exclusive dealing.pdf

8/12/2019 exclusive dealing.pdf

1/41

Exclusive DealingAuthor(s): B. Douglas Bernheim and Michael D. WhinstonSource: Journal of Political Economy, Vol. 106, No. 1 (February 1998), pp. 64-103Published by: The University of Chicago PressStable URL: http://www.jstor.org/stable/10.1086/250003 .Accessed: 09/05/2014 12:40

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Journalof Political Economy.

http://www.jstor.org

This content downloaded from 16 8.176.5.118 on Fri, 9 May 201 4 12:40:20 PMAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/action/showPublisher?publisherCode=ucpresshttp://www.jstor.org/stable/10.1086/250003?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/10.1086/250003?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=ucpress


2/41

Exclusive Dealing

B. Douglas BernheimStanford University

Michael D. WhinstonNorthwestern University, Harvard University, and National Bureau of Economic Research

In this paper, we provide a conceptual framework for understand-ing the phenomenon of exclusive dealing, and we explore the mo-tivations for and effects of its use. For a broad class of models, we

characterize the outcome of a contracting game in which manufac-turers may employ exclusive dealing provisions in their contracts. We then apply this characterization to a sequence of specializedsettings. We demonstrate that exclusionary contractual provisionsmay be irrelevant, anticompetitive, or efciency-enhancing, de-pending on the setting. More specically, we exhibit the potentialfor anticompetitive effects in noncoincident markets (i.e., marketsother than the ones in which exclusive dealing is practiced), and we explore the potential for the enhancement of efciency in asetting in which common representation gives rise to incentiveconicts. In each instance, we describe the manner in which equi-

librium outcomes would be altered by a ban on exclusive dealing. We demonstrate that a ban may have surprisingly subtle and unin-tended effects.

We would like to thank Jean Tirole and seminar participants at the EuropeanScience Foundation conference in Toulouse, George Mason University, the Ken-nedy School of Government, Massachusetts Institute of Technology, Princeton Uni- versity, the University of Chicago, and the University of TexasAustin, as well asan anonymous referee, for their comments on earlier versions of this material. We would also like to thank Claudia Napolilli, Deborah Johnston, and Sharon Kacha-doorian for their assistance in the preparation of this and previous manuscripts.Financial support from the Sloan Foundation and the National Science Foundation(grants SES-89-21996 and SES-91-10211) is gratefully acknowledged.

[ Journal of Political Economy, 1998, vol. 106, no. 1] 1998 by The University of Chicago. All rights reserved. 0022-3808/98/0601-0001$02.50

64

http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp


3/41

exclusive dealing 65

I. Introduction

A manufacturer engages in exclusive dealing when it prohibits a re-tailer or distributor that carries its product from selling certain otherproducts (typically those of its direct competitors). Historically, thecourts have treated exclusive dealing harshly. For example, in one well-known case, Standard Fashion Company v. Magrane-Houston Com- pany (1922), a leading manufacturer of dress patterns (Standard)contracted with a prominent Boston retailer (Magrane-Houston) tosell its patterns on the condition that Magrane-Houston not sell thepatterns of any other manufacturer. 1 Fearful of the foreclosure of competitors from retail outlets, the court struck down the contract,arguing that the restriction of each merchant to one pattern manu-facturer must in hundreds, perhaps in thousands, of small communi-ties amount to giving such single pattern manufacturer a monopoly of the business in such community.

Despite the courts position, many antitrust experts have come tobelieve that exclusive dealing cannot serve as a protable mecha-nism for monopolization and that it should instead be regarded asan efcient contractual form. Commenting on Standard Fashion,Bork (1978, pp. 3067) argues that

Standard can extract in the prices that it charges all that its line is worth. It cannot charge the retailer that fullamount in money and then charge it again in exclusivity that the retailer does not wish to grant. To suppose that it can is to commit the error of double counting. . . . Exclusiv-ity has necessarily been purchased from it, which meansthat the store has balanced the inducement offered by Stan-dard . . . against the disadvantage of handling only Stan-

dards patterns. . . . If consumers would prefer more pat-tern lines at higher prices, the store would not accept Standards offer. The stores decision, made entirely in itsown interest, necessarily reects the balance of competingconsiderations that determine consumer welfare. Put thematter another way. If no manufacturer used exclusivedealing contracts, and if a local retail monopolist decidedunilaterally to carry only Standards patterns because theloss in product variety was more than made up in the cost

saving, we would recognize that decision was in the con-sumer interest. We do not want a variety that costs more

1 For a summary of federal exclusive dealing cases that reached at least the appel-late level, see Frasco (1991).



4/41

66 journal of political economy

than it is worth. . . . If Standard nds it worthwhile to pur-chase exclusivity. . . , the reason is not the barring of entry,

but some more sensible goal, such as obtaining the specialselling effort of the outlet.

In comparison with other vertical restrictions (such as exclusiveterritories and resale price maintenance), exclusive dealing has re-ceived little formal attention (see Katzs [1989] survey; exceptionsinclude Marvel [1982] and Mathewson and Winter [1987], discussedbelow). In this paper, we provide a unied framework for under-standing the motivations for and effects of these contractual provis-

ions. Central to our approach is the view that exclusive dealing isbest understood by studying its costs and benets relative to thoseof common agency (Bernheim and Whinston 1986 a, 1986b ).

Section II studies a simple game in which players bid for represen-tation. Using this game, we exhibit two thematic principles that re-surface repeatedly throughout the paper. The rst principle is that the form of representation (exclusivity or common representation)that arises in equilibrium maximizes the joint surplus of the manu-facturers and the retailer, subject to whatever inefciencies may (or

may not) characterize incentive contracting between the retailer andthe manufacturers. The second principle is that explicit contractualexclusion (as distinguished from a retailers unilateral decision tocarry only one product) will not arise unless common representationinvolves externalities among the manufacturers that cause inef-ciency in incentive contracting.

In Section III, we consider the simplest incentive contractingproblem: the retailer buys and resells the manufacturers products,and these choices are contractible. The model follows closely the

scenario envisioned by Bork. We show that incentive contracting, whether under exclusive or common representation, is always ef-cient in this setting. Our rst general principle therefore implies that the market outcome maximizes the prots of the vertical structure asa whole: one obtains the fully integrated solution. Thus exclusivedealing can arise in this setting only when it is efcient for one prod-uct to be sold. Moreover, we show that, although the outcome neednot be rst-best (since consumer surplus is ignored), banning exclu-sive dealing cannot raise aggregate welfare. All these conclusions are

consistent with Borks analysis.The difculty with this analysis is that it fails to account for theexistence of exclusive dealing. Since there are no contracting exter-nalities in this setting, our second general principle implies that ex-clusionary provisions are superuous; when the fully integrated solu-tion would involve the sale of only one product, this outcome can



5/41


always be supported through nonexclusionary contracts. Conse-quently, in this model, there is no reason either to ban or to permit

exclusive dealing.Given these results, it is natural to ask whether exclusionary provi-sions can ever serve a meaningful purpose. Commentators have sug-gested a number of possible motivations (see, e.g., Scherer 1980; Areeda and Kaplow 1988): some believe that it arises from a desireto foreclose markets and extend market power, whereas others seeit as an efcient contractual device. 2 In Sections IV and V, we providerigorous theoretical foundations for each of these views, by appropri-ately extending the simple model of Section III.

In Section IV, we demonstrate that exclusive dealing can serve asa device for extracting rents from markets other than the ones in which they are employed. We refer to this as a noncoincident market effect. We examine a model in which retail markets develop sequen-tially and in which manufacturers must serve more than one market to achieve important economies. Effective exclusion (i.e., only oneproduct is sold) occurs in the early developing (rst) market when-ever it is jointly optimal for the manufacturers and the rst retailer(a reection of our rst general principle). Exclusion may occur in

this context precisely because it affects the degree of competitionamong manufacturers in the second market and, hence, the extrac-tion of prots from the second retailer (whose prots are not consid-ered in the joint optimization problem that determines representa-tion in the rst market). Moreover, we show that it may be impossibleto achieve the exclusionary outcome in the absence of explicit con-tractual exclusion, precisely because the existence of noncoincident effects may generate contracting externalities for the manufacturersunder common representation (a reection of our second general

principle). In addition, we demonstrate that exclusion may persist even in the presence of a ban on explicitly exclusive deals; indeed,a ban may lead to exclusion through less efcient practices such asquantity forcing or quantity discounts. 3 Thus the welfare implica-

2 According to Scherer (1980, p. 586), For manufacturers, exclusive dealing ar-rangements are often appealing, because they ensure that their products will bemerchandised with maximum energy and enthusiasm.

