TSE‐691 September2016 “CompatibilityChoicesunderSwitchingCosts” Doh‐ShinJeon,DomenicoMenicucciandNikroozNasr Compatibility Choices under Switching Costs∗ Doh-Shin Jeon† Domenico Menicucci‡ Nikrooz Nasr§ August 31, 2016 Abstract We study firms’ compatibility choices in the presence of consumers’ switching costs. We analyze both a model of once-and-for-all compatibility choices and that of dynamic choices. Contrary to what happens in a static setting in which firms embrace compatibility to soften the current competition (Matutes and Régibeau, 1988), when consumer lock-in arises due to a significant switching cost, firms make their products incompatible in order to soften future competition, regardless of the model we consider. This reduces consumer surplus and social welfare. Key words: Compatibility, Incompatibility, Switching Cost, Lock-in JEL Codes: D43, L13, L41. ∗ We thank Paul Hünermund, Joshua Gans, Lukasz Grzybowski, Jorge Padilla, Patrick Rey, Yaron Yehezkel, and the audience who attended our presentations at the Net Institute Conference 2015, Ninth IDEI-TSE-IAST conference on the Economics of Intellectual Property, Software and the Internet, Workshop on the Economics of Network Industries 2016, 14th ZEW Conference on The Economics of Information and Communication Technologies 2016, and 31st Annual Congress of The European Economic Association. We gratefully acknowledge financial support from the NET Institute. Jeon acknowledges financial support from the Jean-Jacques Laffont Digital Chair. † Toulouse School of Economics, University of Toulouse Capitole (IDEI), E-mail: [email protected]. ‡ Dipartimento di Scienze per l’Economia e l’impresa, Università degli Studi di Firenze. Via delle Pandette 9, I-50127 Firenze (FI), Italy. E-mail: [email protected]. § Toulouse School of Economics. Manufacture de Tabacs, 21 allees de Brienne - 31000 Toulouse, France. E-mail: [email protected] 1 Introduction Will the future of the Internet be dominated by incompatible platforms? Back in the 90s’ when the Internet was at its dawn, openness and compatibility seemed to be the future. For instance, during the times, Microsoft was the dominant player in the personal computer market but decided to bring two of its most successful software, Internet Explorer and Microsoft Office, to Macs. However, after the turn of the 21st century, we seem to enter a new era in which platforms try to lock-in customers by making it hard to move data across platforms or by providing some benefits exclusively to those who use all from the same ecosystem. According to Larry Page, a co-founder of Google and the CEO of its parent company, Alphabet Inc.,: "The Internet was made in universities and it was designed to interoperate. And as we’ve commercialized it, we’ve added more of an island-like approach to it, which I think is a somewhat a shame for users."1 If a consumer purchased applications from iTunes (Google Play), she is more likely to choose an iPhone (an Android phone) instead of an Android phone (an iPhone). The same logic applies to services like: iMessage and iCloud versus Hangouts and Google account, One Drive versus Google Drive, iCloud Keychain versus Chrome password syncing. In cloud computing, from the beginning, a major concern was vendor lock-in where due to incompatible technologies, customers would incur a huge cost if they wish to transfer their data from one vendor to another2 . This kind of phenomenon is known as "walled garden". As cloud computing technology becomes ubiquitous, the risk of lock-in in walled gardens of data generalizes. Furthermore, Apple and Google are expanding their businesses outside of the phone industry by leveraging their established smartphone platforms: smart car, smart TV, Google glasses, Apple watch, Apple pay etc. These new hardware and software bring the ultimate lock-in experience. For instance, Android Wear only works with Android phones and tablets. Apple Watch supports iPhones and iPads exclusively. We enter an era where working across multiple ecosystems deprive consumers of real benefit. This is why Drew Houston, CEO of Dropbox, says that consumers are "trapped" in each platform.3 1 http://fortune.com/2012/12/11/fortune-exclusive-larry-page-on-google/ http://fortune.com/2015/10/08/aws-lock-in-worry/ 3 http://vimeo.com/70089044 2 1 In this paper, we attempt to offer a novel insight to understand platforms’ compatibility choices from a dynamic perspective, which is relevant because hardware must be upgraded and subscription to content/application often must be renewed. As the main result, we show that when consumer lock-in arises due to a significant switching cost, firms make their products incompatible in order to soften future competition, which hurts consumers and reduces welfare. More generally, we find a strong conflict between the compatibility regime chosen by the firms and the one maximizing consumer surplus. Our result is opposite to what happens in a static model in which symmetric firms make their products compatible to soften current competition (Matutes and Régibeau, 1988). We extend the model of Matutes and Régibeau (1988) to two periods. They study compatibility choices made by two symmetric firms (A and B) who compete to sell a system of complementary products (x and y). Therefore, under compatibility, four systems are available ((A, A), (A, B), (B, A), (B, B)) while under incompatibility, only two pure systems, (A, A) and (B, B), are available. They study a two-stage game in which the first stage of non-cooperative choice between compatibility and incompatibility is followed by the second stage of price competition. We consider two different dynamic extensions of their setup. In our model 1 of once-and-for-all compatibility choices, firms make their choices only in period one and the compatibility regime determined in period one is maintained in period two. In model 2 of dynamic compatibility choices, firms make their choices every period. Both models are identical to that of Matutes and Régibeau if there were no second period. In the second period of our models, each platform competes to poach consumers by offering prices dependent on past purchase behavior (Chen, 1997 and Fudenberg and Tirole, 2000): there is price discrimination depending on which product or system a consumer bought in the first-period.4 Consider first model 1 of once-and-for-all compatibility choices. When products are compatible, there are four submarkets in period two: the market ij refers to the market composed of the consumers who bought product j = x, y from firm i = A, B in period one. We assume that each consumer discovers the value that she obtains from a product, which is a random draw from a uniform distribution, only after consuming it. We assume for simplicity that all consumers incur the same switching cost s > 0. For instance, if a consumer wants to switch from x of A to x of B, she should incur s and therefore firm A is 4 For instance, apple launched Android device trade-in program to encourage iPhone upgrade: http://9to5mac.com/2015/03/16/applewillofferandroidswitchersgiftcardstotradeinrivalsmartphonesforiphones/ 2 dominant and firm B is dominated in market Ax . When products are incompatible, there are two submarkets (A, A) and (B, B) in period two where market (A, A) is composed of the consumers who bought system (A, A) in period one. If a consumer wants to switch from (A, A) to (B, B), she should incur a switching cost of 2s. When we study how incompatibility affects the firms’ second-period profits from the consumers who bought say (A, A) in period one, we find three thresholds of switching costs (s1 , s2 , s3 ) with s1 < s2 < s3 such that the dominant firm’s profit is higher under incompatibility than under compatibility for s > s1 , the industry profit is higher under incompatibility for s > s2 and the dominated firm’s profit is higher under incompatibility for s > s3 . In other words, when the switching cost is high enough, incompatibility softens second-period competition relative to compatibility. This result can be understood from Hahn and Kim (2012) and Hurkens, Jeon and Menicucci (2016), who extend Matutes and Régibeau (1988) to asymmetric platforms (one is dominant and the other is dominated).5 They consider a static model and find that when the level of dominance is large enough, both the dominant firm and the dominated firm prefer incompatibility since incompatibility softens competition. This mechanism is in place in our model as the level of dominance increases with s. In Section 3.3, following Hurkens, Jeon and Menicucci (2016), we decompose the effect of incompatibility into demand size effect and demand elasticity effect and explain how incompatibility softens second-period competition when switching cost is high. When we study first-period competition for a given regime of compatibility, each firm charges a price smaller than the one in Matutes and Régibeau (1988) by the difference between the dominant firm’s second-period profit and the dominated firm’s one. In other words, each firm dissipates the rent from a locked-in consumer by competing more aggressively in period one. This implies that each firm’s total profit is the sum of the profit in Matutes and Régibeau (1988) and the dominated firm’s (second-period) profit. Therefore, for s > s3 , if the relative weight of the second-period payoff is large enough, the firms choose incompatibility in period one in order to soften future competition even if it 5 Hahn and Kim (2012) consider uniform distribution of consumers on a Hotelling line as Matues and Régibeau do. Hurkens, Jeon and Menicucci (2016) consider the family of log-concave distributions which include the uniform distribution and offer a general intuition based on the property that the distribution of average location is more peaked than the original distribution for this family. In addition, they decompose the effect of pure bundling into demand size effect and the demand elasticity effect and use the insight to build a theory of credible leverage of market dominance. 3 intensifies current competition; for all other parameter values, they choose compatibility. As the firms make compatibility choices mainly to soften competition, we find a strong conflict between the compatibility regime chosen by the firms and the one maximizing consumer surplus, regardless of the model we consider. In particular, in model 1 of onceand-for-all compatibility choices, whenever the firms choose incompatibility, it generates a lower consumer surplus than compatibility. We also find a similar conflict between firm’s choices and welfare. In particular, whenever firms choose incompatibility, it generates a smaller welfare than compatibility, regardless of the model we consider. Welfare is lower under incompatibility as consumers lose the ability to mix and match the products. The analysis of model 2 of dynamic compatibility choices qualitatively confirms the findings from model 1. The key difference between the two models is that in model 2, firms make compatibility choices in period two as well, which implies that when considering a deviation in period one, a firm needs to think about how the deviation will affect the compatibility regime in period two. In a few cases, this generates more powerful deviations than the ones under model 1 such that no pure strategy equilibrium exists in model 2. The analysis of model 2 uncovers some novel mechanisms regarding dynamics of compatibility as follows. First, given first-period market shares, the dynamics of compatibility is asymmetric in that firms are more likely to embrace compatibility tomorrow if products are compatible today but today’s incompatibility does not necessarily imply that the firms are more likely to embrace incompatibility tomorrow. For instance, if the first-period market shares are symmetric, first-period compatibility leads to second-period compatibility no matter what the level of switching cost. In contrast, first-period incompatibility leads to second-period incompatibility if and only if the level of switching cost is high enough. To understand the results, notice that if first-period compatibility is followed by secondperiod incompatibility, there are four submarkets (A, A), (A, B), (B, A), (B, B) in period two. But if incompatibility is maintained in both periods, there are only two submarkets (A, A), (B, B). Hence, first-period compatibility choices dramatically affects the distribution of switching cost in that all consumers switching to (B, B) must incur a total switching cost of 2s under the first-period incompatibility whereas their total switching cost can be 2s or s under the first-period compatibility. As competition in market (A, B) and (B, A) is much stronger under incompatibility than under compatibility, firms choose compatibility in period two if compatibility was chosen in period one. 4 Second, when first-period market shares are endogenous and the relative weight of the second-period payoff is large, we find that the asymmetry in the dynamics of compatibility disappears. For switching costs low (high) enough, firms end up choosing compatibility (incompatibility) in period two no matter what the first-period compatibility regime. In particular, when switching cost is high enough, given first-period compatibility, there exists no symmetric equilibrium in period two and instead there exists a cornering equilibrium in which a firm corners both markets and chooses incompatibility which softens second-period competition compared to compatibility. Finally, we find that there exist two kinds of pure strategy equilibria in model 2: either compatibility is chosen in both periods or incompatibility is chosen in both periods. We also find that if the incompatibility equilibrium arises in model 1, it arises also in model 2. But there are parameter values under which the incompatibility equilibrium arises in the latter while it does not in the former. Symmetrically, if the compatibility equilibrium arises in model 2, it arises also in model 1. But there are parameter values under which the compatibility equilibrium arises in the latter while it does not arise in the former. In summary, when consumer lock-in arises due to high switching cost, firms are likely to make their products incompatible in order to soften the future competition that occurs after consumer lock-in. We also provide an extension to show that incompatibility is more likely to be chosen in a mature market than in a growing market. We merge two different strands of literature, the one on compatibility and the one on poaching. First, as incompatibility is equivalent to pure bundling, our paper is related to literature on bundling or incompatibility. The bundling (or tying) literature can be divided into three categories. The first one includes the papers that view bundling as a price discrimination device for a monopolist (Stigler, 1968, Schmalensee, 1984, McAfee et al. 1989, Salinger 1995, Armstrong 1996, Bakos and Brynjolfsson, 1999, Fang and Norman, 2006, Chen and Riordan, 2013, Menicucci, Hurkens, and Jeon, 2015). The second is about competitive bundling where entry and exit is not an issue (Matutes and Régibeau 1988, 1992, Economides, 1989, Carbajo, De Meza and Seidmann, 1990, Chen 1997, Denicolo 2000, Nalebuff, 2000, Armstrong and Vickers, 2010, Carlton, Gans, and Waldman, 2010, Thanassoulis 2011). The last is about leverage theory of bundling in which the main motive of bundling is to deter entry or induce the exit of rival firms in the competitive segment of the market (Whinston, 1990, Choi and Stefanadis 2001, Carlton and Waldman, 2002, Nalebuff, 2004, Peitz 2008, Jeon and Menicucci, 2006, 2012). 