Predatory Pricing: A New Theory

The issue of predatory pricing is a complex one with a chequered history. Although it has had legal recognition for over a century, it was not until McGee's work in 1958 that it was subjected to economic analysis. His conclusion was that predation would rarely be profitable and consequently would be unlikely to occur. Although his results ran counter to standard intuition, they held sway as late as 1980 since no other theory emerged to refute it. Predatory pricing is a strategic action of temporarily reducing prices in order to increase long-run profit. There is no one unambiguous definition of predation stating how low prices must fall, how rivals will suffer etc. Consequently its identification has been fraught with difficulty as the presence of numerous legal rules suggests. Since the 1980s, however, economic analysis of the subject has taken a new approach yielding models of predation as a profitable and likely strategy for firms. Using this analysis, clearer definitions of predation should be possible making it easier to identify. Drawing on this, the antitrust laws, which developed contrary to McGee's results, could be made more relevant and effective. After giving some insight into antitrust law to date, I will focus on one branch of the new theory and outline the implications arising from it for present antitrust law.

Within the last forty years, economic theory and antitrust law on predatory pricing have evolved largely independently of each other. When Areeda and Turner (1975) introduced their cost-based test to determine if pricing is predatory, no theoretical model of predatory pricing existed. At that time it was in fact "discounted as an economic impossibility". Despite the theory, antitrust law recognised the existence of predatory behaviour, and the arrival of the Areeda-Turner rule, it was believed, brought the needed coherence to the legal analysis of predation which has existed for three-quarters of a century. Attracted by its apparent simplicity, American circuit courts of appeals embraced the rule, and up to the present day this rule and variations of it are still in use. Briefly, the basis for the rule is that a rational firm will not have below-cost prices, unless it is engaging in strategic, predatory behaviour. The rule is, therefore, that any price less than short-run marginal cost (SMC) is predatory and hence illegal. Since difficulties arise in the measurement of SMC, average variable cost (AVC) is used as a proxy. This rule is connected to the theory of perfectly competitive behaviour, where, in equilibrium, Price = SMC. A number of other cost-based tests have since been formulated, some of which allow more for the dynamic nature of predation, unlike the Areeda-Turner rule which focuses on static, or short-run, analysis only.

The motivation for antitrust law is to "promote full and fair competition and to reap the benefits that competition brings with it". The latter half of this statement is particularly important – by reaping the benefits, the welfare of individuals in society will be increased. Increasing welfare in society is a key government objective, and intervention is justified on the grounds of it doing so. Speaking of antitrust law on predation, Joskow and Klevorick note, "a major obstacle to a smooth transition from objective to operational policy is the problematic nature of the ‘offence' with which we are concerned". In particular, predatory and competitive price cuts can be difficult to distinguish since SMC is hard to observe. It is crucial that a legal rule inhibits predatory price cuts only and not competitive, welfare enhancing ones also. Extending the analysis to imperfectly competitive markets, it may be revealed that predatory price cuts do not unambiguously reduce welfare. In this instance, a legal rule inhibiting all predation could violate the very objective of government intervention.

Ideally antitrust law would have small probability of error and a low cost of implementation. Given the nature of predation, the simple tests, although cost effective, are those most likely to be erroneous. Antitrust law faces a trade-off between implementation cost and accuracy of tests. The Joskow and Klevorick rule (1979) is based on a more detailed two-tier approach which aims to minimise the sum of implementation costs and the costs of errors. In the first tier, looking at market characteristics filters out those firms likely to engage in successful and rational predation. In the second tier, these firms are subjected to further scrutiny in the form of a cost-based test (explicitly, P < Average Total Cost). They also deem short-lived price cuts, which may be above ATC, to be predatory.

At present, although the US courts still rely on simple cost-based tests, they require evidence of the possibility for recoupment after predation, and in some cases of intent, also. The extra conditions have created a great level of ambiguity between the courts and the implementation costs have risen. Consequently, fewer and fewer cases of predation have been proven legally, ironically just as the concept has been modelled as an economic possibility. Zerbe and Mumford are critical of this trend and maintain that "perhaps predation should not exist legally, but if so this conclusion should not be derived from incorrect premises". The most recent trend within the EU is a focus on selective predatory price-cutting by dominant firms. There is no unanimity with regard to how low prices must be cut to be illegal, however. Andrews notes that, despite the ruling in the recent Irish Sugar case, ordinarily only those below ATC will be illegal, but in some cases those below ATC but above AVC will not.

