce

the new design and cn < co. The probability of this is /2.

If the price exceeds 1, the buyer will only want to trade for the new design, the

seller will always want to deliver, and the seller will propose it if possible. So by setting

the price slightly above 1, the buyer can commit herself to refusing any trade for the old

design and thus induce the seller to communicate with probability . Since this

guarantees her close to (rn 1) in expected payoff, she will prefer it if

(rn 1) > (rn /2 + - wb)(2wb wb2) + (1 - )(1 wb)wb, (2)

where wb is defined by (1). After using the envelope theorem and doing a bit of algebra,

we can see that (2) is more likely to hold for larger values ofrnand. This makes

intuitive sense. When rnandare large, the new design promises large expected payoffs

and the buyer is willing to pay extra and forego the old design for the chance to cash in

on the new. We can summarize this in

10


11/27

Result C:The Commitment game form has a unique equilibrium and the seller

communicates with probability or/2 depending on whether or not (2) is satisfied.

The Renegotiation Game Form.

The game form with renegotiation is in general much more complicated. The sellers

decision about whether or not to communicate to depends on (co, cn) [0, 1]2, and the

buyers TIOLI offers in turn depends on the posterior cost distibutions induced by the

equilibrium communication strategy. Fortunately, the following Lemma greatly simplify

the analysis.

Lemma: There exists a [-1, 1],such that it is a dominant strategy for the seller to

communicate iffcn co.

Proof: Let wn and wo be the buyers TIOLI offers for the new and old design,

respectively. When the buyer makes her choice between the new and the old design, she

only has the sellers communication decision to condition on. Suppose first that the buyer

selects the new design whenever possible.If the seller communicates, his expected

payoffs are Max{wn- cn, 0}, while they are Max{wo- co, 0} if he does not communicate.

Since the buyers offers reflect only her beliefs about the sellers communication strategy

and do not change with the actual strategy, it is a dominant strategy for the seller to

communicate iff he can produce the new design and cn co < wn

wo. If the buyer does

not select the new design, the seller can avoid revealing any information by setting = 1

and pooling. QED

11


12/27

We can solve the game backwards under the assumption that the sellers

communication strategy is completely characterized by the parameter[-1, 1]. To find

the buyers TIOLI offers in stage 5, we look for the posterior cost distributions as

functions of. There are three possible histories up to stage 5.

History 1: The seller did not propose the new design. This leaves open two cases. One is

that the seller is not able to produce the new design, and the other is that he can produce it

but has chosen not to propose it. To construct the posterior distribution ofco, we start by

finding the posteriors conditional on each of the cases - first assuming that = 0 and then

that = 1. The case where = 0 is easy: If the seller did not communicate because he

can not produce the new design, the probability that w > co is just w. The case where =

1 forces us to look at two different regions. (i) If> 0 the conditional posterior density is

triangular and decreasing, such that the probability that w > cois [2w(1 - ) - w2]/(1 - )2

for 0 < w < 1- . (ii) If< 0 the conditional posterior density is shaped like a square less

a triangle in the upper right corner and the probability that w > cois w/[1/2 -2/2 - ] for

w < - and [w (w +)2/2]/[1/2 -2/2 - ] forw > - . Since the buyer can not tell the

two cases apart, she will make her TIOLI offer based on the unconditional posterior. We

can find the posterior probabilities that the seller can not produce the new design as

() (1 - )/[1 - (1/2 + + 2/2)] for > 0, and

(1 - )/[1 - (1/2 + - 2/2)] for < 0.

Given this, the posterior distribution ofco is

F1(w|) ()w + [1 - ()][2w(1 - ) - w2]/(1 - )2 for > 0 and w < 1 - ,

12


13/27

()w + [1 - ()]w/[1/2 - 2/2 - ] for< 0 and w < - , and

()w + [1 - ()][w (w +)2/2]/[1/2 -2/2 - ] for< 0 and w > - .