5 Our paper is closely related to the second literature, in particular, Matutes and Régibeau (1988), Economides, (1989), the extension of Matutes and Régibeau (1988) to asymmetric firms by Hahn and Kim (2012) and Hurkens, Jeon and Menicucci (2016) and the recent papers on pure bundling when n(> 2) symmetric firms compete (Kim and Choi (2015) and Zhou (2015)). Economides (1989) extends the finding of Matutes and Régibeau (1988) to n > 2 firms. However, Kim and Choi (2015) and Zhou (2015) show that when there more than two firms compete, incompatibility can generate higher profits than compatibility, which contradicts the finding of Economides (1989).6 Second, our two-period game is similar to the games considered in poaching literature (Chen, 1997, Fudenberg and Tirole, 2000). The main difference is that we consider multiproduct firms and their compatibility choices. This generates an interesting dynamics as the firms’ first-period compatibility choices change the distribution of switching costs and thereby affects the second-period compatibility choices and the second-period competition to poach consumers The paper is organized as follows. Section 2 describes the model. Section 3 analyzes second-period price competition given a compatibility regime and introduces demand size effect and demand elasticity effect of incompatibility. Section 4 analyzes model 1 of once-and-for-all compatibility choices. Section 5 analyzes model 2 of dynamic compatibility choices. Section 6 presents the analysis of forward-looking consumers and Section 7 provides the analysis of consumer surplus and welfare. Section 8 presents an extension studying the relationship between compatibility choices and market growth. Section 9 provides the conclusion. All the proofs are gathered in Appendix. 6 Kim and Choi (2015) extend Matutes and Régibeau (1988) by allowing for n ≥ 3 symmetric firms, and assuming that competition occurs on a torus (i.e., the Cartesian product of two circumferences) rather than on Hotelling segments. They prove that if n ≥ 4, then there exists a way to symmetrically locate the firms on the torus which results in pure bundling generating higher profits than independent pricing. Zhou (2015) studies a model with m ≥ 2 symmetric products and n ≥ 2 symmetric firms, but instead of adopting a spatial competition model he uses the random utility framework of Perloff and Salop (1985). He proves that if the number of firms is sufficiently large, then pure bundling increases the firms’ profits with respect to no bundling. For both papers, the results are a consequence of a demand elasticity effect, which makes the demand under bundling less elastic and thus softens competition, as it occurs in the second-period of our model when switching cost is large (see Section 3.3). 6 2 Model There are two firms, i = A, B, who produce two products, j = x, y, at constant marginal cost c. In our model, the two products can be perfect complements or can be independently consumed but our results are the same as long as each consumer obtains a high enough value from each product, which we assume. Since each firm produces each product, four systems are possible: (x, y) = (A, A), (A, B), (B, A), (B, B). We consider two-period models in which consumers have switching costs in the second period. Each firm’s marginal cost of producing a product is c > 0. c is large enough such that if a consumer bought product j = x, y from a firm, he or she has no incentive to buy the same product from the competing firm. Thus, at each period t = 1, 2, every consumer has a unit demand for each product. Given this assumption, we normalize c to zero, and interpret prices as the mark up above marginal cost without loss of generality. Regarding compatibility choices, we consider two different models. In model 1, we consider once-and-for-all compatibility choices and assume that in the beginning of period one, each firm simultaneously and non-cooperatively chooses between compatibility and incompatibility. Compatibility regime prevails only if both firms choose compatibility; otherwise, incompatibility regime prevails. In model 1, the regime determined in period one is maintained in period two. In model 2, we consider dynamic compatibility choices in which each firm makes a compatibility choice at the beginning of every period. In the first period, consumers have heterogeneous costs of learning to use each product as in Klemperer (1995). Precisely, a consumer’ learning cost for product j = x, y is uniformly distributed along a line segment [0, 1]. It is independently distributed across different products and different consumers. For instance, a consumer located at (θx , θy ) ∈ [0, 1]2 incurs a total learning cost of tθx + tθy to use a system (A, A) or tθx + t(1 − θy ) to use system (A, B). In the beginning of period one, every consumer has the same expected valuation for each product, v e . Therefore, depending on the compatibility regime, the first-period utility of a consumer located at (θx , θy ) from purchasing (A, A) is given as follows; under compatibility, it is A U1 (A, A) = 2v e − pA 1,x − p1,y − tθx − tθy ; 7 (1) under incompatibility, it is U1 (A, A) = 2v e − P1A − tθx − tθy , where pi1,j 7 is the price for product j of firm i in period one under compatibility and P1i is the price of the system (i, i) in period one under incompatibility. If there is no second period, our models are identical to that of Matutes and Régibeau (1988). Once a consumer uses firm i’s product j in period one, she discovers her own true value vji for the product, which is ex ante a random draw from a uniform distribution over [v, v], with v > 0. Hence, (v + v) /2 = v e . We assume that the distribution is independent across different products and different consumers. Now let us describe what happens in the second period. Consider for instance a consumer who bought product x from A in period one. Then she has learnt her value vxA . Her choice is either to consume the same product and obtain vxA , or to switch to product x of B. In the latter case, her expected value is v e and in addition she has to incur a switching cost of s > 0. The switching cost includes not only learning cost but also other costs due to a consumer’s product-specific investment. For simplicity, we assume that the switching cost is the same for all consumers and products. In addition, we assume that each firm can engage in behaviour-based price discrimination to poach consumers: the price a firm charges to a consumer in period two can depend on the history of the consumer’s purchase in period one. Suppose now that a consumer who bought (A, A) in the first period, switches to (A, B) in the second period (which is possible only under the second-period compatibility). Then, her second-period utility is given by: B U2 (A, B)|(A,A) = vxA + v e − pA 2,x (A, A) − p2,y (A, A) − s, (2) where pi2,j (i, h) is the second-period price charged by firm i for product j under compatibility to the consumers who bought (i, h) in the first period with i, h ∈ {A, B}. Similarly, if a consumer who bought (A, A) in the first period switches to (B, B) in the second period, her second-period utility under incompatibility is given by: U2 (B, B)|(A,A) = 2v e − P2B (A, A) − 2s, 7 We use lower-case letters for compatibility case and upper-case letters for incompatibility. 8 where P2i (i, h) is the second-period price charged by firm i for its system under incompatibility to the consumers who bought (i, h) in the first period with i, h ∈ {A, B}. All players have a common discount factor δ > 0: δ can be larger than one since it represents the weight assigned to the second-period payoff. All firms have rational expectations. We consider myopic consumers for our exposition and consider forwardlooking consumers in Section 6 and show that if consumers were forward-looking, our results remain unaffected in model 1 and remain qualitatively similar in model 2. We introduce the following assumption to guarantee that a positive measure of consumers switch in each sub-market in period two: Assumption 1: s < 23 ∆v where ∆v ≡ v − v. If the assumption is not satisfied, no switching occurs under second-period compatibility. In model 2 of dynamic compatibility choices, the timing in each period is given by: • Stage 1: Each firm simultaneously and non-cooperatively chooses between compatibility and incompatibility. • Stage 2: After observing the outcome of the compatibility choices, each firm simultaneously and non-cooperatively chooses its prices. • Stage 3: Consumers make purchase decisions. In model 1 of once-and-for-all compatibility choices, Stage 1 occurs only in period one. Notice that there always exist an equilibrium in which both firms choose incompatibility. But we assume that each firm plays its weakly dominant action and therefore compatibility arises if and only if both firms prefer compatibility. In order to solve this two-period model, we first solve for firms’ second-period equilibrium behavior for any given first-period compatibility choice and market shares. We find that equilibrium prices and profits are linearly homogenous in ∆v regardless of compatibility regime. Therefore sometimes it is useful to normalize ∆v to 1; the model of (∆v, s) is equivalent to that of (1, s/∆v). 9 3 Second-period price competition given a regime of compatibility In this section, we first study the second-period price competition to poach consumers for a given regime of once-and-for-all compatibility and show how incompatibility affects the second-period profits. 3.1 Given compatibility Suppose that the firms chose compatibility in the first period. Consider the market composed of the consumers who have bought product j from firm i in the first period. We call it market ij . Although there are four different markets Ax , Ay , Bx , By in period 2, all four markets are alike. Hence, it is enough to analyze just one market ij 8 . We normalize the total mass of consumers in market ij to one. In this market, the firms offer prices pi2,j (ij ) and ph2,j (ij ) with i 6= h and i, h = A, B. To simplify notation, let us define: i − h + i − h + i − h p+ 2 ≡ p2,j (ij ), p2 ≡ p2,j (ij ), d2 ≡ d2,j (ij ) , d2 ≡ d2,j (ij ) , π2 = π2,j (ij ), π2 = π2,j (ij ), where + (respectively, −) refers to the advantage of firm i (respectively, the disadvantage of firm h) due to the switching cost and di2,j (ij ) is the demand for product j of firm i and i (ij ) is firm i’s profit in market ij . A consumer with valuation vji for product j of firm π2,j i is indifferent between staying with firm i or switching to firm h (6= i) if and only if e − vji − p+ 2 = v − p2 − s. −∗ + + − − The equilibrium prices, p+∗ 2 and p2 , maximize p2 d2 and p2 d2 , respectively. We have: Lemma 1. Suppose that compatibility was chosen in both periods. Consider the secondperiod competition in market ij composed of the consumers who bought product j from firm i in period one. We normalize the total mass of consumers in market ij to one. There exists a unique equilibrium, which is characterized as follows: 8 Without loss of generality, we can consider that the poaching price for product x (say) depends only on whether a given consumer bought x from A or B but not on whether he bought y from A or B. 10 (i) The equilibrium prices in period 2 are p+∗ 2 = ∆v s + , 2 3 p−∗ 2 = ∆v s − . 2 3 (3) (ii) The firms’ second period profits from this market are given by: π2+∗ 1 = ∆v ∆v s + 2 3 2 , π2−∗ 1 = ∆v ∆v s − 2 3 2 Corollary 1. π2+∗ (s) + π2−∗ (s) strictly increases with s for s ∈ 0, 3∆v 2 . (4) −∗ +∗ −∗ As socially optimal switching requires p+∗ 2 = p2 but we have p2 > p2 , there always exists excessive switching from a social point of view. The corollary says that the industry profit in a given product market strictly increases with s. We note that in model 2 of dynamic compatibility choices in which each firm makes a compatibility choice every period, Lemma 1 applies to market ij as long as there is compatibility in the second period, independently of the first-period compatibility regime. 3.2 Given incompatibility Suppose that the firms chose incompatibility in the first period. Consider market (i, i) composed of the consumers who purchased both products from firm i in the first period. We normalize the total mass of consumers in this market to one. In this market, firms offer prices P2i (i, i) and P2h (i, i) with i 6= h and i, h = A, B. To simplify notation, let us define: i − h P2+ ≡ P2i (i, i), P2− ≡ P2h (i, i), D2+ ≡ D2i (i, i), D2− ≡ D2h (i, i), Π+ 2 = Π2 (i, i), Π2 = Π2 (i, i). A consumer with valuations vxi , vyi is indifferent between buying (i, i) and (h, h) in period two if and only if vxi + vyi − P2+ = 2v e − P2− − 2s. We have: Lemma 2. Suppose that incompatibility was chosen in both periods. Consider the secondperiod competition in the market composed of the consumers who bought a pure system 11 (i, i) in period one. We normalize the total mass of consumers in market (i, i) to one. There exists a unique equilibrium, which is characterized as follows: (i) The equilibrium prices in period 2 are P2+∗ i i 1 hp 1h p −∗ 2 2 2 2 3 (2s − ∆v) + 8∆v + 5(2s − ∆v) , P2 = (2s − ∆v) + 8∆v − (2s − ∆v) = 8 8 (ii) The firms’ second period profits from this market are given by: Π+∗ 2 = 2 P2−∗ 1− ∆v 2 2 ! P2+∗ , Π−∗ 2 2 P2−∗ = ∆v 2 3 −∗ 3∆v . Corollary 2. Π+∗ 2 (s) + Π2 (s) strictly increases with s for s ∈ 0, 2 As the socially optimal switching in market (i, i) requires P2+∗ = P2−∗ but we have P2+∗ > P2−∗ , there is always a socially excessive switching. In addition, for some consumers the socially optimal switching involves switching only one product. Therefore, the incompatibility generates inefficiency as it forces a consumer either not to switch at all or to switch both products. The corollary shows that the industry profit in market (i, i) strictly increases with s. We note that in the case of dynamic compatibility choices in which each firm makes a compatibility choice every period, Lemma 2 applies to market (i, i) as long as there is second period incompatibility, independently of the first-period compatibility regime. 