Wednesday, July 24, 2024
30.1 C
Los Angeles

How to assess a general-purpose AI model’s reliability before it’s deployed | MIT News

Foundation models are massive deep-learning models that...

El Salvador: Rights Violations Against Children in ‘State of Emergency’

El Salvador’s state of emergency, declared in...

Vietnam: New decree on cashless payments

On 15 May 2024, the Government officially...

Office of Public Affairs | Principal Deputy Assistant Attorney General Doha Mekki of the Antitrust Division Delivers Remarks at GCR Live: Law Leaders Global 2023

OpinionOffice of Public Affairs | Principal Deputy Assistant Attorney General Doha Mekki of the Antitrust Division Delivers Remarks at GCR Live: Law Leaders Global 2023

Remarks as Prepared for Delivery

Thank you very much for the warm introduction, Andrew.

Congratulations to all of the conference organizers for putting together this event. It’s a privilege to be in the company of so many exceptional enforcers, thoughtful scholars, and seasoned practitioners. In particular, I am delighted to join my enforcement colleagues from the Federal Trade Commission and the UK’s Competition & Markets Authority who also are delivering remarks over the next two days. Holly and Colin are terrific public servants and wonderful law enforcement partners.

At a time when competition issues enjoy prominence in the zeitgeist, conversations like these can help advance our debates as well as our understanding of the goals and methods of antitrust enforcement. I commend the organizers for lining up dynamic speakers to discuss some of the most interesting antitrust enforcement and policy issues of the day. From market and monopsony power to potential competition doctrine and self-preferencing, these topics canvas areas of lively debate and capture a range of viewpoints. Perhaps most importantly, they reflect issues that have reached the public consciousness and bear on American life in myriad ways.

I. INTRODUCTION

As antitrust enforcers, we internalize the important role of competition in our broader economic vitality. That philosophy is something of a tradition, perhaps best encapsulated in Justice Thurgood Marshall’s observation that the “[a]ntitrust laws in general, and the Sherman Act in particular, are the Magna Carta of free enterprise.”[1] We are no doubt animated by the notion that Congress enacted the Sherman Act to be a “comprehensive charter of economic liberty.”[2]

Rather than establish a complex regulatory regime or hyper-technical rules to police the industrial relations of firms, Congress wrote a law that was simple in its prohibitions and flexible in its application to civil restraints of trade. It’s an exercise in law enforcement, not regulation. Section 1 of the Act outlaws contracts, combinations, and conspiracies that unreasonably restrain trade.[3] Over more than a century, through interpretations of the statute, the Supreme Court and the lower courts have clarified its meaning and application.

We at the Antitrust Division take seriously our law enforcement duty to administer the antitrust laws. If there is an overarching ethos in the Division today, it is a commitment to understanding market realities and to applying the antitrust laws as Congress wrote them and as the courts have interpreted them.

This work is painstaking but worthwhile. You can see this approach reflected in our enforcement. And it guides our other work too. For example, we are working with the Federal Trade Commission to consider updates to the merger guidelines. And we’re looking at other policy statements and guidance documents to consider whether they are fit for purpose.

Lately, we have been thinking critically about a particular type of Section 1 prohibition– anticompetitive information exchanges, which will be the focus of my remarks today.

Courts have long recognized that the exchange of competitively-sensitive information can subvert the competitive process and harm competition. In some cases, information exchanges can facilitate full-blown criminal conspiracies. When they are a means or method of facilitating price or wage fixing, bid rigging, or market allocation, the information exchanges are part of the conduct condemned under the per se rule, which is to say without elaborate inquiry into its effects.

In other instances, information exchanges can violate the Sherman Act even when they are not part of a per se conspiracy. In such cases, the burden shifting framework of the rule of reason uses a variety of analytical tools to assess the actual or likely effects of this conduct. The Division has set forth directional – but ultimately non-binding – guideposts that set out when, based on the characteristics of a particular industry at a particular period, the exchange of competitively-sensitive information may not warrant the Division’s antitrust scrutiny. But when market realities change such that our guideposts no longer serve their intended purpose or fail to accurately reflect the Division’s enforcement posture, we should revisit them.

