Compatibility, Intellectual Property, Innovation and Welfare in Durable Goods Markets with Network E¤ects Thanos Athanasopoulos1 March 6 20162 Abstract Competition Authorities consider market leaders’refusals to support compatibility with competitors as a potential abuse of dominance with detrimental e¤ects on consumer welfare and competition. A prime example is the 2008 European Commission case against Microsoft regarding its refusal to support compatibility of its O¢ ce software product with those of competitors. This paper investigates dominant …rms’approach towards the compatibility of their durable network goods with that of a future innovative rival in the presence of overlapping generations of forward-looking consumers and the welfare e¤ects from dominant …rms’ refusals to support compatibility. I consider sequential, substitutable product innovations, where a rival can build on a dominant …rm’s existing knowledge. I …nd that in the limiting case when the market leader’s distribution of future innovation statewise dominates the rival’s, dominant …rms commit to support future compatibility with the rivals’future innovative products, while when the probability of the dominant …rm is more innovative than the rival is relatively high, credible dominant …rms currently announce they will support compatibility with the smaller rival in the future. On the normative side, future incompatible networks increase all groups ex-post surplus when the dominant …rm lacks commitment power, while if it currently chooses to be incompatible in the future, this may lead to speculations regarding its lack of future innovative ideas but leads to incompatible future networks 1 Department of Economics; University of Warwick. Email address: A.Athanasopoulos.1@warwick.ac.uk. 2 I would like to express my gratitude to Claudio Mezzetti and Daniel Sgroi, Joseph Harrington, Martin Peitz, Dan Bernhardt, Guillem Ordonez, participants’comments during presentations in the CRESSE Conference 2014, in the Royal Economics Society PhD Meetings and Job Market (2015), in the EARIE 2015 as well as in the Department of Economics at the University of Warwick. All the errors in this work are solely mine. 1 and a higher degree of competition that increases all groups’ expected consumer surplus, especially when network e¤ects are strong. 1 Introduction Should dominant …rms with durable network goods like markets for the applications in the software market industry have the obligation to provide technical compatibility information to direct competitors? This fundamental question lies at the intersection of competition and intellectual property law and di¤erent countries give di¤erent answers. In the European Union, market leaders must provide interoperability3 information to rivals and failure to do so is a potential violation of Article 102 (ex article 82) of the European competition law.4 In a nutshell, a refusal to license Intellectual Property may, in itself, constitute a breach of Article 102 if all the following conditions are met: a) access is indispensable for carrying on a particular business, b) it results in the elimination of competition on a secondary market and c) it may lead to consumer harm.5 A famous example comes from the 2008 European Commission case against Microsoft, which was related to the …rm’s refusal to provide competitors technical information about its O¢ ce suite so that they could craft interoperable software.6 The case followed a complaint7 from …rms-members of the ECIS [European Committee of Interoperable Systems] and was put on hold in December 2009 after Microsoft’s commitment to comply.8 In contrast, in the United States, a more laissez faire competition law is favoured. For example, Thomas Barnett of the United States Department of Justice argues that: "U.S. 3 We will use the terms interoperability and compatibility interchangeably throughout the paper. See http://ec.europa.eu/competition/antitrust/legislation/handbook_vol_1_en.pdf 5 See http://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:52009XC0224(01)&from=EN 6 See http://europa.eu/rapid/press-release_MEMO-08-19_en.htm 4 7 8 See www.techhive.com/article/124813/article.html See http://europa.eu/rapid/press-release_IP-09-1941_en.htm?locale=en 2 courts recognize the potential bene…ts to consumers when a company, including a dominant company, makes unilateral business decisions, for example to add features to its popular products or license its intellectual property to rivals, or to refuse to do so".9 Indeed, the U.S antitrust authorities conclude that "antitrust liability for mere unilateral, unconditional refusal to license patents will not play a meaningful part in the interface between patent rights and antitrust protections".10 In this article I investigate dominant …rms’approaches towards interoperability and the welfare e¤ects of refusals to supply interoperability that may occur in a laissez faire economy. More precisely, my work provides answers to the following questions: When do market leaders block interoperability with a rival future innovator and under what conditions would they support compatibility? If there is incompatibility, does this de-facto mean that it is socially undesirable? Could a laissez faire market converge to interoperability when it is socially e¢ cient? Are consumers better-o¤ in an economy that mandates compatibility? To answer these questions, I built a sequential game in which an innovative rival initially decides whether to invest in product quality. The rival’s choice depends on the importance of her idea and the dominant …rm’s future support or refusal to supply compatibility information due to the leader’s large installed base of consumers. This model …ts a common pattern in durable, technology goods markets, where a smaller rival may have valuable ideas emerging as follow-on, non-drastic substitutable innovations, after the dominant …rm’s invention hits the market in a Schumpeterian scenario of creative destruction. In the latter stage, the dominant …rm chooses its prices for its current product and whether to support compatibility with its competitor’s future product. It also may wish to invest in further improving product quality in the future period, where the …rms’R&D processes are realised and competitors set prices. I …nd that independently of the …rms’ability to exercise price discrimination, the dominant …rm chooses to support compatibility with the competitor’s potentially innovative future product if the probability distribution regarding the innovative step of its future prod9 10 See http://uk.reuters.com/article/2007/09/17/microsoft-eu-usa-idUKN1734095320070917 See http://www.justice.gov/atr/public/reports/236681_chapter7.htm 3 uct dominates the rival’s. A relatively less innovative dominant …rm rejects compatibility but suprisingly, incompatible future networks increase all consumer groups’expected consumer surplus especially when network e¤ects are strong, as future competition is …ercer when incompatibility prevails. Our conclusions cast doubts as to whether mandatory interoperability, while trying to support competition, allow technological advancement and protect consumers from abusive behaviour, may actually distort the market and lead to both socially undesirable results and customers’harm. This article is organized as follows: I next discuss related literature. Section 2 presents the model. In section 3, I characterize equilibrium outcomes when the economy operates under a laissez faire competition law toward Intellectual Property Rights holders. I then look at the e¤ects of dominant …rms’refusals to supply compatibility on consumer surplus. Section 4 discusses applications and concludes. Related Literature