So we can find the buyers TIOLI offer after History 1 as

wo()Argmax(1 w)F1(w|). (3)

We can use standard tools to derive several properties of the buyers expected

payoff and his offer. Note first that are both continuous functions of, that is

decreasing in , andthat the distributions [2w(1 - ) - w2]/(1 - )2, w/[1/2 - 2/2 - ], and

[w (w +)2/2]/[1/2 -2/2 - ], all dominate w on [0, 1]. Sowe can appeal to the

envelope theorem to conclude that the buyers expected payoff is an increasing function

of, and we can use the implicit function theorem to see (after some algebra) that wo is

weakly decreasing in . Neither result is surprising, because larger values ofincreases

the likelihood that the buyer is facing a more favorable cost-distribution.

The effects ofare more complicated. The buyers expected payoff and his offer

are both continuous functions of, but it affects (3) in two different ways. Reflecting the

fact that() is increasing, higher values ofincrease the posterior probability that the

seller had no choice, implying that the buyer more likely faces relatively unattractive

(uniform) cost distribution. However, if the seller had a choice and decided against

communicating, the distribution faced by the buyer is more attractive for higher values of

. The former force, which we will call the bargaining power effect, dominates for small

and the latter for large . So when is close to zero, the buyers expected payoff are

decreasing in , while wo is increasing in . Conversely, the buyers expected payoff are

increasing in , while wo is decreasing in when is close to one.

13


14/27

History 2: The seller proposed the new design but the buyer chose the old. In this case the

posterior density ofcois less attractive than the uniform. If< 0, the density is shaped

like an increasing triangle and the probability that w > co is (w 1 - )2/2 forw > 1 + .

If> 0, the density is shaped like a square less a triangle in the upper left corner. In this

region the probability that w > cois [w + w2/2]/[1/2 -2/2 + ] for 0 < w < 1- , and [1/2

-2/2 + w (1 - )]/[1/2 -2/2 + ] for 1 - < w < 1. If we denote this distribution byF2,

the offer that maximizes expected payoffs for the buyer, call wo, is

wo() = Argmax(1 w)F2(w|). (4)

The expected payoffs are again a continuous function ofand sinceF2 becomes more

attractive as increases, we can use the envelope theorem on (4) to conclude that the

buyers expected payoffs are increasing in .

History 3: The seller proposed the new design and the buyer chose it. If< 0, the

posterior density ofcnis shaped like a decreasing triangle and the probability that w > cn

is 1 (1 + - w)2

/(1 - )2

forw 0, the posterior density ofcnis shaped like

a square less a triangle in the upper right corner. We can find the probability that w > cn

as w/[1/2 -2/2 + ] for 0 < w < , and [w (w - )2/2]/[1/2 -2/2 + ] for < w < 1. If we

denote this distribution byF3, the buyers TIOLI offer will be

wn() = Argmax(rn w)F3(w|). (5)

The expected payoffs are once again a continuous function of. As increases,

the density becomes less favorable implying that the buyers expected payoffsare

decreasing in , while wnis increasing in .

Going now to stage 4, the buyer will choose the new design iff

14


15/27

[rn wn()]F3(wn()|) >[1 wo()]F2(wo()|), (6)

where wn() and wo() are given by (3) and (2), respectively. We have seen that the left

side is decreasing in , while the right side is increasing. So (6) defines a critical value,

call (rn), such thatthe buyer always chooses the new design ifis smaller than (rn),

and never chooses it if> (rn). As the problem has limited interest if the buyer never

chooses the new design, it would be convenient if> 1, such that (6) holds for all values

of. A sufficient condition for this is [rn wn(1)]F3(wn(1)|1) >[1 wo(-1)]F2(wo(-1)|-1),

which reduces to rn2/4 > 4/27. So it follows from rn > 1 that the buyer always chooses the

new design.

Given the above, we can disregard History 2 and conclude that the seller will

communicate in stage 3 iffwn() cn > wo()co, such he prefers History 3 over History

1. According to the Lemma, an equilibrium is defined by a * satisfying

* { wn(*)wo(*), 1}. (7)

The existence of at least one * follows from the facts that wn()

- wo() is continuous,

wn(-1)wo(-1)= 0 > -1,and wn(1)

wo(1)= rn/2 > 0. So there is a * < 1 ifrn< 3.

Furthermore, since wn(0)wo(0)= 1 (1/3)

1/2- < 0, there is a * < 0 if 0 and rn = 1.