II. EXCHANGE OF COMPETITIVELY-SENSITIVE INFORMATION

In United States v. Gypsum, the Supreme Court identified two factors to help detect anticompetitive information exchanges: “the structure of the industry involved and the nature of the information exchanged.”[4] These factors are useful, first-order filters that can help us quickly identify problematic conduct.

In examining the first factor, “the structure of the industry involved,” courts may look to the relative concentration of an industry. An embedded assumption here is that more concentrated markets may be more susceptible to coordination – tacit or otherwise.[5] Notably, courts have never held that only oligopolistic markets can give rise to harmful exchanges of competitively-sensitive information. Indeed, a closer review of the case law reveals that courts have sustained information exchange claims among a relatively substantial number of market participants. In American Column & Lumber Co. v. United States, for example, 365 firms adopted a “Plan” to share sensitive information about their hardwood lumber production and sales.[6] In United States v. Container Corp., 18 firms exchanged sensitive price information concerning specific sales of corrugated containers.[7] And in Todd v. Exxon, 14 major oil and petrochemical companies exchanged employee compensation data.[8]

In analyzing the second factor, “the nature of the information exchanged,” courts have considered the relative freshness or staleness of the information exchanged. The Supreme Court has appropriately highlighted the perniciousness of “[e]xchanges of current price information,” which “of course, have the greatest potential for generating anti-competitive effects and although not per se unlawful have consistently been held to violate the Sherman Act.”[9] It stands to reason that the exchange of forward-looking, competitively-sensitive information should be even more concerning.

Courts also have looked at the degree to which the exchanged data has been aggregated. These decisions considered how, in light of the facts and market realities at the time, the information could facilitate and result in the type of behavior that the antitrust laws condemn. The Second Circuit explained in Todd that “[p]rice exchanges that identify particular parties, transactions, and prices are seen as potentially anticompetitive because they may be used to police a secret or tacit conspiracy to stabilize prices. . . . Courts prefer that information be aggregated in the form of industry averages, thus avoiding transactional specificity.”[10] But facial aggregation of data alone has been held to be insufficient to save otherwise problematic information exchanges. In Todd, the Second Circuit looked beyond data that appeared to be somewhat aggregated to conclude that the defendants had the ability to effectively disaggregate it, raising serious antitrust concerns.[11]

In our experience, both of these factors are helpful screens. But there may be markets and industries where long-held sensibilities about when information exchanges are more benign than harmful are insufficiently sensitive to market developments and thus fail to capture the broader range of harm in the modern economy. The Division’s own experience and learning has made as much clear to us.

For example, we have observed anticompetitive information exchanges in industries that are not characterized by a small number of firms. In July 2022, the Division entered into a civil settlement with poultry processors that allegedly participated in a long-running scheme to exchange competitively-sensitive wage and benefit information. Although the consent decree was entered into by three poultry processors, a data consulting firm, and its president, the Division’s complaint alleges that at least 15 poultry processors participated in the alleged conduct in the past.[12] Notably, the consent decree, if accepted by the court, includes an industry ban for the data consulting firm and its president. In 2018, the Division brought civil enforcement actions to end anticompetitive information exchanges related to television spot advertising among more than twelve broadcast station groups.[13] These cases, among others, are sobering reminders that information exchanges may be easier to facilitate and discipline in concentrated markets, but they can be persistent and harmful in less concentrated markets too.

As then-judge Sotomayor aptly observed in Todd, while concentrated markets make coordination more likely, they are not essential:

[I]t [is] unsurprising that data exchange cases may involve a number of participants that begins to push the boundaries of oligopoly. These players are most in need of such data exchange arrangements in order to facilitate price coordination; a very small handful of firms in a more highly concentrated market may be less likely to require the kind of sophisticated data dissemination alleged in this case.[14]

An overly formalistic approach to information exchange risks permitting – or even endorsing – frameworks that may lead to higher prices, suppressed wages, or stifled innovation. A softening of competition through tacit coordination, facilitated by information sharing, distorts free market competition in the process.

