NET Institute* www.NETinst.org Working Paper #06-04 September 2006 Tying in Two-Sided Markets with Multi-Homing Jay Pil Choi Michigan State University * The Networks, Electronic Commerce, and Telecommunications (?NET?) Institute, http://www.NETinst.org, is a non-profit institution devoted to research on network industries, electronic commerce, telecommunications, the Internet, ?virtual networks? comprised of computers that share the same technical standard or operating system, and on network issues in general. Tying in Two-Sided Markets with Multi-Homing by Jay Pil Choi* September 2006 A bstract This paper analyzes the effects of t ying arrangements on market competition and social welfare in two-sided markets when economic agents can engage in multi-homing, that is, they can participate in multiple platforms in order to reap maxim al network benefits. The model shows that tying induces more consumers to multi-home and makes platform-specific exclus ive content available to more consumers, which is also beneficial to content providers. As a result, tying can be welfare-enhancing if multi-homing is allowed, even in cases where its welfare impacts are negative in the absence of multi-homing. The analysis thus can have important implications for recent antitrust cases in industries where multi-homing is prevalent. JEL Classification: L1, L4. Keywords: tying, two-sided markets, (indirect) network effects, multi-homing. * Departm ent of Economics, Michigan State University, 11 0 Marshall-Adams Hall, East Lansing, Michigan 48824- 1038, T el: 517-353-7281, E-m ail: choijay@m su.edu This research was partially funded by the NET Institute whose financial support is gratefully acknowledged. 30/ 09/ 06 1 I. Introduction This paper analyzes the effects of t ying arrangements on market competition and social welfare in network industries when consumers can engage in multi-homing, that is, consumers can participate in multiple platforms (or purchase multiple products) in order to reap maximal network benefits. The paper is partly motivated by the recent antitrust cases concerning Microsoft. In the European case, for instance, it has been alleged that the company?s tying practice of requiring W indows operating system users to accept its Windows Media Player software is anti-competitive and hurts digital media rivals such as RealNetworks. 1 However, m ulti-homing is common in digital media systems. Many users have more than one media player and many content providers offer content in more than one format, which counteracts the tendency towards tipping and the lock-in effects in industries with network effects. To analyze the effects of tying in m arkets such as digital media, I adopt the framework of platform competition in two-sided markets. The defining characteristic of two-sided markets is indirect network effects or inter-group network externalities that arise through improved opportunities to trade with the other side of the market. In the digital media case, content providers and final consumers constitute the two sides that trade with each other. For instan ce, as more content is available in streaming media the more valuable media player programs become, and vice versa. Other prom inent examples of econom ic importance include auction sites such as eBay and Yahoo where buyers and sellers get together to consummate a deal, credit card payment systems such as Visa and MasterCard 1 On March 24, 2004, the European Union ruled that Microsoft is guilty of abusing the "near-m onopoly" of its Windows PC operating system and fined it a record 49 7 million euros ($61 3 m illion). The case is being appealed by Microsoft. 30/ 09/ 06 2 where both merchants and consumers need to participate in the same system, video game platforms such as PlayStation, X-box a nd Ga meCube where game developers and consumers constitute the two distinct sides, etc.2 In such markets, the need to get all sides of the market to get on board creates a so-called ?chicken and egg? problem (Caillaud and Jullien, 2003) in that mem bers of each group are willing to participate in the market only if they expect m any members from the other side to participate. The literature on multi-sided markets is mainly concerned with the optimal pricing structure to coordinate the demands of distinct groups of customers who need each other in some way.3 More recently, the importance of studying antitrust issues in two-sided markets has been recognized by several authors. Evans (2003) and W right (2003), for instance, provide a general discussion on antitrust policy in two-sided markets and call for caution in applying the traditional one-sided logic to two-sided markets in the antitrust arena. However, their discussion is mainly informal and does not deal with tying arrangements. Rochet and Tirole (2003b) is a notable excep tion in the analysis of tying in two-sided markets. They provide an economic analysis of the tying practice initiated by payments card associations Visa and MasterCard in which m erchants who accept their credit cards were forced also to accept their debit cards.4 They show that in the absence of tying, the interchange fee 2 See Evans (2003) and Ro chet and Tirole (2003 a) for more ex amples of multi-sided markets. 3 See Arm strong (forthcoming) and Jean-Charles R ochet and Jean Tirole (forthcoming). 