3.3 The effects of incompatibility: the demand size effect and the demand elasticity effect Before we compare second-period profits, we here summarize the findings of Hurkens, Jeon and Menicucci (2016), which are useful to understand our main results. They consider competition between a dominant firm and a dominated firm in a static model and decompose the effects of incompatibility (i.e., pure bundling in their paper) into the demand size effect and the demand elasticity effect. We here apply their analysis to the second period in the current model by considering the market composed of the consumers who bought product j (= x, y) from firm A (for instance) in period one. Then, firm A is the 12 dominant firm and firm B is the dominated firm in this market given s > 0. We normalize the mass of consumers in this market to one. Demand size effect. Suppose that under compatibility firm A charges p+ 2 and firm B − A charges p2 . Let vej be the valuation of the consumer who is indifferent between switching and no switching and hence the demand for A’s product is 1 − F (e vjA ) where F (·) is the distribution function of vjA over [v, v], which has a symmetric log-concave density f (·) (including the uniform distribution). In any equilibrium with s > 0, we expect vejA < v e and hence 1 − F (e vjA ) > 1/2. Suppose now that there is incompatibility and that each firm sells its system at a price equal to the sum of the previous prices, that is P2+ = 2p+ 2 and − − i P2 = 2p2 . Then the indifferent consumer has the average valuation equal to vej . Letting F̂ denote the distribution function for the average valuation, we infer that the demand for A’s system is 1 − Fb(e vjA ). Since the distribution of the average valuation is more-peaked around v e than the distribution of the individual location for any symmetric log-concave vjA ) for any vejA < v e . Hence, for given prices, density f (·), we have 1 − Fb(e vjA ) > 1 − F (e incompatibility increases the demand for the dominant firm A and decreases the demand + for the dominated firm B. In particular, these arguments apply when p+ A and pB are equal, respectively, to their equilibrium values under compatibility for any s ∈ (0, 3∆v/2). Demand elasticity effect. After the regime changes from compatibility to incompati− − bility, firms will have incentives to choose prices different from P2+ = 2p+ 2 and P2 = 2p2 . Whether they want to charge higher or lower prices depends on how incompatibility affects demand elasticity, which in turn depends on the valuation of the indifferent consumer and hence on the level of the switching cost. For low levels of switching cost (that is, when vejA is not much smaller than v e ), incompatibility makes the demand more elastic: a given decrease in the average price of a system under incompatibility generates a higher boost in demand than the same decrease in the prices of single products under compatibility because the distribution of the average location is more-peaked around v e than the distribution of individual locations. On the other hand, for high levels of switching cost (that is, when vejA is close to v), incompatibility makes the demand less elastic: because fb(v)/f (v) converges to zero as v tends to v, where fb(·) is the density of the average valuation, for vejA close enough to v, we find that a given decrease in the average price of a system generates a smaller boost in demand than the same decrease in the prices of single products. In summary, incompatibility changes the elasticity of demand such that firms compete more aggressively for low levels of switching costs but less aggressively for high 13 levels of switching costs. Second-period profit comparison. As a result of these two effects, we find the following profit comparison: Corollary 3. Let ∆v = 1 without loss of generality. Then, there are three threshold values of switching cost, s1 , s2 , s3 with s1 < s2 < s3 , such that 1 2π2+∗ (s) R Π+∗ 2 (s) if and only if s Q s ; −∗ 2 2π2+∗ (s) + 2π2−∗ (s) R Π+∗ 2 (s) + Π2 (s) if and only if s Q s ; 3 2π2−∗ (s) R Π−∗ 2 (s) if and only if s Q s . More precisely, s1 = 0.701, s2 = 0.825 and s3 = 1.187. 2.5 Π+2 ∗ 2π2+ ∗ 2 1.5 1 0.5 2π2− ∗ Π−2 ∗ 0 0 0.5 0.701 1 1.187 1.5 s Figure 1: Second-period profits with or without compatibility −∗ Even if each of π2+∗ , π2−∗ , Π+∗ 2 , Π2 depends on s, in what follows we do not highlight −∗ this dependence unless it is necessary. The next graph shows 2π2+∗ , 2π2−∗ , Π+∗ 2 and Π2 14 as a function of s ∈ (0, 3/2) where we set ∆v = 1. There are three thresholds such that +∗ +∗ −∗ +∗ −∗ for s > 0.701 we have Π+∗ 2 > 2π2 , for s > 0.825 we have Π2 + Π2 > 2π2 + 2π2 , and −∗ for s > 1.187 we have Π−∗ 2 > 2π2 . This result can be understood in view of the two effects introduced above. For low level of switching costs (for instance, s close to zero), the demand size effect is negligible relative to the demand elasticity effect and the latter intensifies competition compared to compatibility. Therefore, incompatibility reduces both firms’ profits. In particular, when s = 0, there is only the demand elasticity effect, which explains the result of Matutes and Régibeau (1988) that incompatibility intensifies competition for symmetric firms. For high level of switching costs (for instance, s close to 3/2), the demand size effect is again negligible relative to the demand elasticity effect but now the latter softens competition compared to compatibility. Therefore, incompatibility increases both firms’ profits. For intermediate level of switching costs, the demand elasticity effect might be neutral but the demand size effect is positive for firm A and negative for firm B. Therefore, incompatibility increases A’s profit but reduces B’s profit. These kinds of results were obtained by Hahn and Kim (2012) and Hurkens, Jeon and Menicucci (2016).9 4 Model 1: once-and-for-all compatibility choices In this section, we consider model 1 of once-and-for-all compatibility choices. As the second-period price competition is analyzed in Section 3, we start by analyzing the firstperiod price competition. 4.1 First-period price competition given a regime of compatibility We study the first-period price competition given a compatibility regime. As we mentioned in Section 2, we consider myopic consumers and analyze forward-looking consumers in Section 6. 9 The intuition based on the demand size effect and demand elasticity effect are provided only by Hurkens, Jeon and Menicucci (2016). Both papers find two thresholds of dominance corresponding to s1 and s3 . This suggests that there should be another threshold between the two such that the industry profit is higher under incompatibility if the dominance level is higher than it. 15 4.1.1 Given compatibility Suppose first that compatibility was chosen in the beginning of the first period. For any vector of the first-period prices pi1,x , pi1,y , ph1,x , ph1,y , firm i’s profit is given as follows: Πi = di1,x (pi1,x + δπ2+∗ ) + (1 − di1,x )(δπ2−∗ ) + di1,y (pi1,y + δπ2+∗ ) + (1 − di1,y )(δπ2−∗ ), (5) where di1,j represents demand for product j of firm i in period one (for j = x, y) and is given by 1 1 di1,j = + (ph1,j − pi1,j ). (6) 2 2t We have: Proposition 1. Consider the model of once-and-for-all compatibility choices. Given compatibility, there exists a unique equilibrium, which is symmetric. The equilibrium prices in the first period and the total equilibrium profits are given by +∗ −∗ pi∗ = t − δ π − π ≡ p∗1 , 1j 2 2 for i = A, B and j = x, y. (7) t + δπ2−∗ ≡ π ∗ for i = A, B. (8) 2 Given compatibility, the pricing in (7) is quite intuitive. For δ = 0, each firm charges a price per product equal to t as in a standard Hotelling model. For δ > 0, if firm i attracts a consumer from the rival in the first period in a given product market, its expected profit from the customer in the second period is π2+∗ . But if the customer stays with the rival, its expected profit from that customer in the second period is π2−∗ . Therefore, each firm is ready to pay δ π2+∗ − π2−∗ to attract a consumer from the rival. The profit in (8) can be understood in a similar way. For δ = 0, each firm gets a profit of t/2 as in a standard Hotelling model. For δ > 0, each firm’s profit from the second period is equal to π2−∗ as π2+∗ − π2−∗ is dissipated away during the first-period competition. π i∗ = 4.1.2 Given incompatibility Suppose now that incompatibility was chosen in the beginning of the first period. For any pair of the first-period prices P1i , P1h , the profit functions are as follows i −∗ Πi = D1i (P1i + δΠ+∗ 2 ) + (1 − D1 )(δΠ2 ), i −∗ Πh = (1 − D1i )(P1h + δΠ+∗ 2 ) + D1 (δΠ2 ), (9) 16 where D1i is firm i’s market share in period one and is given by (for P1h −2t < P1i < P1h +2t) D1i 1 = 2 2 1 h 1 i 1 + (P1 − P1 ) − 2 (P1h − P1i ) max{0, P1h − P1i }. 2t 4t (10) We have: Proposition 2. Consider the model of once-and-for-all compatibility choices. Given incompatibility, there exists a unique equilibrium, which is symmetric. The equilibrium prices in the first period and the total equilibrium profits are given by −∗ P1i∗ = t − δ Π+∗ ≡ P1∗ 2 − Π2 Πi∗ = t ∗ + δΠ−∗ 2 ≡ Π , 2 for i = A, B. for i = A, B. (11) (12) Under incompatibility, if δ = 0, it is well-known from Matutes and Régibeau (1988) that each firm charges a price for its system equal to t as in a two-dimensional Hotelling model. For δ > 0, if firm i attracts a consumer from the rival in the first period, its expected profit from the customer in the second period is Π+∗ 2 . But if the customer stays with the rival, its expected profit from him in the second period is Π−∗ 2 . Therefore, each +∗ −∗ firm is ready to pay δ Π2 − Π2 to attract a consumer, which is dissipated away. For this reason, each firm’s equilibrium profit is 2t + δΠ−∗ 2 . 4.2 Compatibility choice From Propositions 1 and 2, each firm chooses incompatibility if and only if 2t + δΠ−∗ 2 > −∗ ∗ −∗ −∗ t t + δ2π2 π , which is equivalent to Π2 − 2π2 > 2δ . In the next proposition, we neglect the incompatibility equilibrium when a compatibility equilibrium exists for the reasons explained in Section 2. We have: Proposition 3. Consider the model of once-and-for-all compatibility choice. −∗ t (i) If Π−∗ 2 −2π2 < 2δ , there is an equilibrium in which both firms choose compatibility. This equilibrium Pareto-dominates the equilibrium in which both firms choose incompatibility. −∗ t (ii) If Π−∗ 2 − 2π2 > 2δ , there is a unique equilibrium and it is such that both firms choose incompatibility. 17 Incompatibility intensifies competition between symmetric firms in a one-period game. Precisely, when δ = 0, incompatibility reduces each firm’s total profit from t to t/2, which is well-known from Matutes and Régibeau (1988). Proposition 3(i) generalizes their result for δt small enough: as long as δt is small enough, the firms choose compatibility because this softens competition in period one and the second-period profits are relatively unimportant. Proposition 3(ii) shows that the finding of Matutes and Régibeau (1988) is reversed if both the switching cost and the weight of the second period are large enough. Precisely, if s > s̄3 , then Π−∗ > 2π2−∗ . Then, the firms choose incompatibility for high δt since 2 incompatibility softens competition in period two. Even if part of the increased secondperiod profit is dissipated away during the first-period competition, each firm retains −∗ δ(Π−∗ 2 − 2π2 ) in terms of increased profit, which more than compensates the reduction in the first-period profit of t/2 if δt is large. Recall that in a standard one-period Hotelling model, t measures products differentiation between firms, and the higher is t, the higher are equilibrium profits. Therefore, the ratio δt measures the importance of the second period relative to the first-period profitability. 5 Model 2: dynamic compatibility choices In this section, we consider model 2 in which each firm makes a compatibility choice in each period. As usual, we start by analyzing the second period. 5.1 5.1.1 Second-period price competition and compatibility choice Second-period price competition Lemma 1 can be applied to the second-period price competition in each market ij as long as there is second-period compatibility, independently of the first-period compatibility regime. Similarly, Lemma 2 can be applied to each market (i, i) as long as there is secondperiod incompatibility, independently of the first-period compatibility regime. Therefore, what remains to be studied is the competition under incompatibility in the market (i, h) composed of the consumers who have bought a hybrid system (i, h) in the first period with i 6= h and i, h = A, B. This competition arises only if the firms chose compatibility in period one but incompatibility in period two. 18 Consider market (i, h). We normalize the total mass of consumers in this market to one. A consumer with valuations vxi , vyh is indifferent between buying (i, i) and (h, h) if and only if (13) vxi + v e − P2i (i, h) − s = v e + vyh − P2h (i, h) − s. Note that s cancels out in (13). It turns out that the demand for firm i’s system is identical to the demand for its system in the market composed of the consumers who bought system (i, i) when s = 0 (see the proof of Lemma 3). Therefore, Lemma 3 below is a special case of Lemma 2. As the two firms are symmetric in this market, we have a symmetric equilibrium. We introduce simplifying notation: i∗ h∗ P20∗ = P2i∗ (i, h) = P2h∗ (i, h), Π0∗ 2 = Π2 (i, h) = Π2 (i, h), where the superscript 0 refers to the fact that no firm has any advantage in market (i, h). We have: Lemma 3. Suppose that compatibility was chosen in period one while incompatibility was chosen in period two. Consider the period-two competition in the market composed of the consumers who bought the hybrid system (A, B) (or (B, A)) in period one. We normalize to one the total mass of consumers in this market. There exists a unique equilibrium, which is symmetric. (i) The equilibrium price for each firm is P20∗ = ∆v . 2 (ii) Each firm’s second-period profit from this market is Π0∗ 2 = ∆v . 4 In the market composed of the consumers who bought a hybrid system in period one, neither firm has any advantage no matter what the level of switching cost. Therefore, neither the equilibrium prices nor the profits depend on s. Note that we have +∗ −∗ Π0∗ 2 = Π2 (0) = Π2 (0). 19 −∗ +∗ −∗ 2 As 2π2+∗ (s) + 2π2−∗ (s) > Π+∗ 2 (s) + Π2 (s) if and only if s < s and both 2π2 (s) + 2π2 (s) −∗ and Π+∗ 2 (s) + Π2 (s) are strictly increasing, the second-period competition in market (A, B) or (B, A) under incompatibility is more intensive than competition in market (A, A) or (B, B) under incompatibility and competition in any product market under compatibility. 