3 In light of this result, there are some noteworthy aspects of a recent lawsuit ledby Virgin Atlantic Airways against British Airways. While British Airways apparently has not attempted to engage any travel agent in an exclusive relationship, it hasoffered travel agent commission override programs, which grant rebates if agentspurchase large quantities or high fractions of their customers travel services fromBritish Airways. Virgin Airways has alleged that these programs effectively amount to exclusive dealing arrangements and that they result in market foreclosure. Seethe decision of U.S. District Judge Miriam Goldman Cedarbaum concerning British Airways motion for dismissal, Memorandum of Opinion and Order, 93 Civ.7270(MGC), December 30, 1994.



6/41


tions of a ban are ambiguous, even when the motive for exclusionis foreclosure.

In Section V, we study the role of exclusive dealing in circum-stances in which common representation involves incentive con-icts. We examine a model in which a risk-averse retailer takes unob-servable actions that inuence the manufacturers sales. In thissetting, common representation entails contracting externalitiesthat produce inefciencies. This can lead to exclusive dealing whenthe associated costs are large relative to the benets of variety (areection of our two general principles). We explore the nature of these inefciencies and the precise circumstances in which exclusive

dealing arises. We also demonstrate that a ban on exclusive dealingmay have surprisingly subtle effects. For example, a ban can reduce welfare even in cases in which no exclusion would have occurred.

Section VI contains concluding remarks. All formal proofs arecontained in the Appendix.

II. Some Unifying Principles

Although this paper investigates the motivations for and effects of

exclusive dealing in a number of distinct models, our analysis is con-nected by several unifying principles. Through an appreciation of these principles, one gains an intuitive understanding of our formalresults. The purpose of this section is to elucidate the unifying princi-ples in the simplest possible setting.

Consider in particular the following three-stage game.Stage 1.Two manufacturers ( j A, B) simultaneously bid for

representation by a retailer. Each bid consists of two numbers: anannounced required payoff for the manufacturer in the event the

manufacturer is represented exclusively ( p e

j for manufacturer j ) anda required payoff for the manufacturer in the event the retailer rep-resents both manufacturers ( p c j for manufacturer j ).

Stage 2.The retailer chooses to represent one manufacturer,both manufacturers, or neither. If it chooses to represent neither,the game ends, and all parties earn zero.

Stage 3.The retailer enters into a contract (or contracts) withthe party (or parties) that it has chosen to represent. Here, we treat this process as a black box, simply assuming that the aggregate

payoffs under common representation are c

and are j

when theretailer represents manufacturer j exclusively. If the retailer has cho-sen to represent both manufacturers, it pays an amount to each man-ufacturer j sufcient to provide that manufacturer with a net payoff of p c j . If the retailer has chosen to represent only manufacturer j , it pays an amount to j sufcient to provide j with a net payoff of p e j .



7/41


Hence, the retailer receives a payoff of c p c A p c B if it serves bothmanufacturers and j p e j if it serves manufacturer j exclusively.

Throughout, we shall assume that the products of manufacturers A and B are substitutes, in the sense that A B c .The models considered in subsequent sections have a similar

structure, except that the contracting process in stage 3 is modeledexplicitly and takes place as part of the bidding process in stage 1. We have adopted a more ad hoc (and clearly less satisfactory) struc-ture here because it allows us to separate simple bidding issues frompotentially complex contracting and incentive issues. It turns out that this simplication does little violence to the underlying eco-

nomic principles that govern the structure of equilibria. As we dem-onstrate in the Appendix, the fundamental principles developed inthis section carry over to a broad range of cases in which contractingoccurs as part of the stage 1 bidding process, including all modelsconsidered later in this paper. 4

The model outlined above can give rise to multiple equilibria. Fol-lowing Bernheim and Whinston (1986 b ), we rene the equilibriumset by treating the retailer as a passive, reactive party and look forequilibria that are Pareto-undominated (within the set of equilibria)

for the manufacturers, on the grounds that they act as rst-movers.5

As a general matter, we can classify equilibria according to whetherthey are exclusive (the retailer contracts with only one manufacturer)or common (the retailer contracts with both manufacturers).

Consider, rst, exclusive equilibria. If each manufacturer j sets p c j , then bidding is reduced to competition to obtain an exclu-sive relationship with the retailer. The following two conditions char-acterize equilibria in which the retailer serves manufacturer j exclu-sively:

j p e j i p e i 0 (1)and

p e j 0 p e i . (2)

Condition (1) has several components: (i) the retailer must earnstrictly positive prots (otherwise i , the excluded manufacturer,could protably deviate to some bid slightly less than i ); (ii) theretailer must not earn less serving j than serving i (otherwise it would

choose to serve i ); and (iii) the retailer must not earn less serving4 The models considered here and in later sections are closely related to, but

not special cases of, the framework of menu auctions developed in Bernheim and Whinston (1986 b ).

5 In the current setting, this is equivalent to requiring that the equilibria be per-fectly coalition-proof (see Bernheim, Peleg, and Whinston 1987).



8/41


i than serving j (otherwise j would lower its bid). Condition (2) re-quires that j receive a nonnegative payoff (otherwise it would with-

draw its bid). Manufacturer i , on the other hand, must receive anonpositive payoff if its bid is accepted (otherwise it would increaseits bid slightly to obtain exclusive representation).

It follows from these conditions that j i ; that is, in any exclu-sive equilibrium, the retailer must serve the manufacturer that gen-erates the highest surplus. Henceforth, without loss of generality, wetake this to be manufacturer A. To see that the existence of an exclu-sive equilibrium is guaranteed, simply set p e A A B and p e B B, and note that both equilibrium conditions are satised. There

are other exclusive equilibria, but none gives a higher payoff to ei-ther manufacturer. Thus the preferred exclusive equilibrium yieldspayoffs of A B to manufacturer A, zero to manufacturer B, and B to the retailer.

Now consider common equilibria. The following conditions char-acterize equilibria in which the retailer serves both manufacturers:

j p e j c p c A p c B 0, (3)

p c j 0, and p e j p c j , j A, B. To understand the rst condition, note

rst that c

p c

A p c

B 0: this expression clearly could not be lessthan zero (or the retailer would decline to represent both manu-facturers); if it was equal to zero, j could protably induce the re-tailer to accept an exclusive contract. 6 Next, observe that one must have j p e j c p c A p c B for each j ; clearly, one cannot have j p e j c p c A p c B (otherwise the retailer would serve j ex-clusively). If j p e j c p c A p c B, then (since c p c A p c B0) manufacturer i could protably deviate by slightly increasing p c i and setting p e i . One obtains the remaining equilibrium con-

ditions as follows: if p c

j

0 for some j , then j would withdraw itsoffer; if p e j p c j , then j could reduce p e j slightly, thereby (in light of [3]) protably inducing the retailer to accept a (more protable)exclusive offer.

Unlike exclusive equilibria, common equilibria do not always ex-ist. It is easy to verify that the set of bids satisfying the equilibriumconstraints is nonempty if and only if c A . Thus exclusive repre-sentation necessarily arises whenever exclusion generates the great-est joint surplus.

6 To establish this claim, note that j p c j 0 for at least one manufacturer j (since A B c ). Suppose that this manufacturer deviates by setting p e j p c j for some small 0. If this exclusive offer is accepted, then j is clearly betteroff, and the deviation is protable. As long as is sufciently small, the retailer would earn j p e j 0 by accepting j s exclusive offer, compared with zero for thecommon offers and at most zero (since this is a common equilibrium) for the exclu-sive offer of j s competitor. Consequently, j s exclusive offer is accepted.