4 This tie-in practice, the so-called ?honor-all-cards? rule, has been challenged recently by major merchants including Walmart in a class action suit. In the class action suit on behalf of thousands of retailers, the stores argued that Vis a and MasterCard unfairly requi red merchants to accept their debit cards, which require a customer' s signature to verify a transaction, to ex clude PIN-based on-line debit cards. The case was originally filed in 199 6. Since then, it was certified as a class action in Febru ary of 2000. The trial was set to commence on April 28, 2003, following the defendants' unsuccessful appeals of the class-certification decision and supplementation of summary jud gment motions. Ho wever, Vi sa and MasterCard each separately agreed to settle the antitrust lawsuit brought against them. Under the settlements, Visa is set to pay $2 .025 billion to merchants over the nex t 10 years and MasterCard is set to pay $ 1 .025 billion over the same period. They also 30/ 09/ 06 3 between the merchant?s and the cardholder?s banks on debit is too low and tends to be too high on credit compared to the social optimum. Tying is shown to be a mechanism to rebalance the interchange fee structure and raise social welfare. Their model, however, is tailored to analyze the paym ent card industry and the recent antitrust suit involving V isa and MasterCard. In particular, the analysis focuses on tying by a non-profit association to reflect the status of credit card associations. I show that tying induces more consumers to multi-home and makes platform- specific exclusive con tent available to more consumers, which is also beneficial to content providers. As a result, tying can be welfare-enhancing if multi-homing is allowed, even in cases where its welfare impacts are negative in the absence of multi-homing. The analysis thus can have important implications for recent antitrust cases and suggests caution in applying the traditional theory of network effects and tipping to markets where multi- homing is prevalent. This paper is also closely related to the literature on the ?leverage theory? of tying. According to the "leverage theory" of tyi ng, a two-product firm with monopoly power in one market can monopolize a second m arket using the leverage provided by its monopoly power in the first market. Whinston (1990), for instance, shows that if the market structure in the tied good market is oligopolistic and scale economies are present, tying can be an effective and profitable strategy to alter market structure by making continued operation unprofitable for tied good rivals. It is important to keep in mind, however, that in Whinston?s basic m odel inducing the exit of the rival firm is essential for the profitability of agreed to drop its "h onor all cards" po licy, which will allow retailers to accept its credit cards without also accepting debit cards. 30/ 09/ 06 4 tying arrangements.5 Thus, if the competitor has already paid the sunk cost of entry and there is no avoidable fixed cost, ty ing cannot be a profitable strategy.6 Choi and Stefanadis (2001) extend the analysis by investigating im plications of tying for innovation incentives. They show that when an incumbent monopolist faces the threat of entry in systems markets consisting of complementary components, tying may make the prospects of successful entry less certain, discouraging rivals from investing and innovating.7 Carlton & Waldm an [2002] are es pecially worth mentioning in relation to this paper. They investigate how the tying of complementary products can be used to preserve and create monopoly positions. Their analysis focuses on two mechanisms through which tying can be used in an anticompetitive way: entry costs and network externa lities. In particular, their model with network externa lities shows that the presence of network externalities for the complementary good can result in the strategic use of tying to deter entry into the primary market. The nature of network effects in their paper, however, is direct and thus does not ex plicitly accounts for the peculiarities of two-sided markets. In addition, none of these papers in the tying literature, however, take into consideration the possibility of multi- homing seriously in the analysis. Carlton and Waldman (2002), for instance, assume that ?if a consumer purchases a tied good consisting of one unit of the monopolist?s prim ary good and one unit of its complementary good, then the consumer cannot add a unit of the 5 Whinston (1 99 0) points out that if the heterogeneity of consumer preferences are allowed for the tying good, tying can also serve as a price discriminating device and ex clusion of the rival firm is not necessary for the profitability of tying. See also Carbajo et al. (199 0). 6 Carbajo, de Meza, and Seidman (199 0) and Chen (1997) provi de an alternative theory of strategic bundling in which bundling plays the role of a product-differentiation device. As in this paper, bundling does not requ ire the ex it of the rival firm to be profitable. Ho wever, bundling is used to segment the market and relax competition. 7 In related papers, Choi (1 9 9 6, 2004) demonstrates that even in the absence of ex it by the rival firm, bundling can be a profitable strategy via its long-term effects on competition through innovation. 