The baseline for testing predatory pricing in the Section 2 Report is average avoidable cost (AAC), together with recoupment as a structural test (Report, p. 65). The AAC test or reasonably close variations, such as average variable cost or short-run marginal cost, seems about right. However, differences among them can become very technical and fine. The Report correctly includes in AAC those fixed costs that “were incurred only because of the predatory strategy, for example, as a result of expanding capacity to enable the predatory sales.” (Report, pp. xiv, 64-65) Such a strategy would make some sense for a predator if the fixed costs in question are easily re-deployed once the predation has succeeded – for example, in the case of an airline whose planes can be shifted to a different route. The test virtually guarantees that in industries that require heavy investment in production capacity that cannot be redployed the test will approach strict average variable cost. In cases where fixed costs are relatively high, an investment of this nature that lasted only through the predatory period and became excess capacity thereafter would not be worth it. Further, if fixed costs are low the market is almost certainly not prone to monopoly to begin with. AVC is probably underdeterrent, but it is also probably the best we can do without chilling procompetitive behavior.
However, when prices are under AVC, then a strict recoupment requirement (see Report, pp. 67-68) is unnecessarily harsh. Proving recoupment requires a prediction about the dominant firm’s prices, costs, and output over a defined future period, which in turn requires a prediction about when new entry will occur, how many firms will enter, and their growth rates. As a result recoupment is much too difficult to prove and does not serve to distinguish aggressive promotional price cuts from those that are anticompetitive. Rather, structural proof should consist of those things that are ordinarily required in a Section 2 case; namely, a dominant share of a properly defined relevant market and high entry barriers. That is, the question should be “is durable monopoly pricing in this market possible,” but not “can predation be predicted to yield a durable period of monopoly pricing with sufficient monopoly returns to pay off the investment in predation.” As a factual matter the former requirement is much more manageable and requires far less speculation. An important additional ingredient is causation in the classical tort sense – namely, can the plaintiff show that the prices below average variable cost were of sufficient magnitude and duration to cause its exit from the market? Sporadic or episodic price drops below AVC are unlikely to meet this requirement.
The biggest concern is with false positives. Are there cases in which prices were below AVC for a substantial length of time to meet the causation requirement and the structural components for monopoly were present, but where we would not want to condemn the conduct because dollars-and-cents proof of recoupment is not possible. I doubt it.
The Report’s test for bundled discounts is based on the same premise as its test for predatory pricing – namely, the test should seek out circumstances when an equally efficient rival cannot match the price. To this end, the ordinary predatory pricing rule should apply when one or more substantial rivals makes all of the goods in the bundle or, presumably, when such a bundle can readily be assembled by rivals (“bundle-to-bundle” situations; pp. 95, 99). The test becomes a little different when no substantial rival makes all of the goods in the bundle. In that case the Report generally follows the recommendation of the Antitrust Modernization Commission, which is a “discount allocation” test. Under that test one attributes the entire discount to the competitive (exclusionary) good and then asks whether the discount drives the price of that good below average variable cost. (p. 99). Suppose a dominant firm sells products A and B at prices of 8 and 5 respectively, but charges a price of 11 for an A/B bundle. No other firm makes both A and B and we are interested in possible exclusion of firms making only B. In that case we attribute the entire discount to B and ask whether an equally efficient producer of B could match it, using ordinary predatory pricing rules.
Although the Third Edition of the Antitrust Law treatise (Volume 3A, ¶749) approves this formulation, I now believe it is too careless, particularly in light of the Ninth Circuit’s dicta in Cascade Health Solutions v PeaceHealth, 515 F.3d 883, 909-910 (9th Cir. 2008), suggesting that bundles that flunk this test are presumed to be unlawful. Rather, an “incremental cost” test is a better opening formulation. That is, one should ask whether the incremental cost of adding B to A is fully covered by the incremental price that is being charged. In the prceeding example the dominant firm was charging 8 for product A alone and 11 for the A/B bundle; so the relevant question would be whether the three additional dollars when product B was included was sufficient to cover the incremental cost of adding B to the bundle.
This test is identical to the attribution test only when there are no joint costs – that is, when no economies of scope in production or distribution apply to joint A/B sales. Most cases of bundled sales will involve at least some joint costs. In PeaceHealth, for example, the defendant was accused of bundling primary, secondary and tertiary medical care. But while these levels of care differ, a hospital delivering all of them will almost certainly use at least some common staff, facilities and durable equipment. Given the size of fixed costs in this market there could be very considerable cost savings to packaged sales that the attribution test does not pick up. For example, secondary and tertiary care might each individually require purchase and maintenance of a costly scanning device, but a hospital offering both levels of care could run one scanner and allocate the fixed costs over both types of care. So the relevant question is: when product B is added to the bundle, is the increased price of the bundle sufficient to cover the increased cost of including it. For example, if the separate costs of offering A and B are A=8 and B=5 but joint production of one A and one B costs only 11, then the attribution test as the AMC defines it can yield false positives. The seriousness of this problem depends on how often joint costs are present. I suggest that they are present most of the time. In general, markets that are subject to monopolization are characterized by significant fixed costs, and for most bundles at least some of these are likely to be common. The same thing is true of distribution costs. See Erik Hovenkamp & Herbert Hovenkamp, Exclusionary Bundled Discounts and the Antitrust Modernization Commission (53 Antitrust Bull. 517 (2008)).
Another problem with the attribution test is that it perversely finds competitive harm most likely in the markets that are most competitive to begin with, because prices in those markets are already close to cost, leaving little room for an attributed discount. For example, the salt injector monopolist who gives a discount to buyers who takes its salt flunks the attribution test if the market price for salt is competitive to begin with. As a result, a serious structural inquiry is necessary, although I would not require “recoupment” as the Report does.
Finally, the attribution test becomes extremely complicated to apply when (a) buyers purchase in variable proportions or (b) the bundles contain more than two products. See Erik Hovenkamp & Herbert Hovenkamp, Complex Bundled Discounts and Antitrust Policy (__ Buffalo. L.Rev. __ (2009)). In general, as the proportion of the competitive products in a bundle increases the requirements of the attribution test are less likely to be met. Alternatively, if the bundle contains more than two goods, whether a package discount excludes depend on how many of the goods in the bundle a particular rival offers.