Rivalry between Strategic Alliances Anming Zhang Sauder School of Business, University of British Columbia Yimin Zhang China Europe International Business School (CEIBS), and Department of Economics and Finance, City University of Hong Kong December 2003 Revised: April 2005 Abstract: Rivalry between strategic alliances is investigated in a model where each alliance member maximizes its own profit and some share of its partner’s profit. A complementary alliance confers a strategic advantage by allowing the partners to credibly commit to greater output, owing to both within-alliance complementarities and cross-alliance substitutabilities. Although rivalry between different alliances can sometimes lead to a Prisoners’ Dilemma for firms, it tends to improve economic welfare. On the other hand, an alliance that arises due purely to the threat of entry may reduce welfare. Keywords: Competing strategic alliances; Partial alliance; Supermodularity; International airline alliances JEL classification: L2; L4; L9 Correspondence: Anming Zhang, Sauder School of Business, University of British Columbia, 2053 Main Mall, Vancouver, B.C. V6T 1Z2, Canada. Phone: (604) 822-8420; fax: (604) 822-9574; e-mail: [email protected]. 1. Introduction Strategic alliances represent an important form of cooperation between two or more business entities. A strategic alliance might be viewed as a lesser form of a merger. It is not a merger per se, since alliance partners remain separate business entities and retain their decision-making autonomy. While merger activities have slowed down significantly since 2000, strategic alliances are increasingly, and widely, used by firms. They are particularly prevalent in network-oriented industries such as airline, shipping, telecommunications, multimodal transportation and logistics industries. Consider the airline industry. The four largest global alliances – namely, Star Alliance, OneWorld, Wings, and SkyTeam – accounted for 57% of the world’s revenue passenger kilometres in 2002 (Airline Business, November 2003). 1 Like alliances in logistics and other transportation industries, an airline alliance is a multi-product network, with each of its products corresponding to travel (either by people or by cargo) in a particular city-pair market. The extent of alliance will affect not only the partners but also the nature of interaction between competing alliances in various markets. The strategic alliance between, for instance, United and Lufthansa – the two major partners of Star Alliance – will thus have a major impact on the alliance decision by American and British Airways in OneWorld, which is competing with Star Alliance in the U.S., European and North Atlantic markets. In this paper, we investigate the issue of strategic alliances and alliance rivalry – where each alliance member maximizes its own profit and some share of its partner’s profit – in the context of both within-alliance and cross-alliance interactions. We focus on the strategic motives of an alliance and on how a strategic advantage is conferred. Our second objective is to investigate whether strategic alliances should be viewed as causes for anticompetitive concerns. Strategic 1 The merger of Air France and KLM was recently approved by EU regulators. As a consequence, KLM is expected to leave Wings and join SkyTeam, whose major partners are Air France and Delta. KLM’s current alliance partner, Northwest, is also expected to join SkyTeam, with Continental (Northwest’s long-time domestic alliance partner) joining as well (Brueckner and Pels, 2005). If the Wings and SkyTeam alliances are consolidated into a single mega-alliance, the three largest global alliances would each have about 20% of the world’s passenger kilometres. 1 alliances allow firms to expand their networks, take advantage of product complementarities, realize economies of scale and scope, and improve product quality and customer service. 2 Despite these potential benefits to firms and/or consumers, antitrust authorities often view strategic alliances with suspicion. In the airline example, most international routes have only a few competitors, so an alliance between any two significant competitors may considerably reduce the degree of competition on certain routes. Here an important consideration is whether government policy should encourage or restrict domestic carriers from expanding their networks through foreign alliances, while taking account of any potential anticompetitive effects. More generally, since alliances are already an important component in each of the transportation modes, 3 further consolidation may raise competition concerns that will depend on the barriers to entry and/or expansion. We examine a complementary alliance in which two firms link up their complementary products in the context of competing strategic alliances. We find that such an alliance confers a strategic advantage by allowing the partners to credibly commit to greater output levels, owing to both within-alliance complementarities and cross-alliance substitutabilities. Even if an alliance creates a negative direct effect on profit, it might be pursued, either because it is a dominant strategy in alliance rivalry or because it would deter entry. Although rivalry between 2 For example, airline alliances enable partners to better coordinate their flight schedules to minimize travelers’ wait time between connecting flights. The partners also benefit from sharing information, research and experiences in different market segments and from linking their existing networks; thereby, reducing the need for developing new hubs: a major cost of entry and particularly so for international aviation due to bilateral treaties and foreign ownership restrictions (Baker and Pratt, 1989; Pitelis and Schnell, 2002). 