30/ 09/ 06 5 alternative producer?s complem entary good to the system (italics added, p. 199).? In other words, either they do not allow the possibility of multi-homing or multi-homing does not arise in equ ilibrium. Doganoglu and W right (2006a) explicitly c onsider multi-homing as a way to reap greater network benefits and analyz e its implications for price competition. Their focus, however, is on the ques tion of whether multi-homing can be a substitute for compatibility. In contrast, my focus is on the effects of tying on competition in the presence of multi- homing. Currently, formal economic analysis of tying that ex plicitly accounts for the possibility of multi-homing is virtually non-exis tent. The analysis in the paper intends to fill this gap in the literature and can have important implications for recent antitrust cases in industries where multi-homing is prevalent. The remainder of the paper is organized in the following way. In section II, I set up a basic model of two-sided markets. In section III, I analyz e the effects of tying arrangements on competition in two-sided markets in the absence of multi-homing. Section IV consid ers the possibility of multi-homing in the analysis of tying. Concluding remarks follow. I I . The Basic Model of Tw o-Sided Markets In this section, I lay out a very simple model of two-sided markets and derive the market outcome in the absence of tying and multi-homing. The model is a modification of the framework developed by Ar mstrong (forthcoming) and Rochet and Tirole (2003a). The analysis in this section will be used as a benchmark to investigate the effects of tying in two- 30/ 09/ 06 6 sided platform markets in section III. In section IV, I will modif y the model to account for the possibility of multi-homing. The model consists of three classes of agents. There are two distinct types of customer groups who interact with each other and intermediaries who provide platforms to enable these two customer groups to ?meet? each other. In the exam ple of streaming multi- media players, the two customer groups can be described as content providers and consumers who download content through the Internet. There are currently three major platform providers: Real Networks, Micros oft and Apple. In m y model, these software/platform providers can be considered as the intermediaries who compete in two- sided platform businesses. Let m e assume that there are two intermediaries index ed by i = A, B. For concreteness, let me label the two customer groups as content providers and consumers as in the streaming media industry. The two intermediaries compete for market share within each group. Let pi and qi denote intermediary i?s charge to content providers and consum ers, respectively, where i = A, B. The intermediaries? costs of serving each co ntent provider and final consumer are given by c and d, respectively. Finally, the num ber of content providers and consumers who participate in platform i are denoted by mi and ni, respectively. As i n Ar mstrong (forthcoming), I consider a situation in which at least one side of the market is characterized by exclusive interm ediation. More specifically, I assume that final consumers ?single-home,? that is, they participate in only one platform. II.1. Content Providers I assume that there is free entry in the market for content provision. Content providers are heterogeneous in their fixed cost of creating c ontent, which is denoted as ?. 30/ 09/ 06 7 The content providers incur this cost twice if they multi-home, i.e., make their content available in digital form on both platforms. I normalize the num ber of potential content providers to 1 and let F(?) be the distribution function for ?. Each content provider gains additional utility (profit) of ? from each consumer who has access to her content. The profit for content providers who create content on platform i is given by ? ni ? pi ? ? when her fixed cost of creating content is ? and the number of final consumers who participate in platform i is ni. A type- ? content provider is willing to create content for platform i if ? < ? ni ? pi Thus, the number of content providers on platform i is given by: mi = F(?ni ? pi) (1 ) This implies that the larger the number of consumers participating in platform i is, the greater the amount of content that will be provided on that platform. II.2. Consumers To analyze the con sumers? choice of platform, I adopt the Hotellin g model of product differentiation. I assume that two platforms, A and B, are located at the two extrem e points of a line with length equal to 1. Consumers, whose size is norm alized to 1, are uniformly distributed along the line. Each consumer?s utility of participatin g in a platform depends on the number of content providers on the same platform. More specifically, the availability of each additional content provider generates additional utility of b. Consumers are also allowed to multi-home, but it will be shown later that they will choose to participate in only one platform. If a consumer located at point x participates in platform A, his utility is given by b mA ? qA ? tx while his utility from participating in 30/ 09/ 06 8 platform B is given by b mB ? qB ? t(1? x). Assum ing that the consumer market is covered, this specification implies that the number of consumers participating in platforms A and B are, respectively, given by: nA = 1 ( ) ( 2 2 A B A Bb m m q q t ? ? ?+ ) and nB = 1 ? nA = 1 ( ) ( 2 2 B A B Ab m m q q t )? ? ?+ (2) II.3. Platform Competition without Tying Platforms compete in prices to attract consumers in both sides of the market. Each intermediary i ?s objective function is given by: ,i ip q Max mi (pi ? c) + ni (qi ? d), (3 ) where mA and nA are jointly determ ined by equations (1) and (2). I mainly focus on the symmetric equ ilibrium in this model. To derive the symmetric equilibrium price for content providers, pA = pB = p*, let me consider a situation where each platform intermediary has market share of ? in the consumer side of the market (nA = nB = ?) and offers consum ers the utility of u = b mi ? qi (gross of transportation costs). Now consider an intermediary?s prof it maxim ization problem given this utility u , maintaining u = b mi ? qi constant. Then by substituting ni = ? and qi = b mi ? u , intermediary i ? s profit can be written as: ?i = mi (pi ? c) + 12 ( b mi ? u ? d) = F( 1 2 ? ? pi) (pi ? c) + 1 2 [ b F( 1 2 ? ? pi) ? u ? d ] (4) The first order condition with respect to pi yields: 30/ 09/ 06 9 ? F?