Here is the remarks as it was prepared for delivery:
I want to thank Eric and Bates White for the invitation to speak with you tonight. The last time I saw Eric in person was shortly before the pandemic at his wedding. I know several of the economists in the audience tonight were there as well. Eric, let me apologize in advance that my remarks tonight will not do justice to the fun we all had at that lovely event.
Tonight, we are here for a different kind of commemoration. In the fond memory of Professor Hal White, this event gives us the opportunity to think deeply about antitrust and economic policy. Professor White’s legacy includes enormous contributions to the field, not least including an approach to examining our assumptions and practices that I believe will have staying power. I am delighted to deliver this keynote in person for the first time since the onset of the pandemic in 2020, and proceed tonight in the hope that the tradition will continue in Hal’s honor for many years.
- Introduction
I will start with two disclaimers. First, the standard DOJ disclaimer. These are my own views and don’t necessarily reflect those of the Department of Justice.
Second, a disclaimer I feel I must give when addressing a room with this many PhDs: I am not an economist. My training in economics comes almost entirely at the hands of the kind members of the Antitrust Division’s Economic Analysis Group who have, over more than 10 years at the division, humored from me a lot of questions about the underlying details of their economic analyses. I am particularly indebted to our former DAAG Nancy Rose, former Economics Director Bob Majure, Jeff Wilder, and the late Wayne Dunham. One of the great things about the expert economists in EAG is their engagement with attorneys, early and often, as we work up cases together.
I first encountered Hal White as an undergraduate physics major. Like many students toiling away at statistical thermodynamics, I was confronted with the dueling challenges of both pronouncing heteroskedasticity, and testing for it. Professor White’s test was fortunately already programmed into Stata for me, so I was able to solve at least one of those. As I think those here know, the White test has traveled far beyond economics into nearly every discipline relying on statistical analysis.
There’s a terrific video on the UC San Diego website in which Professor White talks about his inspiration for pursuing that work. He said he started with the idea that “We should be more up front about the fact that we are really doing exploration of the unknown, and that if we build in assumptions to our models, they could easily be wrong, and we need ways to check them.”[1] From that starting point, he developed his least-squares paper, which led to his heteroskedasticity work and the prodigious research career that followed.
That inspiration could easily describe much of what we’re doing in antitrust policy at the Department of Justice today. When we build in assumptions, they can easily be wrong, and we need ways to check them. In too many respects, models and assumptions we have come to rely on do not fit with market realities. They fail to capture important predictors of harm to competition, and important consequences of harm to competition as well. Far too often, reliance on outdated or incorrect assumptions leads to underenforcement. I’ll get to that in a moment.
We meet tonight in a time of great challenge and great opportunity in antitrust. Corporate power has grown to levels that leave our fellow citizens concerned and confused. On a daily basis, we at the division find ourselves asked how competition law went astray, and what we can do to reinvigorate antitrust enforcement and meet our nation’s challenges.
If we are looking for evidence that our prior assumptions in antitrust have been wrong, I think the broad concern swelling up in communities all around us is compelling. There are also of course more empirical measures, like the increase in price-cost markups,[2] and the decrease in new firm formation, that speak to these issues.[3] And there are concentration measures as well, alongside a robust methodological debate.[4] But when we find ourselves in an era where lay people who hadn’t heard of antitrust ten years ago are calling for greater enforcement, I think we should listen and try to understand what is animating their concerns.
And to correct a common misconception I hear in antitrust circles, the attention to antitrust in Congress and the media are not driving popular interest in reinvigorating competition in the American economy, they are responding to it. The root cause of the popular interest in antitrust is the dissatisfaction felt by many millions of Americans with a rise in corporate power and a corresponding reduction in their economic liberty. Individual Americans feel this in subtle but important ways every day, such as when they face too few choices how to connect with their loved ones, what private data will be extracted by whom, or where to sell their labor.
But the individual’s economic liberty is a foundational premise of our capitalist democracy. That is why those words go so well together — capitalism and democracy are both premised on individual freedom, choice, and opportunity. As Thurgood Marshall said in Topco, the antitrust laws are “as important to the preservation of economic freedom and our free-enterprise system as the Bill of Rights is to the protection of our fundamental personal freedoms.”[5] Those freedoms must be protected.
- Correcting Erroneous Assumptions in Antitrust Policy
Tonight, in the spirit of Professor White, I will ask you to think carefully about four assumptions that have been commonly used in antitrust in recent decades, and to consider how reassessing them can help us more effectively enforce the antitrust laws.