9/41


When c A , the game gives rise to many common equilibria.It turns outsomewhat surprisinglythat there is a unique Pareto-

undominated equilibrium. To establish this point, we actually ndthe equilibrium. Note that there is only one equilibrium satisfying p e j p c j for both manufacturers; it is obtained by substituting theseexpressions into (3) to obtain two linear equations in two unknowns.The solution to these equations is given by p c j c i , i j .In this equilibrium, each manufacturer j earns c i , i j (itsmarginal contribution to total surplus), and the retailer earns A

B c . It is easily demonstrated that no other equilibrium isundominated. 7

Taken together, these results lead to the following conclusions:(i) when c A , in equilibrium the retailer necessarily serves man-ufacturer A exclusively; (ii) when c A , there are both exclusiveand common equilibria, but there is a unique common equilibriumthat Pareto-dominates (for the manufacturers) all other equilibria,both common and exclusive; and (iii) when c A , there is aunique Pareto-dominant (for the manufacturers) payoff vector, but it is achievable through either an exclusive or a common equilib-rium.

From these results, we deduce two general principles that provideunifying themes for the remainder of this paper. The rst principleis that the form of representation (i.e., exclusivity or common representation)is chosen to maximize the joint surplus of the manufacturers and the retailer,subject to whatever inefciencies may (or may not) characterize incentive contracting between the retailer and the represented parties. The nal qualify-ing phrase in the previous sentence is important. In particular, whencalculating joint surplus for this purpose, we do not pretendthat the contracting outcome under common representation (in

stage 3) is necessarily efcient for the manufacturers, that is, equalto the outcome that would arise were the manufacturers to cooper-ate. On the contrary, c may be strictly less that the total surplus, callit c , that could be obtained if the manufacturers selected incentiveschemes cooperatively under common representation.

This observation leads to our second unifying principle. While it is easy to imagine reasons why c might be less than A , it is hardto imagine that c would be less than A . Indeed, since representa-tion of a second manufacturer only expands opportunities, one

would generally expect the opposite. Thus we would expect to have c A only if there is a contracting inefciency resulting from thenoncooperative provision of incentives under common representa-

7 Consider some equilibrium in which p e i p c i . Then i p c i i p e i c p c A p c B; hence p c j c i , i j , for each j .



10/41


tion (i.e., only if c c ). Coupling this with our previous result, we conclude that, in general, explicit exclusive dealing (as distinguished

from a retailers unilateral decision to carry only one product) will not arise unless common representation involves externalities among the manufacturers that result in contracting inefciencies. This principle focuses our at-tention on the efciency of joint incentive contracting as the key factor inuencing the use of explicit exclusive dealing provisions.

III. The Simplest Contracting Problem

We begin our formal analysis of exclusive dealing by studying the

potential for it to arise in a simple setting that corresponds closely to the environment envisioned by Bork. In particular, we assumethat the retailer directly controls the level of retail sales for eachmanufacturer j , henceforth denoted x j . Manufacturer j can observeand verify x j , as well as the nature of j s relation with the retailer(exclusive or nonexclusive); however, j is unable either to observeor to verify the level of retail sales made on behalf of manufacturer

j (x j ).Here and in all subsequent sections, we depart from the simple

ad hoc model of Section II by assuming that rms announce contract offers in stage 1, rather than simple payments. A contract offer formanufacturer j consists of a contingent pair ( P e j , P c j ): P e j is an exclusive contract, which applies if the retailer contracts only with manufac-turer j ; P c j is a common contract, which applies if the retailer contracts with both manufacturers. Each contract is a function that maps x j to a monetary payment. In essence, each manufacturer can offer theretailer a compensation scheme that ties monetary payments to itsown sales, as well as to the nature of its relationship with the retailer,

but cannot tie payments to sales of another manufacturers product.8

Once contract offers are announced, the retailer chooses to repre-sent either or both manufacturers (stage 2). Finally, in stage 3, theretailer chooses sales levels ( x A , x B) (positive values of x j are permit-ted only if the retailer has accepted one of manufacturer j s offers),receives revenues of R (x A , x B) (for convenience, we sometimes writeR (x j , x j )), and makes payments to the manufacturers as required

8 As shown in a previous version of this paper (Bernheim and Whinston 1992),our analysis is essentially unchanged when one permits manufacturers to conditionpayments on each others sales. The central difference is that, under this alternativeassumption, manufacturers can always write nominally nonexclusive contracts that are equivalent to exclusive contracts (e.g., by permitting the retailer to serve othermanufacturers, while penalizing the retailer heavily whenever the sales of anothermanufacturer are positive). Thus the alternative assumption obscures the formaldistinction between exclusive and nonexclusive contracts without adding to the sub-stantive content of the problem.



11/41


by the contract or contracts accepted in stage 2. 9 In the course of producing x j , manufacturer j incurs costs of c j (x j ) (where c j (0) 0).

The formal analysis of this contracting game is developed in detail inthe Appendix.In this setting, a fully integrated vertical structure would choose

to produce and sell

x ** (x ** A , x **B ) argmaxx A , x B

R (x A , x B) j A, B

c j (x j ),

which, for convenience only, we assume to be unique. On the otherhand, were only product j available, a vertically integrated rm con-sisting of the retailer and rm j would select

x * j argmaxx j

R (x j , 0) c j (x j ).

We make the following assumption. Assumption B1.

A R (x * A , 0) c A (x * A ) B R (0, x *B ) c B(x *B ) 0.

Thus product A is the more protable of the two products if only one of them can be sold. 10 We also assume that the two products aresubstitutes, in the sense that product j contributes less in incremen-tal prots when product j is also sold than it does when it aloneis sold.

Assumption B2.

c R (x **) c A (x ** A ) c B(x **B ) A B.

A. Characterization of Equilibria

The following result characterizes undominated equilibria for thismodel.Proposition 1. In any undominated equilibrium, the retailer

chooses x **, manufacturer j earns its marginal contribution to joint prots, c j , and the retailer earns A B c . There isalways a common equilibrium yielding this undominated outcome.

According to proposition 1, undominated equilibria always max-imize the joint payoffs of the retailer and both manufacturers (they generate the vertically integrated outcome). Unless the products are

9 If the retailer rejects both of manufacturer j s offers, no payment is made to orfrom j .

10 If the rst inequality in assumption B1 holds with equality, all our results con-tinue to hold. But there are also exclusive equilibria (possibly with payoffs that aredominated for the manufacturers) in which B is served; see proposition 1.



12/41


perfect substitutes, this typically requires common representation.Moreover, even when an undominated equilibrium outcome entails

no sales for manufacturer B, this outcome can always be achievedthrough nonexclusionary contracts.Proposition 1 is easily understood in light of our general unifying

principles. To see this, consider the characteristics of exclusive rep-resentation and common representation. Imagine rst an intrinsi-cally exclusive setting, in which the retailer must accept an exclusiveoffer from manufacturer j or represent no one at all. In this setting, j s optimal contract offer has the property that j extracts all eco-nomic surplus over and above the retailers reservation payoff. Man-

ufacturer j can, for example, achieve this outcome through a forc-ing contract that requires the retailer to choose x * j and species alevel of compensation such that the retailers participation con-straint just binds. Joint payoffs for manufacturer j and the retailerare then given by j .

Next imagine a setting with intrinsic common representation(Bernheim and Whinston 1986 a ), in which the retailer must accept offers from both manufacturers or from neither. Were the manufac-turers to cooperate with each other in their choice of contracts, they

would induce the retailer to choose x ** (e.g., through forcing con-tracts), and they would extract all economic surplus over and abovethe retailers reservation payoff. Joint prots for the manufacturersand the retailer would then be given by c .

