Dynamic Vertical Foreclosure

Transcription

1 Dynamic Vertical Foreclosure Chiara Fumagalli Massimo Motta First version: March 31, 015 This version: May 1, 015 VERY PRELIMINARY - PLEASE DO NOT CIRCULATE Abstract The aim of this paper is to emphasize that the rational of vertical foreclosure can have a dynamic component. We consider a vertically integrated incumbent that faces the threat of entry in the downstream market in the current period and in the upstream market in the future period. In this setting, we show that the upstream monopolist may have an incentive to foreclose a more efficient downstream rival even though it sacrifices current profits. In the current period, in which only downstream entry can take place, the incumbent would find it more profitable to supply the more efficient downstream competitor, because it would be able to extract sufficient rents from it. However, by engaging in refusal to supply, the vertically integrated incumbent will affect the future market structure and will earn larger future profits. In particular, lack of suitable access to the input may deprive the rival of the critical profits (or, more generally, of the critical scale, customer base, or reputation) it needs in order to be successful in the downstream market. In turn, weakened competition for input procurement, by reducing the post-entry profits of the prospective upstream competitor, may discourage future upstream entry and prevents the incumbent from entirely losing its future profits. The paper also shows that if future entry cannot be discouraged, the incentives to engage in vertical foreclosure are indeed reinforced. In this case foreclosure is motivated by the incumbent s intent to transfer its monopoly power from the upstream to the downstream market. Keywords: Inefficent foreclosure, Refusal to supply, Scale economies JEL Classification: K1, L1 Comments by Luis Cabral, Philippe Choné, Joe Farrell, Laurent Linnemer, Fiona Scott-Morton, David Spector, Ralph Winter, as well as by participants to the UBC Summer Conference on Industrial Organization (Vancouver), and by seminar audience to the Paris School of Economics (Paris) are gratefully acknowledged. Massimo Motta is currently Chief Economist at DG Competition, European Commission. The views expressed here are formulated in a personal capacity, and do not represent the position of the European Commission. Università Bocconi (Department of Economics), CSEF and CEPR ICREA-Universitat Pompeu Fabra and Barcelona Graduate School of Economics

2 1 Introduction Vertical foreclosure refers to situations in which a vertically integrated firm which dominates one market acts in such a way to exclude (or marginalize) rivals in vertically related markets. For example, a monopoly owner of a necessary input may refuse to sell it to the downstream competitors and reserve it all for its own downstream affiliate. The upstream monopolist may also resort to more subtle ways to foreclose the activity of the downstream rivals, for instance by reducing the quality of the input supplied to rivals, by degrading interconnection, or by delaying the input provision. Alternatively, the vertically integrated firm could set a combination of high upstream (or wholesale) prices and low downstream (or retail) prices such that competitors cannot profitably operate in the downstream market, a practice known as margin squeeze. For a long time economists were unable to provide a solid rationale for vertical foreclosure and to find convincing counter-arguments to the serious doubts cast by some scholars the so-called Chicago School critique to its validity. In particular, the Chicago School pointed out that the owner of an essential input has the ability to exclude downstream rivals, but rarely will it have the incentive to do so, especially in the presence of more efficient downstream rivals: this is because the control of the bottleneck input enables the upstream monopolist to earn higher profits by serving efficient downstream rivals, and extracting rents from them, rather than excluding them. The Chicago School view made a valid and important point which had the beneficial effect of forcing industrial economists to put the foreclosure argument on firmer ground. Indeed, it is only since the late 190s that modern industrial organization theory has been able to find rigorous explanations for why the incentive to exclude may exist. The common trait of these theories, that we will discuss more in detail at the end of this Section, is that they all rely on imperfect rents extraction: they have identified the circumstances under which the upstream monopolist is able to extract little rents from the downstream rivals and for this reason it may find it more profitable to foreclose them and to monopolise the final market through the own, though less efficient, affiliate. These theories have a static perspective. The aim of this paper is to emphasize that the rational of vertical foreclosure can have a dynamic component. We consider a vertically integrated incumbent that faces the threat of entry in the downstream market in the current period and in the upstream market in the future period. (However, the same mechanism would apply if the scope for current entry is upstream and future entry may take place in the downstream market.) In this setting, we show that the upstream monopolist may have an incentive to foreclose a more efficient downstream rival even though it sacrifices current profits. In the current period, in which only downstream entry can take place, the incumbent would find it more profitable to supply the more efficient downstream competitor, because it would be able to extract sufficient rents from it. However, by engaging in refusal to supply, the vertically integrated incumbent will affect the future market structure and will earn larger future profits. In particular, lack of suitable access to the input may deprive the rival of the critical profits (or, more generally, of the critical scale, customer base, or reputation) it needs in order to be successful in the downstream market. In turn, weakened competition for input procurement, by reducing the post-entry profits of the prospective upstream competitor, may discourage future upstream entry and 1