( 1 2 ? ? pi) (pi ? c) + F( 1 2 ? ? pi) ? 1 2 b F? ( 1 2 ? ? pi) = 0 (5) The (symmetric) equ ilibrium price for content providers is thus given by p* = c ? 1 2 b + 1? , (6) where ? = 1'( *) 2 1( * 2 F p F p ? ? ? ? ) can be considered as a measure of the price elasticity of content supply. In deriving the symmetric equili brium price for final consumers, qA = qB = q*, I first note a one-to-one relationship between qi and ni from equation (2), given the rival intermediary?s price for final con sumers qj , where i = A, B, and j ? i. qi = qj + t ( 1 ? 2ni) + b[ F(?ni ? p*) ? F(?(1?ni) ? p*)] (7) It turns out to be more convenient to treat ni as a control variable for intermediary i in the consumer side of the market. in Max?i = mi (p* ? c) + ni ( qi ? d) = F(?ni ? p*) (p* ? c) + ni {qj + t ( 1- 2 ni) + b[ F(?ni ? p*) ? F(?(1?ni) ? p*)] ? d } (8) The first order condition with respect to ni yields: ? F?(?ni ? p*) (p* ? c) + {qj + t ( 1 ? 2ni) + b[ F(?ni ? p*) ? F(?(1?ni) ? p*)] ? d } + ni{? 2t + b ? [ F?(?ni ? p*) + F?(?(1?ni) ? p*)] }= 0 (9) At the symm etric equilibrium , the first order condition (9 ) is satisfied at ni = 1/2. This implies that the symmetric equilib rium price qA = qB = q* can be characteriz ed as: q* = d + t ? ? F?( 1 2 ? ? p*) [ b + p* ? c ] (1 0) 30/ 09/ 06 10 Using equation (6), the s ymmetric eq uilibrium price q* can be written as q* = ( d + t) ? ? F ? 1 2 b?F? (11), which can be interpreted as the standard Hote lling price (d + t) adju sted by two terms representing inter-group network externalities in the two-sided markets. The first term (?F) represents direct inter-group extern alities that an additional consumer bestows on content providers. The second term ( 1 2 b?F?) represents the indirect inter-group externalities that an additional consumer gives to other existing c onsumers through the feedback process through which additional content is provided. II.4. Socially Optimal Outcome In this subsection, I conduct a welfare analysis in which I derive the socially optimal outcome and compare it with the market equ ilibrium. Under the Hotellin g model with the assumption of a covered market on the consumer side, social welfare depends only on the price for content providers (pi) that affects the amount of content provided in the market. With the symmetric outcome, social welfare as a function of the price for content providers (p1 = p2 = p) can be written as: W = 2( p ? c) F( 1 2 ? ? p) + bF( 1 2 ? ? p) + 2[ 1 2 0 1( ) 2 p p dF ? ( )? ? ?? ? ?? ] + 2[ 120 txdx? ] (12) The first order condition with respect to p yields the following socially optimal price for content providers: 1 2 op c= ? b (1 3 ) 30/ 09/ 06 11 The socially optimal outcome thus requires belo w cost pricing for content providers to take into account of their positive externalities to consumers. I I I . An Analysis o f Tying in Two-Sided Markets with Single-Homing Consumers To analyze the ef fects of tying on competition in two-sided markets, I assume that intermediary A is also a monopolist in a related market called M with unit production cost of cM. More specifically, to reflect circumstances in the antitrust case against Microsoft concerning the tie-in of Media Player with the Windows operating system, assume that the good/service M in the monopolized m arket (operating systems) is necessary for consumers to participate in the two-sided market (streaming multi-media) analyzed in this pap er. All final consumers have valuation of v (> cM) for product M. It is assumed that entry into market M is not feasible.8 I consider the following two-stage game. In the first stage, firm A (the monopolistic supplier of product M) decides whether or not to tie the two products. A price game ensues in the second stage with the tying decision in the previous stage taken as given. The timing assumption reflects the fact that the tying decision through product design is a longer term decision that cannot be modified easily compared to the price decision. The outcomes are described below and depend on firm A?s tying decision in the first stage. III.1. No Tying If the two products are not bundled, let me assume that consumers buy product M prior to participating in the two-sided market since M is essential for the latter activity. Due 8 Fi rm 1 m ay have a patent or have an installed base that makes entry unprofitable in the presence of switching costs or network ex ternalities (Farrell and Kl emperer, 2001). 30/ 09/ 06 12 to the essentialness of product M, the monopolist can extract consum er surplus from participating in the two-sided market. Let u *= b m*? q* (gross of transportation costs) be the equilib rium utility offered in the two-sided market. Assu me further that (v ? cM) is sufficiently large that it is in best interest of firm M to cover the market. The consumer who has the lowest surplus is the one located in the middle of the line whose surplus is given by u *? 2 t . The monopolist will charge the price of v + u *? 