studio economico parcu & associati

Transcription

1 Via de Quirinale 26, Roma, talia tel Fax Towards an conomic Aroach to Art. 82: The Case for Anti-cometitive Winback Strategies by ANTONO NCTA & PR LUG PARCU DSCUSSON PAPR SUBMTTD TO C DGCOMP PUBLC CONSULTATON ON ART.82 Roma, 23 March 2005

2 xecutive Summary n this discussion aer we outline a concrete case in which an economic aroach to Article 82 of the Treaty may hel in detecting an abuse of dominant osition by an incumbent oerator ursued throughout the so-called winback strategies in relevant markets characterized by entry costs. While traditional aroaches exclusively focused on entry deterrence would be inclined to consider as ro-cometitive the winback strategies enacted by an incumbent oerator, we show how the economic analysis of the effects roduced by those strategies on the market might be anti-cometitive, both harming consumers welfare and efficient entrants. We outline how winback strategies may roduce, under given circumstances, the same anti-cometitive effects of standard raising rivals costs strategies. On the other hand, we argue that the examle of winback strategies suggests that even the adotion of an economic aroach does not hel in solving some of the ersistent antitrust dilemmas such as the choice between short-term equilibria which imrove actual consumers welfare to the detriment of actual cometitors, and long-term cometitive equilibria which sacrifice actual consumers welfare. These dilemmas seem to go beyond the choice of an economic aroach to art. 82, relying on the secific aims ursued by the cometition olicy authorities. 2

3 1. ntroduction Winback actions refer to incumbents strategies aimed at contacting a former customer who has left for a new entrant, for the urose of regaining that customer back. These strategies aly to ost-entry cometition and thus are clearly distinguished from raising rivals costs, redatory actions (redatory ricing, retention/matching rices, exclusivity and switch enalties clauses uon customers, etc) or any other entry deterrence otion (i.e. strategic caacity selection) available to the incumbent rior to the time of entry. While it is generally argued that these strategies are merely the result of cometition to the benefits of customers, some antitrust and regulatory authorities in telecommunications have recently considered them as anticometitive conducts. n Canada, for instance, the Canadian Radio-Television and Telecomunications Commission has adoted in the last five years several winback rules, i.e. rules banning incumbent s winback strategies from six to twelve months after customers switch to new entrants. n UK, France, Sain, The Netherlands, taly and even in some US courts, recent antitrust decisions 1 addressed the issue of the ossible anticometitive nature of incumbent s winback strategies. The above decisions however fail, in our view, to clarify the economic rationale for incumbent s winback strategies and the real meaning of their anticometitive nature. n this discussion aer, we rovide an economic aroach to winback strategies under the lens of Art. 82 of Treaty. We outline the otential anti-cometitive rationale for winback actions as strategies aimed at blocking off ost-entry cometitors exansion to the critical level which enable the entrant to comete on an equal basis against the incumbent. With resect to standard entry deterrence strategies, which generally face roblems of ost-entry commitment and credible threats by incumbent, we show how blocking off cometitors ost-entry exansion could be, in some circumstances, a better strategy in the incumbent-s interest, bringing to a higher degree of effectiveness and rofitability. We name the effect roduced by winback strategy as the holed bucket effect, since entrant s customers are induced to exit like water ouring in a holed bucket, ostoning entrant s ability to cover start u costs. 1 See A. Nicita (2006) Winback Strategies and Post-ntry Blockaded xansion, mimeo. 3

4 The scholarly wide literature on entry deterrence, redation, raising rivals costs and so on has mainly focused on the imact of incumbent s re-entry deterrence strategy on cometitors incentives to enter the monoolistic market. From the Bain-Sylos Labini ostulate on, the attention has been most entirely devoted to analyze the incumbent s ability to make a credible threat not to accommodate after entry has occurred. ven if inducement of exit is taxonomically identical to entry deterrence, only few models have dealt with analytical frameworks in which entry is taken as granted and the focus is on incumbent s strategies to affect cometitors subsequent behaviour. f we look at the realm of cometition olicies, esecially in network industries, exclusionary strategies are for the most adoted against new entrants that have already established their business and served a significant grou of customers. n network industries moreover, entry is somehow granted by the establishment of rights to access to essential facilities and by a regulatory environment which inhibits incumbents to freely set rices or suddenly change their rate card. Under this ersective, it seems interesting to investigate ost-entry exclusionary strategies held by dominant firm in order to freeze cometitors markets shares. The distinguishing feature of a winback strategy with resect to standard exclusionary ractices is that it is alied to (a ortion of) cometitors customers rather than to incumbent s customers. n a sense, winback strategies are a mirror image of standard exclusionary ractices such as exclusivity or retention strategies: while the latter imly a olicy of rebates and discounts made by the incumbent towards his own customers to retain them from switching towards cometitors, the former are based on rebates and discounts made by the incumbent towards cometitors customers in order to induce them to switch and come back to the incumbent s list. n this aer we show how, from a theoretical oint of view, there seems not being any comelling reason to evaluate winback strategies searately from any other raising rivals cost strategies whose ultimate effect is that of generating inefficient exclusion. Of course, the antitrust evaluation of the above ractices may differ among countries and actually it differs between US and U antitrust olicies but we argue that whenever an antitrust authority deems as anticometitive exclusionary strategies like network exclusivity clauses, fidelity or target rebates, selective rice undercutting or discounts, and so on, it has to include also winback in its warning list. 4