3 The three largest strategic alliances in the air cargo sector, namely, New Global Cargo, SkyTeam Cargo, and OneWorld, accounted for 49% of international freight tonne kilometers in 2002. In water transportation, the five shipping alliances – namely, Maersk/Sea-Land, Grand Alliance, New World Alliance, The K-Line/COSCO Group, and United Alliance – accounted for 61% of international freight tonne kilometers in 2000. The same trend has recently appeared in the U.S. domestic airline industry. Over the past few years, virtually all of the U.S. hub-spoke carriers have entered into broad codesharing partnerships, including the United/US Airways alliance that began in January 2003, and the three-way alliance between Northwest, Continental and Delta initiated in June 2003. 2 different alliances can sometimes lead to a Prisoners’ Dilemma for firms, it tends to improve economic welfare because it would, owing to the strategic effect, result in greater output levels than would be found in the absence of the rivalry. On the other hand, an alliance can arise due purely to the threat of entry; such an alliance may reduce welfare. Although there is not a formidable body of theoretical analysis on competition between alliance networks, 4 there have been some relatively recent important empirical contributions to the literature. Brueckner (2001) provides a theoretical analysis of an international airline alliance with a monopoly pair of carriers. In his model, the alliance partners provide complementary products when their respective domestic spoke-to-hub routes are used in combination by connecting passengers. In the common international hub-to-hub route, however, the partners’ outputs are substitutes. Brueckner and Whalen (2000) analyze duopoly-pair alliances on a similar network with the international hub-to-hub route suppressed. Both papers obtained interesting results under the assumptions of linear demand functions, linear marginal costs, and symmetric route structures. In particular, Brueckner points out that the cooperative pricing of trips by alliance partners would increase traffic in the connecting market since portions of a connecting trip are complements. 5 We extend this literature by analyzing generic strategic alliances and developing the welfare results in a more general setting. We also explicitly consider rivalry between alliance networks where potential partners decide not only on whether to form an alliance, but also on the extent of cooperation if they were to form an alliance. 2. The Model 4 This is in sharp contrast to the vast theoretical literature on single-firm, single-product oligopoly and on mergers. See, e.g. Farrell and Shapiro (1990a), Cabral (2003) and Pesendorfer (2005). 5 Oum, et al. (2000) obtained similar results based on restrictive demand and cost specifications and restrictive strategy sets. Empirical investigation of international airline alliances includes USDOT (1994), USGAO (1995), Oum, et al. (1996, 2000) and Brueckner and Whalen (2000). In particular, Oum, et al. and Brueckner and Whalen provide empirical support for the pro-competitive potential of complementary airline alliances. Chen and Ross (2000) explore possible entry-deterrence effects of strategic alliances involving the sharing of production capacity. 3 We consider a four-firm, four-product model that is likely the simplest structure in which strategic alliance rivalry can be addressed. The four firms, denoted 1, 2, 3 and 4, produce goods 1, 2, 3 and 4, respectively. The inverse demand function is written as: p i = p i (q1 , q 2 , q3 , q 4 ) (1) with p12 (Q) > 0, p12 (Q) > 0, p 43 (Q) > 0, p34 (Q) > 0 . (2) In (2), Q ≡ (q1 , q 2 , q3 , q 4 ) denotes the output vector and the subscripts denote partial derivatives, e.g. p 12 (Q) ≡ ∂p 1 (Q ) / ∂q 2 . The first two inequalities show demand complementarity between goods 1 and 2, whereas the other two inequalities show demand complementarity between 3 and 4. Using ci (q i ) to denote total cost, the profit of each firm can be written as: π i (Q) = qi p i (Q) − ci (qi ) , i = 1,2,3,4 . (3) Condition (2) is then equivalent to the following condition: π 21 (Q ) > 0, π 12 (Q) > 0, π 43 (Q ) > 0, π 34 (Q) > 0 (4) since, e.g. π 21 (Q) ≡ ∂π 1 (Q) / ∂q 2 = q1 p 12 by (3). Inequalities (4) say that an increase in the output of firm 2 (firm 4) will increase the profit of firm 1 (firm 3), and vice versa. Further, we assume: π 121 (Q) > 0, π 343 (Q) > 0, π 212 (Q) > 0, π 434 (Q) > 0 (5) that is, changes in the output of firm 2 (firm 4) raise the marginal profit of firm 1 (firm 3), and vice versa. Condition (5) indicates that goods 1 and 2 (3 and 4) are “strategic complements” (Bulow, et al., 1985). Observe that the fact the goods are complements is conducive to their 1 1 strategic complementarities. For example, since π 12 and p 12 > 0 by (2) or (4), it = p 12 + q1 p12 1 1 follows that π 12 > 0 if p12 ≥ 0 . Thus condition (5) will hold, given (2) or (4), if the demand functions are linear. Nonetheless, (5) can hold for more general demand specifications. We shall for simplicity assume that firms 1 and 2 always form a (potential) alliance pair, with 3 and 4 forming the other pair. The situation is depicted in Figure 1, where the former is 4 referred to as the North alliance, while the latter the South alliance. For illustration, “North” may be Star Alliance, in which United and Lufthansa provide complementary products since their respective domestic routes are used in combination by international passengers, while “South” may be OneWorld with American and British Airways providing respective domestic services. Alternatively, one may consider firms 1 and 3 producing (differentiated) peanut