3 prevents the incumbent from entirely losing its future profits. In this case foreclosure is therefore motivated by the incumbent s intent to protect its monopoly position in the upstream market from future entry threats. The paper also shows that if future entry cannot be discouraged (for instance because upstream entry entails very low fixed setup costs relative to the profits the entrant could make even in the worst case scenario), the incentives to engage in vertical foreclosure are indeed reinforced. In this case the incumbent knows that it will lose the upstream monopoly in any event, but it may want to foreclose the downstream competitor in the current period so as to a downstream monopoly in the future and use such position to extract rents from the more efficient upstream entrant. In this case foreclosure is motivated by the incumbent s intent to transfer its monopoly power from the upstream to the downstream market. We will discuss in the Conclusions the crucial ingredients for our theory of harm to be applied and we we will illustrate in the next Section some recent cases in which our theory might rationalize the conduct of the dominant firm. We close the Introduction with a discussion of the related literature. This paper is related to the literature that, since the late 190s, has rationalized the incentive to engage in vertical foreclosure. As we mentioned earlier, in that literature, it is the inability of the upstream monopolist to extract sufficient rents from the more efficient downstream competitors that generates the incentive to foreclose its activity. This inability to extract rents may be due to the presence of sectoral regulation which restricts the upstream monopolist s freedom to contract with downstream rivals. 1 Another source of imperfect rent extraction is the so called commitment problem, first proposed by Hart and Tirole (1990) and recently applied by Reisinger and Tarantino (01) to a context in which a vertically integrated incumbent faces a more efficient downstream rival. One way to explain the intuition behind the commitment problem is the following. As suggested by the Chicago School, the vertically integrated incumbent would like to extract the entire rents produced by the more efficient downstream competitor, through a suitable choice of contracts. For instance it may want the more efficient downstream rival to set all or most of the production; to set (industry) profit maximizing prices and to earn high profits. Such profits would then be extracted by the upstream monopolist by requiring a large (fixed) payment for the input in exchange. However, if the downstream rival feared that the incumbent would in parallel use its downstream affiliate to compete for consumers (i.e. that it would behave opportunistically, reneging on the incumbent s promise to limit its activity in the downstream market) its willingness to pay for the input would decrease, since expected competition from the incumbent s downstream affiliate would decrease the rival s expected profits. In turn, this would limit the ability of the incumbent to extract profits from the independent downstream rival and may make a foreclosure strategy potentially more profitable. 3 1 See Jullien, Rey and Saavendra (01) and Fumagalli, Motta and Calcagno (forthcoming), for models that study the conditions under which regulation of the wholesale price induces a vertically integrated incumbent to engage in margin squeeze. See also the work by O Brien and Shaffer (199), McAfee and Schwartz (199), Rey and Verge (00). See also Rey and Tirole (007) for an insightful review of this literature. 3 Interestingly, Reisinger and Tarantino (01) shows that the inability not to operate the downstream affiliate does not necessarily lead to complete foreclosure of the independent rival. If the efficiency gap between the incumbent s affiliate and the independent rival is not too large, the incumbent engages in partial foreclosure: it does supply the independent rival, but at less favourable terms (i.e. at a higher wholesale price than the own affiliate. Instead, if the

4 The incumbent s ability to extract rents may be significantly reduced also by the presence of uncertainty (for instance regarding the costs of downstream firms) and of risk-adverse downstream firms (see Hansen and Motta, 013). Finally, if the incumbent faces some competition in the provision of the input, then the incentive to deny the input to independent downstream firm may come from the so-called raising rivals cost argument, due to Ordover et al. (1990): the incumbent s withdrawal from the wholesale market will make the downstream rival pay a higher price for its input requirements, because such inputs will be bought from the independent upstream firm, which will enjoy stronger market power over the independent downstream firm once the integrated incumbent commits not to serve the input to it. In this case the downstream competitor is not completed excluded from the market, but it faces higher input costs, which makes it less competitive and aggressive in the downstream market, to the benefit of the incumbent s downstream profits. As we emphasized earlier, we depart from this literature because in our paper the incentive to engage in vertical foreclosure does not stem from imperfect rent extraction. Indeed, in the current period, when entry occurs only in one of the vertically related market, the incumbent sacrifices profits by engaging in refusal to deal. However, vertical foreclosure affects the future market structure and allows the incumbent to make larger profits in the future. This paper is also related to a large part of the recent economic literature on exclusionary practices. A common feature of this literature is that it refers to situations in which a competitor, if deprived of crucial sales, buyers, profits, becomes a poor rival for other buyers, in other markets, in other periods. There are several sources of such a vulnerability. An obvious one is the existence of scale economies from the supply side: the competitor needs to cover a fixed cost or more generally to reach a given scale in order to be successful, and access to a sufficiently large number of buyers/markets is crucial to achieve that scale. In this context, should the competitor be denied access to such a critical set of buyers, then it would desist from entering the market and supplying the remaining buyers or it would supply them at high costs. In this context several works have shown that an incumbent firm can take advantage of this vulnerability and, by securing the critical buyers or markets, it can profitably prevent the entry of a more efficient rival, discourage its expansion beyond some market niche, or induce the rival s exit. These works also have shown that the incumbent can make use of several different practices to attract the critical sales (or buyers, or markets) and to exclude: exclusive dealing (as shown, for instance, by Rasmusen et al. 1991, Bernheim and Whinston 199 and Segal and Whinston 000, followed by several other authors); 5 but also low prices to some early buyers and markets (i.e. predation, as shown by Bolton and Scharfstein 1990, Cabral and Riordan 1997 and Fumagalli and Motta 013), 6 price efficiency gap is large enough, the incumbent finds it optimal to offer to the independent rival a wholesale price which is indeed lower than the one paid by the own affiliate. A similar mechanism applies if scale economies arise from the demand side and the firm needs to achieve a critical base of customers so as to generate sufficient network externalities and make the quality of its product large enough to be an effective competitor. Alternatively, there may be inter-temporal scale economies, whereby limited current production implies little accumulation of learning and causes high production costs in the future. One may also think of financial market imperfections: a firm that does not make enough profits and does not retain enough liquidity may be unable to overcome financial constraints and fails to obtain external funds to finance its investment and to continue its activity in the future. 5 See for instance the extensions to buyer-retailers as in: Fumagalli and Motta (006); Abito and Wright (00); Simpson and Wickelgren (007). See also the experimental paper by Spier and Landeo (009). 6 The possibility to exclude by fixing low prices to early buyers arises also in Bernheim and Whinston (199) and in 3