A. Monopolies are Not Inherently Self-Correcting
The first and probably most problematic assumption has been that monopolies are inherently self-correcting, such that we should intentionally underenforce. This assumption traces most prominently to Judge Frank Easterbrook’s 1984 article on the limits of antitrust. As the argument goes, assuming monopoly is “self-destructive,” “judicial errors that tolerate baleful practices are self-correcting while erroneous condemnations are not.”[6]
The assumption that monopoly is self-correcting has been used to justify underenforcement in a wide range of circumstances, ranging from Trinko[7] to the withdrawn Section 2 report briefly issued by the Antitrust Division in 2008.[8] It even appeared as recently as the 2018 oral arguments for Ohio v. American Express, when Justice Gorsuch asked “why shouldn’t we take Judge Easterbrook’s admonition seriously, that judicial errors are a lot harder to correct than an occasional monopoly where you can hope and assume that the market will eventually correct it.”[9]
The answer to Justice Gorsuch’s question is that we should not take Judge Easterbrook’s admonition seriously because it was wrong — the assumption that markets are inherently self-correcting is not true.[10] Entry barriers are quite real, whether exclusionary, regulatory, or natural features of markets.
The first two of those barriers to entry — exclusionary and regulatory barriers — arise from the fact that monopoly rents, as it turns out, are profitable. A portion of those rents can then be spent to undermine competitors or pay off third parties, either of which create exclusionary barriers to entry. Meanwhile, regulatory barriers to entry may already exist when a monopoly is obtained, or they may be acquired by spending monopoly rents to influence regulatory and political processes. Indeed, the danger of monopoly to our political process is not a novel proposition — it was one of the core concerns animating the Sherman Act.
In addition to those barriers to entry that result from monopoly power, natural characteristics of markets may create significant entry barriers. In the modern economy, network effects are incredibly common. That has enormous benefits for society, but flips Judge Easterbrook’s analysis on its head. A robust literature explains how network effects create barriers to entry and render markets prone to tipping.[11] If anything, markets with network effects skew toward monopoly, not the other way around. If we take error cost analysis seriously, that would augur towards favoring overenforcement, not underenforcement, in these markets.
For my part, I think we should target neutral and just enforcement of the laws. That maximizes not only competition values, but the rule of law values that undergird our constitutional democracy.
B. Digital Networks are Cooperative – They are Not Built Solely by Their Owners.
Trinko incorporates a second problematic assumption — that the principal investment in a monopoly product is always made by its owner, and so we should underenforce the Sherman Act to protect those investments.
The reason Trinko made this assumption is understandable. Trinko was about phone networks, not digital ones. It expressed concern that the incentive to build expensive, physical infrastructure would be undermined by a rule sometimes requiring monopolists to permit competitors access to that infrastructure. It applied but narrowed Aspen Skiing, involving competitors’ access to tickets on the monopolist’s ski lifts.
Writing at the time, Justice Scalia had no reason to foresee that a coming sea change in technology would undermine this logic. One thing Aspen Skiing and Trinko had in common was expensive physical infrastructure. Ironically, both cases involved tall poles with wires strung between them — in one case carrying phone calls, and in the other skiers. In both cases, the investment in that infrastructure had been borne primarily by the monopolist.
Today, many digital platforms are radically and materially different. The wires, poles, and fiber-optic cables that support incremental users are already there. The additional investment in physical infrastructure to add each marginal user of a software service over those pre-existing connections may not be zero, but it is comparatively small. In these markets, the connections from which the platform derives its value are often built by users and third parties with little incremental cost to the platform owner. In the same way that a telephone network’s value increases with each new wire the phone company strings from a pole to a home, the value of a search engine increases with each new page someone else adds to the web. The value of a social network grows with each new user who builds a profile and posts content, and each new third party that invests in pages or widgets. The value of an app store grows with each app someone else contributes to the platform.
You can see the critical difference. In many digital markets, the incremental connections that grow the value of the platform are contributed not by a monopolist owner, but by users and third parties. This has important implications for competition policy. Relative to the physical networks of Trinko, our concern with disincentivizing the monopolists’ investment shrinks, at the same time that our concern should grow that third parties may be disincentivized from investing in the platform.