To clarify the intuition of these models I will take the example of a market with two firms where firm 1's costs are private information, and can take on only two values, high or low. Quantity is the strategic variable. The profits of firm 2 depend on firm 1's costs – the lower these costs are relative to his or her own, the less profitable it will be for firm 2 to continue producing. I assume that firm 2 will exit if firm 1 is a low-cost firm. Low prices signal low costs and high prices signal high costs. By expanding its output and lowering its price firm 1 can signal to firm 2 that it is a low-cost firm, and hence induce it to exit. Once firm 2 has exited, firm 1 has monopoly power in that market. This reduced price needs bear no direct relation to the predator's costs. The price must only be low enough to make firm 2 believe it is unprofitable for it to produce. Where firm 1 has any cost advantage over firm 2, this reduced price is likely to be above firm 1's cost. Predation involves an increase in long-run profit for the predator at the expense of reduced short-run profit. The incentive to signal to the rival will depend on this trade-off which in turn depends on the size of the price cut, and on the actual cost structure of firm 1. In some cases both low- and high-cost types of firm 1 will find it optimal to expand output by the same amount in the short run. This is known as a pooling equilibrium. In this case firm 2 cannot determine whether firm 1 is actually a high- or a low-cost type after the signal has been sent. Alternatively, the optimal short-run strategies of the two firm types may differ in which case there will be a separating equilibrium. In this case the high-cost firm can do better by revealing its type than by mimicking the low-cost firm. Consequently only the low-cost type is prepared to expand output to credibly signal its type. The correct signal is sent and complete information emerges.

Since the two firms are already in the market, signalling models have an added complexity which enriches the final result. With firm 2 in the market the predator's output, (be it a high- or a low-cost predator), depends on what firm 2 is already producing as well as on demand and its own cost structure. But firm 2's output depends on what firm 1 produces, in accordance with its best response function. Unable to directly observe firm 1's type at the outset, firm 2 attaches probabilities to it being low and high cost types. In the case of a separating equilibrium in the first period, firm 2 will base its output decision on the expected output of firm 1, which it derives using these probabilities, since firm 1's actual cost, and hence output, are unknown. Firm 2's output is a decreasing function of both types of firm 1's output. We already know that in a separating equilibrium the low-cost firm 1 will send a signal and produce more than it otherwise would, since it is profitable for it to send a credible signal. The possibility of firm 1 being a low-cost firm raises firm 2's evaluation of expected firm 1 output. This means that firm 2's output will be reduced. But since firm 2's output is lower a high-cost firm 1 will expand its output, even though it is not signalling. This result, originally derived by Mailath, implies that in signalling models the short-run equilibrium has the feature that both types of firm 1 produce more than in the case of symmetric information and firm 2 produces less. The very possibility of firm 1 being a low-cost firm causes firm 2 to reduce its output. Consequently in this situation with two incumbents in the market and asymmetric information the predator has not just one but two motives for predation - to induce exit (as already mentioned), and also to force its rival to reduce its output when it remains in the market.

To date neither the signalling models nor any other of the recently developed theories of predatory pricing have been embraced by antitrust law. Klevorick has lamented that "the lack of impact that the recent equilibrium models of predation have had on the development of antitrust law concerning predatory pricing is unfortunate". In particular, the possibility of asymmetric information appears to have been completely overlooked. As mentioned, a major difficulty for antitrust law is distinguishing predatory from competitive price cuts. In view of this, the courts have favoured cost-based tests for identifying predatory price cuts. (Above-cost price cuts are usually viewed as competitive rather than predatory.) One of the strongest implications to emerge from the recent models, however, is that successful predation does not require prices to fail the Areeda-Turner, or any cost-based test. The new models suggest that, contrary to the arguments of Mastromanolis, identification of above-cost predation is correct, and such test standards should be applied. Given this, it is other factors, such as the relevant market or information structure, that must distinguish predatory from competitive price cuts. As somewhat of an exception, the Irish Sugar decision by the European Commission in 1997 was against selective, although not below-cost, predation. This suggests that EU law may, albeit independently of these models, be developing in the right direction.