5 discrimination including various types of rebates (as shown by Karlinger and Motta 01), and tying between a monopolised good and a complementary product for which there exists potential competition (Carlton and Waldman 00). In this paper we focus on refusal to supply a monopolized input to an independent downstream rival. Within the literature on exclusion based on scale economies, our paper is very close to the dynamic leverage theory of tying, due to Carlton and Waldman (00). In their model, an incumbent firm that is a monopolist of a primary product, and also sells a complementary one, may want to bundle the two products, thereby excluding a rival producer of the complementary product. This is done to protect the monopolistic position in the primary market, where the incumbent faces future entry. We will discuss at the end of Section 3 the additional insights that our analysis of vertical foreclosure brings as compared to Carlton and Waldman (00). The paper proceeds as follows. In Section we will present some recent cases of vertical foreclosure which may be rationalised by our theory. Section 3 sets up our baseline model and illustrates our results. In Section we develop some extensions to the baseline model, while Section 5 concludes the paper and draws some policy implications. Recent Cases.1 Telefónica In Telefónica, the Commission found that the Spanish telecoms incumbent abused its dominant position by way of a margin squeeze in the Spanish broadband market, from September 001 to December Telefonica was the unique operator having a local access network, i.e. a network that reaches final users. Alternative operators wishing to provide services throughout Spain had no other option than buying wholesale services from Telefonica. Access to wholesale services can occur in three different ways, which differ in the intensity of the own investment required to supply the retail market: (i) Local loop unbundling (ULL); it allows alternative operators to use only the very final portion of the incumbent s network, but requires them to make a significant investment in their network. This solution ensures alternative operators to gain flexibility in the provision of the services and ability to differentiate their services from Telefonica s. (ii) Regional wholesale access; it requires alternative operators to invest less as compared to the case of ULL, but the investment is still substantial. Alternative operators can rely on less flexibility in the provision of the services. (iii) National wholesale access. Alternative operators forego most of the network investment, but they do not have any flexibility in the provision of services at retail level. In this context, the Commission identified as relevant retail market the broadband services to the mass market of residential and non-residential users; as relevant wholesale markets the broadband access at regional and national levels. The Commission considered that access through the local loop unbundling was not in the same market as national and regional wholesale access because switching from national and regional access to local loop unbundling would require heavy investments and would be costly and time consuming. The margin squeeze investigation therefore disregarded local Carlton and Waldman (00). 7 Decision 00/C 3/05 [007 OJ C 3/06.