We can summarize the above analysis in:

Result R:The Renegotiation game form has at least one equilibrium. Any equilibrium is

characterized by a * [-1, 1] and the seller communicates iff cnco


16/27

The result combines at least two interesting effects. First, there is more

communication for large values ofrn due to the incentive transfer effect. Because the

buyer wants to be sure that her offer is accepted, she will bid more and thus transfer some

of the incentives to the seller. Secondly, ifis small, the bargaining power effect may

cause the seller to remain silent even ifcn< co. Not communicating allows him to extract

a better offer by hiding behind the 1 - probability that he could not produce the new

design. We get * >0 and more communication when the incentive transfer effect

overwhelms the bargaining power effect.

It is difficult to make a global comparison with the Commitment game form, but

we can characterize the relative intensity of communication in several different regions of

the parameter space. (i) Ifrnandare very large, the probability of communication is

in both game forms. (ii) Ifrn is large and is small, the probability is with

Renegotiation and /2 with Commitment. (iii) Ifrn is small and is large, the probability

is less than /2with Renegotiation and /2 with Commitment. (iv) For intermediate

values ofrnand, Renegotiation implements communication with probability between

and /2, while the probability is /2 in the Commitment game form. (v) For very small

values ofrnand, Renegotiation implements less communication than Commitment.1

The comparison between regions (ii) and (iii) shows that Renegotiation is better when

adjustments are rare but large, while Commitment is better when adjustments are

frequent but small. We can intuitively understand this as a result of the tradeoff between

the incentive transfer effect and the bargaining power effect.

16


17/27

IV. THE SELLER HAS POWER

To see how the effects are driven by the allocation of bargaining power, we now

look at the case in which the communicating party makes the TIOLI offer. We start with

the Commitment game form.

The Commitment Game Form

Again analyzing from the back, we first assume that the price is less than or equal

to1. In this case, the buyer will always want the new design, the sellerwill propose it

wheneverco > cn,and he will want to deliver if the price is larger than the smaller of the

two costs. If the price exceeds 1, the buyer will only want to trade for the new design, the

seller will always want to deliver, and the seller will propose it if possible. When the

seller makes the TIOLI offer, call ws, she will want to set it equal to 1 if

1 (1 - )/2 - EMin{co, cn} > (rn - ), (6)

and otherwise equal to rn. So ws = 1 for small values ofrnand . In such cases, the seller

communicates iff he can produce the new design and cn < co. For larger values ofrnand

, the seller sets ws = rn and then communicates whenever he can make the new design.

Simplifying (6) and summarizing, the probability of communication is /2 or

depending on whether or not rn< 1/3 + 1/(2).

Compared to the case in which the buyer makes the TIOLI offer we find the same

basic effect. When rnand are large, the new design promises a lot of expected surplus

1Renegotiation may be more efficient in some cases when * < 0 because the buyer makes her TIOLI offer

with better information. However, the Commitment game form will be more efficient for sufficiently small

values ofrnand.

17


18/27

and the price is set to ensure that the probability of communication is maximal at . For

smaller values of one or both of the parameters, the probability of communication is

again /2.

The Renegotiation Game Form

Since the seller makes his TIOLI offer with complete information, it is obvious

that the will charge 1 for the old design and rn for the new. This makes the analysis of the

Renegotiation game form much easier. Note first that the Lemma continues to hold.

Assuming that the buyer selects the new design whenever possible, the sellers expected

payoffs from communication are rn - cn, while they are 1 - co if he does not communicate.

It is a dominant strategy for him to communicate iff he can produce the new design and

cn - co< rn -1. So * = Min{rn 1, 1}, and the probability of communication is ifrn > 2,

and (2rn - rn2/2 1) forrn [1, 2].

2

Comparing with the case in which the buyer makes the TIOLI offer we here see a

somewhat different picture. The communication and offer strategies implement the first

best and neither the bargaining power effect nor the incentive transfer effect plays any

role. This is once again intuitively appealing, since all the bargaining power and therefore

all surplus is given to the player who communicates. The distortions discussed in Section

III appear because communication and power are misaligned.