This work contributes to the literature regarding …rms’incentives toward compatibility with their competitors when network e¤ects are present. In a seminal article, Katz and Shapiro (1985) show that …rms with a larger installed base prefer to be incompatible with their rivals. In the same vein, Cremer, Rey and Tirole (2000) analyze competition between Internet backbone providers and predict that a dominant …rm may want to reduce the degree of compatibility with smaller market players. Similar results appear in Chen, Doraszelski and Harrington (2009), who consider a dynamic setting with product compatibility and market dominance. They …nd that if a …rm gets a larger market share, it may make its product incompatible. If, instead, …rms have similar installed bases, they make their products interoperable to expand the market. In a static environment, Viecens (2009) distinguishes between direct and indirect network e¤ects by studying platform competition between two …rms where users buy a platform and its compatible applications. By allowing for applications to be substitutes, complements or independent, she considers compatibility in two 4 dimensions: 1) compatibility of the complementary good, which she calls compatibility in applications, 2) inter-network compatibility with direct network externalities. She …nds that the dominant …rm never promotes compatibility in applications but both …rms …nd internetwork compatibility pro…table. In contrast to this literature, I focus on durable goods with direct network e¤ects. Both durability and network externalities are important features of most software products that are at the heart of this compatibility question. Contrary to the literature, I consider improvements in product quality and …nd that the dominant …rm may support compatibility when its future ideas dominate the competitors’future innovative step. When the probability that the dominant …rm is less innovative than the rival is relatively large, market leaders reject compatibility with the smaller rival. Absent innovative activity and although the literature has generically supported the view that incompatibility reduces welfare, as in Economides (2006), recent work in Chen, Doraszelski and Harrington (2009) has shown that in the long-run, a policy of mandatory compatibility is not preferable for consumers. Viecens (2009) concludes that compatibility in the applications may be harmful for users and social welfare, particularly when asymmetries are strong. Moreover, inter-network compatibility should not be supported by consumers. When …rms can innovate and durable network goods are considered, I …nd that compulsory compatibility results in higher future expected prices because of the lower degree of competition compared to the laissez-faire economy and both current and future consumers bene…t from future incompatible networks, especially when network e¤ects are strong. In the literature on sequential innovation, Scotchmer (1991, 1996), Scotchmer and Green (1996) and others study the case of single follow-on innovations. They focus on the breadth and length of patents needed to secure the initial innovator’s incentives to innovate when a second innovator threatens to innovate as well. They hold the view that patents for the …rst innovator should last longer when a sequence of innovative activity is undertaken by di¤erent …rms than when innovation is concentrated in one …rm. I am mainly interested in the interplay between Intellectual Property Rights protection and …rms’behaviour towards 5 compatibility. In contrast to these articles, I …nd that the …rst innovator will voluntarily o¤er compatibility to rivals because strategic pricing enables him to absorb more of the expected future pro…t when he anticipates that the rival is less likely to come up with an innovative product as valuable as its own. 2 The Model My objective is to provide insights into how dominant …rms’ short-run compatibility and pricing decisions regarding their durable network goods relate both to homogeneous, forwardlooking consumers, and to future products introduced by the rival. The industry I have in mind is the market for computer software applications highlighted in the introduction. Supply In the two-date model, the sequence of events in the supply side is as follows: initially, the dominant …rm has an installed base 0 of consumers and is marketing its product, which improves the level of quality from the previously sold version of quality q0 to a higher level q1 .11 The upgraded version of quality q1 is backward compatible, allowing its purchasers to interact with the users of the old version of quality q0 . In contrast, forward incompatibility prevents users of the initial version from editing and saving documents that are created with the upgrade.12 At the beginning of date t=1, a risk neutral rival can choose to incur a …xed cost F to create a future substitute product of quality q2 , which is distributed according to the commonly known continuous cumulative distribution function R, which is de…ned in the semi-closed interval (q1 ; q2 ]. At the end of date t=1, the market leader sets the prices for his product of quality q1 while he may also potentially innovate in the future period, and his future product quality q2D is distributed according to the privately known distribution 11 12 I follow Ellison and Fudenberg (2000) who also assume quality as a positive, real number. See Ellison and Fudenberg (2000) for a paper with backward compatibility but forward incompatibility. 6 function D; which is de…ned in the closed interval [q1 ; q2 ]; and its investment cost is F0 . If the dominant …rm has commitment power, it may also decide whether to support compatibility of its current and potentially innovative future versions with the rival’s future product, which is a binary decision. Otherwise, any announcement for its future action is perceived as credible by consumers: Following Malueg and Schwartz (2006), I assume that compatibility requires both parties’consent and cannot be achieved unilaterally using converters or adapters. In addition, in order to avoid potentially collusive behaviour, licensing of Intellectual Property is free.13 In a laissez faire market, the dominant …rm’s potential refusal to support compatibility does not, per se, hinder innovation, as there is an alternative route that allows the rival to innovate without using the dominant …rm’s network of existing customers. This assumption accords with the Microsoft O¢ ce case highlighted in the introduction: Microsoft’s refusal to o¤er compatibility did not, per se, prevent rivals from innovating as they could use the Open Document Format, which could allow product innovation. On the other hand, compatibility allows the smaller rival to use the dominant …rm’s network of existing consumers. At date t=2 and if the dominant …rm lacked commitment power, …rms make their compatibility decisions based on the realised states of their R&D processes. The smaller competitor sets its prices for its product while the dominant …rm is free to choose its pricing strategy if it fails to innovate; that is, it may either choose to set prices only at t=1 or change its prices in the future period. Consistent with software applications, I assume that the marginal cost of production is zero for all versions. Also note that I will consider both the cases when …rms have the power to price discriminate between the di¤erent consumer classes and when they cannot. 13 See Malueg and Schwartz (2006). 