butter, whereas 2 and 4 producing (differentiated) jelly. Thus, firms 1 and 2 form one peanut butter/jelly alliance pair, while 3 and 4 form the other pair. Since within each firm pair the products are (strategic) complements, such alliance may be called a “complementary” alliance.6 1 2 North Alliance 3 4 South Alliance Figure 1. Basic Model Structure We examine complementary alliances in a model where each alliance member maximizes its own profit and, to a certain extent, its partner’s profit. The decision problems faced by firms 1 and 2 may be expressed as: 6 Observe, however, that although the complementarity conditions (4) and (5) are here driven by demand complementarities, they can also arise through complementarities on the cost side – e.g. there are economies of scope in the partners’ joint production process. More generally, therefore, we can refer to strategic alliances that satisfy (4) and (5) as complementary alliances. We discuss the issue further in the concluding remarks. 5 Max π 1 + απ 2 ≡ Max π 12 (Q;α ) q1 q1 (6) Max π 2 + απ 1 ≡ Max π 21 (Q;α ). q2 q2 That is, in making its optimal quantity decision a firm incorporates, apart from its profit, a fraction α of its partner’s profit, with 0 ≤ α ≤ 1 . One natural interpretation for this formulation is cross shareholding in equity alliances. A significant proportion of strategic alliances in the airline industry are equity alliances in which one airline buys a share of stock in its partners. 7 An equity alliance tends to yield greater firm values, measured in stock returns, than other types of strategic alliances. 8 Although alliances of this type do not necessarily change production possibilities at any partner firm, they do alter the incentives of the alliance partners. 9 In particular, when α = 0 , each firm owns zero share of the other firm so it maximizes its own profit. Partial alliance occurs when 0 < α < 1 , in which two firms own the same fraction of each other’s shares. Finally, a “full” alliance occurs when α = 1 , where each firm pays full attention to its partner’s profit and so the firms maximize the joint profit. Similarly, the decision problems faced by firms 3 and 4 are expressed as: Max π 3 + βπ 4 ≡ Max π 34 (Q; β ) q3 q3 Max π 4 + βπ 3 ≡ Max π 43 (Q; β ) q4 7 (7) q4 Most recent international examples include the Air France/KLM alliance, the Cathay Pacific/Air China alliance, and the proposed Qantas/Air New Zealand alliance. 8 Oum, et al. (2000) find, in the airline industry, that conditional on a broad scope of cooperation between partners, equity alliances had a greater abnormal return on the day of alliance announcements than non-equity alliances. More generally, it has been argued that equity participation can reduce transaction cost and uncertainty, and can also lesson partner perceptions of opportunistic behavior (e.g. Pisano, 1989; Parkhe, 1993). 9 Farrell and Shapiro (1990b) have made a similar point in the context of joint ownership of assets in Cournot oligopoly. In a different context, Mandy (2001) examines “vertical control” that a vertically integrated upstream monopolist exercises over its subsidiary and its implications for pricing and economic welfare. Vertical control in his model means the extent to which the parent company can align the objective of its downstream affiliates with the objective of the overall firm. This is not dissimilar from the problem that we examine in which one alliance partner attempts to maximize its profit and some fraction of its partner’s profit. 6 where the extent of alliance or partial ownership is captured by parameter β , 0 ≤ β ≤ 1 . 10 In addition to the degree of cross shareholding, parameters α and β may be interpreted as the degree of cooperation between potential partners. When α = 0 , firms 1 and 2 act independently. As α increases, the firms strengthen their cooperation, with α = 1 corresponding to full integration in which the firms act as if they were a single decision-making unit in their joint-profit maximization. The discussion holds similarly for β . Our goal is to explore the implications of conditions (4) and (5) for alliance strategies and economic welfare in an environment of competing alliances. This environment is relevant for studying strategic alliances in a number of network-based markets. For example, an international passenger making an interline trip can usually choose between several airline pairs, either allied or non-allied. For the allied pairs, the competing choices include Star Alliance, OneWorld and SkyTeam. In our model, the competing aspect is reflected by the following assumption: π 1j + π 2j < 0, π i3 + π i4 < 0, j = 3,4 ; i = 1,2 (8) that is, the outputs of the two alliances, North and South, are substitutes. Furthermore, following the standard practice in models of quantity competition (e.g. Tirole, 1988) we assume that an alliance’s marginal profit declines when the output of the rival alliance rises: π 112j < 0, π 221j < 0, j = 3,4 ; π 334i < 0, π 443i < 0, i = 1,2 (9) implying the outputs of the two alliances are “strategic substitutes” (Bulow, et al., 1985). In addition to output decisions, the firms within each of these competing pairs together need to decide on whether to form an alliance and the extent of alliance α or β . Since a strategic alliance (e.g. equity alliance) brings the participants into what is essentially a medium- to long-term contractual relationship, the alliance structure, once decided upon, cannot easily be 10 In (6), (7) we assume a Cournot game in the product-market competition. Brander and Zhang (1990, 1993), for example, find some empirical evidence that rivalry between duopoly airlines is consistent with Cournot behavior. 