Of course, the structure of the game does not permit the manufac-turers to cooperate. Nevertheless, there still exist equilibria that im-plement x ** and generate cooperative payoffs for the manufacturers(holding the retailer to its reservation payoff ). One such equilib-rium involves forcing contracts: the payments to A and B are set at

a level that provides the retailer with its reservation utility condi-tional on choosing x **, and each rm j demands an innite payment for any x c j x ** j . Another involves sellout contracts of the formP c j (x j ) F j c j (x j ), which essentially transfer to the retailer the fullmarginal returns from the sale of each product j in return for xedpayments F j c j . For these equilibria, joint payoffs for themanufacturers and the retailer are c c .

In this setting, one necessarily has c j , and the inequality isstrict provided that the vertically integrated outcome entails positive

output by both manufacturers ( x ** 0). Since the (undominated)equilibrium outcome of the bidding game maximizes the joint sur-plus of the manufacturers and the retailer (our rst general princi-ple), we see that exclusive dealing never occurs in this context, ex-cept in the degenerate case in which there is an equivalent outcome with common representation. This conclusion is also a direct reec-



13/41


tion of our second general principle: in this model, c c , so there are no contracting externalities that could give rise to the need for explicit

exclusion.

B. Policy Implications

The preceding analysis corroborates Borks (1978) argument that exclusive dealing cannot be used protably to foreclose a rival froma market. Because each manufacturer must effectively compensatethe retailer to attract it to an exclusive deal, manufacturers internal-ize the retailers cost from the loss in product variety. As a result,

the market outcome is exactly the one that would arise with a fully integrated vertical structure. Indeed, just as Bork asserts, in equilib-rium each manufacturer extracts a prot exactly equal to the incre-mental value of its product.

In light of our results, it is surprising that, using a model similarto ours in many respects, Mathewson and Winter (1987) reach strik-ingly different conclusions. In their model, producers offer whole-sale contracts to the retailer on a take-it-or-leave-it basis. These con-tracts specify a wholesale price and possibly an exclusive dealing

requirement. Mathewson and Winter show that exclusive dealingarises as the unique equilibrium outcome for a range of parameter values.

The key difference between our model and that of Mathewsonand Winter concerns the set of feasible contracts. 11 In our notation,Mathewson and Winter allow only contracts of the form P e j (x j ) w e j x j and P c j (x j ) w c j x j for constants w e j and w c j . These restrictions create contracting externalities for the manufacturers, and this largely accountsfor the differences between our ndings. Even the exibility to

charge xed fees would, in many instances, restore our results. Theimportance of xed fees is easily understood in the context of Borksargument. If a manufacturer insists on exclusivity, it must compen-sate the retailer for the loss of surplus associated with selling otherproducts. If a xed fee is not available, then the manufacturer cancompensate the retailer only by reducing its wholesale price. How-ever, this form of compensation alters the retailers incentives onthe quantity margin; its value to the retailer is therefore less thanits cost to the manufacturer.

11 There are also some differences in the timing of decisions. In Mathewson and Winter, both rms rst decide whether to insist on exclusivity; if either does, thenboth compete in the offering of exclusive contracts (otherwise, the retailer sells theproducts of both manufacturers). However, if we were to change the timing of con-tract offers in our model while retaining exibility in the form of the contracts, thebasic conclusions of our analysis would be unaltered.



14/41


Proposition 1 implies that the retailer and manufacturers act asan integrated unit. However, contrary to Borks assertion, it does

not follow that the equilibrium maximizes social surplus unless theretailer is able to extract all consumer surplus (say, through perfect price discrimination). From a social perspective, the integrated solu-tion can involve the production of either too many or too few prod-ucts and inefcient retail pricing (Tirole 1988, pp. 1045). Never-theless, for this model, Bork is correct that a ban on exclusive dealingcannot promote social welfare. Formally, we model this prohibitionas the restriction that P c j (x j ) P e j (x j ), so that manufacturer j is pre- vented from conditioning compensation on the retailers decision

to serve

j . The following result demonstrates that a prohibitionon exclusionary contracts leaves the equilibrium outcome unaf-fected. 12

Proposition 2. Suppose that manufacturers are restricted to of-fering contracts that satisfy P c j (x j ) P e j (x j ). Then there is an equilib-rium in which the retailer accepts both manufacturers contracts andchooses x **, and payoffs are exactly as in proposition 1. Further-more, this equilibrium weakly dominates (for the manufacturers)any other equilibrium of this game. 13

Although propositions 1 and 2 appear to conrm much of Borksreasoning, in one important sense they fail to do so: exclusionary provisions are superuous in this model. Whether exclusionary provisions are permissible has no effect on undominated equilibriumoutcomes, which are always achievable through nonexclusionary contracts, even when one manufacturer is effectively excluded (i.e.,makes no sales). Hence, in this model, there is no reason either toban or not to ban these arrangements. Thus the present model may provide a poor framework for understanding the effects of the exclu-

sionary contracts that are observed in practice. In the next two sec-tions, we turn our attention to models in which exclusionary provi-sions serve a meaningful purpose.

12 Proposition 2 does not follow directly from proposition 1, despite the fact that undominated equilibria need never employ exclusive contracts. As a formal matter,the prohibition on exclusionary contracts changes the nature of manufacturersstrategies and could in principle subtly alter their incentives.

13 OBrien and Shaffer (1991) analyze a model that is equivalent to this restrictedgame. They show that in any equilibrium of this restricted game, the quantities cho-sen by the retailer ( x A , x B) must satisfy

x j argmaxx j

[R (x j , x j ) c j (x j ) c j (x j )]

for j A, B. Thus, if integrated prots are strictly concave, all equilibria result ina choice of x **.



15/41


IV. Exclusive Dealing with Noncoincident Market Effects

One frequently cited motive for exclusive dealing is the desire tocreate or enhance market power. Yet in the model of Section III,no such effect could occur: The equilibrium market outcome alwaysmaximized the total prots of the vertical structure, and it achievedeffective exclusivity (where jointly efcient) without any explicit ex-clusionary provisions.

Thus far, however, we have conned our attention to an isolatedset of vertically related parties. Commentators have also expressedconcern that the exclusion of competitors from one market might enhance a rms power in other markets. In this section, we show that the concern over what we shall call noncoincident market effectsdoes indeed have a valid theoretical foundation.

We explore the role of noncoincident market effects for a modelin which two retail markets develop sequentially and in which impor-tant economies can be achieved only by serving more than one mar-ket. 14 As in Section III, effective exclusion occurs whenever it is jointly optimal for the manufacturers and the retailer in the rst market. In this context, exclusion may arise precisely because it re-duces competition in the second market and hence facilitates theextraction of prots from the second retailer (whose prots are not considered in the joint optimization problem that determines repre-sentation in the rst market). Moreover, it may be impossible toachieve exclusivity without explicit contractual exclusion, precisely because the existence of noncoincident effects may generate con-tracting externalities for the manufacturers. We also examine theeffects of banning exclusive dealing and demonstrate that this doesnot always end effective exclusion. Indeed, in the presence of a ban,effective exclusion may be achieved through even less efcient prac-tices.

A. The Model

Suppose that initially there is a single retail market (market 1),served by a single retailer (retailer 1). With time, another retail mar-ket (market 2), again served by a single retailer (retailer 2), becomes viable. Manufacturers and retailers can enter into long-term con-tracts. Thus, prior to the emergence of market 2, manufacturers cancontract with retailer 1 for sales made after the emergence of market

14 Similar noncoincident market effects can arise in other contexts, e.g., when anexclusive contract between a manufacturer and a retailer reduces competition forthe manufacturers inputs.



16/41


2. Manufacturers cannot, however, contract with retailer 2 for salesin market 2 until this market emerges. To isolate the key role played

by long-term contracts in market 1, we suppress all sales in market 1 that occur prior to the emergence of market 2 (one can easily make earlier sales explicit at the expense of some additional nota-tion).