5.1.2 Second-period compatibility choice Compatibility choice when compatibility was chosen in period one. Suppose that both firms chose compatibility in the first period. Let ∆v = 1 without loss of generality. If both firms choose compatibility in the second period, then firm i’s secondperiod profit is: di1,x + di1,y π2+∗ + dh1,x + dh1,y π2−∗ , (14) where di1,j is the first-period market share of firm i for product j, and di1,j + dh1,j = 1. If any of the two firms chooses incompatibility in the second period, then i’s profit is 0∗ i h h i h h −∗ di1,x di1,y Π+∗ + d d + d d 2 1,x 1,y 1,x 1,y Π2 + d1,x d1,y Π2 . (15) Therefore, firm i chooses compatibility in the second period as long as (14) is at least as large as (15). Using Lemmas 1-3, we have the following result: Lemma 4. Suppose that compatibility was chosen in period one (and let ∆v = 1 w.l.o.g.). Then compatibility choices in the second period are as follows: 1 (i) If s is such that 2π2+∗ − Π+∗ 2 ≥ 0 (i.e., s ≤ s ), then both firms choose compatibility for any (di1,x , di1,y ) in [0, 1]2 . (ii) If s is such that 2π2+∗ − Π+∗ < 0 (i.e., s > s1 ), then at least one firm chooses 2 incompatibility if and only if di1,x and di1,y are both close to 1 or both close to 0. For s > s1 firm i prefers incompatibility if di1,x > −∗ +∗ −∗ i 0 2π2−∗ − Π−∗ π2+∗ + π2−∗ − Π02 2 − (Π2 − Π2 − π2 + π2 )d1,x i and d > , 1,y +∗ −∗ +∗ −∗ −∗ +∗ 0 i 0 Π+∗ (Π02 − Π−∗ 2 + π2 − π2 − Π2 2 − π2 + π2 ) + (Π2 + Π2 − 2Π2 )d1,x 20 that is if di1,x and di1,y are close enough to one. This is quite intuitive, as when di1,x and di1,y are both close to 1, the comparison between (14) and (15) approximately reduces to 2π2+∗ +∗ +∗ 1 vs Π+∗ 2 , and we have 2π2 < Π2 given s > s . Likewise, firm h prefers incompatibility if dh1,x and dh1,y are close to 1 (i.e., when di1,x and di1,y are close to zero). For instance, figure 2 shows firm i prefers incompatibility if (di1,x , di1,y ) is above the thin curve, and firm h prefers incompatibility if (di1,x , di1,y ) is below the thick curve, when s = 1.1. 1 0.8 i d1,y 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 i d1,x Figure 2: Second-period compatibility choice when compatibility was chosen in period one for s = 1.1 As s increases, the set of (di1,x , di1,y ) such that at least one firm prefers incompatibility becomes wider, but for no s it includes the point such that di1,x = di1,y = 12 , as the following corollary states. Corollary 4. Suppose that the firms chose compatibility in the first period (and let ∆v = 1 B A w.l.o.g.). In the case of symmetric first-period market shares, that is dA 1,x = d1,x = d1,y = 1 dB 1,y = 2 , both firms prefer compatibility to incompatibility for any s ∈ (0, 3/2). 21 Given symmetric first-period market shares, we find for any s ≥ 0 1 1 0∗ −∗ π2+∗ (s) + π2−∗ (s) > (Π+∗ 2 (s) + Π2 (s)) + Π2 , 4 2 where the L.H.S. (the R.H.S) is each firm’s second-period profit under compatibility (under incompatibility). As we wrote after Lemma 3, competition in market (i, h) under incompatibility is more intensive than competition in (i, i) under incompatibility or under +∗ +∗ −∗ −∗ 1 (Π (s) + Π (s)) for any compatibility. More precisely, Π0∗ < min π (s) + π (s), 2 2 2 2 2 2 s > 0. This force induces the firms to embrace compatibility no matter what the level of −∗ switching cost even if π2+∗ (s) + π2−∗ (s) < Π+∗ (s) + Π (s) /2 holds for s > s2 . 2 2 Compatibility choice when incompatibility was chosen in period one. We now consider the case in which incompatibility was chosen in the first period. This means there is no hybrid system in the first period. Let D1i denote i’s first-period market share. Suppose first incompatibility in both periods. The equilibrium of the competition under incompatibility in each market (i, i) is given in Lemma 2. Hence firm i’s secondperiod profit is h −∗ D1i Π+∗ (16) 2 + D1 Π2 . Suppose now that the first-period incompatibility is followed by the second-period compatibility. Then, as we wrote just after Lemma 1, the firms’ second-period profits in each market ij do not depend on the first-period compatibility regime. Therefore, Lemma 1 applies and firm i’s second-period profit is 2 D1i π2+∗ + D1h π2−∗ . (17) In order to compare (16) and (17), we use D1h = 1 − D1i and see that firm i prefers second-period compatibility if and only if i −∗ D1i 2π2+∗ + (1 − D1i )2π2−∗ > D1i Π+∗ 2 + (1 − D1 )Π2 , which is equivalent to D1i < 10 2π2−∗ −Π−∗ 2 +∗ +∗ −∗ Π2 −Π−∗ 2 −(2π2 −2π2 ) ≡ D.10 Likewise, firm h with h 6= i −∗ +∗ −∗ 3 Notice that the denominator Π+∗ 2 − Π2 − (2π2 − 2π2 ) is positive for each s ∈ (0, 2 ). 22 prefers second-period compatibility if and only if +∗ i D1i 2π2−∗ + (1 − D1i )2π2+∗ > D1i Π−∗ 2 + (1 − D1 )Π2 , which is equivalent to D1i > immediate. +∗ Π+∗ 2 −2π2 −∗ +∗ −∗ −(2π −Π Π+∗ 2 −2π2 ) 2 2 ≡ D. Then the following lemma is Lemma 5. Suppose that the firms chose incompatibility in the first period (and let ∆v = 1 w.l.o.g.). Then compatibility choices in the second period are as follows: +∗ (i) If Π+∗ (that is, if s < s1 ), then D > 1 and D < 0. Hence compatibility 2 < 2π2 emerges in the second period for any D1i ∈ [0, 1]. −∗ +∗ −∗ 2 (ii) If Π+∗ 2 + Π2 ≥ 2π2 + 2π2 (that is, if s ≥ s ), then D ≤ D. Hence incompatibility emerges in the second period for any D1i ∈ [0, 1]. −∗ (iii) If Π2+∗ ≥ 2π2+∗ and Π+∗ < 2π2+∗ + 2π2−∗ (that is, if s1 ≤ s < s2 ), then 2 + Π2 0 ≤ D < 21 < D ≤ 1. Hence compatibility (incompatibility) emerges in the second period if D < D1i < D (if D1i ≤ D, or D1i ≥ D). We know from Corollary 3 that if the dominant firm prefers compatibility (i.e., if Π+∗ 2 < +∗ −∗ −∗ 2π2 ), then also the dominated firm prefers compatibility (i.e., Π2 < 2π2 ). Therefore, as long as the dominant firm prefers compatibility (i.e., if s < s1 ), we have that (16) is greater than (17) for any D1i and both firms choose compatibility in the second period. As −∗ +∗ the switching cost becomes larger than s2 , the inequality Π+∗ + 2π2−∗ holds. 2 + Π2 > 2π Then incompatibility emerges since this inequality means that the sum of the secondperiod profits is greater under incompatibility than under compatibility. Therefore at least one firm prefers incompatibility. Lemma 4(i) and Lemma 5 show that if s < s1 , the firms always end up choosing compatibility in period two regardless of the first-period compatibility regime and market shares. But such strong prediction cannot be made for s ≥ s2 . The following corollary covers the case of equal market shares. Corollary 5. Suppose that the firms chose incompatibility in the first period. In the case of symmetric market shares, i.e., D1A = D1B = 21 , the firms prefer incompatibility to −∗ +∗ −∗ 2 compatibility in the second period if and only if Π+∗ 2 + Π2 ≥ 2π2 + 2π2 (i.e., s ≥ s ). 23 5.2 First-period price competition In this subsection, we study the first-period competition given a first-period compatibility regime. As the case of first-period incompatibility is easier to study than that of the firstperiod compatibility, we start with the former. 5.2.1 Given incompatibility in the first period If incompatibility was chosen in period one, then Lemma 5 applies. As a consequence, for +∗ any s ∈ [0, s1 ) (i.e., when Π+∗ 2 < 2π2 ) and for any pair of the first-period prices under incompatibility P1i , P1h , the firms will choose compatibility in the second period. Hence, we consider the following profit functions for any P1i , P1h : Πi = D1i (P1i + δ2π2+∗ ) + (1 − D1i )(δ2π2−∗ ), Πh = (1 − D1i )(P1h + δ2π2+∗ ) + D1i (δ2π2−∗ ), (18) where D1i is firm i’s market share in period one and is given by (10). −∗ +∗ A similar argument can be applied to any s ∈ [s2 , 23 ) (i.e., when Π+∗ 2 + Π2 ≥ 2π2 + 2π2−∗ ). In this case, for any pair of the first-period prices under incompatibility P1i , P1h , the second-period compatibility regime will be incompatibility. Therefore, for any s ∈ [s2 , 32 ), we consider the profit functions from (9). Proposition 4. Suppose that incompatibility was chosen in the first period. Let ∆v = 1 without loss of generality. +∗ 1 (i) If Π+∗ 2 < 2π2 (i.e., if s < s ), there exists a unique equilibrium, which is symmetric, and such that P1i∗ = t − δ 2π2+∗ − 2π2−∗ for i = A, B. Both firms choose compatibility in the second period. Each firm’s total profit is Πi∗ = t + δ2π2−∗ 2 for i = A, B. −∗ (i.e., if s ≥ s2 ), there exists a unique equilibrium, (ii) If 2π2+∗ + 2π2−∗ ≤ Π+∗ 2 + Π2 which is symmetric. Both firms choose incompatibility in the second period. The first-period price is P1i∗ = P1∗ and each firm’s total profit is Π∗ as described in Proposition 2. 24 Proposition 4 covers s in [0, s̄1 ) and in [s̄2 , 23 ). Precisely, Proposition 4(ii) which considers s in [s̄2 , 23 ) is similar to Proposition 2, as both yield the same equilibrium prices and profits. The difference is that Proposition 2 refers to a setting in which incompatibility has been chosen once-and-for-all in period one while Proposition 4(ii) refers to a setting in which firms can change the compatibility regime in period two. Proposition 4(i) considers s in [0, s̄1 ) and describes a change in the compatibility regime over the two periods, which is implied by Lemma 5(i). However, the equilibrium prices and profits it identifies can be explained by arguments very similar to those provided after Proposition 1 and 2. Finally, as Lemma 5 states, if s is between s1 and s2 , then the compatibility/incompatibility regime in period two depends on whether D1i belongs to the interval (D, D) or not: compatibility emerges if the first-period market shares are not too different. In Figure 3 below we represent D (the thin curve) and D (the thick curve), for s ∈ [s1 , s2 ). Figure 3: (D, D) under first-period incompatibility choice for s ∈ [s1 , s2 ). In case there exists a symmetric pure-strategy equilibrium, then D1i = 21 and thus the firms choose compatibility in the second period. However, in period one a firm may deviate in a way to induce incompatibility in period two, that is in a way such that D1i ≤ D or D1i ≥ D holds, and an equilibrium needs to be robust to this sort of deviations. Next 25 proposition identifies a symmetric equilibrium candidate √ and shows that it is robust to 2(1−D̄)+2·(3D̄−2) √ . Conversely, such deviation if δ/t is not larger then δ̄(s) ≡ +∗ −∗ −∗ +∗ −Π 2(1−D̄) Π ( 2 2 −(2π2 −2π2 )) if δ/t is large then a firm can profitably deviate by choosing a small price in period one in order to gain a large market share, which yields a large profit in period two.11 In such a case no symmetric pure-strategy equilibrium exists, and the next proposition shows that no asymmetric pure-strategy equilibrium exists either. Proposition 5. Suppose that incompatibility was chosen in the first period. Let ∆v = 1 +∗ +∗ −∗ +∗ −∗ without loss of generality. Suppose that Π+∗ 2 ≥ 2π2 and 2π2 + 2π2 > Π2 + Π2 (i.e., s is between s1 and s2 ). (i) If a pure strategy equilibrium exists (symmetric or asymmetric), then P1A∗ = P1B∗ = t − δ 2π2+∗ − 2π2−∗ , as in Proposition 4(i). (ii) For each s ∈ (s1 , s2 ), the prices in (i) constitute an equilibrium if δ/t ≤ δ̄(s). 5.2.2 Given compatibility in the first period If compatibility was chosen in period one, then Lemma 4 applies. Therefore, for s ≤ s1 , compatibility arises also in period two as no firm wants to choose incompatibility in period two regardless of first-period market shares. Hence, we consider the profit function from (5). Then there exists an equilibrium which is outcome-equivalent to the equilibrium described by Proposition 1: Proposition 6. Suppose that compatibility was chosen in the first period. Then, there +∗ t exists a unique equilibrium, which is symmetric, if and only if Π+∗ 2 − 2π2 ≤ δ . In the equilibrium, each firm chooses compatibility in the second period. The equilibrium prices in ∗ the first period are pi∗ 1j and each firm’s total profit from market j = x, y is π as described in Proposition 1. However, for s greater than s1 each firm can induce incompatibility in the second period by choosing low prices such that it corners both markets in the first period. Indeed, for δ large, cornering is the best deviation as there are economies of scale under second-period A A incompatibility. Suppose that firm A deviates by capturing dA 1,x = d1,y = d1 ∈ (1/2, 1] 11 Precisely, the unique candidate equilibrium is such that each firm charges the price P ∗ = t−δ(2π2+∗ − but if firm i deviates with P i = P ∗ − 2t then firm i obtains full market share and profit −t − −∗ t − 2π2−∗ ) + δΠ+∗ if δt is large. 2 . This is greater than the candidate equilibrium profit 2 + δ2π2 2π2−∗ ), δ(2π2+∗ 26 share of consumers.12 Then, under the second-period incompatibility, firm A’s second2 +∗ A 0∗ A 2 −∗ period profit is dA Π2 + 2dA 1 1 (1 − d1 )Π2 + (1 − d1 ) Π2 , which is strictly convex in +∗ −∗ 0∗ dA 1 as Π2 + Π2 − 2Π2 > 0 holds for s > 0. To corner both markets, the deviating firm should charge p∗1 − t in period one and then obtains a profit of Π+∗ 2 in period two, which +∗ +∗ −∗ generates the total profit of δ Π2 − 2 π2 − π2 . Therefore, the cornering deviation +∗ +∗ t is not profitable if Π2 − 2π2 ≤ δ holds. This inequality is obviously satisfied for for +∗ t s ≤ s1 , but for s ∈ (s1 , s3 ) there exist t, δ satisfying Π+∗ 2 − 2π2 > δ , and then we find no pure strategy equilibrium. However, there exists a cornering equilibrium when s > s3 and δ/t is large. In this equilibrium, one firm charges a price below the marginal cost (i.e., p̄ < 0) in each market and the other firm charges p̄ + t such that the first firm gains full market share in both markets in period one and therefore chooses incompatibility in the second period, earning a total profit of 2p̄ + δΠ+∗ 2 . The other firm has zero market share in both markets in period one and earns a total profit of δΠ−∗ 2 . The value of p̄ must be small enough in order to deter the cornered firm from trying to earn a positive market share in period one, and high enough to deter the cornering firm from reducing its period-one market share in order to reduce its period-one losses. Since the possible deviations are somewhat intricate to describe, stating precisely the conditions under which this equilibrium exists is complicated, except that there exists a function δ` such that for each s > s3 , a cornering . It turns out that for each s ∈ (s̄3 , 23 ), the ratio equilibrium exists if and only if δt ≥ δ` (s) t δ` (s) is between Π+∗3/2 and Π+∗16/9 : see the proof in the appendix for the full details. t −2π +∗ −2π +∗ 2 2 2 2 Proposition 7. Suppose that compatibility was chosen in the first period, and assume −∗ Π−∗ 2 − 2π2 > 0. Then there exists an asymmetric equilibrium like the one described above if and only if δt ≥ δ` (s). Precisely, −∗ (i) in the first period, one firm charges price p̄ = γ − 2δ (Π+∗ 2 − Π2 ) in both markets (for a suitable γ in (0, t)), the other firm charges price p̄ + t in both markets, and the first firm corners both markets; (ii) in the second period, incompatibility is chosen at least by the firm who cornered the markets, which earns a total profit of 2γ + δΠ−∗ 2 , whereas the other firm’s total profit is δΠ−∗ 2 . 