7 Demand Consumers are identical and have a per period unitary demand. Initially, there is a mass 0 of customers in the economy who have purchased the dominant …rm’s product of quality q0 at a past date (t=0), and new customers arrive in ‡ows of measure t at dates t=1,2. In order to exclude less interesting cases but also focus on realistic markets, I assume that 0 + 1 >2 2 (*) that is, I consider a relatively mature industry. At the end of date t=1, a measure 1 of new customers observes the dominant …rm’s prices for its products of quality q0 and q1 . Their utility is linear in income and partially dependent on direct network e¤ects captured by a parameter , while they incur a cost c to learn how to use a product. Thus, if they purchase the product of quality q1 and because of its backward compatibility with the old version of quality q0 , they join a network of maximum size at date t=1 and their total discounted utility is: q1 + ( where 0 + 1 0 + 1) c p11 + EV; is the market size at t=1, p11 is the dominant …rm’s price, where the …rst and second subscripts denote the time and the consumers’cohort, respectively and is the common discount factor for all the agents in the economy and thus, EV is their net expected discounted date t=2 utility.14 A measure 0 of old customers owns the product of quality q0 and observes the market leader’s upgrade price p10 for his product q1 : If a customer buys it, her total discounted bene…t is q1 + ( 0 + 1) cu 14 p10 + EV; Note that the utility function may not be necessarily linear in income (any monotonic transformation would su¢ ce). 8 where cu is the additional adoption cost she needs to incur (cu < c), while if she sticks to q0 , her total discounted utility is q0 + 0 x0 + 1 x1 0 + EV : A date t=1 customer’s overall bene…t depends on her forecast of how other customers make purchase decisions. At date t=2, a measure 0 + 1 of old consumers observes the competitors’ prices. If a customer that belongs to this group keeps the product of quality q1 and because of its forward incompatibility with any new product of quality q2 ; her utility is V = q1 + 2 x2 + ( 0 + 0 1 )x1 ; 0 where x2 and x1 are the new and old date t=2 customers’fractions that also own q1 at date t=2, respectively. If she purchases a product of higher quality q2i from the dominant or the rival …rm in the presence of compatibility, her net utility is V = q2i + cu p21 ; i = D; R due to backward compatibility of the product of quality q2 , after normalizing the date t=2 market size to unity where D,R stand for the dominant and the rival …rm, respectively. If compatibility is not supported, her utility by purchasing q2i is q2i + where 1 x2 and 1 2 (1 x2 ) + ( 0 + 1 )(1 0 x1 ) cu 0 p21 ; i = D; R; 0 x1 are the new and old date t=2 customers’fractions that also purchase q2i at date t=2, respectively. Although old consumers’ coordination problem has multiple equilibria, I assume that they coordinate on the Pareto optimum. 9 A measure 2 of new consumers also observes the competitors’prices. If they purchase the dominant …rm’s product of quality q1 , their utility is q1 + 2 x2 + ( 0 where x2 and x1 are the new ( 2 ) and old ( 0 0 + 1 )x1 1) 0+ c p12 ; date 2 customers’fractions that also own q1 , respectively. When the dominant …rm does not support compatibility with the rival’s new product of quality q2 , their bene…t from getting the product of quality q2i is q2i + 2 (1 x2 ) + ( 0 + 1 )(1 0 x1 ) c 0 p22 ; i = D; R; as they form a network only with those consumers who also own it. If they buy the product of quality q2i and compatibility is present, their utility is q2i + c p22 due to the product compatibility with any other product in the market. New customers’ purchasing decision resembles a coordination game and following the literature, I assume that they behave as a single player.15 All consumers make their purchasing decisions simultaneously and prefer a new product rather than an older version when they gain the same net utility from each of these two choices. 3 Market outcome Laissez faire Market Independently of the competitors’ability to exercise price discrimination between the di¤erent consumer groups: 15 See Ellison and Fudenberg (2000). 10 Proposition 1 (a) The dominant …rm currently announces that it will support compatibility if the probability of being more innovative than the rival is relatively large and commits to supporting compatibility in the limiting scenario that it knows with certainty it is more innovative than the rival. (b) When the probability that the dominant …rm is more innovative than the rival is relatively small, the dominant …rm rejects compatibility. Proof. See the Appendix Think …rst of the states when the dominant …rm’s new product is more innovative than the rival’s with certainty. The dominant …rm currently supports compatibility when …rms lack the power to price discriminate and the installed base of consumers 0 is large enough to allow the dominant …rm to currently serve all date t=1 consumers. This happens as the lesser degree of future competition between the competitors’ products when compatibility is present allows the dominant …rm to bene…t from higher future prices, while its current revenue is identical in the presence of compatible or incompatible future networks. The dominant …rm faces a trade-o¤ when its optimal choice is to currently serve only the new date t=1 consumers: although the lower degree of future competition increases its future revenue under compatible networks, its current price to the measure 1 of new date t=1 consumers is higher if it refuses to be compatible with the smaller rival. Thus, it still supports compatibility when the former e¤ect of higher expected pro…t ouweighs the latter of the decreased current bene…t.16 When the smaller competitor is more innovative than the market leader with certainty, the dominant …rm is always better-o¤ by currently refusing to be compatible. In the limiting case when the dominant …rm lacks any innovative ideas, it can bene…t both from higher prices at t=1 and potentially enjoy a positive expected payo¤ at date t=2. More precisely, if both …rms lack the power to price discriminate between the consumer groups, the dominant …rm may currently decide to serve either all or only the new date t=1 consumers, who can pay more than the upgrade price that old date t=1 consumers are ready to pay. For a relatively 16 A similar condition applies when …rms have the power to price discriminate. 11 small number of the installed base of consumers 0; it can currently charge the measure 1 of new date t=1 consumers more under incompatible networks because of the higher expected surplus that results from …ercer competition for any future product quality q2 . For a larger installed base, the market leader can charge all date t=1 consumers more at date t=1 when it refuses to be compatible, as it can currently charge a premium for less innovative future states than in the economy mandating compatibility. 17 18 More precisely, if the dominant …rm changes its prices at date t=2, an old date t=1 consumer’s net surplus in less innovative future states is zero when compatible networks arise, as her optimal future choice either after purchasing the product of quality q1 or after sticking to q0 is to purchase the product of quality q2 and pay the date t=2 equilibrium price. Under incompatible networks, although her optimal future choice after purchasing the product of quality q1 is to stick to it for relatively less innovative future states, her alternative future payo¤ if she currently sticks to the product of quality q0 is lower, as she purchases the product of quality q1 by paying the equilibrium date t=2 price. Her net surplus in these states of the world is positive and leads the dominant …rm to set a higher current price under incompatible networks.19 Note that if the dominant …rm rejects to be compatible and decides to set prices both at t=1 and t=2, it enjoys positive, rather than zero expected revenue under compatible networks, from the measure 2 of new date t=2 consumers for less innovative states. Overall, the dominant …rm commits to support compatibility when the random variable q2D dominates the variable q2R with certainty. Otherwise, its alternatives are either to announce its future support to compatibility or currently follow the path of incompatibility, as it knows that consumers will perceive lack of any announcement as lack of the dominant …rm’s any future innovative ideas. Thus, it currently announces its support to future compatibility when its expected bene…t under compatible future networks for the states it is 17 Obviously, the dominant …rm can bene…t from both the old and new date t=1 consumers if it can price discriminate betwen the di¤erent consumer groups. 