The game unfolds in three phases. In phase 1, manufacturers offercontracts to retailer 1; as in Section III, the retailer then choosesamong contract offers and selects quantities. Production, however,does not occur until phase 3. 15 In phase 2, each manufacturer j hasthe opportunity to invest a xed sum ( K j ) in cost reduction. This

investment reduces the unit cost of production from c j j (where j 0) to c j . In phase 3, having observed each others investment decisions, the two manufacturers engage in a contracting game withretailer 2 (as in Sec. III). Finally, production is carried out and theretailers make the payments required by their contracts. Retailer n srevenues are given by a continuous function R n (x A n , x Bn ), where x jn denotes manufacturer j s sales to retailer n . As in Section III, a re-tailer earns zero if it accepts no contracts, and a manufacturer earnszero in any market in which the retailer rejects its contract.

For the sake of simplicity, we focus on the case in which K A

A 0 and B . In other words, we assume that manufacturer A has no further opportunities to reduce costs and that manufac-turer B cannot produce at all unless the investment is undertaken.These assumptions imply that A may be able to eliminate competi-tion from B in market 2 by excluding B from market 1. However, Bdoes not have a symmetric incentive to exclude A from market 1.

It is convenient to dene for each market n 1, 2 the joint prot levels

c n c n maxx {R n (x ) c A x A c Bx B}

and

j n maxx j

{R n (x j , 0) c j x j }.

As in Section III, these would be the joint payoffs from common andexclusive outcomes were only market n to exist. In parallel to thenotation of Section III, we denote the (unique) solutions to thesemaximizations by ( x ** A n , x **Bn ) and x * jn ( j A, B), respectively, and

15 The retailers choice of quantities can also be delayed without affecting theconclusions, but the game is somewhat easier to solve if this decision is made immedi-ately.



17/41


assume that these quantities are strictly positive. We also assume that assumption B2 holds in each market, so A n Bn c n for n 1, 2.

To focus attention on the cases of greatest interest, we state severalpertinent assumptions. Assumption C1. 0 c 2 A 2 K B. Assumption C2. c 1 c 2 A 1 A 2 K B 0. Assumption C3. A 1 A 2 c 1 j ( c 2 j 2) K B. Assumption C1 states that manufacturer Bs contribution to total

prot in market 2 is positive but strictly less than Bs required invest-ment. Since Bs prots in market 2 (gross of K B and conditional onhaving invested in phase 2) are given by the middle term in assump-

tion C1 (see proposition 1), this condition implies that, if excludedfrom market 1, B will neither invest in phase 2 nor compete against A in market 2 during phase 3; thus assumption C1 creates the poten-tial for the foreclosure of a noncoincident market. Assumption C2indicates that if retailer 2s prots are also considered, aggregateprots are maximized when B participates. If the retailers practiceperfect price discrimination, this implies that Bs participation is so-cially desirable and that Bs exclusion from market 1 is inefcient. Assumption C3 states that the joint payoffs for retailer 1 and the two

manufacturers are higher if B is excluded from market 1 (given Bssubsequent decision not to participate in market 2) than if B makessales to retailer 1. The intuition developed in Section II suggests that this condition is required to generate effectively exclusive outcomesin market 1.

We assume throughout that assumptions C1 and C2 are satised,and we investigate the properties of equilibria contingent on whether assumption C3 holds.

B. Equilibrium Exclusion To understand the properties of equilibria for this model, it is help-ful to build intuition using the principles developed from the simpleanalytic framework of Section II. A small amount of work is rst re-quired before the applicability of this framework becomes evident.

To solve for equilibria, one would begin with phase 3. If B haschosen to invest in phase 2, then phase 3 payoffs for manufacturer j are given by c 2 i 2, i j (see proposition 1). If B has chosen not

to invest, then manufacturer A faces no competition in market 2. Inthat case, A extracts all the potential rents from retailer 2, earning A 2 .

Next consider the phase 2 investment decision of manufacturerB. If retailer 1 has chosen a positive quantity for B, manufacturerB certainly invests (otherwise B would incur innite losses since its



18/41


contract would require it to produce x B1 at innite costs). If retailer1 has an exclusive relationship with A or has simply chosen x B1 0,

then B chooses not to invest (given assumption C1 and proposi-tion 1).Finally, consider the phase 1 contract offers by manufacturers A

and B to retailer 1. Note that the phase 1 problem can be treatedas the type of game considered in Section III, provided that we de-ne payoffs appropriately to reect outcomes on the equilibriumcontinuation paths. In particular, we can solve the phase 1 con-tracting problem with retailer 1 by studying an equivalent single-market model, in which the costs of manufacturers A and B are given

by C A (x A1, x B1) c A x A1 A 2 I (x B1 0)( c 2 B2 A 2 )

and

C B(x B1) c Bx B1 I (x B1 0)( c 2 A 2 K B),

where the indicator function I (x B1 0) equals unity when x B1 0,and zero otherwise. Note that this equivalent single-market modeldiffers from the class considered in Section III in one important

respect: As implicit costs depend on Bs production as well as on As production. The importance of this observation becomes evident below.

It is for this equivalent, single-market problem that one can develop intuition by invoking the principles developed in Section II. As in Section III, we proceed by considering the characteristics of exclusive representation and common representation. Imagine rst an intrinsically exclusive setting involving the retailer and manufac-turer A. Under the optimal exclusive contract the retailer chooses

sales of x * A1, and joint payoffs for the manufacturers and the retailerare given by

R 1(x * A1, 0) C A (x * A1, 0) A 1 A 2 A .Next imagine an intrinsically exclusive setting involving the re-

tailer and manufacturer B. Under the optimal exclusive contract theretailer chooses sales of x *B1, and joint payoffs for the manufacturersand the retailer are given by

R 1(0, x *B1) C B(x *B1) ( c 2 B2 )

B1 j

( c 2 j 2) K B B

(where c 2 B2 denotes the payoff to manufacturer A when A isexcluded from market 1).



19/41


Finally, imagine a setting with intrinsic common representation. Were the manufacturers to cooperate with each other, they would

extract all economic surplus over and above the retailers reservationpayoff, and they would induce the retailer to make the joint prot-maximizing choice among ( x ** A1 , x **B1 ), ( x * A1, 0), and (0, x *B1). It is easy to check that the third choice is always inferior to the rst; conse-quently, cooperative joint prots are given by 16

max c 1 j

( c 2 j 2) K B, A c .

Once again, the structure of an intrinsic common agency gamedoes not permit the manufacturers to cooperate. To be consistent with our earlier notation, we use c to denote the joint payoffs associ-ated with the undominated noncooperative equilibrium of the in-trinsic common agency game. Obviously, c c . Notably, in con-trast to Section III, one cannot rule out strict inequality in thiscontext. We shall explain and elaborate on this point shortly.

Now consider this setting in light of the principles developed inSection II. When assumption C3 holds, A c c , and the coop-erative common outcome in market 1 involves effective exclusion(quantities of ( x * A1, 0)). Exclusion of manufacturer B from market 1 is jointly efcient for retailer 1 and the two manufacturers since joint losses in market 1 from reduced variety, c 1 A 1 , are morethan offset by the joint gain arising from reduced competition inmarket 2, A 2 j ( c 2 j 2) K B. These gains reect the moreeffective expropriation of rents from retailer 2, who loses A 2 B2 c 2. Since the (undominated) equilibrium outcome of the bid-ding game maximizes the joint surplus of the manufacturers andthe retailer (our rst general principle), effective exclusion arises inthis case. Moreover, this is so precisely because of anticompetitiveeffects in the noncoincident market.

In contrast, if assumption C3 is strictly reversed, then c A and the cooperative common outcome yields ( x ** A1 , x **B1 ) (x * A1, 0).Moreover, for this case, one can show that there is an equilibriumof the intrinsic common agency game in which ( x ** A1 , x **B1 ) is sus-tained through forcing contracts, so c c . Intuition based on theframework of Section II therefore suggests that any undominatedequilibrium of the contracting model in this case is a common equi-librium with quantities in market 1 of ( x ** A1 , x **B1 ).

16 The rst term in braces is simply the joint prot level associated with ( x ** A1 ,x **B1 ) and is derived from the expression R 1(x ** A1 , x **B1 ) C A (x ** A1 , x **B1 ) C B(x **B1 ).