12 Lemma 6 in the proof of Proposition 6 shows that the optimal deviation is characterized by symmetric market shares if the deviating firm expects incompatibility in period two. 27 5.3 First-period compatibility choice Up to now we analyzed the first-period price competition given a first-period compatibility regime. We now study the firms’ first-period compatibility choices. Let us normalize ∆v to one without loss of generality. If the switching cost is low enough (i.e., s < s1 ), then by Lemmas 4 and 5 the firms choose compatibility in period two no matter their compatibility choices in period one. As compatibility in period one generates a higher first-period profit than incompatibility (see Propositions 6 and 4(i)), the firms will choose compatibility in both periods. Suppose now that the switching cost is high enough (i.e., s > s3 ). Then, incompatibility in period one leads to incompatibility in period two, with the profit of 2t + δΠ−∗ 2 for each firm (see Proposition 4(ii)). Regarding the scenario of compatibility in period one, +∗ we have to distinguish δ/t small from δ/t large. If δ/t is small (i.e., Π+∗ 2 − 2π2 ≤ t/δ), compatibility in period one leads to compatibility in period two (see Proposition 6). In this case, each firm’s profit is t + δ2π2−∗ . Therefore, the firms choose compatibility for δ/t small. If δ/t is large enough (see the inequality in Proposition 7), compatibility in period one leads to incompatibility in period two through the cornering equilibrium. In such an −∗ t equilibrium, the cornered firm earns δΠ−∗ 2 , which is less then 2 + δΠ2 . Therefore such a firm will choose incompatibility in period one. Summarizing, we have Proposition 8. (first-period compatibility choice) Consider the case in which each firm makes a compatibility choice every period. (i) If either the switching cost is low enough (i.e., s ≤ s1 ) to satisfy 2π2+∗ (s) ≥ Π+∗ 2 (s), 1 or if δ/t is small enough to satisfy δ/t ≤ Π+∗ −2π+∗ (and, furthermore, δ/t ≤ δ̄(s) 2 2 for s ∈ (s̄1 , s̄2 )) then there is a unique equilibrium, which is symmetric. In the equilibrium, the firms choose compatibility in both periods. −∗ (ii) If both the switching cost and δ/t are high enough to satisfy Π−∗ 2 (s) > 2π2 (s) and δ/t ≥ δ` (s) , then there is a unique equilibrium, which is symmetric. In the t equilibrium, the firms choose incompatibility in both periods. 28 5.4 Comparison: Model 1 vs. Model 2 We now compare the results we obtained in Proposition 3 from model 1 of once-and-for-allcompatibility choices with those in Proposition 8 from model 2 of dynamic compatibility choices. The key difference between the two models is that in model 2, firms make compatibility choices in period two as well, which implies that when considering a deviation in period one, a firm needs to think about how the deviation will affect the compatibility regime in period two. In a few cases, this generates more powerful deviations than the ones under model 1 such that no pure strategy equilibrium exists in model 2. For instance, +∗ given first-period compatibility, if δt is large and Π+∗ 2 > 2π2 , it is profitable for a firm to deviate in such a way to gain a large market share in period one in order to induce incompatibility in period two. As a consequence, Proposition 8 covers values of parameters such that a pure strategy equilibrium exists in model 2 given each compatibility regime in period one. With this caveat in mind, when we compare Proposition 3 with Proposition 8, we find that the incompatibility equilibrium arises in model 2 more frequently than in model 1. In order to see why, notice that in model 1 the incompatibility equilibrium emerges if and only if each firm’s profit under incompatibility, 2t + δΠ−∗ 2 , is larger than the one under compatibility, t + δ2π2−∗ . In model 2, the incompatibility equilibrium arises as long as the cornering equilibrium exists given compatibility in period one – see Proposition 8 – since this induces at least one firm to choose incompatibility in period one. The profit under compatibility is not relevant because when s > s̄3 and δt is large, the cornering equilibrium is such that compatibility in period one is followed in period two not by compatibility but by incompatibility. The existence condition for the cornering equilibrium is less −∗ restrictive than the inequality 2t + δΠ−∗ 2 > t + δ2π2 , and therefore the incompatibility equilibrium is found more frequently in model 2 than in model 1. As a consequence, the compatibility equilibrium arises less often in model 2 than in model 1. Precisely, in model −∗ 1 the compatibility equilibrium exists as long as 2t + δΠ−∗ 2 < t + δ2π2 , but in model 2 this inequality does not suffice since the compatibility equilibrium must be robust also to a period-one deviation which induces incompatibility in period two. Summarizing, we have: Proposition 9. (i) If the incompatibility equilibrium arises under once-and-for-all compatibility choices, it arises also under dynamic compatibility choices. But there are param29 eter values under which the incompatibility equilibrium arises in the latter while it does not arises in the former. (ii) If the compatibility equilibrium arises under dynamic compatibility choices, it arises also under once-and-for-all compatibility choices. But there are parameter values under which the compatibility equilibrium arises in the latter while it does not arise in the former. Despite the differences explained above, the two models generate quite similar predic−∗ t tions. If δt and the switching cost are large enough (i.e., Π−∗ 2 − 2π2 > 2δ holds), then the incompatibility equilibrium arises in both models. Similarly, if either the switching cost 1 is low enough ) to satisfy 2π2+∗ (s) ≥ Π+∗ 2 (s) or δ/t is small enough to satisfy n (i.e., s ≤ s o 1 1 2 δ/t ≤ min Π+∗ −2π+∗ , δ̄(s) for s ∈ (s̄ , s̄ ), then the compatibility equilibrium arises in 2 2 both models. 6 Forward-looking consumers Up to now, we have assumed that consumers are myopic. We here show that if consumers were forward-looking, our results remain unaffected in model 1 and remain qualitatively similar in model 2. First notice that our analysis of the second period continues to apply regardless of whether consumers are myopic or forward-looking. Second, in model 2 of dynamic compatibility choices, each single consumer knows in period one that he cannot be pivotal in influencing the firms’ second-period compatibility choices since he is an infinitesimal fraction of the population of consumers. Then, taking as given the compatibility regimes in period one and two, we inquire how forward-looking consumers make their purchases in period one. It is simple to see that if a consumer expects in period one compatibility to prevail period two, then his expected second-period utility does not depend on which products he buys in period one. Precisely, if he buys product j from firm i in period one, then his −∗ e utility in period two is the expectation of max{vji − p+∗ 2 , v − p2 − s}; if he buys product j −∗ e from firm h in period one, then his utility is the expectation of max{vjh −p+∗ 2 , v −p2 −s}. Since vji and vjh are identically distributed, it follows that the two expectations are equal, and the consumer will choose in period one on the basis of his first-period utility, as a myopic consumer does. The same principle holds if incompatibility prevails in period 30 one, and a consumers expects that it will be maintained in period two as well. If a consumer chooses the system of firm i in period one, then his payoff in period two is equal to the expectation of max{vxi + vyi − P2+∗ , 2v e − P2−∗ − 2s}. If he buys the system of firm h in period one, then his payoff in period two is equal to the expectation of max{vxh + vyh − P2+∗ , 2v e − P2−∗ − 2s}. Since the distribution of vxi + vyi is the same as the distribution of vxh + vyh , the two expectations are equal, and the consumer will choose in period one on the basis of his period one utility, as a myopic consumer does. This proves that in model 1 of once-and-for-all compatibility choices, a forward-looking consumer behaves exactly the same as a myopic consumer does. Proposition 10. (forward-looking consumers) Suppose that consumers are forward-looking. In model 1 of once-and-for-all compatibility choices, forward-looking consumers behave in the same way as myopic consumers do. However, things are quite different if compatibility prevails in period one, and a consumer expects incompatibility to prevail in period two. In this case, if a consumer buys both products of firm i in period one, then he obtains a payoff in period two equal to the expectation of max{vxi +vyi −P2+∗ , 2v e −P2−∗ −2s} ≡ uii2 , which is equal to uhh 2 . Conversely, if in period one a consumer buys product x from firm i and product y from firm h, then his payoff in period two is equal to the expectation of max{vxi +v e − ∆v −s, vyh +v e − ∆v −s} ≡ uih 2 , 2 2 hi ii ih which is equal to u2 . The inequality u2 < u2 holds since (as we previously noted in section 5.1.1) second-period competition under incompatibility is more intensive in the market of the consumers who bought a hybrid system than in the market of the consumers who bought a pure system. This implies that given compatibility in period one, a forward-looking consumer who expects incompatibility in period two makes more frequently hybrid purchases in period one than a myopic consumer does. However, given incompatibility in period one, there is no difference in behavior between forward-looking consumers and myopic consumers no matter the second-period compatibility regime which they expect to prevail. Consider now the compatibility equilibrium. In model 1, this equilibrium exists when it is robust to the deviation starting with first-period incompatibility. In model 2 with myopic consumers, the equilibrium exists if it is robust to, in addition to the aforementioned deviation, the deviation in which first-period compatibility is followed by secondperiod incompatibility. Now consider model 2 with forward-looking consumers. Relative 31 to when consumers are myopic, the deviation starting with first-period incompatibility is unaffected but the deviation in which first-period compatibility is followed by secondperiod incompatibility is less profitable. These arguments imply (i) if compatibility arises in model 2 with myopic consumers, it arises also in model 2 with forward-looking consumers; (ii) if compatibility arises in model 2 with forward-looking consumers, it arises also in model 1. Consider now the incompatibility equilibrium. In model 1, this equilibrium exists when it generates a payoff larger than the one when both firms choose compatibility. In model 2 with myopic consumers, for relevant parameter values, first-period compatibility leads to a cornering equilibrium inducing second-period incompatibility. The incompatibility equilibrium exists when it generates a payoff larger than the one obtained by the cornered firm in the cornering equilibrium. In model 2 with forward-looking consumers, relative to when consumers are myopic, first-period compatibility is likely to lead less frequently to the equilibrium in which one firm gains sufficient market shares to choose incompatibility in period two because of uii2 < uih 2 . Therefore, there are two possibilities: the relevant payoff conditional on the first-period compatibility is either the one from symmetric equilibrium leading to second-period compatibility or the one from asymmetric equilibrium leading to second-period incompatibility. In the former case, the incompatibility equilibrium exists in model 2 with forward-looking consumers if and only if it exists in model 1. In the latter case, the incompatibility equilibrium would exist in model 2 with forwardlooking consumers more frequently than in model 1 but less frequently than in model 2 with myopic consumers. 7 Consumer surplus and welfare In both models, we have, as pure strategy equilibrium, either the equilibrium in which compatibility is chosen in both periods or the equilibrium in which incompatibility is chosen in both periods. In this section, we compare, in terms of consumer surplus and welfare, the allocation that competition generates when compatibility is maintained in both periods with the one when incompatibility is maintained in both periods. Without loss of generality, we normalize ∆v to one. 32 7.1 Consumer surplus In period one, consumer surplus is greater under incompatibility than under compatibility for any s and δt for two reasons. The first reason is known from Matutes and Régibeau (1988): for symmetric firms, incompatibility intensifies first-period competition. As we explained in Section 3.3, this has to do with the demand elasticity effect of incompatibility. Even if incompatibility increases transportation costs incurred by consumers, this effect is dominated by the effect from more intensive competition. The second reason is that the difference between the dominant firm’s second-period profit and the dominated one’s profit greater under incompatibility than under compatibility, that is we have Π+∗ 2 (s) − Π−∗ 2 (s) > 2π2+∗ (s) − 2π2−∗ (s) 3 . for any s ∈ 0; 2 Hence, under incompatibility, the firms dissipate more the second-period profit from locked-in consumers by competing more aggressively in period one. Consumer surplus in period two depends on s. For instance, consider a consumer who bought both goods from firm A in period one. Then, for large values of s, we find that −∗ +∗ −∗ P2+∗ > 2p+∗ > 2p−∗ > p+∗ 2 , P2 2 , P2 2 + p2 .since incompatibility softens competition in period two with respect to compatibility. In particular, the first inequality implies that if this consumer buys both goods from A in period two, then he is better off under compatibility than under incompatibility, but in fact we find that for any possible realization of vxA , vyA , this consumer is better off under compatibility. As a consequence, the secondperiod consumer surplus is higher under compatibility for large s (i.e., for s > 0.876). −∗ Conversely, for s close to zero the opposite inequalities hold: P2+∗ < 2p+∗ < 2p−∗ 2 , P2 2 , +∗ +∗ −∗ P2 < p2 + p2 . The second-period consumer surplus is higher under incompatibility for s < 0.876. Let CS I (CS C ) denote the total consumer surplus under incompatibility (compatibility). The previous arguments seem to suggest that CS C > CS I when s > 0.876 and δt is large. But in fact we need to take into account that an increase in δ reduces both 2p∗1 and (especially) P1∗ . Hence, we find CS C > CS I if s > 1.168 and δ/t is above a suitable threshold which we denote `(s). Regarding the effect of equilibrium compatibility choices on total consumer surplus, we find a minor difference between model 1 and model 2. Precisely, Proposition 3 proves 33 that incompatibility arises in the equilibrium of model 1 if and only if s > s̄3 and δ 1 > 2(Π−∗ −2π −∗ , whereas Proposition 8 establishes that incompatibility arises in the equit ) 2 2 9/5 1 librium of model 2 if s > s̄3 and δt > Π+∗9/5 +∗ . We find that +∗ < `(s) < −∗ , Π+∗ 2(Π−∗ 2 −2π2 2 −2π2 2 −2π2 ) 1 and therefore whenever incompatibility emerges in model 1, (i.e., if δt > 2(Π−∗ −2π −∗ ), it 2 ) 2 necessarily reduces consumer surplus. Conversely, in model 2, there exist parameter valδ I C ues (i.e., if Π+∗9/5 +∗ < t < `(s)) such that firms choose incompatibility and CS > CS . 