5 As with many other exclusionary ractices, the antitrust consideration of winback strategies is imortant because it affects the extent to which dominant firms may defend themselves against cometition rather than act to consolidate or even increase their dominance in the market 2. The debate thus falls into two well-known warring cams: the Chicagoan School sketics and the Post-Chicagoan activists : (i) The Chicagoan would robably believe that winback strategies are simly ro-cometitive actions given that when firms cut rices they should not take onto account the effect on cometitors rofits but only the hyothetical harm to customers (thus considering as ro-cometitive any cut rice as long as it is over above the costs threshold for redatory abuse) (Bork, 1978). Since entry has already occurred and some customers in fact have already left the incumbent, winback strategies, even if carried on by incumbent firms, are merely the effect of cometition on the merits. Customers will reorient their choices towards (what they erceive to be) the best seller and cometition will be alive even if cometitors may not. Moreover, since winback strategies should be associated to some economic advantage assed on customers in order to induce them to switch back, consumers welfare is about to be enhanced. Finally, if new entrants are driven from the market as a consequence of winback strategies that means that they were somehow inefficient, i.e unable to match the incumbent winback offers to retain their own customers; (ii) the Post-Chicagoan activists, on the other side, might consider winback strategies as a sort of selective rice undercutting. Since they believe that even above-cost rice undercutting or selective discounts may harm cometition as well as final customers by inhibiting long-term efficient entry, also winback strategies should consequently receive an antitrust scrutiny. n their view, dominance imlies a short-term cometitive advantage by incumbent firm. Thus whenever an entrant as efficient as the incumbent cannot effectively relicate the discount olicies adoted by incumbents (as long as a critical threshold is reached in terms of caacity, 2 See Jones and Sufrin (2001) The oint raised here for winback strategies is thus strictly related to that raised by selective discounting or rice undercutting. 5

6 minimum scale or minimum number of subscribers when switching costs or network effects are resent) exclusionary ractices that raise rivals cost u to the oint to discourage entry or to induce exit should sanctioned by antitrust authorities. We outline the conditions under which (i) incumbent s winback strategy is rational only when exclusionary and (ii) a ban on incumbent s winback may actually increase consumers welfare, even if that encourages short-term inefficient entry. The discussion aer roceeds as follows. Section 2 recalls the results of the Gelman and Salo (1983) model on ost/entry cometition with zero entry cost. Section 3 shows how entry costs may affect incentives to accommodate by the incumbent. Section 4 outlines how, taking entry has granted, the incumbent has strong incentives to accommodate entry followed by winback strategies, then we show the welfare imlications of a ban on winback olicies. Section 5 elaborates some motivation surrounding the behavioural assumtion on entrant s strategy. Sections 6 and 7 comare the exclusionary effect generated by winback with those associated to traditional foreclosure strategies. Section 8 draws the main conclusions. 2. An xamle of Post-ntry Cometition without ntry Costs n this section we illustrate a simle examle of ost-entry cometition without entry costs outlined by the Gelman and Salo (1983) model 3. Let us consider a market in which a leader firm sells the quantity q on the market and a follower firm covers the residual demand on the market selling a quantity q. To the sake of simlicity let us assume that: - firms have identical cost structure and roduction costs equal to zero; - the incumbent firm has unlimited caacity and both the firms roduce an homogenous good with a demand curve given by = 100 Q where Q=( q + q ); 3 We refer here to the reduced form develoed in Shy (1995). 6

7 - all consumers refer the less exensive demand; moreover consumers refer the incumbent s brand name when rices are matched). Let and resectively be rices charged by the leader and the follower on the market, and let k be the caacity invested by (which is assumed as a roxy of the grou of customers served), with k being the observable entry s caacity. The quantities demanded on the market are: q $ 100 % = # 100 % k % if if & > q $ k = # 0 if if < % Market demand is shared between the leader and the follower according to the value of the rices. Let us define k* as the value of entrant s caacity-market share which induces a matching rice strategy by the incumbent. n order to derive the value of k* we have to comare the incumbent rofit in the case of undercutting or rice undercutting U with the rofit level A the one associated with the accommodation strategy. f the accommodation strategy is that chosen by the incumbent firm it must be that: (1) = (100 ) < = (100 k ) U and first order conditions bring to the following results: (2) A A 2 (100 k) (100 k) 100 k = q A = A = As we can see the incumbent rofitability decrease as the ntrant s caacity or market share k increases, d A / dk < 0. Given these values we define the market share k* of the entrant which is comatible with an accommodation strategy uon entry by the incumbent as the value which equalizes is observable by both arties. U and A. We assume that the value of k* 7

8 A 100 k = 4 2 (100 ) k k * 100 k Figure 1 Now let us consider the cometitor s rofit function which is a linear function of market share (caacity) k: (3) = k Thus we have the following three stages game structure. 1. At t=0, the entrants sets rices and starting caacity (market share) k=k ; 2. After entry the incumbent observes k and decides whether accommodate undercutting; When the accommodating equilibrium in the ost-entry scenario strictly dominates the undercutting rice equilibrium as long as the cometitor commits not to enter with a market share greater than k*. When k >k* the market equilibrium will be characterized by a matching rice at the cometitive rice. Then we have the following roosition (Gellman and Salo, 1983) 4 : 4 See also Shy (1995)which the reduced model form is derived. 8