18 A similar argument holds if the dominant …rm is less innovative and sells at t=2 a product of lower physical quality q2 than the rival. 19 A similar argument holds when the dominant …rm chooses not to change its prices at date t=2. 12 more innovative than the rival outweighs its current loss from currently rejecting incompatibility due to the foregone current income. In particular, in a laissez faire market where …rms cannot price discriminate between the di¤erent consumer groups, the dominant …rm decides to announce future compatibility for values of the probability that it is more innovative than the rival above a threshold that satisfy the following inequality: Pr ob(q2D +( 0 + ( where q = q1 + ( 1 )[Pr ob(q2D 0 + 0+ q2R )[ ( < q2R < q )(q1 1 ) Pr ob(q2R 1 ) + cu < q )](q1 20 2. 0 + 1) 2 ]+ q2R + ( 0 q2R + ( 0 + + 1) 1) + cu )) 2 + cu )] > 0 Note that if the dominant …rm currently announces it will support compatibility, date t=1 consumers in a laissez faire economy observe the price currently set by the dominant …rm and can estimate the probability : For example, if the dominant …rm chooses to serve all date t=1 consumers, the measure 0 + 1 of old date t=1 consumers can compare the current price p10 = + (1 where q+ 1 cu + Pr ob(q2R < q )[q1 ) Pr ob[q2D < q2R < q )(q1 q2R + ( q2R + ( 0 + 0 + 1) 1) 2 + cu ] + cu ]; is the probability of the dominant …rm failing to innovate and q = q1 q0 with the price the dominant …rm would have chosen absent any announcement: p10 = q+ 1 cu + Pr ob(q2R < q )[q1 20 q2R + ( 0 + 1) + cu ] There are similar conditions when …rms have the power to price discriminate between the di¤erent consumer groups. 13 and this comparison allows them to understand the value of the probability : In the limit, when the dominant …rm’s future product quality statewise dominates the rival’s with probability 1, its optimal date t=1 choice is to commit to supporting compatibility at t=2. Otherwise, the dominant …rm’s optimal choice is to announce its support to future compatibility and passes the probability of being more innovative into the date t=1 price charged to consumers, who understand the relative distribution functions of the future competitors. When the dominant …rm is less likely to be more innovative than the rival, the dominant …rm refuses to be compatible. Unlike the undisputable decision to currently refuse compatibility when it is less likely to be innovative than the rival, the dominant …rm faces a trade-o¤ when it decides whether to change its prices at t=2. While it bene…ts from setting prices twice because of the current higher revenue from the measure revenue from the measure 2 1 of new date t=1 consumers as well as the future expected of new date t=2 consumers, setting prices only at t=1 allows this …rm to charge the old date t=1 consumers more. In particular, by setting prices both at dates t=1 and t=2, it can exploit the higher degree of future competition in relatively less innovative future states and charge the measure 1 of new date t=1 consumers more in the present market. It also enjoys positive expected revenue from the measure 2 of new date t=2 consumers for less innovative future states. On the other hand, the dominant …rm bene…ts from setting a higher current price to the measure 0 of old date t=1 consumers when it only sets prices at t=1, as a consumer’s alternative expected payo¤ from sticking to the product of quality q0 is lower for less innovative states compared to the case it sets prices twice and thus, the dominant …rm can charge her more today to purchase its version of quality q1 . From the discussion above, it is reasonable to infer that the dominant …rm’s current choice of incompatibility may lead to speculations about its lack of any innovative ideas. But even so, do incompatible networks both when the dominant …rm can commit and when it lacks this power necessarily harm current and future consumers? 14 4 Welfare The following proposition summarises the welfare e¤ects from dominant …rms’ refusals to support compatibility and is independent of the competitors’ability to price discriminate: Proposition 2 (a) When the dominant …rm lacks commitment power, all consumer groups are better-o¤ when the dominant …rm fails to innovate and incompatible networks arise at t=2, (b) When the dominant …rm currently chooses to be incompatible and for strong network e¤ects, (1) incompatible networks increase both the new ( 2 ) and old ( 0 ) date t=2 consumers’ expected surplus when the probability that the rival’s future product quality is above a threshold is relatively high and the dominant …rm decides to set prices both at t=1 and t=2. (2) old date t=1 consumers’ expected surplus may be higher when incompatible networks arise and the dominant …rm decides to set prices only at t=1. Proof. See the Appendix When the dominant …rm currently announces its support to future compatibility, incompatible future networks increase all consumer groups ex-post surplus when the dominant …rm fails to innovate at t=2. The rationale is that independently of the competitor’s product quality, old date t=2 consumers are better-o¤ at t=2 under incompatible networks, as …ercer competition increases their surplus. In particular, old date t=1 consumers are better-o¤ when incompatible networks arise when their date t=2 bene…t from incompatibility surpasses their loss due to the higher price already paid at t=1 compared to the economy mandating compatibility.21 New date t=2 consumers are always better-o¤ under incompatible networks unless when the rival’s product is more innovative and …rms can price discriminate, in which case they are indi¤erent between compatible and incompatible networks. When network e¤ects are strong and the dominant …rm chooses the path of incompatibility at t=1, all consumers’expected surplus may be greater when incompatible networks arise. In particular, if the dominant …rm’s optimal choice is to change its prices at t=2 21 A similar argument holds for the new date t=1 consumers. 