The finding that communication can be induced by an appropriate allocation of

bargaining power is consistent with Farrell and Gibbons (1995). There are, however,

many reasons to doubt the feasibility of this solution: (i) Bargaining power may not be a

control variable, for example if it is tied to market position. (ii) Other forces, such as a

18


19/27

desire to minimize hold up problems, may influence the optimal allocation of bargaining

power. (iii) The parties may be communicating about several issues, suggesting different

allocations of power. (iv) There may be uncertainty on both sides. For example, suppose

that rn equals Rn with probability and otherwise is one. In this case the sellers TIOLI

offer will beRn or 1, depending on the relative magnitudes of(Rn -cn) and 1 cn. If the

latter is larger, he will offer 1 and communicate with probability (iffcn < co). So in this

case nothing is accomplished by reallocating bargaining power.

There is no question that some communication problems can be solved by

reallocation of bargaining power, but a change in game form is an alternative instrument.

A significant advantage of solution-by-game-form is that it can be done on a feature by

feature basis as a contract is designed. A disadvantage is that the solutions are likely to be

imperfect.

V. DISCUSSION

In this Section, we generalize the Commitment game form, argue that it has

several properties in common with an employment contract, and review some empirical

evidence with this in mind.

Aligning Incentives in the Commitment Game Form

One way of thinking about the inefficiencies in the Commitment game form is that

the players gains from adjustment are uncorrelated. The players could reduce the

problems caused by the incentive transfer effect by increasing the correlation, thereby

raising the efficiency of the commitment game form. As we saw in the Result C, if the

2 Compared to the Commitment game form, we see that Renegotiation is better because the price can be

19


20/27

seller has an idea that will increase the buyers revenues, he will still not communicate it

unless it reduces his own costs. However, he will be more cooperative if the contract

gives him a different payment when the buyer has large gains. (This is of course just an

example of the principle that an ex ante contract is more efficient if it is closer to being

complete.)

Even if we maintain the assumption that it is impossible for the players to contract

directly on the nature of the design itself, there are many other ways of inducing

correlation. To illustrate the general idea, we will look at a single example: the widgets

come in units,the contract species a unit price, and the buyer may choose to use more

units of the new design than of the old. Suppose that the buyers unit revenues are o and

nfor the old and new designs, respectively. If she buys the quantity Q() and v is the

unit input price, then w = vQ may let the seller share some of the gains from adjustment

without ex post renegotiation. Specifically, the buyers gains from adjustment are

Q(n)(n- v) - Q(o)(o - v),while the sellers gains are Q(n)(v - cn) - Q(o)(v - co). The

ability to use such contracts allows the Commitment game form to induce more

communication.

Interpretation of the Commtment Game Form as an Employment Contract

Like the seller in the Commitment game form, employees take orders and only

rarely ask to renegotiate their compensation as a direct consequence of a new order. In

contrast, and like the seller in the Renegotiation game form, independent contractors

typically renegotiate prices as a result of change orders.

adjusted to reflect the identity of the traded design.

20


21/27

There are several possible explanations for these stylized facts. The property

rights theory of the firm (Grossman and Hart, 1986; Hart and Moore, 1990; Hart, 1995)

defines the boss by asset ownership and explains the stylized facts by appeal to the power

this confers to him. Employees attempts at renegotiation are bound to fail because they

are powerless, while independent contractors have power and thus the ability to extract

some rents from a change. In contrast, the adjustment-cost theory (Bajari and Tadelis,

2001; Simester and Wernerfelt, 2004; Tadelis, 2002; Wernerfelt, 1997; 2002; 2004)

defines the employment relationship as an implicit contract in which the employee has

agreed to refrain from renegotiation, retaining only the right to quit. It explains the

efficiency of this institution by claiming that frequent renegotiation is costly. On the other

hand, there are also various costs associated with an employment relationship and if these

costs are larger than those of infrequent renegotiation, a market relationship may be more

efficient. A central prediction of this theory is then that relationships needing more

frequent adjustment are more likely to be organized inside the firm. (The difference

between the work done by a secretary and that done by a homebuilder may illustrate this

contrast.) In the present paper, we have added another force to the tradeoff by showing

that the market may implement a different set of adjustments than those implemented in a

firm.3

Phrased in these terms, our central finding is that the market is better when

adjustments are rare but large, while employment is better when adjustments are frequent

but small.