Notwithstanding the serious risks that are associated with unlawful information exchanges, some of the Division’s older guidance documents set out so-called “safety zones” for information exchanges – i.e. circumstances under which the Division would exercise its prosecutorial discretion not to challenge companies that exchanged competitively-sensitive information. The safety zones were written at a time when information was shared in manila envelopes and through fax machines. Today, data is shared, analyzed, and used in ways that would be unrecognizable decades ago. We must account for these changes as we consider how best to enforce the antitrust laws.

Over the nearly 30 years during which those statements have been in effect, we have learned that concerning anticompetitive conduct can nonetheless satisfy many if not all of the safety zones’ factors. We have seen the safety zones be misinterpreted. Sometimes they are misapplied to other contexts or industries that were never contemplated by the guidance. Moreover, markets have evolved well beyond the context in which the safety zones, and some of the guidance more broadly, were articulated.

III. LOOKING AHEAD

I will share one action the Division is undertaking in recognition of these issues and three lessons we can draw from our experience detecting and rooting out anticompetitive information exchanges.

1. The Division will revisit outdated policy statements that no longer reflect market realities.

When policy statements and guidance documents no longer reflect market realities or the Division’s current enforcement posture, it is incumbent on us to revisit them. The Division is undertaking a review of our policy and guidance documents for that reason.

Since 1993, the Division – at times joined by the FTC – has issued policy statements concerning the health care industry: the 1993 Department of Justice and FTC Antitrust Enforcement Policy Statements in the Health Care Area; [15] the 1996 Statements of Antitrust Enforcement Policy in Health Care,[16] which revised and expanded the 1993 guidance; and the 2011 Statement of Antitrust Enforcement Policy Regarding Accountable Care Organizations Participating in the Medicare Shared Savings Program.[17]

Much has changed in the health care industry over the 30 years since these statements were issued. The delivery of health care products and services have changed. In many respects, our understanding of health care economics has evolved for the better. Increasingly health care is a data intensive industry that relies on the power of machine learning, artificial intelligence, and other advanced tools to develop or deliver products or services. Some markets are increasingly multi-sided. These realities affect how buyers or sellers transact business, which may bear on important dimensions of competition in this industry.

Moreover, a wave of consolidation in the health care industry has brought together industry participants who once served distinct or adjacent functions. As just one example, large health insurance companies now own providers, PBMs, health data analytics companies, and acute care clinics. If the concept of industry roll up was not in our lexicon then, it is more commonplace now. In many cases, these combinations and other entanglements may have changed the underlying incentive structures in the industry.

Given these changes and the outdated nature of many statements reflected in those documents, we are no longer confident that the documents fully reflect market realities, the risk of serious competitive harm, or the full scope of liability under the antitrust laws.

These documents also recognize “safety zones” around the exchange of competitively-sensitive information. As I previously discussed, the Division is concerned that the factors do not consider the realities of a transformed industry and, therefore, understate the antitrust risks of competitors sharing competitively-sensitive information.

One factor in the safety zones is the use of a third-party intermediary to facilitate information exchanges. But exchanges facilitated by intermediaries can have the same anticompetitive effect as direct exchanges among competitors. In some instances, data intermediaries can enhance – rather than reduce – anticompetitive effects. Likewise, the suggestion that data that is at least three-months old is unlikely to be competitively-sensitive or valuable is undermined by the rise of data aggregation, machine learning, and pricing algorithms that can increase the competitive value of historical data for some products or services. To use an analogy, maintaining the safety zones would be like developing specifications for audio cassette tapes and applying them to digital streaming. Similarly, the Division’s enforcement actions and the case law itself demonstrate that having five or more participants in an information exchange is no guarantee that the exchange will not harm competition, especially in situations where the companies exchanging the information collectively have significant shares of the relevant market.

At bottom, we believe the health care industry has evolved considerably since the statements were issued and that they no longer serve their intended purposes to provide encompassing guidance to the public on relevant health care competition issues. Accordingly, the Division will withdraw these three statements concerning the health care industry. At this time, the Division does not have immediate plans to replace them.