7 altered in a major way. 11 The alliance relationship is a strategic decision, which might reasonably be regarded as given at the time of output competition. This suggests a two-stage alliance game: In stage 1 each firm pair decides on its alliance structure, α or β ; then in stage 2 the four firms choose their output levels. 12 3. Effects of Alliance We now examine the subgame perfect equilibrium of the alliance game. In the second-stage competition, the four firms choose their quantities to maximize profits, taking the alliance structure (α , β ) as given. Assuming the second-order conditions π iiK < 0 hold, for K = 12, 21, 34, 43, then the Cournot equilibrium is characterized by the first-order conditions, π 112 (Q; α ) = π 11 + απ 12 = 0 (10) π 221 (Q; α ) = π 22 + απ 21 = 0 (11) π 334 (Q; β ) = π 33 + βπ 34 = 0 (12) π 443 (Q; β ) = π 44 + βπ 43 = 0 . (13) Regularity conditions are imposed so that the equilibrium exists and is stable. In the absence of the South alliance, this Cournot game of firms 1 and 2 is supermodular: Roughly speaking, a game is called supermodular if the increment in a player’s payoff due to an increase in its strategy variable is increasing in the strategy variable of any of the other players. Thus, supermodularity may be viewed as the mathematical characterization of the notion of strategic complementarity (e.g. Topkis, 1998; Vives, 2000). With competing alliances, however, our four-firm Cournot game is no longer supermodular, because the outputs of the North and 11 Parkhe (1991) defines a strategic alliance as a “relatively enduring interfirm cooperative arrangement, involving flows and linkages that utilize resources and/or governance structures from autonomous organizations, for the joint accomplishment of individual goals.” A strategic alliance is thus different from an ordinary alliance in that the partners here make a more serious commitment to cooperation. As a result, α ’s or β ’s are chosen, since equity participation is presumably a unilateral decision made by each firm. Extending the analysis to individual α ’s and β ’s would add 12 We do not address here the issue of how firm-specific mathematical complexity; nevertheless, our basic results would continue to hold in that case. 8 South alliances are strategic substitutes. Observe, nevertheless, that the game is supermodular in 34 3 12 12 1 = π 34 + απ 344 > 0 by (5), and π 13 <0, (q1 , q 2 ,−q3 ,−q 4 ) , because π 12 = π 12 + απ 122 > 0 and π 34 π 1412 < 0 , π 3134 < 0 and π 3234 < 0 by (9). Similar inequalities hold for π 21 and π 43 , so the game is supermodular in (−q1 ,− q 2 , q3 , q 4 ) . Thus the strategic complementarity within an alliance, together with the strategic substitutability across the alliances, make the second-stage game supermodular in (q1 , q 2 ,−q3 ,−q 4 ) and (−q1 ,− q 2 , q3 , q 4 ) . Given this result, the effects of alliance variables α and β on the equilibrium quantities, denoted q i (α , β ) , can be derived in Proposition 1. These comparative static effects are central to our subsequent analysis. Proposition 1. Under complementary alliances, qαi > 0, q βj > 0, qαj < 0, q βi < 0, i = 1,2 ; j = 3,4 (14) that is, the strengthening of one of the alliances increases the output of both member firms’ products and decreases that of their rival firms. Proof: Since the game is supermodular in (q1 , q 2 ,−q3 ,−q 4 ) , Topkis (1998, Theorem 4.2.2) indicates that qαi (α , β ) > 0 for i = 1,2 and qαj (α , β ) < 0 for j = 3,4 if the game has increasing differences in α . The latter holds because, by (4), π 112α = π 12 > 0, π 221α = π 21 > 0, π 334α = 0, π 443α = 0 i.e., π iKα ≥ 0 , for K = 12, 21, 34, 43 and i = 1, 2, 3, 4. The other part of the Proposition is shown in a similar way. Q.E.D. Alliance structures therefore influence subsequent output competition, which in turn affects firms’ overall profitability. Taking the second-stage equilibrium output into account, firm i’s profit in the first stage, denoted φ i , is given by: φ i (α , β ) ≡ π i (Q(α , β )) , i = 1,2 ; φ j (α , β ) ≡ π j (Q(α , β )) , j = 3,4 . (15) The alliance equilibrium arises when each firm pair chooses its profit-maximizing alliance 9 structure, taking the alliance structure of the other pair as given at the equilibrium value. Proposition 2. At the alliance equilibrium both firm pairs form a full alliance. Proof: From (15), it follows that π α12 + π α21 = φα1 + φα2 + (π 1 + π 2 ) + α (π α1 + π α1 ) . Substituting π α1 + π α2 = φα1 + φα2 and recollecting the terms yields: π α12 + π α21 = (1 + α )(φα1 + φα2 ) + (π 1 + π 2 ) . Solving for φα1 + φα2 and using the expressions for π α12 and π α21 , we then obtain: φα1 + φα2 = (1 − α )(π 12 qα1 + π 21 qα2 ) + [(π 31 + π 32 )qα3 + (π 41 + π 42 )qα4 ] (16) where π 12 , π 21 are positive by (4) and π 1j + π 2j < 0 by (8), j = 3,4 . Thus, by Proposition 1, ∂ (φ 1 + φ 2 ) / ∂α > 0 . That ∂ (φ 3 + φ 4 ) / ∂β > 0 can be similarly shown. Q.E.D. Proposition 2 shows that a complementary alliance can be both an offensive strategy and a defensive strategy in an alliance rivalry. It improves the alliance’s own profit, given the alliance structure of the rival firms; it defends the allied firms when the rival firms intensify their cooperation. Under the specified conditions a full alliance is, in effect, each firm pair’s dominant strategy. Note furthermore that the effect of a change in α (or, β ) on profit can be split into two parts: i) a direct effect of the shift in the alliance’s profit function, and ii) an indirect effect of the shift in the marginal profits, which in turn changes the equilibrium. The direct effect may be illustrated by considering a monopoly pair of firms (say, 1 and 2) who are unconcerned with entry. In this case, the bracketed term on the right-hand side of (16) vanishes. The