20/41


The following proposition conrms the validity of these intuitivearguments.

Proposition 3. When assumption C3 holds, all undominatedequilibria involve effective exclusion of manufacturer B from market 1 (i.e., x B1 0). When the inequality in assumption C3 is (strictly)reversed, no undominated equilibrium involves effective exclusionof manufacturer B from market 1.

Of course, as we have emphasized, there is an important distinc-tion between effective, noncontractual exclusion and explicit, con-tractual exclusion. Indeed, for the single-market setting of SectionIII, explicit exclusionary provisions were superuous: whenever ef-

fective exclusion (i.e., only one product is sold) was jointly optimalfor the retailer and the two manufacturers, this could be achievedthrough nonexclusive contracts. However, the logic of that ndingdepended on the equality of c and c , which in turn followed fromthe assumption that As costs were independent of Bs sales (i.e., nocost externalities). This assumption permitted the rms to support (x * A , 0) using sellout contracts that transferred all variation in prots, without violating the restriction that a common contract cannot con-dition compensation on a rivals sales. In the current context, there

are cost externalities, since As implicit costs, C A (x A1, x B1), do de-pend on Bs sales; thus, without an ability to condition compensationon Bs sales, A cannot transfer all residual prot variation to retailer1. This implies that contracts between retailer 1 and B may imposeexternalities on A, in which case we might have c c A . Insuch a situation, the undominated equilibrium would still maximize joint prots through exclusion, but this would require explicit con-tractual exclusion of manufacturer B.

As noted above, when assumption C3 is strictly reversed, c c ,

and so the presence of cost externalities does not interfere with theefciency of intrinsic common representation. However, when as-sumption C3 holds, a deviation from the jointly efcient outcome,(x * A1, 0), may benet retailer 1 and manufacturer B precisely becausepositive sales for B impose a negative externality on A. When willthis externality be of sufcient size to justify the deviation? To answerthis question, we dene the following set:

D {x A1|maxx B1

[R 1(x A1, x B1) c Bx B1 I (x B1 0)

( c 2 A 2 K B)] R 1(x A1, 0)}.

In words, x A1 D if and only if retailer 1 and manufacturer B cannot jointly benet by arranging a deviation from ( x A1, 0) to ( x A1, x B1) forany x B1 0. Henceforth, we shall refer to D as the deterrence set. One



21/41


would expect to observe explicit exclusionary practices wheneverx * A1 is not in the deterrence set D , since in this case one cannot sup-

port the efcient exclusionary outcome through common represen-tation ( c c A ). This intuition is conrmed in the followingresult.

Proposition 4. When assumption C3 holds, undominated equi-libria necessarily involve an explicit exclusive dealing provision (andoutput of ( x * A1, 0) in market 1) if and only if x * A1 D .

Thus when assumption C3 holds and x * A1 D , retailer 1 agrees toan exclusive arrangement with manufacturer A to enhance As mar-ket power in a noncoincident market and to capture a share of the

resulting prots. Given assumption C2, this outcome is inefcient in the sense that it fails to maximize total retailer and producer sur-plus. 17 Even under the assumption that retailers can perfectly price-discriminate (which, as explained in the last section, is implicit inBorks analysis), exclusive dealing depresses social welfare. 18 Ouranalysis therefore provides a theoretical foundation for the concernthat exclusive dealing can foreclose markets anticompetitively.

C. The Effects of Banning Exclusive Dealing

The preceding subsection raises the possibility that, under certaincircumstances, exclusive dealing is an anticompetitive practice withadverse consequences for social welfare. This observation suggestsa potential role for antitrust policy. One possibility would be to im-pose a ban on exclusive dealing, which we model as in Section III.However, as we now show, when an inefcient market outcomearises that involves exclusive dealing, the welfare effects of a ban areambiguous; it may make things even worse. Moreover, a ban can

have surprising effects on the distribution of payoffs. We found in Section IV B that our model gives rise to one of threeoutcomes: explicit exclusion, effective exclusion, and common rep-resentation. Each outcome emerges for different parameterizations. Accordingly, we organize our discussion of policy around threecases.

17 Despite the inefciency of equilibrium, opportunities for renegotiation neednot alter our conclusions. Imagine that retailer 1 and manufacturer A have enteredinto an exclusive relation, but B has nevertheless invested in phase 2. Retailer 1 andmanufacturer A should be willing to renegotiate their contract at the start of phase3 to permit sales by B. However, B will typically capture less than 100 percent of the surplus gained through renegotiation. If Bs share is sufciently small, even theanticipation of renegotiation will fail to justify investment by B once A has consum-mated an exclusive contract with retailer 1. Thus, as long as Bs bargaining poweris not too great, exclusive dealing emerges exactly as in proposition 4.

18 Ironically, without perfect price discrimination by retailers, exclusive dealingcould conceivably raise social welfare.



22/41


Case 1. Explicit exclusion.Assumption C3 holds and x * A1 D .Case 2. Effective exclusion.Assumption C3 holds and x * A1 D .

Case 3. Common representation.Assumption C3 is strictly reversed.For cases 2 and 3, it is natural to conjecture that a ban on (explicit)exclusive dealing would be irrelevant. It turns out that this is almost correct. Proposition 5 below demonstrates that, in the presence of a ban, effective exclusion and common representation persist incases 2 and 3, respectively. However, in the course of proving thisresult, we isolate a condition under which, in case 3 (common repre-sentation), the imposition of a ban shifts payoffs from retailer 1 tomanufacturer B (see the Appendix for details). This occurs because

the ban alters out-of-equilibrium alternatives in a way that improvesBs ability to extract rents from retailer 1. Intuitively, if explicit exclu-sion is more protable than effective exclusion, then B need not cede as much surplus to secure representation when (explicit) exclu-sive dealing is proscribed.

Case 1 is of much greater interest. Although A enters into an ex-plicit exclusive deal with retailer 1, it does not necessarily follow that the imposition of a ban on this practice would end the effective ex-clusion of manufacturer B. Although it is impossible in this case to

sustain an effectively exclusive equilibrium wherein A producesx * A1, it may nevertheless be possible to achieve an exclusionary out-come through the use of a contract that induces retailer 1 to choosesome x A1 D . On the basis of our rst general principle (Sec. II),one might expect to obtain such an outcome as long as the joint prots for retailer 1, manufacturer A, and manufacturer B exceedthe joint prots received by these parties when B makes strictly posi-tive sales in market 1.

Following this intuition, we dene

x A1 argmaxx A1 D [R 1(x A1, 0) c A x A1].Retailer 1 and manufacturer A receive higher joint prots in market 1 from x A1 than from any other output level in the deterrence set D . Effective exclusion of B through selection of x A1 maximizes thetotal prots of retailer 1 and both manufacturers whenever the follow-ing assumption holds. 19

Assumption C4.[R 1( x A1, 0) c A x A1] A 2 c 1

j

( c 2 j 2) K B.

We use this condition to dene two subcases.19 Note that assumption C4 is always satised in case 2: since x * A1 D , we have x A1 x * A1, which implies that assumptions C4 and C2 are equivalent. This is not true

in case 1: when x * A1 D , assumption C4 is more demanding than assumption C2.