9 Proosition 1 n the game below there exists a sufficiently limited caacity level k and a rice set by the entrant that ensure that the incumbent will find it rofitable to accommodate entry. t=0 Stay out t=1 M = ( 100 ); = 0 O Price undercutting Choice of k Accommodate k = ( 100 ); = 0 = ; = k 4 Figure 2 Proosition 1 is very helful in exlaining why entry is observed being not blocked by any rice undercutting by incumbent oerators in many industries. However it does not go further in exlaining the nature of ost-entry cometition. n the next section we extend the above model by assuming that entrant has to sustain some entry cost. 3. An xtension: Post-ntry Cometition with ntry Costs Let us now extend the revious model by assuming that: - the incumbent firm has unlimited caacity, while the entrant firm has to sustain some start u costs indicated by h, after this sunk cost has been recoued the follower can comete on an equal basis against the leader, given that both the firms roduce an homogenous good with a demand curve given by = 100 Q where Q=( q + q ); - all consumers refer the less exensive demand; moreover consumers refer the incumbent s brand name till the follower s brand has reached a valuable market reutation (here associated with a given market share, k ). 9

10 Let and resectively be rices charged by the leader and the follower on the market, and let k be the caacity invested by (which is assumed as a roxy of the grou of customers served), with k being the observable entry s caacity and k be the minimum caacity needed in order to relicate incumbent s choices and challenge on an equal basis his leadershi. We assume that k k and k k *, where k* is the value at which the incumbent firm decides whether or not to accommodate entry (see below). We assume that while k and k* are common knowledge, k is s rivate information uon entry. The quantities demanded on the market are now: q $ 100 % 100 % k % = # '(100 % ) 0 if if if if & > = > and k & k and k & k and k > k and k > k q $ k (1 '))(100 ' = # 100 ' 0 ) if if if if < = < & and and and % k k ( k k > k k > k Market demand is shared between the leader and the follower according to the value of the rices as well as to the follower s scale k. When the follower has reached just a small scale k < k (which is corresonds to the associated market share) then the incumbent has full incentives to accommodate and to let the follower cover a quota k 10

11 of the total market 5 at a rice while firm will cover the residual demand at the monoolistic rice. This equilibrium holds only if k lower than k* defined as above as the value of entrant s caacity-market share which induces a matching rice strategy by the incumbent. Now let us consider the cometitor s rofit function: (3b) Since we have assumed that = k h, with d / dh < 0 and d / dk > 0. k < k, for the cometitor to survive in the market and to comete on an equal basis against the incumbent firm it must increase the entry s market share k and reach at least a market share of (4) k = h / a value which is s rivate information. Thus we have the following three stages game structure: 1. At t=0, the entrants sets rices and starting caacity (market share) k=k ; 2. After entry the incumbent observes k and decides whether accommodate or undercutting; 3. At t=1, if the incumbent has accommodated, then the entrant reveals his rivate information on the value of k, i.e. its roject of exanding or not the starting caacity k. 4. At t=2 if the entrant has decided to exand, then the incumbent has to decide whether to accommodate (which means reducing to a his market share) or to react through win-back strategies. Let us assume first that the entrant is able to credibly commit not to increase the starting caacity k (and market share) after entry has occurred. Thus the game lasts just one eriod. At t=0, the entrants sets rices < and starting caacity (market share) k=k and after entry has occurred the incumbent observes k and decides whether accommodate or undercutting =. f he undercuts then the entrants is induced to exit. f the incumbent accommodates then the entrant will serve k customers at a rice 5 We assume that k < (1 )(100 ). 11

12 <, while the residual demand q = 100 k is served by the incumbent at rice. The above is exactly the outcome of the Gelman and Salo (1983) model. The following figure illustrates the Nash equilibrium of this game. t=0 Stay out t=1 M = ( 100 ); = 0 O Price undercutting Choice of k k * Accommodate t=2 = ( 100 ); = h k= k k < k k 100 k = 4 2 ; = k h = # (100 ); = (1 #) (100 ) Figure 3 f = (100 ) < = (100 k ), then the unique equilibrium is one in U A which the incumbent accommodates since by assumtion k <k*. Thus if h=0 we have here one Nash equilibrium in accommodating strategy. Proosition 2 When h=0, the above game has one unique N in accommodating strategy as long as = (100 ) < = (100 k ). U A However when h>0, the entrant maintains strong incentives to increase her caacity after entry in order to reach the long-term caacity that it is needed in order to comete on an equal basis against the incumbent. Let us define g as the robability that attributes to s decision to extend her caacity to the level k, g ( k ) = rob{ k k}. 12

13 Proosition 3 After s entry occurred with ositive entry costs h>0, if g=g( k ) is sufficiently high the above game will have a unique Nash quilibrium associated with s rice undercutting Discussion Proosition 3 clearly shows that when there are entry costs, the entrant s commitment not to exand caacity after entry is not credible. The assumtion of ositive entry costs thus modifies the result of Gelman and Salo (1983), given that the incumbent maintains in the context strong incentives not to accommodate entry and thus to choose rice undercutting. As a consequence, the circumstance that k<k* is neither a necessary nor a sufficient condition to induce accommodation by. What is relevant here is the robability that the entrant may reach a caacity level equal to k, once entry has occurred. The ost-entry rice undercutting equilibrium in roosition 3 derives however from the assumtion that we are taking entry has granted, i.e. that s belief on s decision is not observable by the entrant. f we relax this assumtion and consider g( k ) were observable by before entry we should exect no entry at all. Thus we can actually imagine five different equilibria in the game deicted in figure 3 according to the assumtions made on the level k and on asymmetric information on s caacity and s belief over g: a. blockaded entry when h>0, and when g( k ) is high and observable by with rofit levels = ( 100 ); = 0 ; M b. entry with rice undercutting when h>0 and when g( k ) is high and not observable by with rofit levels O = ( 100 ); = h ; c. entry with accommodating strategy when h=0 or when g( k ) is very low and observable by with rofit levels 100 k = 4 2 ; = k h ; d. entry with accommodating strategy and exansion by to k= k when h>0 and g( k ) is very low and observable by or s entry at k = k with 13