A further implication is that predation will not necessarily reduce welfare. In separating equilibria, complete information emerges and the rival can make its decision without bias, thus enabling the market to function more efficiently – the rival will only remain if it is cost-effective and profitable to do so. For instance, if demand is truly low, it may only be efficient for one firm to supply the market. Again, contrary to Mastromanolis' arguments, any simple rule applied indiscriminately will be erroneous therefore, by preventing all signalling, and thus protecting some inefficient firms. Depending on the cost of these errors, the efficiency gain of eliminating welfare-reducing predation may not be sufficient to warrant antitrust law on predation. Indeed, Milgrom and Roberts note that "it may be best simply to give up on attempts to control predation, even if one believes that it can and does occur".

Ordover and Saloner believe that, despite the possibility of error, simpler, more explicit tests of predation are still preferable, should courts continue to control predation. Clearly the currently ambiguous definitions of predation in both the EU and the US courts could be clarified with the aid of the new models, enabling less erroneous and yet more explicit tests. Saloner suggests improving upon the Joskow and Klevorick rule by introducing a third tier that would examine the information structure of the industry. In this way predation which increases welfare may be distinguished from that which reduces welfare and ‘good' signalling will not be inhibited. From the discussion earlier, simpler, unambiguous tests will be implemented more easily by the courts. The focus now must be on using the new theory to construct usable tests which identify and inhibit welfare-reducing predatory price cuts only. Drawing on signalling models, Scharfstein has made an attempt at this – he models the correct incentive structure for firms, via the governments strategy of search, investigation and fine, to ensure that firms give only correct signals, without discouraging them from engaging in competitive behaviour.

The so-called ‘new' theories of predatory pricing that have developed over the last fifteen to twenty years have reopened the debate on predation. By looking at the signalling models alone several implications for antitrust law emerge. Essentially they serve to highlight necessary changes for antitrust practice in the area, which arguably has lacked a firm economic foundation to date. Cost-based tests may be too limiting and can no longer be relied upon to distinguish between predatory and competitive price-cutting. Furthermore, predation may not always violate government welfare objectives. Since these models do not refute the complexity of predation, they may confirm one argument that it is not cost-effective to control predation. More optimistically, however, they may help to create a less ambiguous definition (and possibly more than one) of predation from which the courts could derive a more effective control of the practice. In particular they provide guidance on more accurate identification of predatory behaviour, and where it is identified, determining its implications for the welfare of individuals in society, and thus whether or not it should be prohibited.

Areeda, P.E. & D.F. Turner (1975) "Predatory Pricing and Related Practices Under Section 2 of the Sherman Act" in Harvard Law Review 88, pp. 697-733.

Farrell, J. & G. Saloner (1986) "Installed Base and Compatibility: Innovation, Product Pre-announcements, and Predation" in American Economic Review 76, pp. 940-955.

Joskow, P.L. & A.K. Klevorick (1979) "A Framework for Analysing Predatory Pricing Policy" in Yale Law Journal 89, pp. 213-270.

Klevorick, A.K. (1993) "The Current State of the Law and Economics of Predatory Pricing" in Antitrust and Industrial Organisation 83(2), pp. 162-167.

Mailath, G. (1984) The Welfare Implications of Differential Information in a Dynamic Duopoly Model, Princeton University.

Mastromanolis, E.P. (1998) "Predatory Pricing Strategies in the European Union: A Case for Legal Reform" in The European Competition Law Review 4, pp. 211-225.

McGee, J. (1958) "Predatory Price Cutting: The Standard Oil (N.J.) Case" in Journal of Law and Economics 1, pp. 137-169.

Milgrom, P. & J. Roberts, (1982) "Limit Pricing and Entry under Incomplete Information: an Equilibrium Analysis" in Econometrica 50, pp. 443-459.

Ordover, J., & G. Saloner (1989) "Predation, Monopolization and Antitrust" in Schmalensee R. and R.D. Willig (eds.) in The Handbook of Industrial Organisation.

Roberts, J. (1986) "A Signaling Model of Predatory Pricing" in Oxford Economic Papers (Supplement) N.S. 38, pp. 75-93.

Saloner, G. (1987) "Predation, Mergers and Incomplete Information" in Rand Journal of Economics 18, pp. 165-186.

Salop, S. & C. Shapiro (1980) "A Guide to Test Market Predation", Mimeo.

Scharfstein, D. (1984) "A Policy to Prevent Rational Test-Marketing Predation" in Rand Journal of Economics 15, pp. 229-243.

Tirole, J. (1988) The Theory of Industrial Organisation. MIT Press.

Zerbe, R. Jr. & M.T. Mumford, (1996) "Does Predatory Pricing Exist? Economic Theory and the Courts after Brooke Group" in The Antitrust Bulletin Winter, pp. 949-985.