23/41


24/41


Among other things, proposition 5 tells us that, in case 1 a , theban on exclusive dealing fails to end Bs exclusion. Rather, when

2

R 1( )/ x A1x B1 0, A engages in nonexplicit exclusion by induc-ing retailer 1 to purchase enough output from A to render Bs partic-ipation unprotable. Thus explicit exclusion is replaced by effectiveexclusion implemented through quantity forcing or quantity dis-counts. The welfare consequences of this response are ambiguous.If retailer 1 practices perfect price discrimination (as assumed im-plicitly by Bork), social welfare declines. If retailer 1 is instead a con- ventional nondiscriminating monopolist, the increase in As output may enhance welfare (unless deterrence of B requires As output to

be sufciently excessive from a social perspective). V. Exclusive Dealing as a Consequence of

Incentive Conicts

In Section IV we saw that the potential for foreclosure of noncoinci-dent markets can provide a coherent motivation for exclusive deal-ing. A commonly expressed alternative view is that exclusive dealingarises in response to a manufacturers fear that common representa-tion would subject the retailer to conicting incentives. In this sec-

tion, we show how exclusive dealing can indeed arise when problemsof incentive provision are introduced.Before proceeding, we should stress that although we focus here

on a model with moral hazard, similar points could be establishedin other settings in which the provision of incentives is costly. Forexample, Marvels (1982) (informal) argumentthat exclusive deal-ing protects manufacturers quasi rentscan be viewed formally asan example of double moral hazard (manufacturers advertise, whereas the risk-neutral retailer can switch consumers among

brands). Since the double moral hazard problem also makes it costly to provide incentives, one can obtain similar results. 23

A. The Model

We consider a situation in which the retailer chooses nonveriableprices for each of the products it carries. 24 We denote the retail price

23 Similar effects also arise in settings in which the retailer possesses hidden infor-mation and either faces an interim individual rationality constraint or is risk-averse.See Martimort (1996) and Stole (1990). Given this fact and the results below, it issurprising that Marvel (1982, pp. 34) argues against the view that exclusive dealingis a device to obtain increased dealer promotional effort.

24 For example, the true price charged by a new car dealer is often unveriablebecause of trade-ins. The retailers price choice in this model could also be inter-preted as the choice of a nonobservable level of service that has a monetary valueto customers equal to its cost of provision. In any case, the basic points developedbelow hold for much more general kinds of nonobservable marketing choices.



25/41


of product j by p j for j A, B. When both products are carriedby the retailer, price choices of ( p A , p B) lead to a stochastic realiza-

tion of demand for each product j , given by x j q j ( p A , p B), where IR is a nonnegative random variable with distribution function( ). We adopt the normalization that E () 1, so that q j ( p A , p B)represents manufacturer j s expected sales level given retail prices( p A , p B). When rm j s product is carried exclusively at retail price p j , its sales are x j q j ( p j , ). Manufacturer j s production costs arec j per unit, and for simplicity we assume that the retailers only costsare the costs of acquiring products from the manufacturers. We alsoassume that q j (c j , ) 0 for j A, B.

Each manufacturer is restricted to offering contracts that condi-tion compensation on sales of only its own product (i.e., not on thesales of its competitor or on prices). Moreover, we restrict these pay-ments to be linear in sales: P j (x j ) F j j x j (actual incentive con-tracts often have this relatively simple structure; for one formal justi-cation, see Rey and Tirole [1986]).

We assume also that the retailer maximizes expected utility andhas a Bernoulli utility function of the constant absolute risk aversionform u (w ) 1 e aw , where a 0. Risk aversion (a 0) makes

incentive provision costly (i.e., the rst-best is not attainable) andthereby introduces the possibility that common representation willlead to contracting externalities (see Bernheim and Whinston1986a , theorems 2 and 3). As in Section III, a manufacturer earnszero if its contract is not accepted, and the retailer earns zero if it rejects both manufacturers offers.

To establish our results, we require one additional technical (but fairly standard) assumption. Let [ P A (q A , q B), P B(q A , q B)] denote theinverse of the function [ q A ( p A , p B), q B( p A , p B)], and suppose that

this inverse is well dened on IR 2

. DeneR (q A , q B; c A , c B)

j A, B

[P j (q A , q B) c j ]q j

(expected variable prots to the retailer as a function of expectedsales), and make the following assumption.

Assumption D1. The function R ( ) is twice continuously differ-entiable and strictly concave in ( q A , q B), and R ( )/ q A q B 0 at

all (q A , q B)

0.Under assumption D1, the mean sales induced by contracts[( F c A , c A ), ( F c B, c B)] are given by continuously differentiable func-tions q c j ( c j , c j ), j A, B, which are nonincreasing in c j and nonde-creasing in c j (strictly so at any ( c A , c B) such that [ q c A ( c A , c B),q c B( c A , c B)] 0).



26/41


B. Equilibrium Behavior

As in previous sections, one can dene j , c , and c to be the levelsof joint prots for the retailer and manufacturers under (respec-tively) exclusive representation of manufacturer j , cooperative com-mon representation, and noncooperative common representation.The general principles articulated in Section II (and formalized inthe Appendix) imply that c c is a necessary condition for exclu-sive dealing to arise in all undominated equilibria: equilibria in theintrinsic common agency game must involve some inefciency. Thenext result shows that this condition always holds when the coopera-tive outcome involves positive expected sales levels of both products.

Proposition 6. Suppose that assumption D1 holds and that [( F * A , * A ), ( F *B , *B )] maximizes the manufacturers joint protsin the intrinsic common agency setting with retailer reservation util-ity U 0. Then if [ q c A (* A , *B ), q c B(* A , *B )] 0, [( F * A , * A ),( F *B , *B )] is not a Nash equilibrium of the intrinsic common agency game. Hence, if all cooperative contracts involve positive expectedsales for both manufacturers, then c c .

Proposition 6 follows because the presence of retailer risk aversionmakes incentive provision costly: with cooperative contracts, themanufacturers retain some risk ( * j c j 0). But this immediately gives rise to an externality: if manufacturer j lowers j , this causesthe retailer to reduce q j and lowers manufacturer j s expectedprots. 25

The fact that c c creates the potential for exclusive dealing.Indeed, we saw in Section II that no common equilibria exist when c max{ A , B}: all equilibria are exclusive. The same principleapplies here. Thus, if max{ A , B} is close to c , we can expect exclu-sive dealing to arise; intuitively, the gain from having both productsavailable were the manufacturers to cooperate in incentive provisionis small relative to the loss due to incentive conicts.

To see this more concretely, consider the limiting case in whichproducts A and B are perfect substitutes with identical costs c A c B c . Obviously, A B c . Hence, as long as c c , exclu-sive dealing must arise. Though we cannot use proposition 6 directly here (because assumption D1 is violated in the limiting case of per-fect substitutes), we nevertheless obtain the expected result.

Proposition 7. Consider the case of perfect substitutes with iden-tical costs of production. Assume that R (q j , 0; j , 0) is twice continu-

25 When the retailer is risk-neutral, the derivations in the proof of proposition 6(see the Appendix) can be used to show that * j c j 0 for j A, B. Hence, thecooperative contracts are sellout contracts, which create no externalities across man-ufacturers. In this case, we would have c c .



27/41


ously differentiable and strictly concave in q j at all q j 0. Then c c A , and all equilibria are exclusive.

Proposition 7 follows because an efcient (common or exclusive)contract requires c . This is the standard consequence of thetrade-off between risk bearing and incentives. In contrast, in an equi-librium of the intrinsic common agency game, Bertrand-like compe-tition between manufacturers drives wholesale prices ( ) to (or be-low) marginal cost.

It is interesting to consider also the opposite limiting case in whichthe demands for products A and B are completely independent (i.e.,q j ( p j , p j ) depends only on p j ). This assumption removes the depen-

dence of manufacturer j s prots on manufacturer

j s choice of j and thereby eliminates contracting externalities. To see this, notethat manufacturer j s prots depend only on p j , j , and F j . Since j chooses j and F j , any externality must be experienced through ef-fects of j s contract on p j . But it is easy to verify that p j dependsonly on j , and not on j . Since there are no contracting externali-ties, the optimal cooperative contracts also form a Nash equilibriumof the intrinsic common agency game, which implies c c . More-over, under our assumptions, ( q A , q B) 0 in any cooperative com-

mon outcome, so j c

for j A, B. Thus we have the followingproposition.Proposition 8. Suppose that products A and B are independent

in demand. Then any undominated equilibrium entails commonrepresentation.

C. The Effects of Banning Exclusive Dealing

We now consider the effect of banning exclusive dealing. For the

case of perfect substitutes with identical costs of production, the next proposition demonstrates that a ban always leads to an inefcient outcome (recall from proposition 7 that, for efcient incentiveschemes, c ).