This withdrawal increases transparency because the guidance provided in these documents no longer reflects the market realities of the modern health care system or the Division’s current enforcement priorities. Recent enforcement actions and competition advocacy provide guidance to the public about our enforcement priorities. Careful consideration of the facts and an enforcement approach that is consistent with the law as Congress wrote it and as the courts have interpreted it will allow the Division to better evaluate mergers and conduct in health care markets that may harm competition.

The Division will continue to revisit out-of-date guidance that no longer provides transparency concerning how the Division presently evaluates potentially anticompetitive conduct.

2. If we conceive of information exchanges with the limits of the physical world, we will miss anticompetitive conduct in new economy markets.

We have to accept that many markets have evolved – in some cases dramatically so – over the last few decades. Not only has the provision of products and services relied more heavily on digital technologies and data at scale, but the impact of machine learning and other advanced tools on how companies use information to compete also may warrant revisiting how we think about the exchange of competitively-sensitive information.

In a bygone era of delayed communications, manual pricing strategies, and imprecise reporting, some of the heuristics we’ve employed to identify the most serious forms of illegal conduct may well have been defensible. Communication and coordination among large groups of disparate firms may have been more challenging, though not impossible. And aggregated, older data may have been less useful than disaggregated current or prospective information.

The modern economy may have solved for these speed bumps. The realities of some markets – and the products and services that are core to them – challenge embedded assumptions about the susceptibility of those markets to concerted action among market participants of varying sizes and geographies. In some industries, high-speed, complex algorithms can ingest massive quantities of “stale,” “aggregated” data from buyers and sellers to glean insights about the strategies of a competitor. Where that happens the distinctions between past and current or aggregated versus disaggregated data may be eroded.

Where competitors adopt the same pricing algorithms, our concern is only heightened. Several studies have shown that these algorithms can lead to tacit or express collusion in the marketplace, potentially resulting in higher prices, or at a minimum, a softening of competition.[18] Our criminal case involving coordination through the use of pricing algorithms in the online wall posters is just one example of how algorithms can be used to facilitate collusion in practice.[19]

Indeed, we are experiencing an inflection point in the use of algorithms, data at scale, and cloud computing. Additional changes are inevitable and likely to come in rapid succession. That is why it is important to revisit outdated guidance before it strays even further from market realities.

In particular, we should reconsider whether our traditional guideposts concerning market structure and the nature of the data exchanged are sufficiently sensitive to detect modern competition harms. To the extent new technologies or other changes in market realities have altered the competitive value of different types of data, our traditional guideposts – at least as originally conceived – may prove unhelpful in answering the ultimate question: is competition likely to be diminished.

3. The Division will closely scrutinize mergers when there is a prior history of collusion.

Our investigations into anticompetitive exchanges of information are not limited to Section 1 of the Sherman Act. Where the Division is presented with a proposed merger in an industry with a history of coordination or collusion, that context will play heavily in our evaluation of the transaction under Section 7 of the Clayton Act. The potential for tacit or express coordination after a merger, a concept that already exists in the current horizontal merger guidelines, is often top of mind during our reviews. But where an industry already has engaged in such coordination, our concerns are not just theoretical. Merging parties will face an uphill battle convincing us that post-merger coordination or collusion is unlikely – even when mergers fall below the structural presumption.

Our concerns only increase when one or more merging parties has engaged in an anticompetitive information exchange. In such cases, where there is evidence that firms engaged in pre-merger information exchange that harmed, or was likely to harm, competition, the coordinated conduct may serve as anatural experiment of a kind about the potential effects of the merger.

4. We need to bring a whole-of-government approach to anticompetitive information exchanges.

The Division is taking a whole-of government approach to competition issues across its programs and enforcement authorities. That includes anticompetitive information exchanges.

Competition and consumer protection statutes dot the United States Code. It was Congress’s design that other executive branch and independent agencies would police competition issues in their sectors alongside the DOJ and FTC. And in some cases, their authority is broader than the Sherman Act. Notably, some of these other statutes contemplate enforcement by the Attorney General. In other cases, agencies are able to refer or delegate their authority to the DOJ for investigation and/or enforcement.

I will share just two examples today.