remaining term, (1 − α )(π 12 qα1 + π 21 qα2 ) , then represents the direct effect of alliance on the pair’s profit. This direct effect is positive so long as α < 1 . The positive sign arises because of the monotonicity properties of the supermodular game between firms 1 and 2. As indicated above, supermodularity captures the notion of complementarity: in our case, complementarity between q1 and q 2 means that the marginal returns to one variable increases in the level of the other variable. Therefore, the term captures the internalization of the familiar “complementarity externality:” the alliance allows partners to internalize the positive effect of an increase in one 10 firm’s output on its partner’s profit. This internalization of demand externalities is similar to the mitigation of the “double marginalization” problem in vertical integration (e.g. Tirole, 1988). Brueckner (2001, 2003) and Brueckner and Whalen (2000) have a clear demonstration of the problem for an international airline alliance, where double marginalization occurs because each carrier of an international interline itinerary tries to maximize the profit from its own segment independently from the other carrier. Consequently, independent carriers typically charge segment fares higher than a single decision maker controlling prices over the joint itinerary would. In other words, cooperative pricing of an alliance enables carriers to internalize part of the double-marginalization externality associated with joint pricing on interline tickets. In our context, both firms use quantities as their strategy variable and we show the result in a general setting using recent developments in supermodular games. The monopoly-pair case serves as a useful base for comparison with the duopoly-pair case. Whilst the term (1 − α )(π 12 qα1 + π 21 qα2 ) captures a “direct” advantage of complementary alliances, the “indirect” effect is unique to competing alliances. This effect is represented by the term (π 31 + π 32 )qα3 + (π 41 + π 42 )qα4 in (16). Since this effect works by indirectly influencing the behavior of the rival firms (which in turn improves own profits), it may be referred to as the “strategic effect” of alliance. Observe that this indirect, strategic effect augments the direct effect – recall that conditions (4) and (5) tend to reinforce each other – that is, parametric shifts dα will shift both the total and marginal profits in the same direction. Our analysis therefore suggests that the rivalry between multiple alliances may, owing to the strategic effect, result in a higher degree of cross shareholding or alliance than would be had in the absence of rivalry, such as in the monopoly-pair case. 13 13 More specifically, let (α , β ) denote the equilibrium duopoly-pair alliance structure and (α , β ) the d d m m optimal monopoly-pair alliance structure. Then, other things being equal, the strategic effect tends to produce αd ≥αm, β d ≥ β m when the alliance yields large negative synergies for the participants. The negative synergies of alliance may arise due to partners’ opportunistic behavior or to their incompatible structures in financial 11 The above discussion is useful in explaining Proposition 1. A basic insight of Proposition 1 is that to obtain the monotone comparative static results on the Cournot equilibrium, the complementarity conditions (supermodularity and increasing differences) are the key ingredient. For illustration, consider the creation of an alliance between one peanut butter firm and one jelly firm. Since an increase in the output of peanut butter will increase the demand for jelly, the peanut butter firm will produce more after being allied with the jelly firm. In other words, this alliance will lead these two firms to internalize demand externalities. With the rising marginal profits for both firms, their reaction functions will shift so as to produce more output, holding the output choice of their competitors constant. Since the outputs of competing alliances are strategic substitutes, the other two firms will react by reducing their output. 14 Although Proposition 2 has established that full alliance will be the dominant strategy for two firms offering complements, it is not always true, shown in Proposition 3 below, that alliance partners are better off if they both form full alliance than if they both choose partial alliance or remain independent. Proposition 3. Rivalry between different alliances can result in a Prisoners’ Dilemma for firms. The proof of Proposition 3 is given in the Appendix. To explain the intuition behind the result, it is useful to first examine the effect of an alliance on the rival’s profit. From (15) and (7), φα3 + φα4 = [π 134 qα1 + π 234 qα2 + π 334 qα3 + π 434 qα4 − β (π 14 qα1 + π 24 qα2 + π 34 qα3 + π 44 qα4 )] [ ] + π 143 qα1 + π 243 qα2 + π 343 qα3 + π 443 qα4 − β (π 13 qα1 + π 23 qα2 + π 33 qα3 + π 43 qα4 ) . Since π 334 = 0 by (12) and π 443 = 0 by (13), rearranging then yields: status, asset composition, labor contracts and management style (e.g. Parkhe, 1993). Further, these negative effects can be fixed effects in the sense that they are independent of the level of output. Since the fixed effects are additive constraints and do not affect most of the results, they are assumed away from the analysis presented in the paper. 14 This is classic behavior in quantity competition: Here, forming an alliance allows the allied firms to be more aggressive. We add to the analysis by looking at the issue when there are four firms interacting with each other, rather than the usual two firms interacting with each other. 