2 −2π2 Similar properties hold with respect to compatibility: in both models, when compatibility arises in equilibrium, it is often the case that CS I > CS C . Proposition 11. (consumer surplus) (i) Consumer surplus under compatibility is CS C = 1 −∗ 2[v e − p∗1 − 41 t + δ(v e − p+∗ 2 + 2 π2 )]. (ii) Consumer surplus under incompatibility is CS I = 2v e − P1∗ − 32 t + δ(2v e − P2+∗ + 2 −∗ Π )]. 3 2 (iii) The inequality CS C > CS I holds if and only if s > 1.168 and δt > `(s) = 5 +∗ +∗ 1 −∗ . Whenever incompatibility arises in equilibrium, in model 6(2π2+∗ −π2−∗ −2p+∗ 2 +P2 −Π2 + 3 Π2 ) 1, it always implies CS C > CS I ; in model 2, it implies CS C > CS I if δt > `(s) and CS C < CS I if δt < `(s). Whenever compatibility arises in equilibrium, CS I > CS C for s < 1.168 regardless of the model. What is remarkable is the conflict between the firms’ compatibility choices and conn −∗ 1 sumer surplus. Except for a very small range of parameters ( s ∈ (1.168, 23 ) and `(s) > δt > 2(Π−∗ 2 − 2π2 whenever the firms choose incompatibility (compatibility), it reduces CS compared to compatibility (incompatibility). This is because firms choose a regime of (in)compatibility in order to soften competition, which hurts consumer surplus. 7.2 Social welfare We start by describing the first-best allocation that maximizes social welfare. In period one, the first–best requires a consumer of location θj for good j (j = x, y) buys good Aj if θj < 21 and buys good Bj if θj > 12 . In period two, the first-best requires a consumer who purchased good j of A (for instance) in period one and observes vjA to keep buying good j from A if vjA > v e − s and to switch to B if vjA < v e − s. In particular, notice that no switching should occur in period two if s ≥ 21 . 34 Under compatibility, the first-period allocation coincides with the welfare-maximizing allocation, whereas the second period allocation is inefficient (as we noticed after Lemma −∗ 1) since p+∗ 2 > p2 implies that too much switching occurs. However, this efficiency loss is small if s is close to zero or if s is close to 23 . In the former case, we have p+∗ 2 close −∗ to p2 . In the latter case, maximizing the second-period welfare requires no switching and the second-period allocation under compatibility is such that the market share of the poaching firm (i.e., the proportion of switching consumers) tends to 0 as s tends to 23 . Incompatibility generates an allocation which is inefficient in both periods. In period one, consumers cannot mix and match, which increases the transportation costs incurred by consumers. In period two, likewise, incompatibility forces consumers to make switching decisions only at a system level, which prevents it from reaching the first-best. However, for some intermediate values of s (i.e., for s ∈ (0.535, 1.152)), the second-period social welfare is higher under incompatibility than under compatibility. This occurs because for these values of s, no switching maximizes social welfare and the market share of the dominant firm is significantly larger under incompatibility than under compatibility precisely because of the demand size effect of incompatibility explained in Section 3.3. As a consequence, total social welfare is greater under incompatibility than under compatibility for s ∈ (0.535, 1.152) if δ/t is large since the superior performance of incompatibility in period two dominates its inefficiency in period one. Conversely, if s < 0.535 or s > 1.152, then social welfare is greater in both periods under compatibility than under incompatibility. From Propositions 3 and 8, we see that s > s̄3 is a necessary condition for incompatibility to emerge in equilibrium regardless of the model. Since s̄3 > 1.152, we conclude that whenever the firms choose incompatibility, this necessarily reduces social welfare. When the firms choose compatibility, this choice generates a higher social welfare except for the case in which s ∈ (0.535, 1.152) and δ/t is large. Summarizing, we have Proposition 12. (welfare) (i) Social welfare under compatibility is SW C = 2[v e − p∗1 + 5 −∗ 3 t + δ(v e − p+∗ 2 + 2 π2 )]. 4 (ii) Social welfare given under incompatibility is SW I = 2v e − P1∗ + 13 t + δ(2v e − P2+∗ + 8 −∗ Π )]. 3 2 (iii) The inequality SW I > SW C holds if and only if s ∈ (0.535, 1.152) and δt > 1 . Whenever incompatibility arises in equilibrium, SW I < +∗ +∗ −∗ +∗ 5 −∗ 6(Π+∗ 2 + 3 Π2 −P2 −(2π2 +3π2 −2p2 )) SW C holds regardless of the model. When compatibility arises in equilibrium, SW C > 35 SW I holds except when s ∈ (0.535, 1.152) and 8 δ t > 1 . +∗ −∗ +∗ +∗ 5 −∗ 6(Π+∗ 2 + 3 Π2 −P2 −(2π2 +3π2 −2p2 )) Extension: growing vs mature market We here consider a simple extension of model 1 (once-and-for-all compatibility choices) in order to study how the compatibility choices are affected by whether the market is growing or mature. For this purpose, we assume that in the second period, a measure σ > 0 of new consumers appears. As the firms can apply behavior-based price discrimination, the firms can offer targeted prices to these consumers. Therefore, each firm’s profit from the new consumers is σt under second-period compatibility, is σt/2 under second-period incompatibility. Hence, the firms will choose compatibility in period one if t + δ 2π2−∗ + σt > t/2 + δ Π−∗ + σt/2 and choose incompatibility otherwise. Therefore, 2 we have: Proposition 13. Suppose that the firms make once-and-for-all compatibility choices in period one and that there is a measure σ > 0 of new consumers who arrive in the second period. Then the firms choose compatibility if t + δ 2π2−∗ + σt > t/2 + δ Π−∗ 2 + σt/2 holds and choose incompatibility otherwise. Therefore, compatibility is more likely to be chosen in a growing market than in a mature market. The above proposition generates an interesting prediction that platforms are more likely to embrace compatibility in a growing market than in a mature market. 9 Conclusion Our paper captures well Larry Page’s claim of an "island-like" Internet. Consumers’ data are stocked in platforms’ data center and consumers access them through cloud. As platforms typically own such data and moving data across platforms is hard, consumer lock-in arises very naturally.13 Then, our theory predicts that platforms embrace incompatibility today in order to soften future competition after consumer lock-in even if this intensifies today’s competition. Therefore, incompatibility hurts consumers. It also reduces welfare by making switching at product level impossible. 13 See Schneier (2015) for a good introduction to "the hidden battles to collect your data and control your world". 36 Our results are derived when any second-period rent from locked-in consumers is completely dissipated away during the first-period competition. In reality, such full passthrough of rent to consumers seems unlikely. If pass-through is partial and hence some rent is retained, this will further expand the regime of incompatibility. 10 References Armstrong, M. (1996). ”Multiproduct Nonlinear Pricing”, Econometrica 64, pp. 51-76. Armstrong, M. and Vickers, J. (2010). ”Competitive Non-linear Pricing and Bundling”, Review of Economic studies, 77: 30-60. Bakos, Yannis and Eric Brynjolfsson (1999). “Bundling Information Goods: Pricing, Profits and Efficiency” Management Science, 45, 1613-1630 Carbajo, Jose, David De Meza and Daniel J. Seidmann (1990). "A Strategic Motivation for Commodity Bundling", Journal of Industrial Economics, 38: 283-298 Carlton, Dennis W., Joshua S. Gans, and Michael Waldman (2010). "Why Tie a Product Consumers Do Not Use?," American Economic Journal: Microeconomics 2: 85105 Carlton, Dennis, W. and Michael Waldman, Michael (2002). "The Strategic Use of Tying to Preserve and Create Market Power in Evolving Industries," RAND Journal of Economics, 194-220 Chen, Yongmin, (1997). “Equilibrium Product Bundling,” Journal of Business, 70, pp. 85-103. Chen, Yongmin, (1997). "Paying Customers to Switch", Journal of Economics and Management Strategy, 6: 877-897. Chen, Yongmin and Michael Riordan (2013). “Profitability of Bundling,” International Economic Review, 54 (1): 35-57 Choi, Jay Pil and Stefanadis, Chris (2001). “Tying, Investment, and the Dynamic Leverage Theory”, RAND Journal of Economics, pp. 52-71. Denicolo, Vincenzo (2000). “Compatibility and Bundling With Generalist and Specialist Firms", Journal of Industrial Economics, 48: 177-188 Economides, Nicholas (1989). “Desirability of Compatibility in the Absence of Network Externalities”, American Economic Review, 79, 1165-1181. 37 Fang, H. and P. Norman (2006). ”To Bundle or Not to Bundle”, RAND Journal of Economics 37, pp. 946-963. Fudenberg, Drew and Jean Tirole (2000). "Customer Poaching and Brand Switching", The RAND Journal of Economics, 31: 634-657. Hahn, Jong-Hee and Sang-Hyun Kim (2012). “Mix-and-Match Compatibility in Asymmetric System Markets”, Journal of Institutional and Theoretical Economics, 168, 311-338. Hurkens, Sjaak, Jeon, Doh-Shin, and Menicucci, Domenico, (2016). "Leveraging Dominance with Credible Bundling," CEPR Discussion Paper. Kim, Sang-Hyun and Jay Pil Choi (2015). "Optimal Compatibility in Systems Market", Games and Economic Behavior, 90, 106-118 Klemperer, Paul (1995). "Competition when Consumers have Switching Costs: An Overview with Applications to Industrial Organization, Macroeconomics, and International Trade", Review of Economic Studies, 62: 515-539 Jeon, Doh-Shin and Domenico Menicucci (2006). "Bundling Electronic Journals and Competition among Publishers", Journal of the European Economic Association, vol.4 (5): 1038-83 Jeon, Doh-Shin and Domenico Menicucci (2012). "Bundling and Competition for Slots", American Economic Review, 102, 1957-1985. Matutes, Carmen and Régibeau, Pierre. (1988). ” ”Mix and Match”: Product Compatibility without Network Externalities,” RAND Journal of Economics, vol. 19(2), pages 221-234. Matutes, Carmen and Régibeau, Pierre. (1992). “Compatibility and Bundling of Complementary Goods in a Duopoly,” Journal of Industrial Economics, 40(1): 37–54. McAfee, R.P, McMillan, J., Whinston, M. (1989) ”Multiproduct Monopoly, Commodity Bundling, and Correlation of Values”, Quarterly Journal of Economics 104, pp. 371-384. Menicucci, Domenico, Sjaak Hurkens, and Doh-Shin Jeon (2015). "On the optimality of pure bundling for a monopolist," Journal of Mathematical Economics, 60(C), 33-42. Nalebuff, Barry, (2000). “Competing Against Bundles.” In Incentives, Organization, Public Economics, edited by Peter Hammond and Gareth D. Myles, London: Oxford University Press, pp. 323-336. Nalebuff, Barry (2004). “Bundling as an Entry Barrier.” Quarterly Journal of Economics, 119, 159-188. 38 Peitz, Martin (2008). "Bundling May Blockade Entry," International Journal of Industrial Organization, 26: 41-58. Salinger, Michael A. (1995). “A Graphical Analysis of Bundling.” Journal of Business, 68, 85-98. Schmalensee, Richard L. (1984). “Gaussian Demand and Commodity Bundling.” Journal of Business 57. S211-S230. Schneier, Bruce (2015). Data and Golliath: the Hidden Battles to Collect Your Data and Control Your World, W. W. Norton & Company, New Yorl · London. Stigler, George J. (1968). “A Note on Block Booking.” in The Organization of Industry, edited by George J. Stigler. Homewood, Illinois: Richard D. Irwin. Thanassoulis, John, (2011). “Is Multimedia Convergence To Be Welcomed?”, Journal of Industrial Economics 59(2), 225-253. Whinston, Michael D. (1990). “Tying, Foreclosure and Exclusion.” American Economic Review, 80, 837-59. Zhou, Jidong (2015). "Competitive Bundling", Mimeo, Yale University. 11 Appendix A Proof of Lemma 1 The demand for product j of firm i in market ij is given by: 1 if + − + 1 1 d2 = + ∆v s + p2 − p2 if s − 2 0 if s + ∆v 2 ∆v 2 + p− 2 − + p2 p+ < s− 2 + ≤ p2 ≤ s + < p+ 2 ∆v 2 ∆v 2 + p− 2 − + p2 (19) + and the demand for product j of firm h is d− 2 = 1 − d2 . + + −∗ − − (i) By definition of equilibrium, p+∗ 2 maximizes p2 d2 and p2 maximizes p2 d2 . Hence, + + − − the following F.O.C. need to hold (they are sufficient as p+ 2 d2 is concave in p2 , and p2 d2 is concave in p− 2 ): p+ 2 + d+ = 0⇔ 2 ∆v p− − 2 + d− = 0⇔ 2 ∆v − ∆v + + s + p− 2 − 2p2 = 0 2 ∆v + − s − 2p− 2 + p2 = 0 2 39 −∗ s ∆v s ∆v This implies p+∗ 2 = 2 + 3 , p2 = 2 − 3 . (ii) Using (19), and the equilibrium prices, we have π2+∗ = 1 ∆v ∆v 2 + s 2 , 3 π2−∗ = 1 ∆v ∆v 2 − s 2 . 3 Proof of Lemma 2 The demand function for firm h’s system is given by (for −∆v < P2+ − P2− − 2s < ∆v) D2− = 2 1 1 ∆v − 2s − P2− + P2+ − (P + −P2− −2s) max{0, P2+ −P2− −2s}, (20) 2 2(∆v) (∆v)2 2 and D2+ = 1 − D2− is the demand for the system of firm i. (i) The equilibrium prices maximize D2+ P2+ and D2− P2− , which implies the F.O.C. 2 P2+ = 2s − ∆v + 3P2− , and −8(P2− ) − 2(2s − ∆v)P2− + ∆v 2 = 0.14 Hence, P2+∗ = i i 1 hp 1h p (2s − ∆v)2 + 8∆v 2 − (2s − ∆v) 3 (2s − ∆v)2 + 8∆v 2 + 5(2s − ∆v) , P2−∗ = 8 8 (ii) Substituting the equilibrium prices in (20) yields the equilibrium demands and eventually the equilibrium second-period profits. Proof of Proposition 1 Given compatibility, the total profit of firm i from product j is: di1,j pi1,j + δ di1,j π2+∗ + (1 − di1,j )π2−∗ where di1,j = 21 + order condition 1 (ph 2t 1,j (21) − pi1,j ) is the first-period demand. From (21) we obtain the first di1,j − 1 i (p + δπ2+∗ − δπ2−∗ ) = 0 2t 1,j A similar first order condition,dh1,j − 2t1 (ph1,j + δπ2+∗ − δπ2−∗ ) = 0, needs to hold for firm h.15 Solving then, we find the first-period equilibrium prices in Proposition 1, and the equilibrium profits are obtained by replacing equilibrium prices in (21). 14 Since D2− is such that ∂D2− /∂P2− D2− is negative and decreasing in P2− (i.e., D2− is log-concave in P2− ), satisfying the first order condition relative to P2− D2− suffices to maximize P2− D2− with respect to P2− . A similar remark applies to the maximization of P2+ D2+ with respect to P2+ . 15 As in Lemma 1, these first order conditions are sufficient since the profit functions are concave. 40 Proof of Proposition 2 Assume without loss of generality that the equilibrium first-period prices, P1i∗ , P1h∗ , satisfy P1i∗ ≥ P1h∗ . Then given the profit functions in (9), with D1i in (10), P1i∗ , P1h∗ solve the F.O.C. given by16 1 h i −∗ 1 + (P1 − P1 ) P1i + δ Π+∗ − 2tD1i = 0 2 − Π2 2t 1 h −∗ i − 2tD1h = 0 1 + (P1 − P1 ) P1h + δ Π+∗ 2 − Π2 2t −∗ 17 The unique solution is P1i∗ = P1h∗ = t − δ Π+∗ − Π . From (9), the equilibrium profit 2 2 +∗ −∗ +∗ −∗ 1 1 for each firm is 2 (t − δ Π2 − Π2 + δΠ2 ) + 2 (δΠ2 ) = 2t + δΠ−∗ 2 . Proof of Proposition 3 The proof is omitted as it is straightforward. Proof of Lemma 3 The demand for firm i’s system is given by (for P h − ∆v < P i < P h + ∆v) 2 1 1 h i h i ∆v − P P (i, h) + P − (i, h) − P (i, h) (i, h) max{0, P2h (i, h)−P2i (i, h)} 2 2 2 2(∆v)2 (∆v)2 2 (22) and D2h (i, h) = 1 − D2i (i, h) is the demand function for firm h’s system. Notice that (22) coincides with (20) when s = 0. Hence, Lemma 3 is a corollary of Lemma 2 obtained by setting s = 0 into Lemma 2. D2i (i, h) = Proof of Lemma 4 (i) Using (14) and (15), we see that firm i prefers compatibility if and only if 1 1 i i −∗ +∗ −∗ −∗ i i +∗ 2π2−∗ − Π−∗ 2 ≥ d1,x d1,y (Π2 + Π2 − ) + (d1,x + d1,y )( − Π2 − π2 + π2 ) 2 4 (23) 16 Since D1i (D1h ) is log-concave in P1i (P1h ), the same given in footnote 14 applies here to establish that F.O.C. are sufficient. 