No matter what the primary reasons for the existence of firms, they are clearly

subject to the informational inefficiencies of the commitment game form, while markets

3Another attractive feature of the story is that it suggests that the more informed party should have more

power, i. e. be the boss.

21


22/27

suffer from the problems of the renegotiation game form. We would expect certain kinds

of information to be withheld in firms, and other kinds to be withheld in markets. Put

differently, we would expect that firms and markets implement different sets of

adjustments. So we can use firms and markets as examples of the commitment and

renegotiation game forms and thereby throw empirical light on the model.

Empirical Evidence

It is not hard to find evidence consistent with the bargaining power effect. At an

informal level, one often hears that firms keep secrets, to enhance their short-term

negotiating positions or to protect long-term market power. More systematically,

Simester and Knez (2001) study a company in which eighteen different parts are made

both by employees and by independent suppliers. Consistent with the prediction that

concerns for bargaining power lead to less communication between firms, they find that

external suppliers give the company less information about delivery and production

schedules and fewer design suggestions. They even report that the company restricts

communication between its engineers and those of outside suppliers.

It is harder to find systematic evidence consistent with the incentive transfer

effect. Most people know stories in which employees withhold information because its

release would make their job harder slipping deadlines and maintenance needs come to

mind (Rotemberg, 2001). The sociology literature documents individual cases in which

employees make certain tasks seem harder than they really are (Roy, 1955), and cases in

which information is hidden from management (Crozier, 1964). However, I have not

been able to identify a large sample study of this phenomenon.

22


23/27

We may be able to get some sense of the balance between the two effects by

comparing incremental and radical new product development. When improvements are

incremental, the new products revenues will be only marginally higher than those of the

old product, and one could argue that commitment (a.k.a. integration) should be best. For

more radical changes, the new product may command much higher revenues, and

renegotiation (a.k.a. the market) will be more efficient. This general prediction, that more

radical innovations tend to occur between, rather than within firms, is in fact supported

by literature in the area of innovation management (Freeman, 1991; Hagedoorn, 1995;

Powell et al.,1996).

VI. CONCLUSION

The paper has contributed to contract theory by showing that different types of

information is communicated and withheld under ex ante versus ex post negotiation. In

particular, when gains from trade are private information, concerns for bargaining power

in ex post negotiation may cause players to withhold information. On the other hand, ex

post negotiation may provide incentives for communication because it allows the players

to share the resulting gains. On balance, we find that ex post negotiation is better when

adjustments are rare but large, while ex ante negotiation is better when adjustments are

frequent but small.

On a more applied level, we have contributed to the literature on incomplete

contracts by showing how the possibility of renegotiation influences information transfer

between traders. This may help explain why some contracts have more features left up to

renegotiation than others. Finally, because adjustments do not lead to renegotiation in

23


24/27

employment contracts, the results also allow us to suggest that the information

communicated in firms differ from that communicated in markets.

24


25/27

REFERENCES

Aghion, Phillippe, and Patrick Bolton, Contracts as a Barrier to Entry, American

Economic Review, 77, no.3, June, pp.388-401, 1987.

Aghion, Philippe, and Jean Tirole, Formal and Real Authority in Organizations,Journal of Political Economy, 105, no.1, February, pp.1-29, 1997.

Akerlof, George A., The Market for Lemons: Quality Uncertainty and the MarketMechanism, Quarterly Journal of Economics, 84, no. 3, August, pp. 488-500, 1970.

Bajari, Patrick, and Steven Tadelis, Incentives versus Transaction Costs: A Theory ofProcurement Contracts, RAND Journal of Economics, 32, no. 3, Autumn, pp.387-407,

2001.

Cremer, Jacques, Arms Length Relationships, Quarterly Journal of Economics, 110,no. 2, May, pp. 275-95, 1995.

Crozier, Michel, The Bureaucratic Phenomenon, Chicago: University of Chicago Press,1964.

Dewatripont, Mathias, and Eric Maskin, Contractual Contingencies and Renegotiation,RAND Journal of Economics, 26, no. 4, Winter, pp.704-19, 1995.