15 and …rms lack the power to price discriminate between the di¤erent consumer classes, both the measures 0 and 2 of old and new future expected consumers’surplus is higher under incompatible networks when there is a relatively high probability of reaching more innovative future states, which allow these customers to enjoy higher surplus when compatibility is rejected because of the …ercer future competition, while compatible networks increase their surplus for less innovative states. When …rms can exercise price discrimination, incompatible networks still increase 0 customers’surplus in more innovative future states and new date t=2 consumers are always better-o¤ provided that less innovative states are reached with a non-zero probability. If the dominant …rm announces its support to future compatibility but incompatible networks arise, all consumer groups are ex-post better-o¤. Thus, both old and new date t=2 consumers may be better-o¤ when incompatible networks arise. These results are very important, as they indicate that a policy of mandatory compatibility results in consumer harm for all consumer classes when network e¤ects are strong and the probability of reaching more innovative future products is relatively large. So, when network e¤ects are not negligible like in software markets, a laissez faire economy is likely to bene…t all consumer groups. In fact, when the probability distributions of future innovations place a relatively high weight on rather "important innovation", future and current consumers both bene…t from incompatible networks. When network e¤ects are weak, I obtain similar results although it is no longer straightforward that future consumers are always better-o¤ when incompatible networks arise. In particular, all consumer groups’surplus is higher in more innovative future states, as competitive forces decrease equilibrium future prices but for a smaller inventive step, compatibility leads to higher surplus. 16 5 Applications/ Discussion/ Future Research This article analyses dominant …rms’behaviour towards the compatibility of their durable network goods and the welfare e¤ects from their refusals to support compatibility. By using a sequential game, we give a smaller rival the ability to build on innovations previously introduced by the market leader. The …rst new result is that …rms with innovative ideas with more important relative magnitude compared to their competitors, currently embrace compatibility when their bene…t from the lower degree of future competition that allows them to charge more in the future outweighs the potential current loss from the lower price set to existing consumers. Dominant …rms with less innovative future ideas choose the path of incompatibility. In the limit, dominant …rms who lack any future ideas, may decide to change their prices in the future. Our normative analysis yields new results in industries where network e¤ects play an important role: I showed that when there is an alternative process that allows product innovation, incompatibility bene…ts both current and future consumers when network e¤ects are strong. The rationale behind this result is that future consumers bene…t from the …ercer degree of future competition that arises under incompatible network. But this is not the whole story. Current consumers also bene…t from incompatibille future networks when the smaller competitor’s future product quality is likely to be above a threshold. In such a case, current consumers’future expected surplus is higher under incompatibility when their bene…t for more innovative future states outweighs their loss in less innovative future states. The model applies in the scenario highlighted in the introduction: Microsoft sold Microsoft O¢ ce in 2007 and denied to support compatibility with its competitor’s products. Corel’s substitute product followed as well as the European Commission’s mandate to the technology giant for compatibility. I showed that when network e¤ects are strong, as it is considered in software markets, this and similar mandates may harm current and future consumers. The policy implication of these …ndings is that Competition Authorities should investigate 17 whether mandating compatibility may not necessarily bene…t consumers or even harm them. Instead, markets that allow unilateral refusals to supply interoperability information improve consumers’ welfare in industries that involve network externalities. More precisely, in an economy where network e¤ects are strong, all consumers are more likely to bene…t from incompatible networks, as the higher degree of future competition increases their surplus in more innovative states. Nevertheless, there are a number of issues that are important and are not addressed in this article. Perhaps the most intriguing would be to see how the results are generalised in a scenario when …rms’annoucements are perceived by consumers as credible with some probability and test empirically the predictions of the model. It may also be interesting to study the competitors’interoperability/investment decisions in the presence of stochastic demand and innovation. 18 6 Appendix 6.1 Proof of Proposition 1 (a) Strong network e¤ects/ The dominant …rm’s future product is more innovative than the rival’s. Compatibility/Price Discrimination Incompatibility No price discrimination Compatibility At t=2, the dominant …rm sells a product q2D , which is superior than the rival’s q2R : New date t=2 consumers purchase the product q2D and they are ready to pay a price such that: q2D + c p22D q2R + c p22R where p22D ; p22R are the dominant and the rival’s date t=2 prices, respectively. An old date t=2 consumer purchases the product q2D when q2D + cu p22D maxfq2R + cu p22R ; q1 + ( 0 + 1 )g: When …rms lack the power to price discriminate, the dominant …rm’s date t=2 price is p22D = 00 q = q2D q2R (p2R = 0) and all consumers purchase it. Going back to period t=1, an old date t=1 consumer decides to purchase the dominant …rm’s product of quality q1 instead of sticking to q0 when: q1 + ( 0 + 1) cu + q 2 + cu p22D 19 p10 q0 + 0 + q2 + cu p22D ; 00 where p22D = q : A new date t=1 customer purchases the product of quality q1 , given the price p10 set by the dominant …rm when: q1 + ( + 0 1) c + q2D + cu p22D p10 0: Thus, the dominant …rm may choose to serve all date t=1 consumers by setting a price p10 = q+ 1 cu (where a price p10 = q1 + ( 0 + q = q1 1) q0 ) or only the date t=1 new customers by setting cu : A su¢ cient condition for the dominant c + q2R + …rm setting a price such that all date t=1 consumers purchase the product of quality q1 is: ( 0 + 1 )( 0 q + 1 cu ) 1 [q1 which is satis…ed when the installed base + ( 0 + 1) c + q2R + cu ]; (1) is very large. 0 Incompatibility At t=2, the dominant …rm sells a product q2D , which is superior than the rival’s q2R : New date t=2 consumers purchase the product q2D and they are ready to pay a price such that: q2D + ( 0 + 1 )(2x1 1) c p22D q2R c p22R where p22D ; p22R are the dominant and the rival’s date t=2 prices, respectively. An old date t=2 consumer purchases the product q2D even when all other old date t=2 consumers purchase q2R or stick to q1 if: q2D + q2D + 2 cu p22D q2R + ( cu p22D q1 + ( 0 0 0 + + 1) 20 p22R ; 1 ): Thus, the dominant …rm’s optimal price at t=2 is p22D = and all date t=2 consumers purchase q2D : cu 00 q + 2 ( 0 + 1 ) (p22R = 0) Going back to period t=1, an old date t=1 consumer decides to purchase the dominant …rm’s product of quality q1 instead of sticking to q0 when: q1 + ( 0 + 1) cu + q2 + 00 where p22D = q + ( 2 cu 0 + 1 ): p22D p10 q0 + 0 + q2 + cu p22D ; A new date t=1 customer purchases the product of quality q1 , given the price p10 set by the dominant …rm when: q1 + ( 0 + 1) c + q2D + cu p22D p10 0: Thus, the dominant …rm may choose to serve all date t=1 consumers by setting a price p10 = q+ cu (where 1 a price p10 = q1 + ( 0 + q0 ) or only the date t=1 new customers by setting q = q1 1) c + q2R + 2 ( 0 + 1) cu : A su¢ cient condition for the dominant …rm setting a price such that all date t=1 consumers purchase the product of quality q1 is: ( 0 + 1 )( 0 q + 1 cu ) 1 q1 + ( which is satis…ed when the installed base 0 0 + 1) c + q2R + 2 ( 0 + 1) cu : (1) is very large. (b) The dominant …rm lacks innovative ideas (a)Strong network e¤ects ( 2 cu ) Price discrimination At t=2, the rival sets the same price to any user, who holds the dominant …rm’s either q0 or q1 : Compatibility Think about a consumer that belongs to the measure 0 of old date t=1 consumers when the dominant …rm can change its price at date t=2. This consumer decides