We need to recognize that monopoly maintenance can be an effective and problematic strategy that entrenches dominance at the same time that it discourages innovation. Threats of exclusion, discrimination, or retaliation for competing are of greater concern when they reduce third parties’ incentives to build the platform out further. These and other forms of moat building, such as nascent competitor acquisitions, can have profoundly negative effects on the dynamism of our markets.[12]
C. Coordinated Effects are a Problem Posed by Oligopolies
The third assumption I want to highlight is embedded, like heteroskedasticity, in certain models. This is the assumption that mergers to oligopoly do not inherently create problems because coordinated effects simply do not exist.
We often find this assumption in merger simulations submitted to the department to suggest that efficiencies will measurably outweigh the harms of a transaction. I am sure this audience is familiar with the fact that, for all their complexities, merger simulations are theoretical unilateral effects calculations. On the harm side of the ledger, they only even attempt to assess one type of harm, leaving the very real problem of parallel accommodating conduct aside. Then they fold in marginal cost reductions and purport to show the merger will, on balance, lower consumer prices. Not all merging parties take this approach, but many do.
I have to say that this strikes me a little like building a plane based on a flight simulator that leaves out wind resistance. How can we purport to build an accurate simulation that does not even attempt to interact with one of the foundational forces driving the exercise? I would not get on that plane.
This assumption survives, as I understand it, because the magnitude of potential coordinated effects from a reduction in the number of competitors can be difficult to measure. We do have some post hac measurements, such as the recent paper by Miller, Sheu, and Weinberg, retrospectively assessing how coordination strategies in the beer industry were exacerbated by consolidation. Their results obtain even in the presence of marginal cost efficiencies sufficient to offset unilateral effects.[13]
Somehow, we morphed from disregarding coordinated effects because of challenges quantifying them to just assuming they do not exist. In anything but the most strident approach to intentional underenforcement, ignoring the unquantifiable invites systematic error.
All along, the law has provided a simple but effective tool for stopping consolidation before the risk of oligopolistic interdependence arises. The structural presumption set out by the Supreme Court in Philadelphia National Bank does just that.[14] Neither the agencies, nor the courts, nor the limits of models have the authority to rewrite that legal standard unilaterally. As the Supreme Court said in PNB, Congress designed the Clayton Act and the Anti-Merger Act of 1950 to “prevent undue concentration,” and therefore a merger resulting in a firm with control of undue share of a relevant market is unlawful absent rebuttal evidence.[15] In the nearly 60 years since, courts have consistently upheld the structural presumption as a sufficient basis to condemn a transaction, even standing alone. This was most recently reiterated by the Third Circuit in the Hackensack opinion earlier this year.[16]
The approach of recognizing that the risk of coordinated effects rises as the number of firms falls has also been increasingly validated as models of coordination have evolved in recent decades. For example, Simon Loertscher and Leslie Marx made a useful contribution in the Journal of Law and Economics last fall.[17] Their theoretical model shows how a reduction in the number of firms materially increases the risk of coordinated effects. Of course, no heuristic is a perfect predictor. But when a merger creates or aggravates an oligopoly market structure, both the law and sound policy demand we shift burdens to the merging parties to show that the merger does not risk substantially lessening competition.
D. Workers, and Labor Markets, are Important
I will finish with a fourth assumption that I think you are already aware no longer holds. That is the assumption that harms to competition in labor markets are not worthy of investing prosecutorial resources.
For too long, antitrust enforcers, and the antitrust community in general, paid too little attention to labor market issues. But a robust literature has been developed by Elena Praeger, Eric Posner, Ioana Marinescu, and others, demonstrating the degree of market power employers exercise over employees in many markets.[18] Workers’ wages depend on the degree to which employers compete for them — everyone who has negotiated a salary knows that. Yet in too many markets employer power suppresses worker wages.
As we review the merger guidelines, we are thinking carefully about ensuring a sufficient framework is in place to evaluate the labor market impacts of mergers. Under the Clayton Act, an unlawful risk of harm in any relevant market renders a transaction unlawful, whether downstream or upstream, and notwithstanding the size of that market relative to the price of the transaction.
- Conclusion
Let me stop there — there are other assumptions to think carefully about, but I promised Eric I would finish before the desert service. If it’s not clear, we are taking a hard look at the assumptions underlying antitrust practice. The division has an obligation to ensure that the tools we apply reflect the law and help us ask the right questions about modern markets. It’s hard work but it’s interesting, fun, and important. And in the spirit of my current and departed friends in EAG, it makes for collegial and enlightening discussion. It is an incredibly exciting time to be thinking about antitrust policy, and we look forward to the debates and developments to come.