12 φα3 + φα4 = [(π 13 + π 14 )qα1 + (π 23 + π 24 )qα2 ] + [(1 − β )π 34 qα3 + (1 − β )π 43 qα4 ] . (17) Noticing π 13 + π 14 < 0 , π 23 + π 24 < 0 by (8) and π 43 > 0 , π 34 > 0 by (4), so by Proposition 1 the four terms in (17) are all negative. That φ β1 + φ β2 < 0 can be similarly shown, leading to: Proposition 4. The strengthening of an alliance by partners reduces the profit of their rival firms. Now consider the possibility of an alliance between a peanut butter firm and a jelly firm. This alliance will make the two firms increase their output; as a consequence, the competing peanut butter and jelly firms will reduce theirs (Proposition 1). The end result of increased production of the allied firms and reduced production of the rival firms will, according to (17), reduce the profit of the rival firms (Proposition 4). This negative pecuniary externality on the rival firms is not taken into consideration by the alliance firms when they decide on the degree of alliance between them. When both pairs choose full alliance, the strategic gains tend to offset each other while industry output tends to rise beyond the levels produced under partial alliances or independence. If the resulting prices are sufficiently low, then full alliances reduce overall industry profits relative to partial alliance or independence, giving rise to a Prisoners’ Dilemma. Next, we investigate the effect of alliances on price and total surplus. For this purpose, we assume that products 1 and 3 (similarly, 2 and 4) are perfect substitutes; thus there are two markets: a (say) peanut butter market consisting of output q1 + q3 with price p1 , and a jelly market consisting output q 2 + q 4 with price p 2 . The vector of changes in total equilibrium output in each market can be denoted as: ∂q~ ∂ ⎡ q1 + q 3 ⎤ ∂ ⎡ q1 ⎤ ∂ ⎡q 3 ⎤ ≡ ⎢ ⎥= ⎢ ⎥+ ⎢ ⎥. ∂α ∂α ⎣q 2 + q 4 ⎦ ∂α ⎣q 2 ⎦ ∂α ⎣q 4 ⎦ (18) We have the following result (the proof is given in the Appendix): Proposition 5. Assuming homogeneous products 1 and 3 (2 and 4), a complementary alliance 13 increases total output, and hence reduces price, in at least one market. In the case of symmetric firms where four firms face symmetric demands both within a market and across the markets and have the same costs, the alliance increases total output and reduces price in both markets. Thus consumers in at least one market are better off (in terms of lower price) following a complementary alliance, and are better off in both markets if firms are reasonably symmetric. To examine the effect on total surplus, we consider a partial equilibrium framework in which consumer demand is derived from a utility function that can be approximated by the form u (Q) + z , where z is expenditure on a competitively supplied numeraire good. The welfare 4 function is then given by W = u (Q ) − ∑ ci ( qi ) . Differentiating W with respect to α and using i =1 ∂u / ∂qi = p i , we obtain: 4 Wα = ∑ ( p i − ci' )qαi . (19) i =1 The signs of the mark-up terms, p i − ci' , are positive by the first-order conditions, while the qαi terms are either positive or negative depending on i. As a result, the sign of Wα is in general ambiguous, depending on the magnitude of two positive ( p i − ci' )qαi terms ( i = 1,2 ) relative to that of two negative ( p j − c 'j )qαj terms ( j = 3,4 ). However, Wα > 0 for symmetric firms: Proposition 6. Assuming homogeneous products and symmetric firms, a complementary alliance improves social welfare. Although the rivalry between complementary alliances will likely improve economic welfare, a complementary alliance might be used to deter entry; as such, it can reduce welfare. Specifically, suppose now that a pair of firms, say, 1 and 2, have an exogenously given opportunity to choose its alliance structure prior to the entry (and alliance) decision of firms 3 and 4, and that there exists a sunk entry cost. The entry cost can include market-specific irrecoverable research, advertising and promotional expenditures, as well as investments in setting up the network and other initial operations. Then Proposition 4 indicates that there exists 14 a range of entry costs, such that rival entry will be pre-empted if and only if the incumbents form an alliance. An incumbent pair might then ally if it is better off with alliance and no entry than with no alliance and entry. Indeed, numerical examples show that the threat of entry alone can result in a complementary alliance, in the sense that forming an alliance is not profitable (owing, for example, to negative synergies) in the absence of entry. 15 4. Concluding Remarks We have examined alliance incentives for competing strategic alliances. A complementary alliance confers a strategic advantage by allowing alliance partners to credibly commit to greater output, owing to both within-alliance complementarities and cross-alliance substitutabilities. The strategic effect of a complementary alliance tends to augment its positive direct effects on profit that arises from the mitigation of the double-marginalization problem. Even if an alliance creates other negative synergies, it might be pursued, either because it is a dominant strategy or because it would deter entry. Although sometimes it may be profit dissipating, rivalry between complementary alliances tends to enhance economic welfare, because the strategic effect results in a higher degree of alliance, and hence greater output levels, than would be the case in the absence of such rivalry. 