Proposition 9. Consider the case of perfect substitutes with iden-tical costs of production, and let * denote the lowest j among ac-cepted contracts. If exclusive dealing is banned, then * c and F j 0 in any contract accepted by the retailer.

In this case, the welfare consequences of a ban are simple. Con-

sumers benet because a lower wholesale price (

) leads to lowerretail prices. Manufacturers earn zero regardless. The costs of inef-cient incentive provision are borne entirely by the retailer, whosepayoff falls.

The second case considered in the previous subsection might at rst seem entirely straightforward. Since exclusion does not occur



28/41


with independent demand, one might well expect a ban to be incon-sequential. Caution is warranted, however; recall from Section IV C

that a ban can alter equilibrium payoffs even in cases in which exclu-sion would not occur. In the current instance, the effect of a ban iseven more surprising.

Proposition 10. Suppose that products A and B are independent in demand. If exclusive dealing is banned, no pure strategy equilib-rium exists.

We suspect (but have not veried) that the existence of mixed-strategy equilibria is generally assured. But if the manufacturers joint maximization problem is strictly concave, mixed strategies can-

not be second-best efcient. In that case, a ban cannot be Pareto-improving and may even reduce payoffs for all market participants.Thus, through subtle strategic channels, a ban on exclusive dealingcan reduce the efciency of economic activity even in cases in whichno exclusion occurs.

VI. Conclusions

In this paper, we have attempted to provide a conceptual framework

for analyzing the motivations for and effects of exclusive dealing. Insimple settings, our analysis corroborates Robert Borks evaluationof the practice: in particular, exclusion (whether explicit or not)occurs only when it is efcient (when we abstract from issues con-cerning imperfect extraction of consumer surplus). However, in that model, explicit exclusionary provisions are also superuous: ban-ning them is not harmful.

By introducing additional features, we generate models in whichthese provisions serve meaningful functions. We provide formal the-

oretical foundations for the view that exclusive dealing may beadopted for anticompetitive reasons (to enhance market power innoncoincident markets) and for the view that it efciently amelio-rates the incentive conicts associated with common representation. We use these formal models to study the consequences of a ban onthe practice. In either case, the welfare effects of a ban are complex.For example, even when exclusive dealing is used anticompetitively,a ban may simply lead to even less efcient forms of nonexplicit exclusion.

While these models do not encompass all possible motivations forexclusive dealing, our framework should be useful for studying theoperability and consequences of other motivations. For example, as we have already suggested, Marvels (1982) concernthat manufac-turers might free-ride with common representationcan be cap-tured in our framework.



29/41


In practical settings, it can be difcult to determine the motiva-tions for exclusive dealing. For example, in his discussion of the Stan-

dard Fashion case, Marvel (1982) argues that Standard was at-tempting to prevent competitors from free-riding by copyingpatterns that had proved to be popular. Yet Marvels characterizationof the facts is also consistent with the two motivations modeled inthis paper. For example, he attributes Standards poor performanceafter the decision to a competitors new innovation and new entry, without acknowledging that both of these developments may havebeen stimulated (as modeled in Sec. IV) by the courts proscription.Likewise, our model of incentive conict (in Sec. V) easily accounts

for Marvels observation that Standards wholesale prices were sig-nicantly above its marginal costs prior to the decision, as well asfor evidence indicating that manufacturers increased xed fees(charges for display equipment and catalogs) following the courtsdecision. Plainly, there is insufcient evidence to resolve Standardsmotivations.

Our models have two notable limitations. First, we have assumedthroughout that there is no incumbent manufacturer with a preex-isting contract. This reects reality in many, although not all, set-

tings. Aghion and Bolton (1987) and Rasmusen, Ramseyer, and Wi-ley (1991), for example, study the use of exclusivity provisions (ortheir cousin, stipulated damage provisions) when one manufacturerhas a rst-mover advantage (see also Segal and Whinston 1996). Sec-ond, and perhaps more important, we have restricted our focus tomarkets served by a single retailer. This is often unrealistic sinceexclusive dealing rarely precludes rival manufacturers completely from reaching consumers in a market. The extension of our analysisto such circumstances is an important area for future research. Re-

cent papers that make a start in this direction include Besanko andPerry (1993, 1994) (who follow Mathewson and Winter [1987] inrestricting attention to the simple wholesale price contracts) andMartimort (1996).

Appendix

For the sake of brevity, many of the following proofs have been abbreviatedthrough the omission of some details. A more detailed version is available

from the authors on request. We begin by proving some results for a general contracting game that subsumes all the specic models considered in Sections IIIV. The gameinvolves a retailer and two manufacturers ( j A, B), and the contractingprocess consists of the same three stages described in Section III. Contractsare arbitrary functions mapping observable outcomes to payments. The sets



30/41


31/41

exclusive dealing 93U e min{U : i i (U ) j i (U ) 0 for some i and j i } and has prots of j j (U e ) for the manufacturer who is served and j i (U e ) for the excludedmanufacturer.

As we show below, one can characterize common equilibria with refer-ence to an associated intrinsic common agency game, wherein the retailer isrestricted to serve both manufacturers or neither (Bernheim and Whinston1986a ). One obtains this game by imposing the restriction that e j for j A, B and by assuming that the manufacturers receive arbitrarily largenegative payoffs if the retailer rejects both offers. Let c (U ) denote thehighest aggregate payoff earned by the two manufacturers in any equilibriumof an intrinsic common agency game with retailer reservation utility U , andlet E c (U ) c A c B c denote the (set of) associated equilibriumchoices. Assumptions A1A4 imply that if ( P c

A , P c

B, c ) E c (U ), then

(P c A K , P c B K , c ) E c (U ).Lemma A1. Suppose that assumptions A1A4 hold. Then, for any ( P c A ,

P c B, c ), there exists ( P e A , P e B) such that [( P e A , P c A ), ( P e B, P c B), c ] is a commonequilibrium of the contracting game only if ( a ) u c (P c A , P c B, c ) U 0;(b ) ( P c A , P c B, c ) is an equilibrium of the associated intrinsic commonagency game in which the retailer has reservation utility u c (P c A , P c B, c ); and(c ) c j (P c A , P c B, c ) j j (u c (P c A , P c B, c )) for j A, B. If conditions a c holdand we also have ( d ) c j (P c A , P c B, c ) i j (u c (P c A , P c B, c )), then such a(P e A , P e B) exists.

Proof. Necessity is easily veried. For sufciency, we argue that if a d holdfor some ( P c A , P c B, c ), then there is a common equilibrium of the form[( P e A , P c A ),( P e B, P c B), c ] in which max j j u j (P e j , j ) u c (P c A , P c B, c ) for j A, B. Note, rst, that assumption A4 implies that exclusive contractsexist that satisfy this equality. Now, if condition a is satised, the retaileris willing to accept both manufacturers offers. Moreover, with exclusivecontract P e j being offered, any deviation by manufacturer j that causes theretailer to continue to accept manufacturer j s offer must give the retailera payoff of at least u c (P c A , P c B, c ). Condition b therefore implies that thereis no protable deviation for j that has the retailer accept both manufactur-

ers offers, whereas condition c implies that there is no protable deviationfor j that has the retailer accept only manufacturer j s offer. Finally, condi-tion d implies that no deviation that causes the retailer to reject manufac-turer j s offer can raise j s payoff either (since the retailer would then accept i j s offer). Q.E.D.

The models in Sections IIIV satisfy three further conditions that helpus characterize equilibria.

Assumption A5. There exist constants ( c , A A , B A , BB, A B) and a strictly increasing function g (U ) with g (U 0) 0 such that, for all U U 0, c (U ) c g (U ), j j (U ) j j g (U ) for j A, B, and j i (U )

j i for j A, B, i j . Assumption A6. For some j , j j 0 j , and A A BB max{ A , B,

c } 0. Assumption A7. For j A, B and i j , j i min{ 0i , j i (P e j , j )} for all

(P e j , j ).Given assumption A5, for j A, B, we can als

exclusive dealing.pdf

Documents