14 duoolistic market sharing rofits = # (100 ); = (1 #) (100 ). Thus, with full information and rational agents, we should exect two alternative equilibria: (i) no entry or (ii) if entry is actually observed, we should exect a case in which entry occurs immediately at the level k = k bringing to duoolistic market sharing. As a consequence, the circumstance in which we observe markets with entry costs characterized by accommodation by couled with entrant s inability to exand her caacity should be quite difficult to understand in our framework as the outcome of rational agents in a strategic game with full information and common knowledge. Nonetheless this is recisely what we observe in many ost-liberalized network industries at least in uroe. How we can thus conciliate the fact of observing entry by rational agent characterized by ost-entry blockaded exansion in her caacity? Next section tries to address this oint. 4. Post-ntry Cometition with ntry Costs and Winback Strategies n this section we extend the above framework by introducing two further assumtions: (i) we extend the strategy toolkit of the incumbent, conceding that the incumbent may adot, after s choice to exand caacity over k, some exost strategy in order to block the exansion of the entrant towards the break-even value k ; (ii) we assume that entrants, for some reason, cannot anticiate at t=0 incumbent ost-entry decisions to block the entrants exansion towards the break-even value k or believe that incumbent will not or cannot influence, after entry has occurred, entrants ability to exand own caacity (we will rovide in section 6 some ossible economic arguments to justify this assumtion). n articular we analyze here the effect of winback strategies as exclusionary ractice. A winback strategy is here defined as follows. 14

15 Definition 1 - A winback is a strategy enacted by the dominant firm to regain former customers, i.e. customers who already have switched to and are receiving service from another cometitor. n fact, an incumbent s winback it as a strategy of selective rice discrimination towards cometitors customers, imlemented by a olicy of discounts, rebates, romotional rices and so on. As a consequence for a winback strategy to be imlemented we should assume that in some way the incumbent knows the rofile of the customers to be regained. The distinguishing feature of a winback strategy is that it is alied only towards the cometitors customers and it is not extended to the incumbent s customers, while standard exclusionary strategies generally imly a olicy of rebates and discounts made by the incumbent towards his own customers to retain them from switching towards cometitors. n order to outline the emergence and the economic rationale for winback strategies held by an incumbent oerator we should add another stage (t=3) in our revious game structure. That means that in t=3, given that the entrant has revealed her decision at t=2 to exand the caacity towards k = k, the incumbent has to decide whether to accommodate or adot a winback strategy. Here, the adotion of a winback strategy requires the incumbent to aly selective rice undercutting or discounts to regain the roortion of cometitor s customers which is deemed to be decisive in order for the entrant to reach a caacity k = k. The incumbent s ayoff associated with a successful winback strategy is then given by the following equation: (5) = ( 100 k ) + ( k k ) W # where (100 k ) is the ayoff coming form the actual residual demand served by the incumbent at t=3, while the comonent ( k k ) identifies the additional returns to the incumbent coming from winning back the ( k k ) grou of customers, with 0 < 1m at the matching rice undercutting the entrant s rice. We define the 15

16 successful incumbents winback olicy on regaining ( k k ) customers as the holed bucket effect. The idea is that when winback olicy is at stake, entrant s customers our like water in a holed bucket: as long as winback strategies inhibit the entrant to stably reach and tresass the critical threshold k = k, the incumbent cannot relicate on an equal basis the incumbent olicy. Under this assumtion the entrant s rofit are given by: (6) = [ k #( k k )] h Thus we have the following roosition. Proosition 4 winback as raising rivals cost strategy Under incumbent s winback strategy the entrant s critical threshold to comete against winback is endogenously ut forward with resect to the ex-ante level k = k. Proof. From (3), (4), (5) and (6) it derives that the ex-ost entrant critical threshold to comete on an equal basis against the incumbent is given by: (7) k ˆ = k + ( k k ) holed bucket effect Corollary 1 exclusionary effect of a winback strategy A successful winback strategy imlies the adotion of a rate of regainˆ such that k( ˆ) < kˆ is satisfied. The above corollary clearly outlines that for a winback strategy to roduce anticometitive effects, it is not sufficient that the incumbent is just regaining some customers back, rather it must be that the rate of regaining has a dimension large enough to substantially block off cometitor exansion. 16