16 We also found that a complementary alliance can arise purely for the purpose of entry deterrence; such an alliance might reduce total surplus since it creates negative synergies for incumbents and it reduces market competition. In these cases, the socially optimal level of alliance would balance the social costs with the social benefits of internalizing demand externalities. Partial alliance (or limited cross shareholding/control) may arise because of 15 This result has a strong tie to Dixit’s (1980) insight on the role of capital investment in entry deterrence. Essentially, incumbent firms that choose to fully ally with each other are trading off higher fixed costs for higher total and marginal profit effects arising from complementarity externalities. The higher marginal profit credibly commits the alliance to produce more, thereby reducing the profitability of entry. 16 It is worth noting that these results are quite close to those of Brander and Spencer (1983) and Spencer and Brander (1983). In particular, firms seek commitment devices to increase output in Cournot rivalry, but this, while raising welfare, can make firms worse off in equilibrium. 15 external restrictions imposed by regulators on the interfirm relationship. We have concentrated our analysis on complementary alliances which, apart from the aspect of entry deterrence, generally lead to improved welfare. In contrast, horizontal alliances, which are formed by firms who produce substitutes, tend to be anti-competitive. Indeed, we can show that a horizontal alliance may reduce competition not only in the market where prior competition between the partners takes place, but also in other markets of the alliance network – and hence may be restricted by regulatory agencies. Furthermore, a real-world alliance is likely to have both complementary and horizontal elements; consequently, such a “hybrid” alliance may have both pro- and anti-competitive effects. For example, an international airline alliance is likely to reduce competition in the hub-to-hub markets in which both partners operate and usually dominate, while increasing competition in the connecting markets to which the partners provide cheaper airfares. The welfare implications are generally less definitive for hybrid alliances than for the other two alliance types. Existing work by Brueckner and others focuses on specific structures, and a more complete treatment of hybrid alliances in a general setting of competing alliances remains a subject of future research. In this article, the results of complementary alliances are driven by the demand complementarities underlying (4) and (5). However, (4) and (5) can also arise from cost complementarities in producing multiple outputs. Our results will continue to hold in this case. Cost interaction brings in another important aspect of strategic alliances: alliance partners usually coordinate the use of inputs. It would be interesting to examine the impact of strategic alliances on partners’ productivity and their competitiveness in product-market competition. The framework of this paper may be useful in addressing strategic network rivalry in other situations. One potential area of application is in supply chain management and logistics. Supply chain management typically lies between full integration (when one firm manages the entire material and information flow) and independent operation of each channel member. Therefore, coordination to reduce costs and risks between the various players in the chain is key to its 16 effective management. Often the result is a mutually beneficial logistics partnership or strategic alliance. It would be interesting to analyze the rivalry of one supply chain against another supply chain, in a way similar to our analysis of the rivalry between different strategic alliances. Acknowledgement: We are very grateful to two anonymous referees and especially the editor (Simon Anderson) whose comments have led to a significant improvement. Earlier versions of this paper have been presented at University of British Columbia, University of Calgary, American Economic Association’s Transportation and Public Utility Group Conference, Kobe University, and the 7th Air Transport Research Society Conference. We thank the seminar participants and Xuan Zhao for helpful comments. Financial support from the Social Science and Humanities Research Council of Canada (SSHRC) is gratefully acknowledged. 17 References Baker, S.H. and Pratt, J.B. (1989), “Experience as a barrier to contestability in airline markets,” Review of Economics and Statistics, 71, 352-256. Brueckner, J.K. (2001), “The economics of international codesharing: An analysis of airline alliances,” International Journal of Industrial Organization, 19, 1475-1498. Brueckner, J.K. (2003), “International airfares in the age of alliances,” The Review of Economics and Statistics, 85, 105-118. Brueckner, J.K. and Whalen, W.T. (2000), “The price effects of international airline alliances,” Journal of Law and Economics, XLIII, 503-545. Brueckner, J.K. and Pels, E. (2005), “European airline mergers, alliance consolidation, and consumer welfare,” Journal of Air Transport Management, 11, 27-41. Brander, J.A. and Spencer, B.J. (1983), “Strategic commitment with R&D: The symmetric case,” Bell Journal of Economics, 14, 225-235. Brander, J.A. and Zhang, A. (1990), “Market conduct in the airline industry,” Rand Journal of Economics, 21, 567-583. Brander, J.A. and Zhang, A. (1993), “Dynamic behaviour in the airline industry,” International Journal of Industrial Organization, 11, 407-435. Bulow, J.I., Geanakoplos, J.D. and Klemperer, P.D. (1985), “Multimarket oligopoly: Strategic substitutes and complements,” Journal of Political Economy, 93, 488-511. Cabral, L.M.B. (2003), “Horizontal mergers with free-entry: Why cost efficiencies may be a weak defense and asset sales a poor remedy,” International Journal of Industrial Organization, 21, 607-623. Chen, Z. and Ross, T.W. (2000), “Strategic alliances, shared facilities, and entry deterrence,” Rand Journal of Economics, 31, 326-344. Dixit, A. (1980), “The role of investment in entry deterrence,” Economic Journal, 