2 t . In the two-sided market I am interested in, the analysis in the previous section applies. III.2. Tying Suppose that the monopolist bundles the two products and charges a price of % Aq for the bundled product on the consumer side.9 I assume that v (> cM) is sufficiently large that firm A will price the bundled good so that every final consumer purchases it. With n A = 1, the number of content providers on platform A then is given by F(? ? ? Ap ) when the tying firm charges content providers ? Ap . This implies that the bundled good price is set at the price such that the final consumer located at x =1 gets zero surplus: % Aq = v + b F(? ? ? Ap ) ? t (15) The tying firm?s profit maxim iza tion problem can be written as: ? Ap Max?A = mA ( ? Ap ? c) + nA ( % Aq ? d) = F(? ? ? Ap ) ( ? Ap ? c) + [ v + b F(? ? ? Ap ) ? t ? d] (16) 9 Variab les corresponding to tying are denoted with a tilde. 30/ 09/ 06 13 The first order condition with respect to ? Ap is given by ? F?(? ? ? Ap )( ? Ap ? c) + F(? ? ? Ap ) ? b F?(? ? ? Ap ) = 0 (17) Thus, the eq uilibrium bundle price for final consumers ( % Aq *) and the price for content providers under tying ( ? Ap *) are characterized as: ? Ap * = c ? b + % 1 ? , where %? = ?? '( *) ( * A A F p F p ? ? ? ? ) and % Aq * = v + b F(? ? ? Ap * ) ? t (18) III.3. Welfare Analysis In this subsection, I compare the market outcomes under tying and no tying and provide a welfare analysis. There are three channels through which tying can affect social welfare due to the monopolization of both side s of the market. First, all consum ers patronize th e tying firm?s platf orm. This implies that there is less variety in the market. As a result, there are less desirable matches between the consumers and platforms, leading to higher overall ?transportation costs.? Second, content is provided only on the tying firm?s platform, whereas the same content was produced on both platforms in the absence of tying. Thus, there are savings in duplication costs under tying. Third, the number of entrants in the content side of the market that determines the availability of content can differ across regimes. The first effect is negative while the second effect is positive. The sign of the third effect is ambiguous. The coordination of consumers on the tying firm?s platfor m enhances the incentive to enter the content side of the market. Howev er, the tying firm?s pricing decision in that side of the market can offset this positive effect. 30/ 09/ 06 14 To conduct a more explicit welfare analys is, I will analyze th e effect of tying assuming that ? is uniformly distributed on [0,1]. With the uniform distribution, it can easily be verified from equa tions (6) and (18) that p* = 2 4 4 c b ?? + , ? Ap * = 2 2 2 c b ?? + This implies that the number of content providers under each regime is given by: m* = F( 1 2 ? ? p*) = 4 4 2 b c? + ? (1 9 ) ? *Am = F(? ? ? Ap * ) = 2 2 2 b c? + ? (20) Thus, in the case of the uniform distribution, there is more variety of content available with tying ( ? *Am > m*). To explore the welfare implications of ty ing, I note that welfare under tying can be written as: ?W = ( ? Ap * ? c) F(? ? ? Ap * ) + bF(? ? ? Ap * ) + [ ? ? * 0 ( *) ( Ap Ap dF ? )? ? ?? ? ?? 10 txdx?] + [ ] With the uniform distribution, it can be verified that social welfare under each regime are given by: W = 23 ( 2 ) 16 4 tb c? + ? ? , ?W = 23 ( ) 8 2 tb c? + ? ? Thus, the social welfare change due to tying can be written as: ?W = ?W ? W = 2 23 [( ) 2 ] 16 4 tb c? + ? ? (21 ) The result thus suggests that the welfare implications of tying depend on the relative magnitude of inter-group exte rnalities (? and b) and the extent of pr oduct differentiation. If 30/ 09/ 06 15 the ex tent of inter-group exte rnalities (? and b) is significant compared to that of product differentiation (t), tying can be welfare-enhancing since the benefit from internalizing the inter-group network externalitie s outweighs the loss of product variety. Otherwise, tying reduces welfare. IV. Competition in Two-Sided Market s with Both Sides Multi-Ho ming The analysis above considers situations in which the consumer side of the market is characterized by exclusive interm ediation. Howe ver, this assumption is at odds with the prevailing condition in many two-sided markets such as the digital media and the payment card industries. In the digital media case, many users have more than one media player and many content providers offer content in more than one format. The payment card portrays a similar picture with consumers carrying more than one payment card and merchants accepting multiple payment cards. In this section, I modify the basic model to exp licitly analyze the possibility o f multi-homing on both sides. IV.1. Content Providers In the previous sections, I assumed that there is free entry in the market for content provision and did not make a distinction on whether content available across platforms is the same or different. With the assumption of single-homing by consumers, all that matters for consumers is the amount of content available for each platform. Howeve r, once multi- homing is allowed on the consumer side, it makes a difference; in the symmetric equilibrium with the same amount of content available for each platform, consumers would not have any incentive to multi-home and the equ ilibrium identified in section II continues 30/ 09/ 06 16 to be an equ ilibrium even if multi-homing is allowed on the consumer side as long as the same content (not just the sam e amount) is provided across the platforms. However, if the content is different across the platforms, the previous equilibrium may not survive with the possibility of multi-homing on the consumer side. In order to make the possibility of multi-homing on the consumer side play a role, assume that there are two types of content