17 If we consider P1i > P1h , then D1i < D1h , which implies that not both equalities can be satisfied. 41 Likewise, firm h prefers compatibility if and only if 1 1 i i +∗ −∗ i i +∗ +∗ −∗ 2π2+∗ − Π+∗ 2 ≥ d1,x d1,y (Π2 + Π2 − ) + (d1,x + d1,y )( − Π2 + π2 − π2 ) 2 4 (24) If s ≤ s̄1 , then we prove that (23) holds for each (di1,x , di1,y ) ∈ [0, 1]2 by showing that the maximum of the right hand side, considered as a function of (di1,x di1,y ) ∈ [0, 1]2 , is not larger than 2π2−∗ − Π−∗ 2 , the left hand side. To this purpose, notice that the Hessian matrix of the right hand side is indefinite, hence no maximum point exists in the interior of [0, 1]2 , and the four edges of the square [0, 1]2 have to be examined. For instance, if we consider the edge such that di1,x = 1, then the right hand side of (23) is −∗ −∗ +∗ −∗ 1 1 i i di1,y (Π+∗ 2 + Π2 − 2 ) + (1 + d1,y )( 4 − Π2 − π2 + π2 ): this is a linear function of d1,y i i which is smaller than 2π2−∗ − Π−∗ 2 both at d1,y = 0 and at d1,y = 1 . The other three edges of the square are dealt with similarly. The proof for (24) follows the same lines. (ii) The proof follows from immediate manipulations of (23)-(24). Proof of Lemma 5 The proof is omitted as it is straightforward. Proof of Proposition 4 (i) Given s < s̄1 , we will have second period compatibility regardless of first period market shares, by Lemma 5(i). Given the profit functions in (18), with D1i in (10), we can argue as in the proof of Proposition 2 to find the first period equilibrium prices P1i∗ = P1h∗ = t − δ 2π2+∗ − 2π2−∗ . Therefore the equilibrium profit for each firm is t + δ2π2−∗ . 2 (ii) Given s > s̄2 , we will have second period incompatibility regardless of first period market shares, by Lemma 5(ii). Then the proof of Proposition 2 applies. Proof of Proposition 5 Suppose that s ∈ (s̄1 , s̄2 ). We prove three claims: (i) no equilibrium exists such that incompatibility arises at stage two; (ii) the only candidate for equilibrium is such that P1i∗ = P1h∗ = t − δ 2π2+∗ − 2π2−∗ ; (iii) no profitable deviation exists if and only if δt is not too large. 42 Proof of claim (i) There exists no equilibrium such that D < Di < D̄ for some firm i We first prove that in no equilibrium we have Di ≥ D̄. Suppose that P1i∗ , P1h∗ is an equilibrium with D1i = D. Then the profit of firm i if P i < P̄ i is Πi in (9), and its derivative with respect to P i needs to be non-negative at dDi −∗ i i i P i = P̄ i , that is dP 1i (P1i∗ + δ(Π+∗ 2 − Π2 )) + D ≥ 0. The profit for firm i if P > P̄ is Π 1 in (18) and its derivative with respect to P i needs to be non-positive at P i = P̄ i , that is dD1i (P i∗ + δ(2π2+∗ − 2π2−∗ )) + D ≤ 0. However, these inequalities cannot both hold since dP i 1 dD1i dP1i −∗ +∗ −∗ < 0 and Π+∗ 2 − Π2 > 2π2 − 2π2 . Suppose that P1i∗ , P1h∗ is an equilibrium such that D < D1i < 1, which implies P1i∗ < P1h∗ . Then the profit functions are (9) in a neighborhood of (P1i∗ , P1h∗ ), and from these profit functions. We obtain first order conditions as in the proof of Proposition 1(i), which rule out P1i∗ < P1h∗ . Suppose that P1i∗ , P1h∗ is an equilibrium with D1i = 1. Then the derivative of the profit function of firm i with respect to P i is equal to one at P i = P i∗ . Therefore it is profitable for firm i to increase slightly P i above P̄ i . Now we rule out that Di ≤ D. If there exists an equilibrium such that Di ≤ D, then Dh ≥ D̄ and we can rule out this case as we have ruled out above that Di ≥ D̄. Proof of claim (ii) The only possible equilibrium is such that P1i∗ = P1h∗ = t − δ 2π2+∗ − 2π2−∗ We know that if P ∗i , P ∗h is an equilibrium, then D < Di < D̄. Hence the profit functions are given by (18) in a neighborhood of (P ∗i , P ∗h ), and from these profit functions we obtain first order conditions as in the proof of Proposition 4(i), which imply P1i∗ = P1h∗ = t − δ 2π2+∗ − 2π2−∗ . Proof of claim (iii) The prices P1i∗ = P1h∗ = t − δ 2π2+∗ − 2π2−∗ constitute an equilibrium if and only if δ/t is sufficiently close to 0 Given P1i∗ = P1h∗ = t − δ 2π2+∗ − 2π2−∗ , it is immediate that for no firm there exists a profitable deviation which leads to compatibility in period two, hence we consider here deviations which lead to incompatibility in period 2, that is such that Di ≤ D, or Di ≥ D̄. Deviations with large P Given that firms are symmetric, it suffices to consider the point of view of firm i at stage one, given incompatibility at stage one. Given P1h∗ = P1∗ = t−δ(2π2+∗ −2π2−∗ ), let P be the price of firm i such that D1i = D (recall that for s ∈ (s̄1 , s̄2 ), 43 P ∗ −P we have 0 < D < 21 ). Precisely, P belongs to (P1∗ , P1∗ +2t) and is such that 12 (1+ 12t )2 = √ D, that is P = P1∗ + 2t − 2t 2D. Since P1i = P induces incompatibility in period two, the √ profit of firm i from playing P1i = P is D t − δ(2π2+∗ − 2π2−∗ ) + 2t − 2t 2D + δΠ+∗ + 2 −∗ (1 − D)δΠ2 , that is p +∗ −∗ (3 − 2 2D)Dt + δ Π+∗ 2 − 2π2 + Π2 Π+∗ −2π +∗ 2 after using D = Π+∗ −Π−∗2 −(2π+∗ −∗ . The difference between the equilibrium profit 2 −2π2 ) 2 2 −∗ δ2π2 and (25) is (25) t 2 + p 1 −∗ − (3 − 2 2D)D t + δ(2π2+∗ + 2π2−∗ − Π+∗ 2 − Π2 ) 2 √ √ √ √ which is positive for each D ∈ (0, 12 ) since 21 −(3−2 2D)D = (1+2 2D)( 12 2− D)2 > 0 −∗ 2 and 2π2+∗ + 2π2−∗ − Π+∗ 2 − Π2 > 0 as s < s̄ . In fact, firm i can achieve incompatibility in the second period by choosing any P i ∈ −∗ −∗ (P , P ∗ + 2t], and in such a case the profit of firm i is D1i (P i + δ(Π+∗ 2 − Π2 )] + δΠ2 , with derivative dD1i i −∗ i (P + δ(Π+∗ (26) 2 − Π2 )) + D1 dP i Now consider the profit under compatibility in the second period, which is D1i [P i +δ(2π2+∗ − dDi 2π2−∗ )]+δ(2π2−∗ ), and its derivative with respect to P i , dP 1i [P i +δ(2π2+∗ −2π2−∗ )]+D1i , which dD1i +∗ −∗ −∗ +∗ −∗ i i is larger than (26) because Π+∗ 2 −Π2 > 2π2 −2π2 . Since dP i [P +δ(2π2 −2π2 )]+D1 < 0 for each P i > P1∗ ,18 in particular for P i > P , it follows that (26) is negative. Deviations with small P Given P1h∗ = P1∗ = t − δ(2π2+∗ − 2π2−∗ ), let P̄ be the price of firm i such that D1i = D̄ (recall that for s ∈ (s̄1 , s̄2 ), we have 21 < D̄ < 1). Precisely, ∗ P̄ belongs to (P ∗ − 2t, P ∗ ) and is such that 1 − 12 (1 − P 2t−P̄ )2 = D̄, hence P̄ = P ∗ − 2t + p 2t 2(1 − D̄). Since P1i = P̄ induces incompatibility in period two, the profit of firm i from p playing P1i = P̄ is D̄ t − δ(2π2+∗ − 2π2−∗ ) − 2t + 2t 2(1 − D̄) + δΠ+∗ + (1 − D̄)δΠ−∗ 2 2 , that is q 2 2(1 − D̄) − 1 D̄t + δ2π2−∗ (27) 18 This follows from the log-concavity of D1i : see footnote 16. 44 2π −∗ −Π−∗ 2 after using D̄ = Π+∗ −Π−∗2−(2π+∗ −∗ . The difference between the equilibrium profit 2 −2π2 ) 2 2 −∗ δ2π2 and (27) is q 1 − 2D̄ 2(1 − D̄) + D̄ t 2 t 2 + 2 which is positive for each D̄ > 1 2 since 1 2 − 2D̄ p 2(1 − D̄) + D̄ = 8D̄+1) (2D̄−1) (√ 2(1+2D̄+4D̄ 2(1−D̄)) > 0. In fact, i can achieve incompatibility in the second period by choosing any P i ∈ −∗ −∗ [P ∗ − 2t, P̄ ), and in such a case the profit of firm i is D1i (P i + δ(Π+∗ 2 − Π2 )] + δΠ2 , with derivative (26) which is equal to 3 i 2 4(∆π)2 δ 2 − 7t2 + 2∆Πtδ + 4∆πtδ + 4∆Π∆πδ 2 i − (P ) − (2t + ∆Πδ + 4∆πδ)P − 2 2 (28) +∗ −∗ +∗ −∗ with ∆Π = Π2 − Π2 , ∆π = 2π2 − 2π2 . Notice that if (28) is positive or zero at P i = P̄ , then we can conclude that the profit from deviation is increasing with respect to P i for P i ∈ [P ∗ − 2t, P̄ ), hence no profitable deviation exists, since we have already proved that deviating with P i = P̄ is unprofitable. Precisely, at P i = P̄ we find that p 2(1 − D̄), which is non-negative if and only if (28) is equal to −2 + 3D̄ + t+2∆πδ−∆Πδ 2t δ/t ≤ δ̄(s). 1 4t2 Proof of Proposition 6 Consider an equilibrium such that for each firm i, the profit in (14) is strictly greater than the profit in (15) at the equilibrium values for di1,x , di1,y . Then compatibility arises in both period and the profit functions are given by (5) in a neighborhood of first period equilibrium prices. Then we can argue as in the proof of Proposition 1 and find first period equilibrium prices and profits p∗1 , π ∗ in (7)-(8). Moreover, thanks to Corollary 4 we know that given di1,x = di1,y = 12 , both firm choose compatibility in the second period. Given the equilibrium candidate we mentioned above, now we show that no profitable +∗ t deviation exists if and only if Π+∗ 2 − 2π2 ≤ δ . (a) First, by the virtue of Proposition 1, we know that there exists no profitable deviation which induces compatibility in the second period. Therefore, no profitable +∗ 1 deviation exists if Π+∗ 2 ≤ 2π2 (that is, if s ≤ s̄ ) because then Lemma 4(i) reveals that in any case compatibility emerges in the second period. > 2π2+∗ . Then, by Lemma 4(ii), there exist (b) Now consider the case in which Π+∗ 2 45 deviations of firm i which lead to incompatibility in the second period: They are such that di1,x , di1,y are both close to 1, or both close to 0. We show that no profitable deviation +∗ t exists for firm i if and only if Π+∗ 2 − 2π2 ≤ δ . Given incompatibility in the second period, the total profit of firm i is Πi = di1,x pi1,x + di1,y pi1,y i h 0∗ i h 0∗ h h −∗ + δ di1,x di1,y Π+∗ 2 + d1,x d1,y Π2 + d1,y d1,x Π2 + d1,x d1,y Π2 i −∗ = di1,x pi1,x + di1,y pi1,y + δBdi1,x di1,y + δ Π0∗ − Π (d1,x + di1,y ) + δΠ−∗ 2 2 2 (29) −∗ 0∗ 0∗ where di1,j = 21 + 2t1 (p∗1 − pi1,j ) for j = x, y, and B = Π+∗ 2 − Π2 − Π2 + Π2 > 0. We now prove that to maximize (29) with respect to pi1,x , pi1,y , we can restrict to considering pi1,x = pi1,y : Lemma 6. Suppose that the firms chose compatibility in the first period and that a deviating firm expects that the incompatibility would prevail in the second period. Then, without loss of generality, we can restrict attention to deviations in the first period such that pi1,x = pi1,y . Proof of the lemma. Suppose that firm i deviates with pi1,x 6= pi1,y . Then consider p̂i1,x , p̂i1,y such that p̂i1,x = p̂i1,y = 21 (pi1,x + pi1,y ) ≡ p̂i1 . Then the demand for the product of firm i is dˆi1 = 21 (di1,x + di1,y ) in each market and i’s profit is19 i −∗ 2dˆi1 p̂i1 + δB(dˆi1 )2 + δ Π0∗ 2dˆ1 + δΠ−∗ 2 − Π2 2 which is larger than (29) because (i) 2dˆi1 p̂i1 = 2 12 (di1,x + di1,y ) 12 (pi1,x + pi1,y ) = di1,x pi1,x + di1,y pi1,y + 12 (pi1,x − pi1,y )(di1,y − di1,x ) = di1,x pi1,x + di1,y pi1,y + 4t1 (pi1,x − pi1,y )2 > di1,x pi1,x + di1,y pi1,y ; (ii) δB(dˆi1 )2 > δBdi1,x di1,y because δB > 0 and (dˆi1 )2 > di1,x di1,y , given that dˆi1 = 12 (di1,x +di1,y ) i i −∗ −∗ ; (iii) δ Π0∗ 2dˆ1 = δ Π0∗ (d1,x + di1,y ). 2 − Π2 2 − Π2 Given a symmetric deviation such that pi1,x = pi1,y ≡ p, we have di1,x = di1,y = 1 − Notice that if incompatibility emerges in the second period given di1,x , di1,y , then it also emerges given the market shares dˆi1 , dˆi1 in each market, given the shape of the set described in Lemma 4(ii), and immediately after the lemma. 19 46 π2+∗ − π2−∗ − δ 2t 1 p 2t ≡ d and −∗ + δΠ−∗ Πi = 2dp + δd2 B + 2δd Π0∗ 2 − Π2 2 δ 2δ = −4t 1 − B d2 + 4t 1 − A d + δΠ−∗ 2 , 4t 4t (30) −∗ +∗ −∗ where A = −Π0∗ 2 + Π2 + π2 − π2 . Then we need to consider a few cases, as a function of A and B. Suppose that 1 − 4tδ B ≤ 0, which implies that Πi is convex in d. Then Πi is maximized at d = 0 or (and) at d = 1. Precisely, the profit at d = 0 is δΠ−∗ 2 , which is smaller than +∗ +∗ −∗ the profit at d = 1, equal to δ Π2 − 2π2 + 2π2 . The latter is not larger than the +∗ t candidate equilibrium profit t + 2δπ2−∗ if and only if Π+∗ 2 − 2π2 ≤ δ . Now suppose that 1 − 4tδ B > 0, which implies that Πi is concave in d Then the maximum point for Πi is either d = 0, or d = 1, or d = 1 δ − 4t A 2 δ 1− 4t B if 0 < 1 2 We have already dealt with the cases of d = 0 and d = 1. When 0 < the deviation profit is: 2 1 δ − A Πi = 4t 2 4tδ + δΠ−∗ 2 . 1 − 4t B − 4tδ A < 1 − 4tδ B.20 1 2 − δ A 4t <1− δ B, 4t The difference between the candidate equilibrium profit t + 2δπ2−∗ and the deviation profit is " 2 # δ δ 1 δ 1 t(1 − B) + δ(1 − B)C − 4t − A t + 2δπ2−∗ − Πi = 4t 4t 2 4t 1 − 4tδ B δ δ 2 B = A − + C − (A + BC) , 4 4t 1 − 4tδ B ∆v where C = 2π2−∗ − Π−∗ 2 . In case s = 0 we have A = B = 0 and C = 4 , hence the deviation is not profitable. Otherwise, the inequalities 12 − 4tδ A ≤ 1 − 4tδ B, 0 < B − A 20 verify whether d = d= 1 δ 2 − 4t A is the point where the derivative of Πi in (30) vanishes. δ 1− 4t B 1 δ 2 − 4t A induces incompatibility in period two, but we prove below δ 1− 4t B Notice that d = 1 δ 2 − 4t A δ 1− 4t B is smaller than the candidate equilibrium profit. 47 In fact, we should that Πi in (30) at imply δ 4t ≤ 1 , 2(B−A) and since BC + A2 > 0 we find that A− B δ B A2 + BC + C − (A2 + BC) ≥ A − + C − 4 4t 4 2(B − A) (31) Since the right hand side in (31) is positive for each s ∈ (0, 32 ), it follows that t + 2δπ2−∗ − Πi > 0. Proof of Proposition 7 We are considering the equilibrium candidate in which firm i corners the other firm, h, in the first period in both markets. Precisely, firm i plays price p̄ in each market, and firm h plays price p̄ + t in each market, with p̄ to be determined. As a consequence, the −∗ profit of firm i is 2p̄ + δΠ+∗ 2 , and the profit of firm h is δΠ2 . Step 1 First, we focus on deviations given second-period incompatibility: this requires that the market share of the deviating firm is either close to 1 in both markets, or close to zero in both markets. Precisely, consider a deviation of firm i such that it plays price pi in both markets, which implies that the quantity sold by firm i in each market is di = 12 + 2t1 (p̄+t−pi ), and hence the profit of firm i is f i (di ) = 2di pi +δB(di )2 +2δγdi +δΠ−∗ 2 , +∗ −∗ −∗ 0 0 i i with B = Π2 − 2Π2 + Π2 and C = Π2 − Π2 . Using p = 2t + p̄ − 2d t, we obtain +∗ i f i (di ) = (Bδ − 4t) d2 + 2 (2t + p̄ + Cδ) d + δΠ−∗ 2 ; of course, f (1) = 2p̄ + δΠ2 . Since we are considering δt > B4 , it follows that f i is convex. Hence no deviation of firm i which induces incompatibility in the second period is profitable for i if and only if f i (1) ≥ f i (0), +∗ −∗ −∗ δ that is if 2p̄ + δΠ+∗ 2 ≥ δΠ2 , or − 2 (Π2 − Π2 ) ≤ p̄. Now consider firm h and a deviation such that firm h plays price ph in both markets, which implies that the quantity sold by the firm in each market is dh = 21 + 2t1 (p̄ − pi ), h h and hence its profit is f h (dh ) = 2dh ph + δB(dh )2 + 2δCdh + δΠ−∗ 2 . Using p = t + p̄ − 2d t −∗ h we obtain fh (dh ) = (Bδ − 4t) (dh )2 + 2 (t + p̄ + Cδ) dh + δΠ−∗ 2 (of course, f (0) = δΠ2 ), and since f h is convex (as δt > B4 ), it follows no deviation of firm h which induces incompatibility in the second period is profitable for h if and only if f h (0) ≥ f h (1) that −∗ is if p̄ ≤ t − 2δ (Π+∗ 2 − Π2 ). Therefore, no deviation which induces incompatibility is profitable for any firm if and only if δ δ +∗ −∗ −∗ − (Π+∗ (32) 2 − Π2 ) ≤ p̄ ≤ t − (Π2 − Π2 ) 2 2 48 Step 2 −∗ In view of (32), consider p̄ = γ − 2δ (Π+∗ 2 − Π2 ), with γ ∈ [0, t]. We need some −∗ 1 preliminaries in order to state the result precisely. In addition to B = Π+∗ 2 + Π2 − 2 , +∗ −∗ +∗ −∗ +∗ +∗ −∗ 1 1 + define E = Π+∗ 2 + π2 − 4 − π2 > 0, F = Π2 + π2 − 4 − π2 > 0, ∆ = Π2 − 2π2 > 0, −∗ ± + − ∆− = Π−∗ 2 − 2π2 > 0, ∆ = ∆ − ∆ > 0. Compatibility in period two arises if and only if the inequalities (23)-(24) in the proof of Lemma 4 are satisfied, and notice that they can be indifferently written as a function of dix , diy , or of dhx , dhy . Here we use the latter variables: F (dhx + dhy ) − Bdhx dhy ≥ ∆− E(dhx + dhy ) − Bdhx dhy ≥ ∆+ (33) The set of dhx , dhy which satisfies (33) is represented in figure 2 (after changing the labels of the axes to dhx , dhy ). Notice that • in case that firm h plays the same price in both markets, we have dhx = dhy ≡ dh , and then both inequalities in (33) are satisfied if and only if dh is included between a suitable lower bound d` and a suitable upper bound du , that is d` ≤ √ √ √ 2 − 2 + 2 + dh ≤ du with d` = max{ F − F B−B·∆ , E− EB−B·∆ } = E− EB−B·∆ and du = √ √ √ 2 − 2 + 2 − min{ F + F B−B·∆ , E+ EB−B·∆ } = F + F B−B·∆ ; • the southwest border of the feasible set is the graph of the curve dhy = q(dhx ), with q(dhx ) = ∆+ − Edhx E − Bdhx for dhx ∈ [0, ∆+ ] and q(d` ) = d` (34) E We now define several values of δ: (4t + 2γ)d` − 2γ − 4td2` δ̃(s, γ) ≡ , ∆− + ∆± d` δi (s, γ) = max{δ̃(s, γ), δ̄(s, γ)} p 2 (2t + γ) ∆± + 8t∆− − 4 2t (γ∆± + 2t∆− ) ∆+ δ̄(s, γ) ≡ (35) (∆± )2 2d` + 2 γt d` − 4d2` δh (s, γ) ≡ t (36) ∆− + ∆± d` Finally, for each s ∈ (s̄3 , 23 ), let γm (s) denote the unique γ ∈ (0, t) such that δi (s, γ) = δh (s, γ) (existence and uniqueness are proved in Step 1.3 of the proof below). The function δ` we have mentioned in Proposition 7 is such that δ` (s) = δh (s, γm (s)). Now we can state Proposition 7 in detail: 49 Proposition 7 Suppose that compatibility was chosen in the first period, and suppose that s ∈ (s̄3 , 32 ). Then a cornering equilibrium exists if and only if δt ≥ δh (s,γtm (s)) . Precisely, −∗ in the first period, one firm charges price p̄ = γm (s) − 2δ (Π+∗ 2 − Π2 ) in both markets, the other firm charges price p̄ + t in both markets, and the first firm corners both markets. In the second period, incompatibility is chosen at least by the firm who cornered the markets, which earns a total profit of 2γm (s) + δΠ−∗ 2 , whereas the other firm’s total profit δ (s,γ (s)) −∗ 3 m and Π+∗16/9 . is between Π+∗3/2 is δΠ2 . For each s ∈ (s̄3 , 2 ), the value of h t −2π +∗ −2π +∗ 2 2 2 2 In the proof we give below we consider first (in Step 2.1) deviations such that each firm plays the same price in both markets. In Step 2.2 we show that no profitable deviation exists even if the deviating firm can charge different prices in different markets. Step 2.1: For each s ∈ (s̄3 , 32 ), for no firm there exists a profitable deviation which induces compatibility in period two and such that the firm chooses the same price in both markets if and only if δ ≥ δh (s, γm (s)) Step 2.1.1: No profitable deviation exists for firm i if and only if δ ≥ δi (s, γ); moreover, δi = δ̄(s, γ) if γt ≥ 0.05921 Given the price pi played by firm i in both markets, the demand for product ij is di = 21 + 2t1 (p̄ +t − pi ) (for j = x, y), and inverting this relation we obtain pi = 2t + p̄ −2tdi . Hence, the profit of firm i is g i (di ) = 2di (pi + δz) + 2δπ2−∗ = −4t(di )2 + (4t + 2γ − δ∆± )di + 2δπ2−∗ for di ∈ [d` , du ]. Notice that a necessary condition for no profitable −∗ deviation to exist for firm i is g i (d` ) ≤ 2γ + δΠ−∗ 2 (2γ + δΠ2 is the equilibrium profit of firm i), which is equivalent to δ ≥ δ̃(s, γ). . We need to distinguish two cases, depending on the value of γt . • Case A: 0.05921 < γt ≤ 1. The derivative of g i is g i0 (di ) = −8tdi +4t+2γ −δ∆± ,.and ` g i0 (d` ) ≤ 0 if and only if δ ≥ 4t+2γ−8td ≡ δ̂(s, γ) . It turns out that δ̃(s, γ) < δ̂(s, γ) ∆± γ for each s ∈ (s̄3 , 23 ) since t > 0.05921.21 Therefore, when δ ≥ δ̂(s, γ), no profitable deviation exists for firm i. Now consider δ between δ̃(s, γ) and δ̂(s, γ). Then the profit maximizing di is ± 4t+2γ−δ∆± , which is larger than d` but smaller than du .22 We have that g i ( 4t+2γ−δ∆ )= 8t 8t Precisely, δ̃ < δ̂ is equivalent to 2(1 − 2l)DY + γt DX ≥ 2(l)2 DZ (,). 22 For each s ∈ (s̄3 , 32 ), we have that 43 < du and g i0 ( 43 ) , 2γ − 2t − δDZ < 0. 21 50 1 16t 2 (4t + 2γ − δ∆± ) + 2δπ2−∗ , which is smaller than 2γ + δΠ−∗ 2 if and only if −(∆± )2 δ 2 + 4 (2t + γ)∆± + 4t∆− δ − 4(2t − γ)2 ≥ 0 (37) At δ = δ̃(s, γ), the left hand side of (37) is negative since g i (d` ) = 2γ + δΠ−∗ 2 and 32t g i0 (d` ) > 0. At δ = δ̂(s, γ), the left hand side in (37) has value ∆± (γ∆+ + 2t∆− − 4td` ∆− − 2td2` ∆± which has the same sign as δ̂(s, γ) − δ̃(s, γ), hence it is positive. As a consequence, in case A no profitable deviation exists for firm i if and only if δ is at least as large as the smallest solution to (37), which is δ̄(s, γ), a number between δ̃(s, γ) and δ̂(s, γ). • Case B: 0 ≤ γt ≤ 0.05921. In this case δ̂(s, γ) ≤ δ̃(s, γ) for some s ∈ (s̄3 , 23 ), and then no profitable deviation exists for firm i if and only if δ ≥ δ̃(s, γ) (in this case δ̄(s, γ) < δ̃(s, γ)). If for some s ∈ (s̄3 , 23 ) the inequality δ̃(s, γ) < δ̂(s, γ) holds, then we can argue as for case A above that no profitable deviation exists for firm i if and only if δ ≥ δ̄(s, γ). Hence, in case B no profitable deviation exists for firm A if and only if δ ≥ max{δ̃(s, γ), δ̄(s, γ)}. Joining the two cases A and B, we conclude that no profitable deviation exists for firm i if and only if δ ≥ max{δ̃(s, γ), δ̄(s, γ)}, which we have defined as δi (s, γ) in (36). Step 2.1.2: No profitable deviation for firm h exists if and only if δ ≥ δh (s, γ) Given the price ph played by firm h in both markets, the demand for product hj is dh = 12 + 2t1 (p̄ − ph ) (for j = x, y), and inverting this relation we obtain ph = t + p̄ − 2tdh . Hence the profit of firm h is g h (dh ) = −4t(dh )2 +(2t+2γ −δ∆± )dh +2δπ2−∗ for dh ∈ [d` , du ]. Notice that a necessary condition for no profitable deviation to exist for firm h is g h (d` ) ≤ −∗ δΠ−∗ 2 (δΠ2 is the equilibrium profit of firm h), which is equivalent to δ ≥ δh (s, γ). The derivative of g h is g h0 (dh ) = −8tdh + 2t + 2γ − δ∆± , and g h0 (d` ) ≤ 0 if and only ` ` . It turns out that δh (s, γ) ≥ 2t+2γ−8td , hence no profitable deviation for if δ ≥ 2t+2γ−8td ∆± ∆± firm h exists if and only if δ ≥ δh (s, γ). Step 2.1.3: For each s ∈ (s̄3 , 23 ), there exists a unique γ which minimizes max{δi (s, γ), δh (s, γ)} For given s and γ, from Steps 1.1 and 1.2 it follows that no profitable deviation exists for firm i and firm h if and only if δ ≥ max{δi (s, γ), δh (s, γ)}. Then, for a given s, we choose γ to minimize max{δi (s, γ), δh (s, γ)}. It turns out that (i) δi (s, 0) > δh (s, 0) and δi (s, t) < δh (s, t) for each s ∈ (s̄3 , 23 ); (ii) δi (s, γ) is decreasing in γ and δh (s, γ) is increasing in γ. Hence, for each s ∈ (s̄3 , 23 ) we conclude that max{δi (s, γ), δh (s, γ)} is minimized 51 by the unique γ ∈ (0, t) such that δi (s, γ) = δh (s, γ). We denote this value with γm (s), and precisely γm (s) is the smaller solution to the equation w1 γ 2 +w2 γ +w3 = 0, with w1 = (∆− ∆± )2 , w2 = 2∆− ∆± (td` (1 + 2d` )(∆± )2 + 2t(1 + 2d` )∆− ∆± + 4t(∆− )2 ) − 8t∆± ∆+ (∆− 2 +∆± d` )2 , w3 = (td` (1 + 2d` )(∆± )2 + 2t(1 + 2d` )∆− ∆± + 4t(∆− )2 ) − 16t∆− ∆+ (∆− + 3 t), which implies that δi (s, γm (s)) = max{δ̃(s, γm (s)), δ̄(s, γm (s)) ∆± d` )2 . It turns out that γm (s) ∈ ( 92 t, 11 δ̄(s, γm (s)). Step 2.2: For each s ∈ (s̄3 , 32 ), let γ be equal to γm (s). Then for no firm there exists a profitable deviation which induces compatibility in period two, even though each firm can charge different prices in different markets Step 2.2.1: For each s ∈ (s̄3 , 23 ), let γ be equal to γm (s) and δ ≥ δi (s, γm (s)). Then for firm i there exists no profitable deviation which induces compatibility in period two, even though firm i can charge different prices in different markets Let pix , piy denote the prices charged by firm i in period one and dix , diy the demands for the products of firm i in period one. Arguing as in Step 1.1 we find pix = 2t + p̄ −2tdix and piy = 2t + p̄ −2tdiy , hence the profit of firm i under compatibility is δ δ Gi (dix , diy ) = −2t(dix )2 + (2t + γm (s) − ∆± )dix − 2t(diy )2 + (2t + γm (s) − ∆± )diy + 2δπ2−∗ 2 2 Notice that Gi is a concave function to maximize in the feasible set we denote with F: see i∗ i figure 2 in section 5.1.2. Hence, if there exists a critical point (di∗ x , dy ) of G in F (a point i i∗ where both partial derivatives of Gi vanish), then (di∗ x , dy ) is a global max point for G . ± ± 4t+2γm (s)−δ∆ (s)−δ∆ i∗ The proof of Step 1.1 reveals that (di∗ , 4t+2γm8t ) is a critical x , dy ) = ( 8t i∗ point for Gi in F if δ ∈ [δi (s, γm (s)), δ̂(s, γm (s))], but since di∗ x = dy , we know that it does not constitute a profitable deviation for firm i. However, for δ > δ̂(s, γm (s)) there exists no critical point for Gi in F. This suggests to inspect the boundaries of F, but it is immediate that the north-east boundary does not contain any maximum point, by the virtue of the remark in footnote 22. About the southwest boundary, consider any point (dix , diy ) which belongs to this boundary for which we know that G(dix , diy ) < 2γ + δΠ−∗ 2 when δ = δ̂(s, γm (s)). For δ > δ̂(s, γm (s)), the same inequality holds because the derivative of the right hand side with respect to δ, Π−∗ 2 , is greater than the derivative of the left hand side with respect to δ, 2π2−∗ − 12 (dix + diy )∆± . Therefore, for each δ ≥ δi (s, γm (s)), no profitable deviation exists for firm i even though 52 firm i can charge different prices in different markets. Step 2.2.2: For each s ∈ (s̄3 , 23 ), let γ be equal to γ(s). Then for firm h there exists no profitable deviation which induces compatibility in period two, even though firm h can charge different prices in different markets Let phx , phy denote the prices charged by firm h in period 1 and dhx , dhy the demands for the products of firm h in period one. Arguing as in Step 1.2 we find phx = t + p̄ − 2tdhx and phy = t + p̄ − 2tdhy , hence the profit of firm h under compatibility is δ δ Gh (dhx , dhy ) = −2t(dhx )2 + (t + γm (s) − ∆± )dhx − 2t(dhy )2 + (t + γm (s) − ∆± )dhy + 2δπ2−∗ 2 2 From the proof of Step 1.2 we know that there exists no critical point of Gh in F, hence each global maximum point of Gh belongs to the southwest boundary of F,23 which is the B·∆+ −E 2 0 curve dhy = q(dhx ) described in (34) Notice that q 0 (dhx ) = (E−Bd h )2 < 0 with q (d` ) = −1, and q 00 (dhx ) = by 2B(B·∆+ −E 2 ) 3 (E−Bdh x) x < 0. Hence firm h’s profit along the southwest border is given δ δ gq (dhx ) = Gh (dhx , q(dhx )) = −2t(dhx )2 +(t+γm (s)− ∆± )dhx −2t(q(dhx ))2 +(t+γm (s)− ∆± )q(dhx )+2δπ2−∗ 2 2 + for dhx ∈ [0, ∆E ]. First, we prove below that no profitable deviation exists for firm h when δ = δh (s, γm (s)). When δ = δh (s, γm (s)), we prove that gq is maximized at dhx = d` since gq0 (d` ) = 0 and gq is concave. Therefore no profitable deviation exists for firm h since gq (d` ) = δΠ−∗ 2 . Precisely, we find that δ δ gq0 (dhx ) = −4tdhx + t + γm (s) − ∆± − 4tq(dhx )q 0 (dhx ) + (t + γm (s) − ∆± )q 0 (dhx ) 2 2 and gq0 (d` ) = 0 since q(d` ) = d` and q 0 (d` ) = −1. Moreover, δ gq00 (dhx ) = −4t − 4t(q 0 (dhx ))2 − 4tq(dhx )q 00 (dhx ) + (t + γm (s) − ∆± )q 00 (dhx ) 2 (B · ∆+ − E 2 )2 B(∆+ − Ed)(B · ∆+ − E 2 ) δ ± 00 h = −4t − 4t − 8t + (t + γ (s) − ∆ )q (dx ) m (E − Bd)4 (E − Bd)4 2 23 The northeast boundary cannot include any global optimum point because δ 2 DZ, which is negative for each dhx ≥ 12 . A similar property holds for 53 h ∂G ∂dh y . ∂Gh ∂dh x = −4tdhx +t+γm (s)− and at δ = δh (s, γm (s)) we have (t + γm (s) − 2δ ∆± )q 00 (dhx ) < 0 because t + γm (s) − − +2t(d )2 ∆± δh (s,γm (s)) ± ` ∆ = (t+γm (s))∆ > 0 and q 00 (dhx ) < 0. The other terms in gq00 (dhx ) have 2 ∆− +∆± d` the same sign as −(E − Bdhx )4 − (B · ∆+ −E 2 )2 + 2B(E 2 − B · ∆+ )(∆+ −Edhx ). For any 3 1/3 + ·E E fixed s, this is a concave function of dhx which is maximized dhx = B − B1 E −B·∆ 2 and the maximum value of the expression is negative. Therefore gq00 (dhx ) < 0. + We have established above that gq (dhx ) ≤ δΠ−∗ for each dhx ∈ [0, ∆E ] when δ = 2 holds for δh (s, γm (s)). When instead δ > δh (s, γm (s)), the inequality gq (dhx ) < δΠ−∗ 2 + each dhx ∈ [0, ∆E ] because the derivative of the right hand side with respect to δ, Π−∗ 2 , −∗ is greater than the derivative of the left hand side with respect to δ, which is 2π2 − 1 ± h ∆ dx − 12 ∆± q(dhx ). 2 Proof of Proposition 8 The proof is straightforward and is omitted. Proof of Proposition 9 The proof is omitted as we explained the results in the main texts. Proof of Proposition 11 (i) In the compatibility equilibrium, the second-period consumer surplus in market j is given by CSj 1 = ∆v ˆ v̄ (vji v − i p+∗ 2 )dvj 1 − ∆v ˆ −∗ v e +p+∗ 2 −p2 −s v (vji − i p+∗ 2 )dvj −∗ ˆ ve +p+∗ 2 −p2 −s 1 i + (v e − p−∗ 2 − s)dvj ∆v v e 1 −∗ +∗ −∗ v e + p+∗ 2 − p2 − s − v (v − p2 − p2 − s + v) 2∆v e 1 −∗ −∗ v e + p+∗ + 2 − p2 − s − v (v − p2 − s) ∆v 2 1 ∆v ∆v −∗ 2 1 −∗ −∗ e +∗ +∗ −∗ = v − p2 + + p2 − p2 − s = v e − p+∗ (d2 ) = v e − p+∗ 2 + 2 + p 2 d2 2∆v 2 2 2 1 −∗ = v e − p+∗ 2 + π2 . 2 = v e − p+∗ 2 − 54 The first-peirod consumer surplus in market j is 2 Hence, the total consumer surplus in market j is v e − p∗1 − ´1 2 0 (v e − p∗1 − tx)dx = v e − p∗1 − 4t . t 1 −∗ + δ(v e − p+∗ 2 + π2 ). 4 2 (ii) In the incompatibility equilibrium, the second-period consumer surplus is given by 1 CS = (∆v)2 1 − (∆v)2 1 + (∆v)2 e = 2v − − 1 (∆v)2 ˆ ˆ v̄ v̄ (vxi + vyi − P2+∗ )dvxi dvyi v v ˆ 2v e −2s−P2−∗ +P2+∗ −v ˆ 2v e −2s−P2−∗ +P2+∗ −v ˆ v +∗ P2 ˆ 2ve −2s−P2−∗ +P2+∗ −v v (2v e − P2−∗ − 2s) + (∆v)2 e (vxi + vyi − P2+∗ )dvxi dvyi v v ˆ 2v e −2s−P2−∗ +P2+∗ −vxi ˆ 2v e −2s−P2−∗ +P2+∗ −vxi (2v e − P2−∗ − 2s)dvxi dvyi v (2v e − 2s − P2−∗ + P2+∗ − vxi − v)(2v e − 2s − P2−∗ − P2+∗ + vxi + v)dv 2v e −2s−P2−∗ +P2+∗ −v (2v e − 2s − P2−∗ + P2+∗ − vxi − v)dvxi v P2+∗ = 2v − (2v e − 2s − P2−∗ + P2+∗ − 2v)2 1 e −∗ +∗ e −∗ − − (2v − 2s − P2 + P2 − 2v) + (2v − 2s − P2 ) 2(∆v)2 3 −∗ +∗ 2 e (2v − 2s − P2 + P2 − 2v) + (2v e − 2s − P2−∗ ) 2(∆v)2 = 2v e − P2+∗ 1 (2v e − 2s − P2−∗ + P2+∗ − 2v)2 (2v e − 2s − P2−∗ + P2+∗ − 2v) + 3 2(∆v)2 2 2 = 2v e − P2+∗ + D2−∗ P2−∗ = 2v e − P2+∗ + Π−∗ . 3 3 2 The first-period consumer surplus is ˆ 1 ˆ 2 0 0 1−x 2 (2v e − P1∗ − tx − ty)dydx = 2v e − P1∗ − t. 3 55 Hence, the total consumer surplus is e 2v − P1∗ 2 −∗ 2 e +∗ − t + δ 2v − P2 + Π2 . 3 3 (iii) As enough details are given in the main text, the proof is omitted. Proof of Proposition 12 The welfare is simply obtained by adding consumer surplus to profits, which we presented in previous propositions. Regarding (iii), as enough details are given in the main text, the proof is omitted. Proof of Proposition 13 The proof is omitted as it is a minor adaptation of the proof of Proposition 3. 56
© Copyright 2024 Paperzz