6 loop unbundling and focused on a comparison between wholesale and retail broadband access prices.,9 Telefonica was found dominant in all these markets: it was found to be a monopolist in the wholesale regional access market and to have over % of the wholesale national market (competitors gaining access at regional level could or through ULL can make a national wholesale offer to other internet service providers). At the retail level, Telefonica s share of end-users ranged between 5% and 5% during the period of the abuse found (its market share by revenue consistently exceeded 60% according to the Commission). Looking at the regulatory environment, Telefonica retail prices were regulated up until November Following that date, retail prices were liberalised. As for wholesale products, national wholesale access was not price-regulated during the period of abuse. Regional wholesale access was regulated by price caps set on a retail-minus basis (i.e. Telefonica could not charge for wholesale more than (100-x)% of what it charged at the retail level for the same product, where x was the proxy for the incremental cost of providing downstream services and any network elements sought by the access-seeker). 11 Telefonica s retail prices remained fixed (in nominal terms) throughout the whole period from September 001 to December 006; however broadband speeds were upgraded for each retail service class (e.g. basic package, premium package). The Commission (borrowing the remark from the Spanish regulator) also noted how a key competition tool in the retail market were promotional offers, such as discounts or waivers of connection fees, subscriptions fees, equipment fees or promotional gifts. Likewise, bundled offers started to emerge in the market: double play (fixed line and internet) and triple play (which also included TV over broadband). The former was offered by both the incumbent and the competitors, the latter mostly by Telefonica, since competitors lacked the necessary infrastructure to provide it nationwide. To establish whether there was a margin squeeze, the Commission applied the as efficient competitor test: it checked whether the margin between Telefonica s retail price and wholesale price would allow an equally efficient competitor to cover the downstream LRAIC, i.e. to cover the average additional costs that one has to bear to operate downstream, on top of access costs. LRAIC included costs for additional network elements needed to provide retail services, recurrent costs of Internet Service Providers, customer acquisition costs (advertising, incentives and commission to the sales network) and a share of common costs (i.e. costs associated to the commercial structure). 1 The By contrast, according to the Commission, whether wholesale regional access was in the same market as wholesale national access would not have changed the conclusion of its assessment. 9 Telefonica, instead, argued that local loop unbundling should be in the same market as regional and national wholesale access. 10 Telefonica had to propose a retail price to the regulator for approval. 11 One may wonder how margin squeeze can occur given that the wholesale price is determined using a retail-minus system. The Commission argued that regulated prices had been determined in 001 on the ground of ex-ante estimates of costs that resulted lower than the costs actually incurred. Moreover, it argued that the price indicated by the Spanish regulator was a maximum price that Telefonica was free to decrease. Hence, Telefonica had the responsability to check whether the combination of the retail prices and regulated price was replicable by an equally efficient competitor and to take into account that the forecasts used ex-ante to determine the regulated wholesale price was not confirmed by the actual evolution of the market. One may therefore read the proceedings against Telefonica as way of the Commission making up for an incorrect regulation. The regulator started imposing cost-oriented offers on both wholesale products from The details of the exact determination of the LRAIC were the object of hot debates. For instance, Telefonica argued that the costs related to the commercial structure of the company should be excluded from the LRAIC. 5

7 Commission opted for the current cost accounting method (i.e. attributing a value to an asset equal to what an entrant using current technology would pay to build it). To assess the viability of an as-efficient competitor, the Commission engaged in two profitability tests: first, it sought to establish whether the difference between the incumbent s retail and wholesale pricews allowed to cover (downstream) costs, on an annual basis. Second, it applied discounted cash flow analysis, considering total discounted revenues and costs over the period of the abuse. This methodology allows to account for the fact that the provision of a new service, such as broadband services, might entail initial losses. 13 Using either test, the Commission concluded that a margin squeeze had occurred, and imposed a fine of 151.9m euros on Telefonica. In the implementation of the test the Commission compared separately the margin and the downstream costs of an equally efficient competitor relying entirely on national wholesale services and those of an equally efficient competitor relying entirely on regional wholesale services. Telefonica objected that this would not have been the profit-maximising strategy for an entrant: entrants would typically cherry-pick customer categories and regions and would choose an optimal mix of regional access, national access and of local loop unbundling. The Commission justified its approach on the ground of the following argument. As shown by the experience of other countries in Europe and in the world, the most profitable entry strategy for an alternative operator is the investment in the own network combined with local loop unbundling. However, due to the risks involved in such a high sunk investment, alternative operators are likely to follow a step-by-tep approach to continuously expanding their customer base and infrastructure investments. In particular, when climbing up the investment ladder, alternative operators seek to obtain a minimum critical mass, in order to be able to make further investments. (Para. 39 of the Decision) The first step of the investment ladder is occupied by an operator whose strategy consists in targeting a mass market (thus involving considerable marketing and advertising expenditure), but who is merely acting as a reseller of the ADSL access product of the vertically integrated provider (the incumbent). As its customer base increases, then the alternative operator makes further investment. In a further step, it may even seek to connect its customers directly (local loop unbundling). Thus the progressive investments take the alternative operator progressively closer to the customer, reduce the reliance on the wholesale product of the incumbent, and increasingly enable it to add more value to the product offered to the end-user and to differentiate its service from taht of the incumbent. (Para. 17 of the Decision) This argument explains why the Commission decided to make a margin squeeze test separately for regional and national wholesale access: It is therefore necessary that there should not be any margin squeeze in relation to any step of the ladder, i.e. in relation to any wholesale product. If there was such a margin squeeze, new entrants that are climbing the ladder of investment would be foreclosed. 13 The Commission expressed some reservations about the latter methodology, though, as inter alia finding a positive net present value is consistent with the incumbent making short-term losses and large long-term gains, achieved through exclusionary behaviour and an increase in market power. 6