available. One type of content is m ore suitable for one of the two platforms (formats) whereas the other type of content is suitable for both formats. To simplify the analysis, let me assume that when content is of the first type, that is, it is more suitable for one of the two formats, it is not economically feasible to encode in the other format. More specifically, the total measure of content potentially available for each format is normalized to 1. Am ong them, the proportion ? is of the first type and thus can be encoded only for a particular format whereas (1 ??) can also be encoded in the other format. The existence of exclus ive content available for each format creates incentive for consumers to multi-home. When the second type of content is encoded for both formats, content providers are said to multi-home. IV.2. Consumers The consumer side of the market is the same as in the previous sections. The only modification is that consumers are now allowed to multi-home. As a result, th ere are three choices for consumers, assuming that the market is covered as before. Consumers can choose to either single-home or multi-home. If they decide to single-home, they choose one of the two platforms in which to participate in. IV.3. Market Equilibrium in the Two-Sided Market with Multi-Homing 30/ 09/ 06 17 I am interested in an equ ilibrium in which both content providers and consumers multi-home. Imagine a situation in which each platform has available exc lusive content of measure ? and nonexclusive content of m easure (1 ??). In other words, the nonexclusive content is available for both formats. Consider a consumer located closer to platform A who would thus choose to participate in platform A in a symm etric equilibrium if he chooses only one platform. Now I analyze th e consumer?s incen tive to multi-home, that is, to participate in platform B in addition to A. If a consumer located at point x participates in platform A, his utility is given by = b m( , )A AU q x A ? qA ? tx as before. With the assumption that mA =1, I have = b ? q ( , )A AU q x A ? tx . If the consumer multi-homes, his utility is given by = b m ? q( , , )AB A BU q q x A ? tx ? qB ? t(1-x), where m is the total amount of content available to consumers who multi- home. Since each platform has duplicative content of measure (1 ??), I have m = 1 +?. As a result, the utility from multi-homing is given by = b (1 +?) ? (q( , , )AB A BU q q x A + qB ) ? t. The location of the consumer who is indifferent between single-homing A and m ulti-homing is given by: x = 1 ? Bb q t ? ? Similarly, the location of the consumer who is indifferent between single-homing B and multi-homing is given by: y = Ab q t ? ? This implies that the number of consumers who single-home platform i is as follows (see Figure 1). 30/ 09/ 06 18 = 1 ? in jb qt ? ? , where i = A, B and j ? i. (22) The number of consumers who multi-home is given by Mn = y ? x = 2 ( ) 1A Bb q q t ? ? + ? (23 ) Let AN and denote the total number of consumers who participate in platform A and B, respectively. Then, we have BN AN = y = An + Mn = A b q t ? ? , = 1 ? x = + BN Bn Mn = Bb qt ? ? (24) Ax n= B 1 1 By n= ?0 A BNAN A Only Consum ers Who Multi-Hom e B Only Figure 1. The Choice of Consum ers On the content provider side, the incentives to participate in each platform depend on the configuration of consumers on the other side of the market. Let m e assume a situation in which the consumer side market is covered and some consumers multi-home, that is, AN + > 1 with BN Mn (= AN + ? 1 ) multi-homing consumers. Exclusive c ontent for platform A will be p rovided if ?N BN A ? pA ? 0. For nonexclusive conten t, the incentives to encode in format A depend on whether the sam e content is provided for the other format. If 30/ 09/ 06 19 it is already provided for format B, the condi tion for a content provider to multi-home, that is, to encode in duplicative format A is given by ?nA ? pA ? 0. With multi-homing on the consumer side, NA > nA. This implies that platform A can either charge ?NA and attract only ? exclus ive content providers or charge ?nA and attract both excl usive and nonexclusive content providers. Let m e analyze platform A?s profit m axi mization problem assuming that it serves both exclusive and nonexclusive content provi ders with nonexclusive content providers multi-homing. Conditions for such behavior to constitute an equilibrium will be derived later. In such a case, we have = 1 ? An Bb qt ? ? . Notice that with consumers multi- homing, the number of consumers single-homing A ( ) depends only on the other platform?s price charged to consum ers (q An B). Thus, the optimal price for platform A on the content provider side depends on qB under the configuration I consider, and is given by pA* = ?nA = ? ( 1 ? Bb q t ? ? ) (25 ) This implies that under the configuration in which both consumers and content providers multi-home, each platform?s opti mal price for each side is independent of the other. On the consum er side, platform A solves the following problem : Aq Max ( qA ? d) NA = ( qA ? d) Ab q t ? ? (26) Thus, the optimal price on the consumer side is given by qi* = 2 b d? + , i = A, B (27) 30/ 09/ 06 20 The optimal price on the consumer side implies that NA = NB = 2 b d t ? ? . For this to be consistent with the consumer side multi-homing, we need NA + NB > 1, that is, b d t? ? > , which I assume to hold: A1 . b d t? ? > This condition means that for multi-homing to occur on the consumer side, the amount of exclusive content and the network benefits for consumers should be sufficiently high compared to the cost and ?transportation? parameters. In addition, for the multi-homing configuration postulated above to constitute an equilibrium , I need to have ?ni > ? ?Ni , that is, attracting both exclusive and nonexclusive content providers yields a higher payoff for platforms than attracting only excl usive content providers. With the equilib rium price of qi* = 2 b d? + on the consumer side, I have ni = 2 ( ) 2 t b d t ?? ? and Ni = 2 b d t ? ? . Thus, the condition can be written as follows: A2 . (1 )( ) 2 b d t? ?