independently 21 of other customers any time that she potentially makes purchasing decisions. Thus, she purchases the product of quality q1 when her total expected discounted payo¤ if she purchases q1 , given the price p10 ; exceeds her total expected payo¤ from sticking to q0 : q1 + ( where 0 + 1) 0 cu + q 2 + q1 and p21 = q = q2 cu 0 q + p21 p10 q0 + 0 + q2 + cu p21 ; cu; is the date t=2 price they will face to purchase 2 q2 . Thus, the date t=1 price set to these consumers is: p10 = Think now of the measure 1 q+ 1 cu : of new date t=1 consumers. The dominant …rm’s optimal price for its product of quality q1 is such that their total discounted expected payo¤ by purchasing it is at least equal to their outside opportunity: q1 + ( where p21 = 0 q + 2 0 + 1) c + q2 + cu p21 p11 0; cu and thus, the dominant …rm’s optimal price choice is: p11 = q1 + ( 0 + 1) c + q1 + ( 0 + 1 ): Note that the dominant …rm receives no revenue from the measure 2 of new date t=2 consumers. Also note that the prices above also hold for the case when the dominant …rm only charges prices at t=1. Incompatibility Let’s …rst think of the case where the dominant …rm sets prices both at t=1 and t=2. A consumer that belongs to the measure 0 of old date t=1 consumers purchases the product 22 of quality q1 at date t=1 even if all others stick to q0 when, given the dominant …rm’s price 0 p10 , her total expected discounted payo¤ is at least equal to her total expected discounted payo¤ from sticking to q0 : q1 + ( 1) cu + 0 + (q )(q1 + where q = q1 + ( 0 0 + q0 + + states [q2 < q = q1 + ( 1) (q )(q1 + ) + [1 cu + cu 0+ 2: 1 ) + cu (q )][q2 + p12 ) + [1 0 cu (q )](q2 + 0 p21 ] p10 0 cu p21 ); Note that if she sticks to q0 and for less innovative 2 ], she knows that everyone else keeps or purchases the product of quality q1 and thus, she purchases it at t=2, by paying the equilibrium price set to the measure 2 of new date t=2 consumers p12 = q1 q2 + ( 0 + 1 ). Thus, the dominant …rm’s optimal price choice at date t=1, p10 ; is p10 = q+ cu + 1 (q )[(q1 q2 + ( 0 + 1) + cu ]: Note that a su¢ cient and necessary condition for this price to be higher than the price under mandatory compatibility is that the probability of reaching the relatively less innovative future states is non-zero. Think of the measure 1 of new date t=1 consumers. The dominant …rm’s optimal price p11 is such that their total discounted expected payo¤ by purchasing q1 at t=1 is at least equal to their outside opportunity: q1 + ( where p21 = 0 + 1) 0 q + c+ (q )(q1 + ) + [1 ( 2 0 + 1) (q )](q2 + cu p21 ) p11 0; cu ; and thus, the maximum price the dominant …rm can charge them is: p11 = q1 + ( 0 + 1) c+ (q )(q1 + ) + [1 23 (q )][q1 + 2 ( 0 + 1 )] = q1 + ( + 0 1) c + q1 + ( 0 + 1) + (q ) 2 + [1 (q )] ( 0 + 1 )): Note that this price is always greater than the optimal price the dominant …rm can charge under an economy mandating compatibility (p11 = q1 + ( 0 + 1) c + q1 + Let’s think now of the dominant …rm’s optimal price to the measure 0 ( 0 + 1 )). of old date t=1 consumers when it only sets prices at t=1. A consumer that belongs to this group purchases the product of quality q1 even if all the other members of her group stick to q0 if: q1 + ( q0 + 0 0 + 1) + 0 where p21 = cu + (q )[q1 + ( (q ) maxfq0 ; q1 + ( 0 q+ 2 ( 0 + 1) 0 0 + + 1 )] 1) + [1 cu (q )][q2 + 0 p10 g + [1 0 cu p21 ] (q )][q2 + cu 0 p10 0 p21 ] cu . Note that the di¤erence compared to the case where the dominant …rm sets prices both at t=1 and t=2 is that the measure 2 of new date t=2 consumers always purchase the product of quality q2 (their net surplus by getting q1 with the date t=1 price p11 is negative). Thus, and although the measure the measure 0 + 1 2 always purchases q2 ; of old date t=2 consumers may or may not purchase q2 :Thus, the date t=1 optimal price set by the dominant …rm is: p10 = q+ 1 cu + (q )[ q + ( 0 + 1 )]; It is straightforward to see that a necessary and su¢ cient condition for this price to be higher than the equilibrium price under compatibility is that the relatively less innovative states 0 are reached with non-zero probability [ (q ) > 0]: Think of the measure 1 of new date t=1 consumers. The dominant …rm’s optimal price p11 is such that their total discounted expected payo¤ by purchasing q1 at t=1 is at least 24 equal to their outside opportunity: q1 + ( 0 + 0 where p21 = 1) c+ 0 q + [q )[q1 + ( 2 ( 1) c+ 0 + + 0 1 )] + [1 (q )](q2 + cu 0 p21 ) p11 0; cu ; and thus, the maximum price the dominant …rm 1) can charge them is: p11 = q1 + ( 0 + q1 + ( 0 + 1) (q )[q1 + ( c + q1 + ( 0 0 + + 1 )] 1) + [1 + [1 (q )(q1 + 2 ( (q )] ( 0 + 0 + 1 )) = 1 )): Note that a necessary and su¢ cient condition for the price to be higher than the optimal price under compatibility is that more innovative states are reached with non-zero probability. Also note that the optimal price set to the measure 1 of the new date t=1 consumers when the dominant …rm only sets prices at t=1 is smaller than the scenario when the dominant …rm charges prices twice when the less innovative states are reached with non-zero probability. When the dominant …rm sets prices both at t=1 and t=2, its expected second period revenue from the measure 2 of new date t=2 consumers is positive, as it receives posi- tive revenue for less innovative future states, and thus, when these states are reached with non-zero probability, the dominant …rm is better-o¤ by not supporting compatibility. The dominant …rm is also better-o¤ by setting prices both at t=1 and t=2 instead of only at t=1 to these consumers, as when it only sets price at t=1, it receives no revenue from these consumers. Thus, the dominant …rm is better-o¤ by setting price both at t=1 and t=2 when the following condition is satis…ed: 2 (q )[q1 q2 + ( 0 + 1 )] + 1 (q ) 2 + 0 (q )[q0 q2 + cu ] > 0; or equivalently, it chooses to do so when its expected bene…t from the measures 25 2 and 1 outweighs the loss from the lower price to the measure 0: No price discrimination Under an economy mandating compatibility, the rival …rm’s 0 only choice when network e¤ects are strong is to set a price p2 = q + 2 cu and both the new and old date t=2 consumers purchase it. Thus, the date t=1 optimal choices for the dominant …rm are identical to its choices when the rival has the ability to price discriminate. Thus, its optimal choices are: p10 = q+ 1 cu and p11 = q1 + ( for the measures 0 and 1 0 + 1) c + q1 + ( 0 + 1 ); of old and new date t=1 consumers, respectively. When incompatible networks arise, there is also one price choice for the rival …rm, as its optimal price choice is: p2 = 0 q + ( 2 0 + 1) cu for the more innovative future states, and thus, it gains positive expected payo¤ by serving all date t=2 consumers. Thus, the date t=1 optimal prices set by the dominant …rm are identical to its price choices when the rival has the power to price disciminate between the di¤erent consumer groups. If the dominant …rm sets prices only at date t=1, the optimal prices set at date t=1 to the measures 0 and 1 of olds and new date t=1 consumers are already calculated. (b) When network e¤ects are weak ( to the measure 0 < cu ), think …rst of a consumer who belongs 2 of old date t=1 consumers when compatibility is present. Given the dominant …rm’s current price, p10 , this consumer purchases the product of quality q1 instead of sticking to the product of quality q0 if: q1 + ( 0 + 1) cu p10 + 0 (q )[q1 + ( 0 26 + 1 )] + [1 0 (q )][q2 + cu p21 ] q0 + 0 + 0 (q )[q2 + 0 where q = q1 + cu 0 p22 = q ; p21 = cu 0 p22 ] + [1 (q )][q2 + cu p21 ] or equivalently, after substituting the equilibrium date t=2 prices, 2; 0 q + cu ; the dominant …rm’s optimal price at date t=1 is 2 p10 = Now think of the measure q+ 0 cu + 1 (q )(cu 2 ): of new date t=1 consumers. Given the dominant …rm’s date 1 t=1 price, p11 ; they decide to purchase the product of quality q1 if:q q1 + ( 0 + 1) c p11 + 0 (q )[q1 + ( 0 + and after substituting the date t=2 price, p21 = 1 )] 0 + [1 0 q + (q )][q2 + cu p21 ] 0 cu ; the dominant …rm’s optimal 2 date t=1 price choice p11 is: p11 = q1 + ( 6.1.1 0 + 1) c + q1 + ( 0 + 1 ): Incompatibility Think of a consumer that belongs to the measure 0 of old date t=1 consumers. Given 0 the dominant …rm’s price p10 ; she currently purchases the product of quality q1 instead of sticking to the product of quality q0 if: q 1 + ( 0 + 1 ) cu + q0 + 0+ where q (q )[q1 + ]+ [ (q ) (q )[q1 + = q1 + ( 0 cu p12 ]+ [ (q ) 0 + 1 ); q = q1 + ( (q )][q1 + ( 0 + 1 )]+ [1 (q )][q1 + ( 0 + 1 )]+ [1 0 + 1 )+cu 2 (q )][q2 + and p12 = q1 q2 + ( dominant …rm’s date t=2 equilibrium prices to the measure 27 0 (q )][q2 + 2 0 0 cu p21 ] p10 0 cu p21 ] 0 + 1) is the of new date t=2 consumers 0 and p21 = measure ( q + 0 + 1) ( 2 0 + cu is the smaller competitor’s date t=2 price to the 1) of old date t=2 consumers. After substituting these prices in the inequality above, the equilibrium date t=1 price p10 is: 0 p10 = q+ cu + 1 q+ (q )[q1 cu + 1 (q )(cu q+ as q = q1 + ( 0 + 1) q2 + ( + cu 2 Think now of the measure 1 + 2) 1) + cu ] + [ (q ) + [ (q ) 0 cu + 1 0 (q )(cu 0 > q = q 1 + cu (q )](cu (q )](cu 2) 2) 2) > > = p10 ; 2: of new date t=1 consumers. Given the dominant …rm’s date t=1 price, p11 ; they purchase the product of quality q1 and the highest price the dominant …rm can currently charge them is such that: q1 + ( 0 + 1 ) c+ (q )[q1 + ]+ [ (q ) (q )][q1 + ( 0 + 1 )]+ [1 (q )][q2 + 0 0 cu p21 ] p11 0; or equivalently, after substituting the date t=2 prices, 0 p11 = q1 + ( 0 + 1 ) c+ q1 + 6.1.2 (b) 1 ( 0 + 1 )+ (q ) 2+ [1 (q )] ( 0 + 1 ) > p11 = q1 + ( 0 + 1 ) c+ q1 + can postpone their purchase Think of the measure 1 of new date t=1 consumers when network e¤ects are strong and the dominant …rm sets prices both at t=1 and t=2. When compatibility is present, the highest price that the dominant …rm can charge them to purchase the product q1 instead of waiting until the future period to make their purchasing decision satis…es the equation: q1 + ( 0 + 1) c + q2 + cu 28 p21 p11 = (q2 + c p22 ); 0 where, given the price p11 ; p21 = q + date t=2 consumers and p22 = q + ( measure 2 cu is the equilibrium date t=2 price to the old 2 0 0 + 1) is the equilibrium date t=2 price to the of new date t=2 consumers. Thus, after substituting these prices in the above equation, one gets that the optimal price choice at date t=1. p11 , is given by the expression: p11 = q1 + ( 0 + q1 + ( 0 1) + c + [q1 + ( 1) c+ ( 0 0 + + 1] (q1 + 1) 2 2 c) = + c: Note that if the dominant …rm sets prices only at t=1, these customers’expected payo¤ from postponing their purchasing decision is zero, as there is no price competition at t=2 and their surplus by purchasing the product of quality q2 tomorrow is zero. Thus, the dominant …rm can set them a higher price if it only sets a price at t=1. This price is: p11 = q1 + ( 6.1.3 0 + 1) c + [q1 + ( 0 + 1 ]: Incompatibility Think of the measure 1 of new date t=1 consumers when network e¤ects are strong and the dominant …rm sets prices both at t=1 and t=2. When compatibility is not present, the highest price that the dominant …rm can charge them to purchase the product q1 instead of waiting until the future period to make their purchasing decision satis…es the equation: q1 + ( 0 + 1 ) c+ (q )(q1 + )+ [1 (q )(q1 + 0 where p11 = q1 q2 + ( 0+ c 1 ); (q )]fq2 + p12 ) + [1 p22 = 0 q + ( 0 cu [ q + 2 (q )]fq2 + 0+ 1 ); p21 = c 0 q + 0 ( 0 + 1 ) cu ]g p11 = p22 g; 2 ( 0+ 1) cu are the equilibrium date t=2 prices set to old and new date t=2 consumers. After substituting these prices in the equality above yields that the optimal date t=1 price that the dominant 29 …rm can set to 1 consumers is: 0 p11 = q1 + ( 0 + 1 ) c+ q1 + ( 0 + 1) c+ ( (q )(q1 + )+ [1 0 + 1) 2+ (q )][q1 +2 ( 0 + 1 )] c+ f (q )(q1 q2 + [ (q )(q2 + 2 +c)+[1 2 (q )] ( c)+[1 0 + 1 )g Note that these consumers’alternative payo¤ by waiting is higher when incompatible networks arise and may outweigh their higher actual payo¤ by purchasing the product q1 at date t=1. So, the price under compatibility may be higher when the probability of the relatively less innovative states is relatively large. If the dominant …rm sets prices only at t=1, the measure 1 customers’expected payo¤ from defering their purchasing choice to t=2 is zero, as there is no price competition at t=2 between the competitors’ products. Thus, the dominant …rm can charge these consumers more if it sets prices only at t = 1: This price is: p11 = q1 + ( 6.2 0 + 1) c+ (q )(q1 + ) + [1 (q )][q1 + 2 ( 0 + 1 )]: Proof of Proposition 2 (a) If …rms can price discriminate, the dominant …rm announces compatibility and incompatible networks prevail unambiguously at t=2 (when the dominant …rm fails to innovate and q2R < q ), the next expressions give the di¤erence between consumer suplus for the di¤erent customers’groups when incompatible networks arise and when compatibility is mandated: CS( 0 inc ) CS( p10 0 c) = q1 + ( 0+ p21 ] = fq1 + ( 1) 0 + cu + q 1 + 1) 0 p10 cu + q 1 + Pr ob(the d.f fails) Pr ob(q2R < q )(q1 30 [q1 + ( ( q+ q2 + ( 0 + 0+ 1) cu + q 2 + 1 cu g 1) + cu )g cu (q )](q1 + 2 fq1 + ( [q0 + 0 Pr ob((q2D < q2R < q )(q1 q2 + ( + ( q+ 1) cu + q 2 + cu + 1 1) 2 + cu )g cu ) [ q+ cu ]g = 2 Pr ob(the d.f fails) Pr ob(q2R < q )(q1 q2 + ( 0 + 1 ))] [q0 + 0 +q1 + Pr ob((q2D < q2R < q )(q1 = 0 q2 + ( 0 + 1) Pr ob(the d.f fails) Pr ob(q2R < q )(q1 2 Pr ob((q2D < q2R < q )(q1 q2 + ( 0 1) ( 0 + 1 )] + cu )g = 2 q2R + ( + 0 +q1 + 0 2 + 1) + cu ) + cu )g: Note that when the date t=2 bene…t from the larger network under incompatible networks outweighs the higher price incurred yesterday, old consumers are better-o¤ when incompatible networks arise.22 New consumers are always better-o¤ at t=2 under incompatibility, as the di¤erence in their surplus is: CS( q1 + c 2 inc ) [q1 CS( 2 c) = q1 + q2R + ( 0 q2R + 2 + c 1 )] c 0 p12 [q2R + (q1 + 2 [q2R + c c 0 ( q + ( p22 ] = 0 + 1 )] = c) > 0: Similarly, date t=1 consumers are better-o¤ when incompatible networks arise in the scenario when the dominant …rm refuses to be compatible after failing in its R&D e¤ort and the competitor sells a product q2D > q : Old date t=2 consumers are better-o¤ because of the …ercer competition and new date t=2 customers are indi¤erent. If …rms cannot price discriminate, all date t=2 consumers are still better-o¤ when incompatible networks arise. (b) When the dominant …rm chooses the path of incompatibility at t=1, think of the 22 Similarly, new date t=1 consumers’surplus is higher when incompatible networks arise at t=2. 31 measure of new date t=2 consumers. When …rms can price discriminate and under an 2 economy mandating compatibility, these consumers’total expected surplus is: (q2 + p22 )= [q2 + c q 0 ( + 0 1 )] = (q1 + c); where p22 = 2 0 q + ( 0 + c 1) is the date t=2 price to purchase q2 : Under incompatible networks and when the dominant …rm’s optimal choice is to set prices twice, these consumers’expected surplus is: p22 ) = (q )fq1 + = c [q1 (q )(q2 + q2 + ( c)+ [1 2 0+ (q )(q1 + 1 )]g+ (q )](q1 + [1 2 c p12 )+ [1 (q )][q2 + c (q )](q2 + q 0 ( 0+ c 1 )] = c): It is straightforward to see that a su¢ cient condition for incompatible networks to bene…t this class of consumers is that the less innovative states are reached with a non-zero probability. Think now of the measure of old date t=1 consumers. Under an economy mandating 0 compatibility, their expected total surplus is: q1 + ( p21 = q1 + ( q0 + 0 0 + q1 + + ( 1) 0 cu + ( q+ 0 + 1) cu cu ) + q2 + 1 p10 + q2 + 0 cu ( q + cu cu ) = 2 1 ): Under incompatible networks, these consumers’expected total surplus when the dominant …rm’s optimal choice is to set prices twice is: q1 + ( ) + [1 (q )][q2 + q1 + ( [1 q0 + ( 0 + 0+ 1) cu (q )]fq2 + 0 1) + + [1 (q )[ q 0 + p21 ] = f q+ 1 (q )(q2 + 2 ( 0 cu + 0 [ q + 0 + 2 (q )[q1 ( cu ) + [1 2 (q )] ( + 0 (q )(q1 + 1 ) + cu ]g + (q )(q1 + )+ 1) cu p10 + 0 cu cu 0 + 1) 0 + q2 + ( 1) 0+ cu ]g = (q )][q1 + 2 ( 0 + 1 )] = q0 + 0 + q1 + 1 )+ cu ]: It is also straightforward to see that old date t=1 consumers bene…t from incompatible networks for relatively more innovative states and they are overall better-o¤ when incompatible networks arise when their expected bene…t in the future relatively more innovative states o¤sets their loss in the less innovative future states. 