90, 95-106. Farrell, J. and Shapiro, C. (1990a), “Horizontal mergers: An equilibrium analysis,” American Economic Review, 80, 107-126. Farrell, J. and Shapiro, C. (1990b), “Asset ownership and market structure in oligopoly,” Rand Journal of Economics, 21, 275-292. Mandy, D.M. (2001), “Price and Vertical Control Policies for a Vertically Integrated Upstream Monopolist When Sabotage is Costly,” Working Paper No. 01-13, Department of Economics, University of Missouri. Oum, T.H., Park, J.H. and Zhang, A. (1996), “The effects of airline codesharing agreements on firm conduct and international air fares,” Journal of Transport Economics and Policy, 30, 187-202. Oum, T.H., Park, J.H. and Zhang, A. (2000), Globalization and strategic alliances: The case of the airline industry, (Pergamon, London). Parkhe, A. (1991), “Interfirm diversity, organizational learning and longevity in global strategic 18 alliances,” Journal of International Business Studies, 22, 579-601. Parkhe, A. (1993), “Strategic alliance structuring: A game theoretic and transaction cost examination of interfirm cooperation,” Academy of Management Journal, 36, 794-829. Pesendorfer, M. (2005), “Mergers under entry,” Rand Journal of Economics, forthcoming. Pisano, G.P. (1989), “Using equity participation to support exchange: Evidence from the biotechnology industry,” Journal of Law, Economics, and Organization, 5, 109-126. Pitelis, C.N. and Schnell, M.C.A. (2002), “Barriers to mobility in Europe’s civil aviation markets: Theory and new evidence,” Review of Industrial Organization, 20, 127-150. Spencer, B.J. and Brander, J.A. (1983), “International R&D rivalry and industrial strategy,” Review of Economic Studies, 50, 707-722. Tirole, J. (1988), The theory of industrial organization, (MIT Press, Cambridge). Topkis, D. (1998), Supermodularity and complementarity, (Princeton University Press, Princeton, New Jersey). USDOT (1994), A study of international airline codesharing, Gellman Research Associates, Inc., commissioned by the U.S. Department of Transportation, December 1994. USGAO (1995), Airline alliances produce benefits, but effect on competition is uncertain, The U.S. General Accounting Office, GAO/RCED-95-99, April 1995. Vives, X. (2000), Oligopoly pricing: Old ideas and new tools, (MIT Press, Cambridge, Massachusetts). Zhang, A. and Zhang, Y. (1996), “Stability of a Nash equilibrium: The multiproduct case,” Journal of Mathematical Economics, 26, 441-462. 19 Appendix Proof of Proposition 3: Consider the auxiliary function, φ 1 (α , α ) + φ 2 (α , α ) , which is, by (15), φ 1 + φ 2 = π 12 (Q(α , α );α ) − απ 2 (Q(α , α )) + π 21 (Q(α , α );α ) − απ 1 (Q(α , α )) . (A1) Totally differentiating (A1) with respect to α and applying (10) and (11) yields: d (φ 1 + φ 2 ) dq1 dq 2 dq 3 dq 4 = (1 − α )π 12 + (1 − α )π 21 + (π 31 + π 32 ) + (π 41 + π 42 ) dα dα dα dα dα (A2) where dq i (α , α ) / dα represents total differentiation, i = 1,2,3,4 . The signs for the terms in front of dq i / dα terms are fixed (i.e. either positive or negative). It remains to see whether we can sign dq i / dα . Totally differentiating the first-order conditions (10)-(13) with respect to α after replacing β with α , solving for dQ / dα and substituting, we obtain: [ ] dQ 12 21 34 43 d ∏ = −( I − R ) −1 diag π 11 , π 22 , π 33 , π 44 dα dα [ (A3) ] where dQ / dα is a column vector, and d ∏ / dα = π 12 , π 21 , π 34 , π 43 is a column vector and its four elements are, by (4), all positive. In (A3), I is the identity matrix and ⎡ 0 ⎢ 21 R R ≡ ⎢ 21 ⎢ R3134 ⎢ 43 ⎢⎣ R41 R1212 R1312 0 R3234 R2321 0 R4243 R4343 R1412 ⎤ ⎥ R2421 ⎥ R3434 ⎥ ⎥ 0 ⎥⎦ with RijK ≡ −(π iiK ) −1 π ijK , for i ≠ j . Note, e.g. Rij12 is the derivative of firm 1’s reaction function. By (5), (9) and the second-order condition, matrix R has the following sign pattern: ⎡0 ⎢+ R=⎢ ⎢− ⎢ ⎣− + − 0 − − 0 − + −⎤ − ⎥⎥ . +⎥ ⎥ 0⎦ (A4) We now determine the sign of [I – R]-1. The stability of the Cournot-Nash equilibrium implies that the magnitude of the eigenvalues of R must be less than unity (Zhang and Zhang, 1996). Hence, by Neumann Lemma, [I – R]-1 exists and [I − R ] = I + R + R 2 + L + R n + L . Hence, −1 by (A4), [I – R]-1 must have the following signs: 20 ⎡+ ⎢+ −1 [ I − R] = ⎢ ⎢− ⎢ ⎣− + + − − − − + + −⎤ − ⎥⎥ . +⎥ ⎥ +⎦ (A5) Given the sign pattern of [I – R]-1, each element of dQ / dα will be the difference of two positive numbers and, hence, will in general have an ambiguous sign. It then follows from (A2) that the sign of d (φ 1 + φ 2 ) / dα is ambiguous, and in the cases where (such numerical examples exist): φ 1 (1,1) + φ 2 (1,1) < φ 1 (α , α ) + φ 2 (α , α ) , φ 3 (1,1) + φ 4 (1,1) < φ 3 (α , α ) + φ 4 (α , α ) for 0 ≤ α < 1 , a Prisoners’ Dilemma will rise. Q.E.D. Proof of Proposition 5: Differentiating (10)-(13) with respect to α yields: Φ ∂Q (α , β ) ∂ ∏ + =0 ∂α ∂α (A6) where 12 ⎡π 11 ⎡ q1 (α , β ) ⎤ π 1212 π 1312 π 1412 ⎤ ⎢ 21 ⎥ ⎢ 2 ⎥ π π 2221 π 2321 π 2421 ⎥ ⎢q (α , β )⎥ , , Q α β ( , ) ≡ Φ ≡ ⎢ 21 34 ⎢π 31 ⎢ q 3 (α , β ) ⎥ π 3234 π 3334 π 3434 ⎥ ⎢ 43 ⎥ ⎢ 4 ⎥ 43 43 43 ⎣⎢q (α , β )⎦⎥ ⎣⎢π 41 π 42 π 43 π 44 ⎥⎦ ⎡π 12 ⎤ ⎢ ⎥ ∂ ∏ ⎢π 21 ⎥ . ≡ ∂α ⎢ 0 ⎥ ⎢ ⎥ ⎣⎢ 0 ⎦⎥ Partition matrix Φ into: ⎡Φ Φ = ⎢ 11 ⎣Φ 21 Φ 12 ⎤ Φ 22 ⎥⎦ where Φ ij is a 2-by-2 matrix. Solving the second equation in (A6) and substituting then yields: ∂ ∂α 1 ⎡q 3 ⎤ 34 ∂ ⎡ q ⎤ ⎢ 4 ⎥ = R12 ⎢ ⎥, ∂α ⎣q 2 ⎦ ⎣q ⎦ where R1234 ≡ −Φ −221 Φ 21 . By the stability condition the norm of matrix R1234 must be less than unity (Zhang and Zhang, 1996). If follows that either q α3 < q α1 or q α4 < q α2 , and by Proposition 1, ∂q~ / ∂α must have at least one positive element. In the case of symmetric firms, dq 1 = dq 2 and dq 3 = dq 4 , thereby implying ∂q~ / ∂α > 0 and hence prices fall as α increases. 21 Q.E.D.
© Copyright 2026 Paperzz