The GSK France case in search of a

The GSK France case: in search of a consistent story Michele Polo (Bocconi University and LECG) 2007 ACE Conference, Toulouse

The case • Very complete summary in Choné’s presentation • I do not enter into relevant market issues (interesting comments in Avenel’s presentation) • I focus on the predatory pricing issue: – Identify the anticompetitive story proposed by the CC; – Build an alternative competitive explanation – Which factual elements allow to discriminate between the two? – Did the decision run a proper evaluation of these factual elements?

The anticompetitive story of the CC • • Actors: Glaxo France (predator); Flavelab (prey); Merck, Panpharma, Lilly, …(other firms) Markets: pharmaceutical markets at the turning point when patent protection is expiring; supply to hospitals (bidding markets). – Market A: Acyclovir; – Market B: cephalosporins (2 molecules: cefuroxime and cefamandole, but only the former is considered!) Qualitative argument: the patent holder establishes a reputation of aggressive behaviour in one (small) market and deters entry also in other markets where patent protection is going to expire. Predation strategies establish a link across otherwise different and separate markets Reference to two models of predatory pricing: – Reputation: below cost pricing in market B (cefuroxime) where Glaxo is not dominant with the aim of constructing a reputation of aggressive player and deter entry in market A (acyclovir) where Glaxo is dominant; – Financial Predation: aggressive pricing to induce negative performance and inability of the entrant to rely on external finance to cover losses

The anticompetitive story of the CC (2) • 1999 -2000: Glaxo prices Zinnat® (market B: cefuroxime) below purchasing price in hospital procurements where it faces Flavelab as a competitor>> selective price cuts; • Glaxo establishes a general reputation of aggressive incumbent towards entrants (also in other markets, including market A) • Flavelab exits in 2000 and goes bankrupt in 2001 • Glaxo maintains a dominant market share in market A (thanks to reputation) after patent protection expires

A competitive story • Supply to hospitals: procurements, bidding markets; • Bertrand-type competition if more than one participant; price = AVC • Price not greater than the reservation value of the hospital (bargaining? repeated interaction? ) when only one participant; price > AVC • Efficient competitors survive (even if modest margins) while inefficient ones exit. • No (strategic) link between market A and market B.

A comparison • The anticompetitive story: – Entry of competitors – Selective price cuts: • price below cost in procurements where the competitor participates; • price with positive margins in procurements where only Glaxo participates; – the competitor exits and Glaxo remains monopolist in market B or restarts predation against new entrants; – Glaxo maintains dominance in market A due to reputation • The competitive story – Entry of competitors – Prices driven by the bidding processes in hospital procurements: • price close to average variable costs when Glaxo and the competitor participate; • prices higher than AVC in procurements where only Glaxo participates; – Efficient competitors survive and inefficient competitors exit; – Market A: an independent market whose evolution is not affected by market B

Predictions The two stories give two different empirical predictions: 1. Evolution of market B after a period of low prices: – – Anticompetitive story: Glaxo remains monopolist in market B or restarts predation against new entrants; Competitive story: efficient competitors survive in market B while inefficient competitors exit 2. Level of low prices: – – Anticompetitive story: p < AVC Competitive story: p ~AVC

1. On the evolution of the market Market B 1999 2000 2001 2002 2003 2004 2005 Glaxo 81. 4 75. 0 73. 0 51. 0 35. 0 22. 0 18. 0 Flavelab 16. 0 16. 5 Panpharma 2. 6 8. 5 27. 0 49. 0 65. 0 78. 0 82. 0 Market A 1999 2000 2001 2002 2003 2004 2005 Glaxo 99. 5 90. 5 80. 9 57. 4 65. 0 46. 0 52. 0 Mercks 0. 5 9. 5 19. 1 42. 6 33. 5 44. 4 25. 4 1. 5 9. 6 22. 6 (cefuroxime) Others