17 Proosition 5. Winback as a rational choice A winback strategy ˆ ( k k ) is rational only when k( ˆ) < kˆ is satisfied. Proof. Suose not. Then, if k( ˆ) kˆ, that means that the cometitor ha unlimited caacity to exand her market share at the rice settled. n other words, the cometitor is able to immediately match every winback strategy with the result that the incumbent s ayoff will be raidly equal to that of market sharing at rice = ) S #(100 < U < w which is lower that the undercutting rice strategy uon entry. Since by our assumtion the undercutting strategy is available just after entry in t=0, it would be irrational for the incumbent to select a strategy to obtain available in the first instance. S when another strategy brining to U was Corollary 2 From roosition 3, 4 and 5 it derives that a winback strategy is an incumbent s rational choice only when it roduces exclusionary effects on cometitors. The above corollary imlies that when winback is not effective it would be simly irrational for an incumbent to adot it, given that it would roduce a continuous rebate on market rices. f the entrant were able to match any rice roosed by the incumbent through winback actions, any winback would simly decrease the rice of entrant s customers, attracting new customers towards the entrant and thus raidly decreasing market rice in a contagious way. f entrants match any discount roosed as a winback strategy, incumbent s rice cuts no longer roduce the holed bucket effect as they would otherwise. As a consequence, when winback strategies result to be ineffective to generate exclusion they would never be adoted by a rational incumbent. We are now ready to analyze the three stages game of ost-entry cometition when winback strategies are in lace. f at t=2, the cometitor has exanded her caacity so as to reach k = k, at t=3 the incumbent will react in two ways: accommodating or enacting 17

18 winback strategies. When the conditions for a successful winback ˆ ( k k ) are satisfied, this strategy will be chosen if the ayoffs associated to successful winback (thus imlying k < k ) are greater than that associated to duoolistic market sharing: (8) = ( 100 k ) + ( k k ) > W # #(100 ) = S Since the above always holds by definition 6, then at t=3 winback is a dominant strategy in the subgame. n order to analyze whether the winback strategy is a erfect subgame N, we should move from t=3 to ost-entry at t=0 and consider the exected value of the ayoff under the robability g of entrant choosing to exand ost-entry market share to k = k. n the following figure, ayoffs are defined as follows: M = re-entry monoolistic rofits; U = ost-entry undercutting monoolistic rofits; A = ostentry accommodating rofits; W = winback rofits; C = ost-entry duoolistic rofits. n the following game we have: > > > 0. Let us define M k ĝ as the value according to which (1 gˆ) (100 k ) + #( k k ). 4 Then we have the following roosition. W A U C Proosition 7. Winback as a Nash equilibrium in the ost-entry game For a sufficient high robability g gˆ of entrant choosing to exand ost-entry market share to k = k, the winback decision ˆ ( k k ) is a dominant strategy in the ost-entry game. Proof. From the game structure below it is easy to show by backward induction (and recalling the assumtion on s inability to forecast winback strategies before and after entry) k that, since ĝ is such that (1 gˆ) (100 k ) + #( k k ) then winback 4 6 t is sufficient to show that when k = k, by assumtion = and thus (100 k ) = (100 ). 18

19 strategy t=3 dominates the accommodate strategy at t=2 and ost-entry rice undercutting after t=0. t=0 Stay out t=1 M = ( 100 ); = 0 O Price undercutting Choice of k k * Accommodate t=2 = ( 100 ); = h k= k k < k k t= k A = ; A = k h < 0 4 Accommodate Winback = # (100 ); = (1 #) (100 C C ) W = (100 k ) + #( k k ); W = k h < 0 Figure Discussion Proosition 6 outlines that in a ost-entry game, when there are entry fixed costs and when a successful winback is ex-ost ossible for the incumbent, the rational behaviour by an incumbent firm is always that of accommodating entry, since for any value of caacity chosen by the entrant the associated ayoff for the incumbent are always greater that those associated to rice undercutting uon entry. This is a remarkable conclusion since it shows that an incumbent may have strong incentives to accommodate entry having anti-cometitive uroses to exclude cometitors and/or to maintain a monoolistic or leadershi osition on the market. Thus observing entry in market characterized by monoolistic or dominant osition should not be automatically deemed as an evidence of cometitive entry, esecially when cometitors face difficulties in exanding their market enetration. 19

20 What about the welfare imlications of winback strategies? Under our assumtions (equal roduction costs for both the incumbent and the entrant and homogenous roducts), consumers welfare is enhanced at the lowest ost-entry rice. That means some ambiguity in the results: when ost-entry rice is the same uon entry and at t=2 and at t=3 then it is indifferent for the consumers if the entrant remains or not in the market. However, if ost-entry rice equilibrium in the market, after the entrant has reached her critical threshold k = k are lower than those associated with matching rice equilibrium, then consumers welfare is enhanced with entry and with a ban on winback strategy. n the game structure outlined above, consumers welfare is ranked in the following way: W < W < W < W W. Thus consumers welfare is inversely related to incumbent s M A W U C ayoffs. n articular, consumers welfare under winback strategies is dominated by consumers welfare under rice undercutting equilibrium and ex-ost duoolistic cometition. Proosition 8. Consumers welfare under a ban on winback strategies Given the game structure outlined above, a ban on winback strategy is welfareenhancing for any value of the entrant s caacity or market share k. Proof. With a ban on winback strategies there are only two ossible outcomes: rice undercutting equilibrium or market sharing equilibrium. Given that W < W < W < W W, indeendently on the conditions which induce one or the M A W U C other equilibrium, consumers welfare is enhanced in any case. Proosition 8 shows a remarkable result. ndeendently of the effective decision on k by the entrant, a ban on incumbent s winback strategies (under the assumtion of substantial entry costs) always enhance consumers welfare because in one case (rice undercutting) the exit of the cometitor is associated to a monoolistic configuration with lower rices with resect to re-entry rices; in the other, ex-ost rices bring to a 20