8 This argument also provides a possible rational for Telefonica s incentive to engage in margin squeeze, consistent with the dynamic theory we are proposing in this paper. Only if it obtains a critical size in the retail market, an alternative producer (relying on national or regional access) would be able to make the investment necessary to reach customers directly through local loop unbundling and to gain independence from the services provided by the incumbent. By engaging in margin squeeze, the incumbent is preventing alternative operators from achieving the critical size that would justify investment in their own infrustructure thereby discouraging them from investing upstream. Margin squeeze can therefore be interpreted as a defensive strategy adopted by the incumbent to protect its dominant position in the upstream market. In order to assess the incentives to exclude, one should also check whether the incumbent s strategy was likely to be successfull. On the one hand Telefonica had a dominant position in the retail market which facilitates the success of the margin squueze strategy. On the other hand, the broadband market was a growing market at the time of the decision. Exclusion of downstream rivals is less likely to be successful when future demand is expected to grow, because it is less likely that future demand is insufficient to make entry profitable. Moreover, some independent operators were affiliates of international telecoms groups (such as France Telecom and Deutche Telecom) enjoying dominant positions in their respective national markets. One might argue that alternative operators backed by profitable groups should be able to match Telefonica s retail prices, sustain losses for a period and achieve the critical mass of users, thereby making Telefonica s exclusionary strategy uneffective. In the decision the Commission provides evidence that only companies with a sufficiently strong financial backing have been able to survive and grow (slightly) in the mass market. (Para. 57) These operators did not expand significantly and possibly did not achieve the critical mass of users, while none of the other operators achieved a market share of 1% in the relevant period. On appeal by Telefonica, the General Court and then the Court of Justice upheld the Commission s decision. 1 The General Court rejected Telefonica s argument that the Commission should have carried out a margin squeeze test based on an optimal mix of available wholesale products (i.e. including access to the unbundled local loop). The General Court also confirmed that Telefonica s conduct was likely to reinforce barriers to entry and expansion of competitors in the Spanish retail broadband market.. Genzyme A well-known UK abuse of dominance case is Genzyme. 15 assessment of the Office of Fair Trading (OFT), were broadly as follows. The main facts of the case, based on the Genzyme was the only producer of Cerezyme, which at the time of the case was the only drug available for the treatment of Gaucher disease (a rare metabolic disorder): in the early 000s only 10 UK patients were undergoing treatment. Genzyme was selling Cerezyme through the National Health Service (NHS). Treating a Gaucher patient with Cerezyme cost the NHS, on average, about 100,000 per year over the lifetime of the patient. 1 Cases T-336/07 and C-95/1 P, Telefónica v. Commission. The Kingdom of Spain appealed the Commission s decision too to the General Court, on the grounds that the Commission violated its duty to cooperate with the Spanish regulator and that it breached the principle of legal certainty, given that Telefonica was subject to sectoral regulation. However this appeal too was dismissed (see Case T-39/07, Spain v. Commission). 15 Decision No. CA9/3/03 - Exclusionary behaviour by Genzyme Limited. 7

9 As of the date of the decision by the OFT (March 003) there was one drug (Zavesca) which had just received authorisation to be marketed and which, the OFT considered, may have provided competition to Cerezyme going forward, although only to a limited extent. 16 TKT may have entered the market with a competing drug, but such entry - if at all - was not expected in the short-run. Genzyme s retail price for Cerezyme charged to the NHS included the home delivery and homecare services by Genzyme Homecare. (By contrast, Zavesca was an oral drug which would not have required the provision of home delivery and homecare services.) Up until 000 Genzyme used Heathcare at Home as its exclusive distributor and provider of homecare services for Cerezyme. Following this contract termination, Healthcare at Home, in order to continue to offer the delivery/homecare service, it had to purchase Cerezyme from Genzyme first, and make a retail offer to the NHS which would have also included the cost of delivery and homecare. 17 Genzyme set the same price in both cases (.975 per unit): that is, this was the amount charged both to the NHS for the drug including delivery and homecare services to the patient, and to Healthcare at Home just for Cerezyme. Healthcare at Home kept offering delivery and homecare services of Cerezyme to NHS patients past the year 000, but complained to the OFT. The OFT found that Healthcare at Home was sustaining losses due to Genzyme s pricing policy and that it was only a matter of time until Healthcare at Home would have had to leave this market. The OFT concluded that Genzyme had engaged in an anti-competitive margin squeeze. From an economic perspective, this was similar to setting the downstream price at the upstream price level. This was alleged to leave no potential scope for downstream competition (i.e. in home delivery service). The Competition Appeal Tribunal (CAT) confirmed the margin squeeze finding by the Office of Fair Trading. 1 As for the underlying economics of the case, prima facie, one could read it in the spirit of the single monopoly profit theory (or Chicago School argument): (i) selling the drug and (ii) delivering/administering it were perfect complements, consumed in (near) fixed proportions. Hence a monopolist over the drug could not increase its profits by leveraging its market power over the ancillary (complementary) service. Yet the standard Chicago argument would typically look at the market in a rather static way and would assume that the monopoly position over the first market (drug supply) is exogenously set and not contestable. It was also not clear what the rationale was for Genzyme to set the upstream and the final price at the same level. In fact the OFT noted that in addition to restricting the extent of competition in Cerezyme delivery/homecare services, Genzyme s behaviour - by preventing viable independent provision of delivery/homecare services for Cerezyme (and potentially other drugs) - also raised barriers to entry into the (upstream) market for the supply of drugs for the treatment of Gaucher disease. 19 According 16 This was due to clinical reasons, including the fact that at that time Zavesca was only prescribed to patients who could not tolerate Cerezyme. 17 Cerezyme was also sold to for administration at hospitals, where the tax treatment was different from the home channel. 1 Case No: 1016/1/1/03, [00 CAT. The Office of Fair Trading also found abusive tying, but the Tribunal dismissed this. 19 See paragraph 331 of the OFT decision and the summary made by the CAT at para. 0: [As a result of Genzyme s conduct it is more difficult for competitors to enter the upstream market for the supply of drugs for the treatment of Gaucher disease. Since the supply of Homecare Services is effectively tied to Genzyme Homecare, a new competitor