Dewatripont, Mathias, and Jean Tirole, Advocates, Journal of Political Economy, 107,

no. 1, February, pp. 1-39, 1999.

Farrell, Joseph, and Robert Gibbons, Cheap Talk about Specific Investments, Journal

of Law, Economics, and Organization, 11, no.2, October, pp. 313-34, 1995.

Freeman, Christopher, Networks of Innovators: A Synthesis, Research Policy, 20, pp.

499-514, 1991.

Gertner, Robert H., Coordination, Dispute Resolution, and the Scope of the Firm,

manuscript, The University of Chicago, 1999.

Grossman, Sanford, and Oliver D. Hart, The Costs and Benefits of Ownership, Journalof Political Economy, 97, no. 4, August, pp. 691-719, 1986.

Hagedoorn, John, Technology Partnering During the 1980s: Trends, Networks, and

Corporate Patterns in Non-Core Technology, Research Policy, 24, pp.207-31, 1995.

Hart, Oliver D., Firms, Contracts, and Financial Structure, Oxford UK: Oxford

University Press, 1995.

25


26/27

Hart, Oliver D., and John Moore, Property Rights and the Nature of the Firm, Journalof Political Economy, 98, no. 6, December, pp. 1119-58, 1990.

Hermalin, Benjamin E., and Michael L. Katz, Judicial Modification of Contracts

between Sophisticated Parties: A More Complete View of Incomplete Contracts andTheir Breach, Journal of Law, Economics, and Organization, 9, no.2, October, pp. 230-

55, 1993.

Powell, Walter W., Kenneth W. Koput, and Laurel Smith-Doerr, Interorganizational

Collaboration and the Locus of Innovation: Networks of Learning in Biotechnology,

Administrative Science Quarterly, 41, no. 1, March, pp.116-45, 1996.

Rogoff, Kenneth J., The Optimal Degree of Commitment to an Intermediate Monetary

Target, Quarterly Journal of Economics, 100, no.4, November, pp. 1169-89, 1985.

Rotemberg, Julio J., and Garth Saloner, Leadership Style and Incentives, ManagementScience, 37, pp.1299-1318, 1993.

Rotemberg, Julio J., Endogenous Altruism in Buyer-Seller Relations and its

Implications for Vertical Integration, manuscript, Harvard University, 2001.

Rothschild, Michael, and Joseph E. Stiglitz, Equilibrium in Competitive Insurance

Markets: An Essay on the Economics of Imperfect Information, Quarterly Journal of

Economics, 90, no. 4, November, pp. 629-49, 1976.

Roy, D., Efficiency and the Fix: Informal Intergroup Relations in a Piece-workMachine shop", American Journal of Sociology, 60, pp. 255-66, 1955.

Simester, Duncan I., and Marc Knez, Direct and Indirect Bargaining Costs and theScope of the Firm, Journal of Business, 75, no. 2, April, pp.283-304, 2002.

Simester, Duncan I., and Birger Wernerfelt, Determinants of Asset Ownership: A Study

of the Carpentry Trade, Review of Economics and Statistics, 86, forthcoming, 2004.

Spence, A. Michael, Job Market Signaling, Quarterly Journal of Economics, 87, no.3,

August, pp. 355-74, 1973.

Spier, Kathryn E., Incomplete Contracts and Signalling, RAND Journal of Economics,

23, no. 3, Autumn, pp. 432-43, 1992.

Tadelis, Steven, Complexity, Flexibility, and the Make-or-Buy Decision, American

Economic Review, 92, no.2, May, pp. 433-37, 2002.

26


27/27

Wernerfelt, Birger, On the Nature and Scope of the Firm: An Adjustment-Cost Theory,

Journal of Business, 70, no. 4, pp. 489-514, 1997.

Wernerfelt, Birger, Why Should the Boss Own the Assets? Journal of Economics and

Management Strategy, 11, no. 3, fall, pp. 473-85, 2002.

Wernerfelt, Birger, Governance of Adjustments, Journal of Business, 77, no. 2, Part 2,

April, pp. S3-S24, 2004.

.

Comm. and Reneg. SSRN-Id581721_2

Documents

Transcript of Comm. and Reneg. SSRN-Id581721_2