+ ? < I assume that both A1 and A2 hold in the rem ainder of the paper. IV.2. Tying As in the previous section, I assum e that v (> cM) is sufficiently large that firm A will price the bundled good so that every final consumer purchases it. Given that every consumer has product A, I analyze inc entives for consumers to multi-home, that is, to participate in platform B in addition to A. Given that all consum ers already have A, nonexclusive content providers have less incen tive to decode the content in duplicate for 30/ 09/ 06 21 format B. I thus consider an equilibrium in which all nonexclusive content is provided only for platform A. For m ulti-homing to take place under tying on the consumer side, it is necessary that exclusive content for p latform B be provided. When there exists ? amount of exclus ive content for platform B, the additional benefit of multi-homing for a consumer located at x from platform B is given by b? ? tx. This implies that the number of multi- homing consumers is given by Mn% = = BN% B b q t ? ? % , where is the price charged to consumers by platform B. The m axi mum price platform B ca n charge to content providers when Bq% Mn% = consumers multi-home is BN% *Bp% = ? . As a result, platform B?s profit maxi mization is given by: BN% Bq Max % ?(? ? c) + ( BN% Bq% ? d) = ?(? BN% Bb qt ? ? % ? c) + ( Bq% ? d) Bb qt ? ? % (28) The first order condition for the above problem yields = *Bq% ( ) 2 b d? ?? + (29 ) The number of consumers who participate in platform B and thus m ulti-home is given by *=BN% Mn% * = ( ) 2 b t d? ?+ ? (3 0) IV.3. Welfare Analysis To explore the welfare implications of ty ing, I note that welfare under no tying and tying can be written as: W = (1 + Mn ?) b ? (1+ Mn ) d ? [ 1 0 BN txdx ?? + 10 AN txdx?? + Mn t ] + [ ? ( AN + ) + (1 ??)] ? ? 2 c, (31) BN 30/ 09/ 06 22 where NA = NB = 2 b d t ? ? and Mn = 1b dt ? ? ? ?W = (1 + Mn% ?) b ? (1+ Mn% ) d ? [ + 1 0 Mn txdx ?? % Mn% t ] + [ ? ( 1+ ) + (1 ??)] ? ? (1+?) c, (3 2) BN% where Mn% = = BN% ( ) 2 b d t ? ?+ ? Thus, the social welfare change due to tying can be written as: ?W = ?W ? W = ( Mn% ? Mn )[[ ? (? + b) ?d)] + (1 ??) c ? { [ + 1 0 Mn txdx ?? % Mn% t ] ? [ + + 10 BN txdx?? 10 AN txdx?? Mn t ] } (33) Notice that Mn% ? Mn = ( ) 22 b d t t ? ? ? + + > 0 by assum ption A2, which implies that tying induces more consumers to multi-home and makes platform-specific ex clusive content available to more consumers, which is also beneficial to content providers. The first term in equation (33) thus represents th e net beneficial effects of wider availability of exclusive content due to tying. There are two channels. F irst, tying induces all consumers to have access to ex clusive content for platform A. Second, the number of consumers who have access to ex clusive content for platform B also increas es. This can be seen from the comparison of NB = 2 b d t ? ? and = BN% ( )2 b t d? ?+ ? . In addition, non-exclusive content providers need to participate in only one platform with tying rather than both ones since every consumer on the other side participates in platform A. The second term represents 30/ 09/ 06 23 such cost savings associated with non-duplication of the same content.10 Both effects are positive. However, tyin g may increase overall ?transportation costs?, which is represented by the third term. Nonetheless, the simple structure of the model yields an unambiguous answer concerning the welfare effects of tying. To see this, I manipulate equation (33) as ?W = ?W ? W = ( Mn% ? Mn )[[ ? (? + b) ? d ? t ] + (1 ??) c ? { [ 1 0 Mn txdx ?? % ? [ 10 BN txdx?? + 10 AN txdx?? ] } (34) The first-term in equatio n (34) is stil l positive by assumption A1. In addition, the express ion in the curly bracket is negative since Mn% = ( ) 2 b d t ? ?+ ? > 2 b d t ? ? = NA = NB . Therefore, ?W = ?W ? W > 0, that is, ty ing is welfare-enhancing in this simple model. To explore the role of multi-hom ing in the model, it is instructive to consider the welfare effects of tying in a situation where tying prevents consumers from multi-homing.11 Without multi-homing, all consumers will use only the tied product in the two-sided market. This implies that all content is provided for format A and exclusive con tent for format B will be no longer available. The welfare level under tying in the absence of multi-homing is given by = ? + b ? c ? d ? (3 3 ) ? SHW 1 0 txdx? 10 In this sense, tying induces more multi-homing on the consumer side but less multi-homing on the content provider side. 11 This would be the case if the monopolist engages in technical tying in which it designs its product in a way that a competitor?s product cannot interoperate with the tying product. My model suggests that such a practice can be anti-competitive. 30/ 09/ 06 24 If the marginal costs of serving additional customers are small on both sides (i.e., c ? 0, d ? 