1. On the evolution of the market (2) • The CC argues that Glaxo predates Flavelab in market B until Flavelab exists (2001) and therefore establishes a reputation of aggressive incumbent. • However, in the same market Panpharma enters in 1999 and grows steadly reaching 82% of the market in 2005. Panpharma grows also in the parallel market of cefamandole cornering Lilly. This is inconsistent with the reputation model, that predicts to predate every time entry occurs. • At the same time, the CC claims that Glaxo maintains a high market share in market A and interprets it as the result of successful predation in market B • The evolution of market shares in market A and B in the interpretation of the CC is inconsistent with the reputation story: – – Glaxo is claimed to remain dominant in market A due to successful predation in market B, but Glaxo tolerates to leave market B to Panpharma, inconsistently with a successful reputation building strategy. • Panpharma prices are aligned to the lowest offers by Glaxo (in the So. O): consistent with an efficient competitor surviving (and growing!). • Since patent protection expires in 1999 in market B and in 2002 in market A, it is not obvious that the two markets are following different patterns: in both cases after 4 years the old patent holder has around 50% of the market!

2. The level of low prices • The second key element in identifying the explanation consistent with facts refers to the level of prices in the competitive episodes: p<AVC or p~AVC in procurements where Glaxo faces a competitor • This requires a careful assessment of AVC • Since Glaxo France purchases the cephalosporine from a production unit of the same group, Adechsa, CC uses the transfer price as a measure of the AVC • A key point is whether an internal transfer price can be considered as the average variable cost for the downstream unit.

Transfer prices vs AVC • The argument put forward by CC is formalistic: § 201 “…Glaxo has significant autonomy to determine its sale price, in particular with respect to the other companies in the group or the parent company. Its purhcase prices, paid to any of its suppliers, are indeed relevant costs in applying the cost test” • Significant autonomy has not a clear economic meaning: – it means that the downstream unit (Glaxo France) is the subject setting the final price (through its hospital sales section). – but it does not imply that Glaxo France, in its price setting decision, is maximizing the downstream unit profits rather than the group profits. – For instance, Glaxo France might set the final price under the constraint that p ≥ AVC, where AVC is the average variable cost of the production unit. – Behaving this way, Glaxo France would be autonomously maximizing the group’s profits!

Transfer prices vs AVC (2) • • • The Akzo test is intended to identify a lower threshold (e. g. AVC) below which a firm would not set the price in a competitive environment, >> minimize probability of type-I errors. It uses the predator’s costs to ensure that an “as efficient” competitor would make losses only below this threshold. When we compare competing firms with different size and architecture, as a large multinational company as Glaxo and a small firm, we want to check whether Glaxo pricing would prevent the rival from competing when this latter produces the drug at the same level of efficiency of Glaxo. Hence, the proper measure is Glaxo AVC in the production unit, that corresponds to the equivalent process of the competitor, and not the transfer price (TP), that may reflect a policy of the large company in allocating the fixed costs (not incurred by the producer of generics) across units. The CC fails to construct, or to require Glaxo to produce, these data.

Transfer prices vs AVC (3) Running the Akzo test based on TP is problematic: • p<TP is consistent with p<AVC (i. e. predation) if TP ≤ AVC, ------p---TP----AVC------but • may be also consistent with p>AVC if AVC < TP, i. e. if the transfer price fully covers the average variable production costs with a margin for the fixed costs (not implausible). In this case, as it is with Bertrand competition (competitive story), the equilibrium prices should be lower than TP. -------AVC---p---TP------ Hence, using TP in the Akzo test when TP>AVC induces a type I error according to the competitive story, and a test based on TP does not allow to identify predation.

Conclusions I’m not convinced of the arguments put forward by the CC on two grounds: 1. The evolution of market shares tells us that in mkt B Glaxo was almost replaced by Panpharma while in mkt A Glaxo retains a dominant position. The evolution of mkt B is inconsistent with successful predation in mkt B but the evolution of mkt A is explained by successful predation: a contradiction! 2. The key quantitative test to discriminate between the competitive and anticompetitive stories is not properly run: transfer prices rather than AVC are used.