32 The measure 1 of new date t=1 consumers are indi¤erent between compatible and incompatible networks, as their surplus is zero in both cases. If the dominant …rm’s optimal choice is to set prices only at t=1, the measures 2 1 and are indi¤erent between compatible and incompatible networks. Think of the measure of the old date t=1 consumers when incompatible networks arise 0 and the dominant …rm decides to set prices only at t=1. These consumers’expected total surplus is: q1 + ( 0 + where p10 = 1) q+ cu 1 0 p10 + cu + (q )[q1 + ( 0 + 1 )] (q )[ q + ( 0 + 1 )]; + [1 (q )][q2 + 0 0 p21 = q + cu ( 2 0 p021 ]; + 1) cu and thus, these consumers’expected surplus is: q0 + 0 + (q )q0 + [1 (q )][q1 + 2 ( 0 + 1 )]: (a2)When …rms lack the power to price discriminate, future consumers’expected surplus under compatibility is: [q2 + 1) c p22 ) = [q2 + 0 c ( q + cu )] = [q1 + ( 2 0 + c + cu ]: When there is incompatibility, future consumers’expected surplus is f (q )(q1 + f (q )[q1 + c p12 ) + (1 (q )(q2 + c (q1 q2 + ( 0 + 1 ))]+(1 (q )(q2 + f (q )[q2 + (q )(q1 + 2 ( 2 c] + (1 cu 0 c ( q+ 0 + 1) p021 )g = 2 ( 0 + 1 ) cu ))g = c + cu g: Note that for more innovative states, these consumers are better-o¤ when incompatible networks arise, while compatibility bene…ts them for less innovative future products. Think about the measure 0 of old date t=2 consumers. For installed base of consumers 33 0 above the threshold de…ned in (1), the dominant …rm’s optimal t=1 price choices are p10 = q+ 1 0 cu ; p10 = q+ cu + 1 (q )[q1 q2 + ( + 0 1) + cu ]; under compatible and incompatible networks, respectively. Then, their expected surplus under the two regimes is identical with a1 above.23 (b) When compatibility is present, the measure (q2 + 0 c) (q ) q (q1 + 0 expected surplus is: 0 [1 c) 2 ’s 0 (q )][ q + ( 0 [1 (q )] ( 0 + 0 + 1 )] = 1 ): These consumers’expected surplus under incompatible networks is: (q )fq1 + c [q1 q2 + ( 0 + 1 )]g+ [ (q ) (q1 + c) + (q )[ q 0 ( 0 (q )]fq2 + + 1 )] [1 2 c [ q (q )] ( 0 0 ( 0 + 1 )]g+ [1 + 1 ): 0 As q > q ; consumers bene…t from incompatible networks for more innovative states but for less innovative states (reached with under incompatible networks), they are better-o¤ (q when compatibility is madnated. Similarly, old date t=1 consumers’ expected consumer surplus when compatibility is mandated is: q1 + ( 0 + 1) cu q0 + p10 + 0 0 (q )[q1 + ( + q1 + ( 0 0 + 23 + 1) 1 )] + [1 0 (q )(cu 0 (q )][q2 + cu p21 ] = 2 ); Similarly, these consumers’expected payo¤ in more innovative states is larger when incompatible networks arise and the installed base of consumers is relatively smaller. 34 (q )]fq2 + while their expected surplus when incompatible networks arise is: q 1 + ( 0 + 1 ) cu + q0 + 0+ q1 + (q )[q1 + ]+ [ (q ) ( 0 + 1 )+ 0 (q )[ q + (q )](q1 + ( 0 + 1 ))+ [1 ( 0 + 1 ) cu ] 2 [ (q ) (q )][q2 + (q )](cu 0 2 )+ [1 Note that for more innovative future states, these consumers are better-o¤ when incompatible networks arise, while compatibility is bene…cial for relatively less innovative future states. References Cabral, L. and Polak, B (2007), "Dominant Firms, Imitation, and Incentives to Innovate", Working Papers 07-6, New York University, Leonard N. Stern School of Business, Department of Economics Cabral, L. and Salant, D. (2007), "Evolving Technologies and Standards Regulation", International Journal of Industrial Organization 36 (2014), 48–56. Carter, M. and Wright, J. (2003), "Asymmetric Network Interconnection", Review of Industrial Organization 22, pp. 27-46. Chen, J., Doraszelski, U. and Harrington, J. E. Jr. (2009) "Avoiding Market Dominance: Product Compatibility in Markets with Network E¤ects", RAND Journal of Economics Vol. 40, No. 3, Autumn 2009 pp. 455–485. Church, J. and Gandal, N. (2000), "System Competition, Vertical Merger and Foreclosure", Journal of Economics and Management Strategy, vol. 9, 1, pp. 25-51. Clements, M. T. (2004), "Direct and indirect network e¤ects: are they equivalent?", International Journal of Industrial Organization, vol. 22, pp. 633-645. 35 0 cu p21 ] p10 = (q )] ( 0 + Crémer, J., Rey, P. and Tirole, J. (2000), "Connectivity in the Commercial Internet", Journal of Industrial Economics, Blackwell Publishing, vol. 48, 4, pp. 433-72. Doganoglu, T. and Wright, J. (2006) "Multihoming and Compatibility", International Journal of Industrial Organization, vol. 24, pp. 45-67. Economides, N and White, L.J. (1994), "Networks and Compatibility: Implications for antitrust", European Economic Review, Vol. 38, pp. 651-662. Economides, N. (2006). "Public Policy in Network Industries", NYU, Law and Economics Research Paper No. 06-49. Economides, N. (1995), "The Economics of Networks", International Journal of Industrial Organization, Elsevier, 14, 673-699. Ellison, G. and Fudenberg, D. (2000), "The Neo-Luddite’s Lament: Excessive Upgrades in the Software Industry", RAND Journal of Economics, 31, 253-272. Graef, I. (2013), "How can Software Interoperability be achieved under European Competition Law and Related Regimes?", Journal of European Competition Law & Practice (2014) 5 (1): 6-19. Green, J and Scotchmer, S. "On the Division of Pro…t in Sequential Innovation", RAND Journal of Economics, Vol.26, (1995), p20. Farrell, J. and Saloner, G. (1985), "Standardization, Compatibility and Innovation", The RAND Journal of Economics, vol. 16, 1, pp.70-83. Katz, M. and Shapiro, C. (1985), "Network Externalities, Competition, and Compatibility", The American Economic Review, Vol. 75, 3, pp. 424-440. La¤ont, J.J., Rey, P. and Tirole, J. (1998), "Network Competition: I. Overview and Nondiscriminatory Pricing", The RAND Journal of Economics, vol. 29, 1, pp.1-37. 36 Lévêque, F. (2005), "Innovation, leveraging and essential facilities: Interoperability licensing in the EU Microsoft case", 28 World Competition, Issue 1, pp. 71-91. Llobet, G. and Manove,M. (2004), "Sequential Innovation, Network E¤ects and the Choice of Compatibility". Malueg, D. and Schwartz, M. (2006), "Compatibility Incentives of a Large Network Facing Multiple Rivals", The Journal of Industrial Economics, vol. 54, 4, pp. 527-567. Välimäki, M. (2005), "Software Interoperability and Intellectual Property Policy in Europe", European Review of Political Technologies, Volume 3, 2005. Samuelson, P. (2010), "Copyright Interoperability Rules in the European Union and the United States", 34(3) European Intellectual Property Review (2010). Scotchmer, S. "Standing on the shoulders of giants: Cumulative Research and the Patent Law", Journal of Economic Perspectives, Vol 5 (1991), pp 29-41. Scotchmer, S. "Protecting Early Innovators: Should Second Generation Products Be Patentable?", RAND Journal of Economics, Vol.27 (1996), p.322. Scotchmer, S. Innovation and Incentives. Cambridge, Mass.: MIT Press, 2005. Viecens, M. (2009), "Compatibility with …rm dominance", Review of Network Economics, Volume 10, Issue 4, Pages –, ISSN (Online) 1446-9022. 37