0) as in the case of digital media systems, I can approxim ate ?W = ? W as ? SHW ?W = ? W ? ? SHW 2( )( ) 4 2 b t b tb t t ? ?? ? ?? ?? + + ?? ?? ? < 2 2( )( ) 4 4 b t b tb t t ? ?? ? ?? ?? + + ?? ?? ? t = ( ) (4 3 ) 4 b t b t t ? ?? ??? + ? < 0 (34) Thus, tying is unambiguously welfare-reducing if multi-homing is not allowed. This result is in sharp contrast to the result obtained with the assumption of multi-homing and highlights the importance of explicitly consider ing the role of multi-homing in the antitrust analysis of network industries. V. Concluding Remarks I analyze the effects of tying arrangements on competition in markets with indirect network effects by using the framework of two-sided markets. In particular, I develop a model of network competition that ex plicitly incorporates the possibility of multi-homing. I consider two possible cases, one in which only the content side is allowed to multi-home with the consumer side of the market being characterized by exclusive interm ediation, and the other in which both content providers and final consumers are allowed to participate in both platforms. My analysis is motivated by the prevailing condition in the digital media market in which content providers encode in multiple formats and consumers use multiple media players. Multi-homing has potential to counteract the tendency towards tipping and the 30/ 09/ 06 25 lock-in effects in industries with network effects. As a resu lt, tying does not automatically foreclose competing products. Even in cases where tying leads to the foreclosure of competing products as in the model with excl usive intermediation on the consumer side, the welfare implications of tying can be subtle and ambiguous. Therefore, we need be cautious in applying the traditional theory of network effects and tipping to two-sided markets. I conclude by mentioning a couple of avenues along which the current analysis can be extended. First of all, I have assumed that there is an exogenous am ount of exclusive content available for each format. This can be just ified if the two platforms are technically differentiated and some content is more appropriate for one particular type of technology. However, exclusivity can be endogenously crea ted through the use of ex clusive contracts. A recent pap er by Doganoglu and W right (2006b) analyzes th e ability of an incumbent to use exclus ive contracts to deter entry by a more efficient entrant in a market characterized by network effects. They find that exclusive contracts can be anticompetitive if consumers can join on ly a single firm. With the possibility of multi-homing, however, they find that contracts that only require consum ers to commit to purchase from the incumbent is not anticompetitive unless they prevent consumers from also buying from the entrant. In my model, an interesting ques tion would be if exclusive contracts can be used by non-tying firms to create incentives for consumers to multi-home when the monopolist ties. In addition, many network industries are in dynamic and technology-driven high- tech fields. Despite the cen tral position innovation occupies in the performance of such industries, the model in this paper has been mainly concerned with pricing implications of 30/ 09/ 06 26 tying in network industries. One im portant extension would be to analyze how the possibility of multi-homing shapes the incentives to innovate in network industries. 30/ 09/ 06 27 References Ar mstrong, Mark, ?Competition in Two-Sided Markets,? Rand Journal of Economics, forthcoming. Caillaud, Bernard an d Bruno Julli en, ?Chicken and Egg: Competition among Intermediation Service Providers,? Rand Journal of Economics, Summer 2003, pp. 309-328. Carbajo, Jo se, David De Meza, and Daniel J. Seidman. "A Strategic Motiv ation for Commodity Bundling," Journal of Industrial Economics, March 1990, 38, pp. 283- 298. Carlton, Dennis, W . and Michael Waldman. ?The Strategic Use of Tying to Preserve and Create Market Power in Evolving Industries,? Rand Journal of Economics, Summer 2002, pp. 194-220. Chen, Yongm in, ?Equilibrium product bundling,? Journal of Business, 70, 1997, pp. 85- 103. Choi, Jay Pil, ?Preemptive R&D, Rent Dissipation, and the ?L everage Theory?,? Quartely Journal of Economics, November 1996, 1153-1181. Choi, Jay Pil, ?Tying and Innovation: A D ynamic Analysis of Tying Arrangem ents,? Economic Journal, 2004. Choi, Jay Pil and Christodoulos Stefanadis, ?Tying, Investment and the Dyna mic Leverage Theory,? Rand Journal of Economics, Spring 2001, 52-71. Doganoglu, Toker, and Julian Wright , ?Multihoming and Compatibility,? International Journal of Industrial Organization, 2006a, pp. 45-67. Doganoglu, Toker, and Julian W right, ?Exclu sive Dealing with Network Effects,? unpublished manuscript, June 2006b. Evans, David S., ?The Antitrust E conomics of Multi-Sided Platform Markets,? Yale Journal on Regulation, Summer 2003, 20, pp. 325-381. Farrell, Jos eph and Paul Klem perer, ?Coordination and Lock-In: Com petition with Switching Costs and Network Effects,? manuscript prepared for Handbook of Industrial Organization, 2001. Rochet, Jea n-Charles and Jean Tirole, ?Platform Competition in Two-Sided Markets,? Journal of the European Economic Association, June 2003a, 1, pp. 990-1029. 30/ 09/ 06 28 Rochet, Jean-Charles and Jean T irole, ?Tying in Two-Sided Markets and the Impact of the Honor All Cards Rule,? unpublished m anuscript, September 2003b. Rochet, Jean-Charles and Jean Tirole, ?Two-Sided Markets: A Prog ress Repo rt,? Rand Journal of Economics, forthcoming. Whinston, Michael D. ?Tying, Fo reclosure, and Exclusion?, American Economic Review, September 1990, 80, 837-859. Wright, Julian, ?One-Sided Logic in Two-Si ded Markets,? AEI-Brookings Joint C ent for Regulatory Studies, 2003. 30/ 09/ 06 29