21 duoolistic market equilibrium also characterized by lower rices with resect to reentry rices. What is imortant here to stress is that this result, and the consequent efficiency of a ban on incumbent s winback strategies, is areciable both by Chicagoan Sketics (defending consumers welfare indeendently of the degree of cometition in the market) and by Post-Chicago Activists (defending consumers welfare through an increase in the degree of cometition). However, as Armstrong and Vickers (1993) have outlined in a similar framework, a ban on selective rice discrimination in the form of winback strategies, may induce too much entry in the market, i.e. inefficient cometitors might be induced to enter the market, inducing the incumbent to accommodate at a rice higher than that associated with efficient entry. Another imortant conclusion to outline here is that the same result here obtained alies in industries with consumers searching costs or with network effects because also in those cases the entrant has to reach a minimum amount of customers in order to be able to fully relicate the incumbent s olicies. 5. Understanding ntry Decisions with Winback Strategies Proosition 7 strictly deends on the assumtions we made on entrant s behaviour. n articular, we have taken entry has granted and focused on ost-entry cometition. According to the assumtion made, entrants cannot anticiate at t=0 incumbent ostentry decisions to block the entrants exansion towards the break-even value k or believe that incumbent will not or cannot influence, after entry has occurred, entrants ability to exand own caacity. t is for this reason that entry occurs and winback is adoted by the incumbent after having accommodated. We are aware this is an ad hoc assumtion. n some resect, this assumtion recalls the long-debated and criticized hyothesis of limit ricing models under which the rosective entrant was assumed to believe that the incumbent would have maintained the same behaviour after entry. n this section we try to clarify the motivations surrounding this ad hoc assumtion. A first motivation relies on the idea that even if cometitors know that winback strategies are ossible (and legitimated under antitrust scrutiny) they may have 21

22 favourable beliefs on the rate of winback: they could exect that the rate at which the ut water on their bucket is higher than that at which water is oured out. One reason for that could be the incumbent s inability to erfectly observe rate of new entrants, reacting thus with a lag to entrant s exansion. k = k or the enetration Another motivation relies on entrant s belief that the incumbent s inability to erfectly observe k = k may induce a rudential attitude not to start a war rice and to accommodate towards a duoolistic market sharing collusion. From corollary 2, we know that with uncertainty over the real value of k = k and on the entrant s comared efficiency the incumbent may rudentially accommodate and signal his intent to move towards a duoolistic collusive equilibrium. Finally, an additional motivation to focus on ost-entry strategy, regardless of entrant s incentives to enter the market in the first instance, might be based on the idea that at least in some liberalized industries, incumbent s rice undercutting olicies just after entry or winback strategies are deemed as an abuse or dominance. n this case it is the institutional framework that induces entry by cometitor in the first instance. The next section comares the effect generated by winback with that roduced by standard exclusionary strategies. 6. Comaring Winback and Standard xclusionary Practices Winback strategies are just one of the ossible exclusionary ractices that an incumbent may adot in order to affect ost-entry cometition and blocking off entrant s ability to exand her caacity towards the level k = k. n this section we show the conditions under which standard exclusionary strategies roduce the same effect of winback in the market. n articular we focus here on those strategies aimed at enforcing re-emtion or retention of incumbent s customers through 7 : (a) the adotion of legal/contractual rules such as exclusivity or breach enalty clauses uon customers; 7 We are not considering here below cost redatory ricing 22

23 (b) the design of incentives to customers aimed at increasing the oortunity costs of switching towards cometitors, through the assignment of rebates, discounts and rice cuts which selectively retain marginal customers. These actions may roduce, under certain conditions, horizontal foreclosure effects against cometitors and, when couled with selective discounts or rice cuts, vertical discriminatory effects among different grous of incumbent s customers. ven if the conditions under which the above ractices are deemed to be anticometitive may differ from one country to another (esecially between US and U 8 antitrust olicies), it is ossible to define a minimum set of conditions according to which these ractices have been considered as an abuse of dominance which harms both cometitors and customers. a. Pre-emtive xclusivity Clauses When exclusivity clauses are signed between the incumbent and his clients, they can take the form of long-term exclusivity contracts including a enalty clause for clients contractual breach and switch to cometitors. That means that in order to relicate the incumbent rice, the entrant has to decrease her rice so as to reay the enalty to the switching customer (Brodley and Ma, 1993). n the case of total foreclosure, is set at a highly rohibitive level so that any efficient entry is actually inhibited. n other cases, efficient entry could be reduced (Aghion and Bolton, 1987). Any decrease in entrant s rice induced by the emergence of exit enalties imlies an endogenous extension of the minimum caacity or market share necessary for the entrant to comete on an equal basis against the entrant. Under the framework outlined above, that means that with exclusivity clauses entrants ayoffs are given by: (9) = ( ) k h The enalty thus generates an endogenous extension of the minimum caacity or market enetration for the entrant (a raising rivals cost strategy), for given values of h and as in (4): 8 See Hovenkam 1994; Facey et al 2003; Motta, 2005; Parcu,