10 to two expert witnesses: Professor Cox [... expresses the view that changing homecare provider in circumstances where he was considering switching treatment could definitely affect the choice of treatment, especially in the case of vulnerable patients requiring infusion assistance, particularly since a very intense relationship can be built up between patients and their homecare providers. Dr Mehta [... also stresses that prescribing decisions have to take into account the patient s viewpoint. In Dr Mehta s view, if there is a change not just of the drug, but also of the arrangements for treatment from the delivery driver that he or she meets each time, to the assisting nurse with whom a relationship may have been built up and with whom the patient is content, this is not an insignificant matter. CAT Judgment, at para 635. Although there are features of the case which are consistent with the dynamic leveraging model illustrated in this paper, it would have been helpful for the OFT decision or the CAT judgment to provide further information about the real chances of successful upstream entry. In its dominance analysis, the OFT emphasised the importance of upstream barriers to entry, but obviously the higher the barriers to entry, the less threatening the possibility of upstream entry and thus the less convincing the need for Genzyme to monopolise the downstream market in order to deter upstream entry. 0 3 The Model An indispensable input is sold by a monopolist seller, U I, which is the upstream affiliate of the vertically integrated firm I. Firm I operates also in a downstream market through the downstream affiliate D I which uses one unit of the input to produce one unit of a final product. Upstream and downstream production is characterised by constant marginal costs: c UI = c DI = c > 0. 1 Market demand is given by Q = 1 p. We analyse a two-period game. In period 1 a rival firm D considers entry in the downstream market, while an upstream competitor U E can enter at a subsequent period, i.e. in period. We use a different index E for the upstream entrant to stress that the two entrants are not necessarily vertically integrated. We are focusing here on the case in which immediate entry can take place downstream while upstream entry can occur only in the future, but one would obtain similar results if immediate entry is possible in the upstream market, whereas downstream entry is feasible in a later period. would face the additional hurdle of persuading the patient to switch not only to a new drug, but also to a new homecare services provider. 0 The CAT seemed to share the OFT s views: Genzyme itself saw the creation of Genzyme Homecare as a strategy which pushes out competition, by providing a shopping basket of tailor made services. In our view, it is a reasonable inference that Genzyme considered that the creation of Genzyme Homecare would make it more difficult for competitors to Cerezyme to enter the market. (para. 637) However, the CAT also expressed some doubts, summarising its position as follows: Our overall conclusion, on the balance of the evidence, is that if Genzyme were to succeed in monopolising the downstream supply of Homecare Services, that would probably have some adverse effect on the ability of a new treatment for Gaucher disease to establish itself in the United Kingdom over a reasonable timescale, but the additional foreclosure effect in the upstream market is unlikely to be as great as that suggested by the OFT in the decision. (para. 639) 1 The assumption that the incumbent s affiliates have equal marginal costs simplifies the exposition. We will analyze this extension in Section.. 9

11 Upstream firms and (respectively) downstream firms sell perfectly homogeneous inputs and (respectively) outputs. We also assume that potential entrants both upstream and downstream are more efficient than the incumbent: 0 = c UE = c D = 0 < c = c UI = c DI. However, entrants have to pay a fixed entry cost, respectively F E and F. Fixed costs need to satisfy the following restrictions: F E < c(1 c) (1) c(1 c) (1 c) < F < c(1 c) + (1 c) Note that we do not impose any lower bound on the fixed costs in the upstream market; indeed, it may be the case that no fixed cost has to be paid to enter upstream. () Instead, it is crucial for our mechanism to work that fixed costs in the downstream market are sufficiently large. ( Finally, we ) impose the following restriction to the incumbent s marginal cost: c UI = c DI = c 5 17, 1. 3 The timing of the game is as follows: 1. Period 1, stage 1: The incumbent decides whether to commit to refusal to supply or, alternatively, to supply the downstream rival.. Period 1, stage : Firm D, after observing the incumbent s decision, decides whether to enter (and pay fixed sunk cost F ) or not; 3. Period 1, stage 3: If D is active, with probability 1/, the incumbent makes a take-it-or-leave-it offer to D. It offers the contract T (q ) = wq + F F. With probability 1/, it is D that makes a take-it-or-leave-it offer to the incumbent. We assume that the incumbent can credibly commit not to operate the downstream unit.. Period 1, stage : If D is active, the contract offer is accepted/rejected. Then active downstream firms choose final prices p and p I, firm D orders the input to satisfy demand, paying accordingly, and transforms one unit of the input into one unit of the final product. 5. Period, stage 1: Firm U E decides whether it wants to enter the upstream market (and pay F E ); D can still enter if it did not enter in period Period, stage : With probability 1/ active upstream firms make take-it-or-leave-it offers; with probability 1/ active downstream firms do. 7. Period, stage 3: Contract offers are accepted/rejected. The active downstream firms set final prices p I and p, orders are made, payments take place and payoffs are realized. 3 The assumption that c < 1/ ensures that p m (0) > c - i.e. that the monopoly price having zero marginal costs higher than the marginal cost of the less efficient vertical integrated firm. This assumption ensures that p I = p = c is an equilibrium in the final market when one firm has zero marginal cost and the other marginal cost equal to c. Moreover, the assumption c > 5 ensures that both the new entrants make positive profits at the equilibrium in 17 which entry occurs both upstream and downstream. 10

12 The timing of the game is summarised by Figure 1: I decides on refusal to supply Entry decision by D (sinking F ) Contract offers Final prices Input orders Entry decision by U E (sinking F E ) and by D if not taken in period 1 Contract offers Final prices Input orders PERIOD 1 PERIOD Figure 1. Timeline Before solving the model by backward induction, let us discuss the role of some assumptions that we have imposed. First, assume that the incumbent is able to (publicly and irreversibly) commit not to serve the independent downstream firm. This is a crucial assumption so as to have a strategic effect. One possible way to credibly commit to refusal to supply to design the input is such a way that it is compatible with the downstream affiliate only (see Choi and Yi, 000 and Church and Gandal, 000). Second, we assume that the incumbent can credibly commit not to operate the downstream unit, thereby ruling out the possibility for the incumbent to engage in opportunistic behavior. We do so because we want to highlight a new rational for vertical foreclosure, other than imperfect rents extraction. Third, we assume that upstream firms and downstream firms have equal bargaining power in the negotiation for contract terms. This assumption simplifies the analysis without loss of generality. For the result it is crucial to exclude only the case in which all the bargaining power is upstream, i.e. the case in which the probability that the downstream firm makes the offer is zero. In that case the downstream rival would obtain zero profits in the first period, if it entered the market. Vertical foreclosure in the first period could not be motivated by the intent to deprive the rival of the profits that are key to cover fixed costs and make entry profitable. 3.1 The incumbent did not commit to refusal to supply Let us start from subgame in which the incumbent decided not to engage in refusal to supply. The following Lemma summarises the post-entry payoffs in period, depending on the configurations of active firms. Lemma 1. Post-entry payoffs The following table indicates the payoff of the incumbent and the payoffs of the the independent firms (gross of the entry costs), in the different configurations of active firms, when the incumbent does not engage in refusal to supply: 11

13 D, U E Active Not Active [ π I = 0 π I = 1 (1 c) + 1 Active Π D = c(1 c) Not Active Proof. See Appendix A.1. Π UE = c(1 c) [ π I = 1 (1 c) + 1 Π D = 0 [ (1 c) Π UE = (1 c) [ (1 c) Π D = (1 c) Π UE = 0 π I = (1 c) Π D = 0 Π UE = 0 There are two aspects to discuss concerning Lemma 1. First, the incumbent is better off when only one independent firm is active (either in the upstream or downstream market) as opposed to the case in which no independent firm is active. In the latter case the incumbent monopolises the final market by using the the own less efficient upstream and downstream technology. Hence, the incumbent makes the monopoly profits associated to a marginal cost equal to c. Instead, when only one independent firm is active, say the independent firm in the downstream market, the incumbent is left with that same payoff when it does not make the offer (which occurs with probability 1/); however, when it makes the offer, the incumbent extracts the entire monopoly profits associated to the use of the more efficient downstream technology, i.e. associated to a marginal cost equal to c. It is only when both independent firms are active that the incumbent is worse off as compared to the case in which none of them is active, because it makes zero profits. Second, each independent firm makes larger profits when the independent firm in the vertically related market is active and competition in that market intensifies. Consider the independent downstream firm. When the independent upstream firm is not active, the profits of firm D are entirely extracted by the incumbent, when the incumbent makes the offer; firm D is able to appropriate the increase in monopoly profits due to the use of its more efficient downstream technology when, with probability 1/, it makes the contractual offer. Instead, when firm U E is active, each independent firms obtains half of the duopoly profits produced in the final market when a firm with marginal cost equal to zero competes with a less efficient rival having marginal cost equal to c. An implication of the above result is that the fact that an independent firm is active in a market facilitates entry in the vertically related market, as the following Lemma indicates: Lemma. Entry decisions in Period. If firm D entered in period 1, then firm U E enters in period for any feasible value of the entry cost F E. If firm D did not enter in period 1, then the continuation equilibria depend on the level of the entry costs F and F E : (i) if c(1 c) (1 c) < F c(1 c) and F E (1 c), then the unique continuation equilibrium is such that both the upstream and the downstream firm enter the market; 1

14 (ii) if c(1 c) < F c(1 c) + (1 c) + (1 c) and F E (1 c), then the unique continuation equilibrium is such that firm D does not enter the downstream market whereas firm U E enters the upstream market; (iii) if c(1 c) < F c(1 c) + (1 c) + (1 c) and (1 c) < F E c(1 c), then the unique continuation equilibrium is that neither firm enters the market. (iv) if c(1 c) (1 c) < F c(1 c) and (1 c) < F E c(1 c), then both the upstream and the downstream firm entering the market is a continuation equilibrium. [ A continuation equilibrium in which neither firm enters the market also exists unless c , ) 7 1 and c(1 c) (1 c) < F (1 c). Proof. If firm D entered the downstream market in period 1, then firm U E will earn π UE = c(1 c) F E if it enters and π UE = 0 otherwise. By assumption A1, it decides to enter. Consider now the case in which firm D decided not to enter in period 1. In case (i) entering the upstream market is a dominant strategy for firm U E. The best reply of firm D is to enter the downstream market, since F c(1 c). In case (ii) entering the upstream market is a dominant strategy for firm U E, while not entering the downstream market is a dominant strategy for firm D. In case (iii) not entering the downstream market[ is a dominant strategy for firm D. The best reply of firm U E is not to enter. In case (iv), when c , 7 1 ), then c(1 c) (1 c) < (1 c). As a consequence, there exist feasible values of F, namely c(1 c) (1 c) < F (1 c), such that entering the downstream market is a dominant strategy for firm D. It follows that both firms entering the market is the unique continuation equilibrium. Otherwise, entering the market is the best reply to the entry of the potential entrant in the vertically related market, whereas not entering is the best reply to the decision of the potential entrant in the vertically related market not to enter. Hence two continuation equilibria may arise: either both firms enter the market or firms fail to coordinate and neither of them enters the market. We can now study the entry decision taken by firm D in the first period, as illustrated by the following proposition: Lemma 3. Entry decision at period 1: If the incumbent does not engage in refusal to deal, then firm D enters downstream in the first period and firm U E enters upstream in the second period. Proof. Firm D anticipates that, if it enters in period 1, then firm U E will enter in period and its total profit is π 1+ D = (1 c) + c(1 c) F. Note that, Firm D also anticipates that, if it decides not to enter in period 1 then, as established by Lemma 1, two possibilities may arise: (i) post-entry profits earned in period alone are insufficient to cover the entry cost; in this case firm D does not enter in period either, and its total profit is zero; (ii) post-entry profits earned in period alone are large enough to cover the entry cost; in this case firm D enters in period and its total profit is π 1+ D = 0 + c(1 c) F. In both cases, firm D finds it more profitable to enter in period 1: (1 c) + c(1 c) F > max{0, c(1 c) F } (3) 13

15 In fact, by assumption A, the total post-entry profits earned by firm D when it enters in period 1 are large enough to cover the entry cost. Then, period-1 entry is more profitable than no entry in either period (case i). Moreover, when entry in period is profitable (case ii), entry in period 1 is more profitable because it allows firm D to earn positive profits for an additional period. 3. The incumbent committed to refusal to supply Let us analyse now the subgame in which the incumbent decided to engage in refusal to supply, starting from the post-entry payoffs in period. Lemma. Post-entry payoffs The following table indicates the payoff of the incumbent and the payoffs of the independent firm (gross of the entry cost), in the different configurations of active firms, when the incumbent engages in refusal to supply: D, U E Active Not Active Active Not Active π I = 0 Π D = c(1 c) (1 c) Π UE = c(1 c) + (1 c) π I = (1 c) + (1 c) Π D = 0 Π UE = (1 c) π I = (1 c) Π D = 0 Π UE = 0 π I = (1 c) Π D = 0 Π UE = 0 Proof. See Appendix A.. Note that the decision of the incumbent to engage in refusal to deal limits the profits that firm D earns if it enters the market as compared to the case in which there is no commitment to refusal to supply. Trivially, when the independent firm in the upstream market is not active, under refusal to deal firm D cannot obtain the input and makes zero profits. However firm D s profits are reduced also when the independent firm in the upstream market is active, because refusal to deal by the incumbent removes competition between the upstream supplier and makes firm D more dependent from firm U E. In fact, in such a case the distribution of the duopoly profits obtained in the final market is more favourable to firm U E. Then, the decision to engage in refusal to deal affects the entry decisions, as we show in what follows. Lemma 5. Entry decisions in period. If firm D entered in period 1, then firm U E enters in period. If firm D did not enter in period 1, then it does not enter in period either. Firm U E enters in period if and only if F E (1 c). This is the same effect that arises Ordover et al. (1990). However, differently from their model, here refusal to deal benefits firm U E but not the incumbent. This is because there is not product differentiation in the final market. 1