First, the Packers and Stockyards Act. Congress enacted this statute “to assure fair competition and fair trade practices, . . . to safeguard farmers and ranchers[,] consumers . . . and . . . members of the livestock marketing and meat industries from unfair, deceptive, unjustly discriminatory and monopolistic practices.”[20] In July 2022, the Division brought a Packers and Stockyards Act claim against poultry processors who engaged in deceptive practices as a part of the Tournament System. It is the first time in recent memory that the Antitrust Division brought a claim under the Packers & Stockyards Act.

Second, our colleagues at the Department of Transportation have the power to challenge unfair and deceptive practices and unfair methods of competition in the airline industry.[21] This includes conduct that “causes or is likely to cause substantial injury, which is not reasonably avoidable, and the harm is not outweighed by benefits to consumers or competition”[22] Just last year, the Department of Transportation issued additional guidance – borrowing heavily from FTC precedent – explaining how it intends to identify such conduct.[23]

These are just two examples of robust competition statutes that are vested in our sister executive branch agencies. They are important competition enforcement partners. We are grateful for the support they can provide – and have already provided – in both our investigations and enforcement actions.

We are continuing to enhance our whole-of-government approach to competition. We are providing education on areas where illegal mergers and conduct can arise and working to strengthen our already strong relationships with agencies and departments across federal and state governments. At the end of the day, we are all working toward a common objective – vibrant, free markets that are free of anticompetitive conduct and consolidation. Such markets yield the benefits of competition, benefiting our economy and Americans across the board.

I thank you again for the privilege of being with you today.

Thank you.

[1] United States v. Topco Assocs., Inc., 405 U.S. 596, 610 (1972).

[2] N. Pac. Ry. Co. v. United States, 356 U.S. 1, 4 (1958).

[3] 15 U.S.C. § 1; Standard Oil Co. of New Jersey v. United States, 221 U.S. 1, 59–60 (1911) (the Sherman Act does not “restrain the right to make and enforce contracts, whether resulting from combinations or otherwise, which d[o] not unduly restrain interstate or foreign commerce,” but only acts to bar those contracts and combinations that amount to an “undue restraint”); State Oil Co. v. Khan, 522 U.S. 3, 10 (1997) (“Although the Sherman Act, by its terms, prohibits every agreement ‘in restraint of trade,’ this Court has long recognized that Congress intended to outlaw only unreasonable restraints.”).

[4] United States v. U.S. Gypsum Co., 438 U.S. 422, 441 n.16 (1978).

[5] See, e.g., Todd v. Exxon Corp., 275 F.3d 191, 208 (2d Cir. 2001).

[6] Am. Column & Lumber Co. v. United States, 257 U.S. 377, 391-93 (1921).

[7] United States v. Container Corp. of Am., 393 U.S. 333, 342 (1969) (Marshall, J., dissenting).

[8] Todd, 275 F.3d at 208.

[9] Gypsum, 438 U.S. at 441 n.16.

[10] Todd, 275 F.3d at 212.

[12] Complaint ¶ 48, United States v. Cargill Meat Solutions Corp., Dkt. No. 1:22-cv-01821, ECF No. 1 (D. Md. July 25, 2022).

[13] Complaint ¶ 2, United States. v. Sinclair Broadcast Group, Inc., Dkt. No. 1:18-cv-02609, ECF No. 48 (D.D.C. Aug. 1, 2019).

[14] Todd, 275 F.3d at 209.

[20] House Report, Livestock Marketing and Meatpacking Industry—Fair Trade Practices, H.R. Rep. No. 85-1048 at 1 (85th Cong. Aug. 9, 1957).

[21] 49 U.S.C. § 41712(a).

[22] 14 C.F.R. § 399.79(b)(1).

[23] U.S. Dep’t of Transp., Guidance Regarding Interpretation of Unfair and Deceptive Practices, 87 Fed. Reg. 52677 (Aug. 29, 2022).

Author:

Story from www.justice.gov

Disclaimer: The views expressed in this article are independent views solely of the author(s) expressed in their private capacity.

Check out our other content

Ad


Check out other tags:

Most Popular Articles