24 ~ h h (10) k = > k =. Moreover, as Rasmusen, Ramseyer and Wiley (1991) have shown, in order to foreclose the market the incumbent need not to sign an exclusive deal with all customers but it is sufficient to block off a ortion of customers just sufficient to inhibit the entrant s ability to reach k = k. Thus the blocking off enalty clause ( k ) is given by that value that imlies ( ) < 0. They show that actual customers, being unable to coordinate their behaviour, would accet to be locked in such a contract. Aghion and Bolton (1987) show how customers may bargain exclusivity clause in exchange of a reduction in actual rices. The actual incumbent sacrifice, reresented by alying a rice ~ <, is rational whenever: (11) ~ ~ (100 ) (100 ) > #[ (100 )] which always holds by definition. ven if some remarkable differences exist between the cometition olicies adoted by US and U antitrust authorities, network exclusivity clauses have been generally considered as an abuse of dominance when the effect roduced on the market has been that of increasing efficient rivals costs to a level sufficient to deter entry or to induce exit. b. Rebates, Price undercutting, Selective discounts, Most favoured customers clauses Another exclusionary strategy that the incumbent may adot consists, rather than on imosing enalties on exit, on assigning aroriate incentives to induce customers retention through a olicy of rebates. n the case of fidelity rebates, the incumbent assigns fidelity rebates as rewards or discounts to customers who urchase all or a secified ortion of their requirements for a given roduct or service from a dominant firm. They may include also sales target-based over-rider discounts and so on. n the case of target rebates the incumbent assigns rebates conditional on a comany meeting a sales target that is higher than revious urchases. Both the tyes of discounts have been in some case deemed as anticometitive when they seemed exclusively aimed at 24

25 excluding cometitors rather than at transferring efficiency imrovements to customers (as in the case of quantity rebates). Whatever is the nature of incumbent s rebates, formally they simly equal the effect of a rice undercutting strategy with incumbent setting uon entry an effective discounted rice =, which, according to the assumtions made in section 3, will block off entry or induce exit. As we have outlined above, in some cases, it is not necessary for the incumbent to adot a rice undercutting towards all the customers served. might be sufficient to adot a selective rice undercutting towards the ortion of customers actually contested by the entrant. This form of retention requires the incumbent having access to information on the identity of customers contested by an entrant. One way for the incumbent to enforce such a strategy without sustaining information costs is that of including in contracts the socalled most favoured customer clause or nglish clause, often deicted as a de facto exclusivity 9, which imlies that the customer is allowed to switch suliers without enalty if the dominant undertaking cannot or will not match more favourable terms offered by another sulier. n this case, the incumbent alies a selective discount or rice undercutting strategy just to the marginal customers with a ayoff given by = ( 100 ) k where k are the customers contested by the entrant. When however the incumbent has the information sufficient to match just the roortion of customers sufficient to block off the entrant, then we can have the same ayoff of winback strategies = ( 100 k ) + #( k k ). Again the strategy of selective rice undercutting is rational for the incumbent as long as: (100 k ) + $ ( k k ) (100 ) k (100 ) > #[. (100 Such forms of rebates have been valued as anticometitive, esecially in uroe, when the existence of start u costs sustained by efficient entrants inhibited cometitors ability to relicate rice undercutting so as to induce the needed amount of customers to switch. uroe selective rice undercutting by an incumbent have been considered as anticometitive because they roduced discrimination against cometitors and against )] 9 J. M. Lave, The law and economics of de facto exclusive dealing, Antitrust Bulletin; Sring 2005; 50, 1 g

26 customers. n US the aroach has been more rudential, ointing out the absence of substantial consumer harm, given that rice undercutting, even when selective, generally enhances consumers welfare. Some US scholars, however, have reeatedly ointed out the anticometitive nature of above-cost rice undercutting (dlin, 1997; 2002). n order for these strategies to be exclusionary it must be that they increase rivals cost of entry. f that is not the case, because for instance the entrant can immediately relicate any discount or rebate, then it would be rational for the incumbent just to accommodate entry. Thus raising rivals costs strategy are held in an economic environment in which there is a clear cometitive advantage held by the dominant firm and an asymmetry in the cometitive caacity between the incumbent and the entrant. Whatever is the antitrust attitude towards the above strategies, the oint that we would raise here is that they roduce the same anticometitive effect of winback strategies, thus there seems not being any comelling reason not to treat winback strategies, as we have defined them above, in the same way in which we deal with standard exclusionary ractices whose ultimate effect is that of generating inefficient exclusion. As a consequence, when a cometition authority considers as an abuse of dominance incumbent s strategies such as exclusivity clauses, fidelity or target rebates, selective rice undercutting and so on, it should also include winback strategies in its warning list. 7. Winback vs. Non Discriminatory Price Undercutting Strategies Proosition 8 has shown how, under the game structure outlined in the aer, a ban on winback strategy is welfare-enhancing for any value of the entrant s caacity or market share k. However, given the structure of the game, a ban on winback strategy simly imlies that the equilibrium in the ost-entry game outlined in figure 5 will be one characterized by rice undercutting, thus with inducement of cometitor s exit. From the consumers welfare oint of view, with a ban on winback we have certainly an imrovement with resect to the equilibrium associated with winback strategies. However, it could also be interesting to comare the welfare associated with rice undercutting equilibrium with that associated with entry and duoolistic market sharing 26

