Comments on Updating the Merger Guidelines
Of course, the Merger Guidelines need to be updated. Except for efficiencies, they haven’t been updated in 17 years. Lawyers and economists with a regular antitrust practice may not require an update in light of their knowledge of the 2006 Commentary, speeches and agency experience. But, the rest of the antitrust world does. The most obvious audience is the courts, who should know what the agencies believe is best practice. Moreover, as FTC Commissioner Kovacic has stressed, the world marketplace for antitrust ideas needs to have the guidance of the US enforcement agencies. They should not have to ferret it out from commentaries and speeches. The Merger Guidelines have been the most emulated feature of US antitrust enforcement worldwide. It would be shame to squander the leadership role.
The “most important” revision is harder to identify. I am caught here between market definition and unilateral effects – and, clearly, revisions to either issue will have important implications for how the agencies and practitioners approach the other issue. Of course, coordinated effects and entry could use some work too. But, let me stick just with the first two and just a sampling of issues there.
The hypothetical monopolist SSNIP test is an elegant but complicated and imperfect methodology. I think it should be explicitly conceded in the Guidelines that market definition is a very imperfect exercise in practice in which the “appropriate” market definition is often controversial and there may not convincing evidence for a single “most appropriate” market. That concession would continue the process of downgrading the perceived importance of market definition in antitrust and focusing the courts more on the bottom line issue of competitive effects. After all, market definition is mainly valued for helping to analyze competitive effects. The imperfections of market definition and the implications for merger analysis have been made by Katz & Shelanski and should be made explicit in the update. In fact, I’d like the Guidelines to start with a “first principles” overview of competitive effects and fit market definition into that framework, not the reverse.
The SSNIP test definitely needs some renovation after all these years. There is considerable controversy today over the proper way to implement the test. For example, standard critical loss analysis often used by economic experts in testimony generally focuses just on the profitability of a SSNIP, whereas the Guidelines focus on whether a SSNIP is profit-maximizing. It also may ignore relevant information. Articles by Katz & Shapiro and O’Brien & Wickelgren (“KSOW”) have proposed a methodology that takes into account the information about demand elasticity contained in the price/cost margins of profit-maximizing Bertrand competitors. This is a step in the right direction.
This KSOW methodology does tend to lead to narrower markets than often are found now. So, the agencies need to decide whether this is the desired outcome or whether they should raise the SSNIP level, or downgrade the importance of market shares and concentration, to compensate.
The “smallest market principle” should be erased. First, the SMP erroneously suggests that a careful application of the ssnip test will identify one and only one relevant market. That outcome is not the case in practice. Depending on the starting point (and there is no unique starting point), the algorithm does not lead to a unique “smallest” market. And, once price discrimination is taken into account (in a negotiated price market, for example) the smallest market may well be a single customer.
The principle also makes no sense as a matter of policy. Suppose there were a narrow market for premium baby food comprised of just Gerber, Beechnut, and organic brands. That finding should not imply that a merger between Gerber and Heinz necessarily would be free of significant competitive issues.
Moving on to unilateral effects, it is clear that the Unilateral Effects section also needs work. One need go no further than the garbled description of the impact of a 35% combined market share. After this part is revised and clarified, I will mourn the lost opportunity of reading it to my students. What a fabulous example of drafting by a committee in conflict! But, the offsetting benefit is that my students, the courts and others may actually understand the enforcement intentions of the agencies, including the issue of whether or not there is a safe harbor.
The Unilateral Effects section also needs to clarify the role of the repositioning. The 2006 Merger Commentary suggests that repositioning is seldom, if ever, sufficient to eliminate concerns. Coupled with tight agency standards towards cognizable efficiency benefits, this has led to a longstanding concern among commentators that virtually every differentiated products merger could be said to raise a significant danger of adverse unilateral effects. I think that this is mainly an evidence issue. If there has been repositioning in the pre-merger market, claims of a post-merger repositioning constraint are more credible. Otherwise it needs to be explained why repositioning would not been observed as the market has been evolving over time.
More attention also must be paid to gauging the magnitude of the predicted unilateral effects, both at the screening stage and in the ultimate inquiry. Greg Werden, O’Brien & Salop, and Farrell & Shapiro each have proposed variants of “upward price pressure indices” (“UPPIs”). These UPPIs rely on information about diversion ratios, margins and efficiency benefits. They could be made an explicit part of the analysis. They also could form the basis for an initial screen that would make more economic sense for unilateral effects than the HHI, which at best is related to coordinated effects. I have heard criticism that the UPPIs cannot be derived rigorously from a general oligopoly model without making certain additional limiting assumptions. True enough, but neither can the HHI!
Formulating an initial screen or safe harbor UPPI will take some work. Should there be an explicit “efficiency credit?” Or, should the agencies simply set a critical level for the UPPI, based on expertise, experience and policy preferences? (In this regard, the HHI thresholds of 1000 and 1800, and the Delta-HHI threshold of 100 were never really justified.) If there is a formal credit, how should it be formulated – as a fraction of cost or price or both? Should the efficiency credit be a “standard deduction” or a “personal exemption?” That is, should cognizable efficiencies proven by the parties be added to the credit (as with a personal exemption), or replace the credit (as itemized deductions replace the standard deduction)?
And, how should the efficiency credit take into account optimal deterrence, including the goal of allowing well-functioning markets for the sale of property and corporate control, two general efficiencies noted in the previous versions of the Guidelines? In addition, what does “incipiency” mean in today’s world — does it mean that the agencies and courts should set standards tilted more towards a concern with false negatives? In my view, optimal deterrence is the most important issue of all, but clearly one of the hardest and most controversial.
The UPPIs also need to be connected up to market definition. The UPPI is a close cousin of the KSOW form of critical loss analysis. Indeed, this connection serves as a reminder that market definition generally has reduced importance in unilateral effects. This does not mean that market definition is irrelevant. Market definition analysis provides a focus on closeness of substitutes, which is central to unilateral effects analysis. The importance of the cross-elasticity of demand was even noted by the Court in its notorious DuPont decision. Market definition also is relevant for coordinated effects – to define a possible coordinating group of firms. And, of course, Section 7 makes it clear that a market must be defined. So, it is clear that market definition cannot be dispensed with, but only downgraded in importance.