27 equilibrium. f we assume that the matching rice selected by the incumbent is equal to the level associate to duoolistic market sharing equilibrium, consumers welfare will be exactly the same in the two cases. However, since by assumtion the entrant has to cover start u costs uon entry, it might be the case for the rice matched by the incumbent to be higher than that, say ~, fixed by duoolistic firms after the entrant has reached her crucial threshold k. When this is the case, in order to maximize consumers welfare a ban on winback strategies should be couled with a ban on undercut ricing (since ban on winback is equivalent to a ban on discriminatory selective ricing a joint ban here simly imlies forbidding any form of rice rebates aimed at foreclosing the market). Proosition 9. The case for a joint ban on winback and on ost-rice undercutting Let us assume that, once the entrant reaches a caacity like equilibrium rice on the market k = k, the duoolistic ~ is lower than that ( ) associated to ost-entry matching rice by the incumbent, with ~. Then the most efficient consumers < welfare configuration requires both a ban on incumbent s winback strategies a ban on ost-rice undercutting. Proof. t is easy to see how, in the game in figure 6, with ~ <, U W C W <. However, since U > C 0, by backward induction, with a ban on winback strategies the incumbent is induced to undercut uon entry. Thus in order to enforce an aggregate outcome like W C a ban on winback should be couled with a ban on ost-entry rice undercutting. However, when, as in the uroean antitrust tradition, cometition authorities encourage entry so as to ensure a long-term mechanism of cometition in the market even at the exenses of short-term benefits on customers, then a joint ban would robably be selected, indeendently of the differential between rices ~,. 27

28 On the other hand, when there is uncertainty on the rice level associated with the duoolistic equilibrium 10 the joint ban outlined by Proosition 9 could be equally associated with a rice level higher than that determined by rice undercutting strategies. n that case a joint ban would reduce consumers welfare. Besides, it should be ointed out that some antitrust authorities are considering as anticometitive ant form of selective rice undercutting or matching because of their discriminatory and foreclosing nature. As a consequence they are inclined to judge as ro-cometitive discounts, rebates and rice undercut which are destined to the whole set of customers. Under this resect it would be roblematic to coule a ban on winback with a ban on rice undercut. However, in uroe, there have been some antitrust cases in which rice undercutting or matching even when non discriminatory among customers (and thus uniformly alied to all customers) have been nonetheless judged as anticometitive. dlin (2002) rooses a theory based on the anticometitive effect of rice undercutting even when rices are above costs. For a critique see lhauge (2003). 8. Conclusions The examle of winback strategies suggests that the adotion of an economic aroach may hel in detecting an abuse of dominant osition by an incumbent oerator ursued throughout the so-called winback strategies in relevant markets characterized by entry costs. f from one side, these strategies are clearly the result of cometition to the benefits of customers, from the other, as some antitrust and regulatory authorities recently ointed out, they may generate substantial anticometitive effects. We have argued how ursuing an economic based aroach to art. 82 it is ossible to show how incumbent s winback strategy is rational only when it is exclusionary. Under given conditions, winback strategies might block off cometitor s exansion after entry, thus generating the same outcome raised by standard exclusionary ractices. We have finally discussed the consumers welfare effects of a temorarily ban on incumbent s winback against new entrants. However, the discussion on consumers 10 dlin (1997) surveys the literature on the anticometitive effects of rice matching, esecially of selective discriminatory rice matching, showing the ambiguous results associated with the rice level of rice matching which will deend on ad hoc assumtions. 28

29 welfare has outlined how even the adotion of an economic aroach to art. 82 does not hel in solving some of the ersistent antitrust dilemmas such as the choice between short-term equilibria which imrove actual consumers welfare to the detriment of actual cometitors, and long-term cometitive equilibria which sacrifice actual consumers welfare. These dilemmas seem to go beyond the choice of an economic aroach to art. 82, relying on the secific aims ursued by the cometition olicy authorities. 29

30 References Aghion, P. and P. Bolton Contracts as a Barrier to ntry. American conomic Review 77: Antitrust Law Journal Symosium on Post-Chicago conomics. Antitrust Law Journal 63: Armstrong, Mark & Vickers, John, Price Discrimination, Cometition and Regulation, Journal of ndustrial conomics, Blackwell Publishing, vol. 41(4), ages , December. Baker, J The Case for Antitrust nforcement. Journal of conomic Persectives 17: Bork, R The Antitrust Paradox: A Policy At War With tself. New York: Basic Books. Brennan, T Understanding Raising Rivals Costs. Antitrust Bulletin 33: Brodley, Joseh F. & Ching-to Albert Ma (1993), Contract Penalties, Monoolizing Strategies, and Antitrust Policy, Stanford Law Review, 45 Coate, M. and J. Fischer Can Post-Chicago conomics.survive Daubert? Akron Law Review 34: Crandall, R. and C. Winston Does Antitrust Policy mrove Consumer Welfare? Assessing the vidence. Journal of conomic Persectives 17: Dewey, D nformation, ntry and Welfare: The Case for Collusion. American conomic Review 69: asterbrook, F The Limits of Antitrust. Texas Law Review 63: dlin A Do Guaranteed-Low-Price Policies Guarantee High Prices, and Can Antitrust Rise to the Challenge, Harvard Law Review, Vol. 111 dlin, A Stoing Above-Cost Predatory Pricing. Yale Law Journal 111: lhauge, Why Above-Cost Price Cuts to Drive Out ntrants Are Not Predatory and the mlications for Defining Costs and Market Power. Yale Law Journal 112: Facey, B. and D. Assaf Monoolization and Abuse of Dominance in Canada, the United States, and the uroean Union: A Survey. Antitrust Law Journal 70: Gelman, J. and S. Salo Judo conomics: Caacity Limitation and Couon Cometition. Bell Journal of conomics14: