Pleading (in) the Fifth (Circuit)
Way back in late January, I wrote a piece called “Lina’s Lingering Legacy?” Lina Khan—at that time, Commissioner Khan, and the week before that, Chair Khan—had not yet left the Federal Trade Commission (FTC) building. But she had been replaced as chair by presidential fiat (as per Section 1 of the FTC Act) and she had announced that she would leave the FTC by close of business Jan. 31.
And she did, in fact, depart Jan. 31. But I had gotten ahead of myself, at least a little. There were matters afoot, and in court, and Khan’s track record slid a bit further during her last week at the commission. The same day that Khan departed the commission, Judge Charles Eskridge of the U.S. District Court for the Southern District of Texas denied the FTC’s motion for a preliminary injunction against Tempur Sealy’s acquisition of Mattress Firm.
The FTC had challenged the merger in its internal Part 3 process (that is, before itself), issuing an administrative complaint in July 2024, while Khan was still chair. As is typical in such matters, the FTC had also filed a motion in federal district court seeking a temporary restraining order and preliminary injunction (a federal court can enjoin or temporarily restrain the consummation of a merger before a decision on the merits, but the FTC cannot).
That’s another one in the loss column for Khan’s merger-challenge record—and that of the Biden administration, given its “whole of government” approach that at times diminished and at times seemed to obliterate the independence of the “independent agency.”
The district court decision does not, in itself, end the Part 3 process; it’s a decision on the preliminary injunction and restraining order, not a final decision on the merits of the underlying antitrust case. But the decision is significant all the same, not least because it rests on a strong preliminary assessment of the underlying case. That’s typical of such parallel district court decisions. The government—in this case, the FTC—needs to show a substantial likelihood of success on the merits.
That’s a lower burden than the plaintiff would face in a trial on the merits, but it’s a real one. And it’s often a pretty good signal of how the court is viewing the merits of the underlying complaint. And on that question, Judge Eskridge leaves no doubt: he doesn’t accept the FTC’s market definition; he doesn’t think the evidence suggests that the FTC’s foreclosure theory will hold water; he doesn’t think it likely that the merger will prove anticompetitive; and he does think that, pending a more thorough hearing, any “lingering concerns” should be remediable.
I won’t argue that the FTC had no case. For one thing, it’s a rule-of-reason case and I have not seen all the evidence. I haven’t even seen the whole complaint: the version that’s publicly available is heavily redacted. For another, while most vertical mergers are procompetitive or benign, they are not—and should not be—per se lawful. And while the FTC’s foreclosure theory looked like tough sledding, it was not logically deficient as a matter of law or economics. Third, there seemed to be some soft (that’s not to say convincing, much less definitive) evidence in favor of an intent to foreclose. Reasons two and three are acknowledged in Commissioner Melissa Holyoak’s statement on the case.
For me, several aspects of the case appear challenging. Indeed, on market definition alone—and leaving aside the precise boundaries of the “premium mattresses” market definition highlighted by Judge Eskridge—several things seemed challenging. This was a case about a large mattress manufacturer acquiring a large retail chain; that is, Tempur Sealy’s acquisition of Mattress Firm, a chain of more than 2,300 retail, brick-and-mortar mattress stores. So far, so good—that’s a substantial acquisition, a very large number of stores, and many, many consumers buy mattresses at brick-and-mortar stores. But “premium” aside, that market definition highlights a question of local competition.
I live in Northern Virginia. I don’t drive to Roanoke, much less Riverside, California, when I shop for mattresses. I suppose I might purchase something online without worrying about the location of the shipper, except to the extent that it might be reflected in shipping charges. But the market definition didn’t (and the FTC did not want to) count online sales.
Brick-and-mortar retail competition is local. Mattress Firm might be a large national chain—it might be the largest such chain nationally—but that doesn’t tell us much about their market share—much less their market power—in any given local geographic market.
A quick search on the Mattress Firm website reveals a few of their stores within a short drive of my home, and about 20 in the larger (and very large) metropolitan area in which I live (about 6.3 million people, according to the U.S. Census Bureau). And another quick search reveals many more local outlets in which I might shop for mattresses, including “premium” mattresses.
Mattress Firm outlets are not my only choice—not even close. Tempur Sealy products seem to be available at quite a few local outlets; and Serta Simmons (identified in the FTC complaint as Tempur Sealy’s largest competitor) seem to be available at quite a few outlets, including various department stores and, as it turns out, Pottery Barn (owned by Williams Sonoma).
Would a combined Tempur Sealy/Mattress Firm really be able to foreclose Serta Simmons and other competing manufacturers from the Northern Virginia retail market, whatever its precise geographic boundaries? The greater Washington D.C. metropolitan area? Philadelphia? Chicago? Los Angeles? Which, if any, are the local markets from which Tempur Sealy, the manufacturer, might effectively foreclose competing manufacturers by purchasing Mattress Firm, the retail chain?
If there are specific, local retail markets at-risk, might the merger not be cured, by agreement, in those specific markets? And while we’re at it, what are the barriers to entry for building and operating a brick-and-mortar mattress outlet? Entry is not exactly free, but we’re not talking about extraordinary fixed costs or, for example, the sorts of regulatory barriers that come with entry into a hospital market.
The FTC’s case faced other challenges. Some—not all—are highlighted in the district court decision. But I’m not trying to litigate the case in a blog post, much less to do so on limited evidence. I am highlighting another loss in the Biden administration’s antitrust-enforcement program and, it seems to me, the need to be careful in choosing and pleading vertical-merger cases. (On vertical mergers more generally, see, e.g., the International Center for Law & Economics (ICLE); Francine Lafontaine and Margaret Slade; and James C. Cooper, Luke M. Froeb, Dan O’Brien, and Michael G. Vita, among others.)
All in all, Khan’s FTC had a tough time in the 5th U.S. Circuit, pace the agency’s settlement with Welsh Carson—where there’d been a partial, but not wholesale, setback. There was the setback for Chair Khan’s ambitious noncompete rule, with Judge Ada Brown of the U.S. District Court for the Northern District of Texas issuing a decision and order vacating the rule in August 2024, which I wrote about here. And there was the Jan. 27 decision by the 5th U.S. Circuit Court of Appeals striking down the FTC’s Combatting Auto Retail Scams (CARS) Trade Regulation Rule.
The FTC has appealed the noncompete decision, not that I expect them to find the circuit court, or the U.S. Supreme Court, sympathetic to that appeal. It’s not that such agreements cannot be anticompetitive; they can be, under various circumstances (here’s ICLE, and here’s me separately). But the FTC adopted an overly broad rule that they cannot plausibly enforce, based on very little enforcement experience, a slanted (and inadequate) evidentiary foundation, and dubious authority. And these are not halcyon days for agency deference.
I expect that the agency will drop the appeal once the majority has shifted. Chairman Andrew Ferguson and Commissioner Holyoak both dissented from the commission’s adoption of the rule in the first place. But as far as I know, the appeal remains live, for now.
For Want of a Process…
The CARS decision is a tricky one, and in some ways unfortunate, as it’s a rule that addresses a legitimate area of enforcement concern under both the FTC Act and the Dodd–Frank Wall Street Reform and Consumer Protection Act, which authorized the FTC to adopt rules regulating auto dealers specifically. The FTC’s final published rule:
…prohibits motor vehicle dealers from making certain misrepresentations in the course of selling, leasing, or arranging financing for motor vehicles, requires accurate pricing disclosures in dealers’ advertising and sales communications, requires dealers to obtain consumers’ express, informed consent for charges, prohibits the sale of any add-on product or service that confers no benefit to the consumer, and requires dealers to keep records of certain advertisements and customer transactions.
At least some of the rule’s provisions address types of conduct that can violate the consumer-protection prong of Section 5’s prohibition of unfair or deceptive acts or practices in or affecting commerce. Perhaps no reader will be shocked by the suggestion that some auto dealers have engaged in deceptive or materially misleading marketing. And at least some of the regulatory prohibitions appeared to be relatively low-cost ways to guard against conduct that is not justified by countervailing benefits to consumers and competition.
I’ll stop short of a line-by-line endorsement of the rule or its attendant cost-benefit analysis. Still, for enforcement purposes, there was a there there. And moreover, FTC staff had at least adequate grounds to consider a rulemaking. Indeed, the 5th Circuit opinion acknowledged the FTC’s enforcement experience in the area and took no issue with the substance of the rule.
The court’s problem, rather, had to do with the procedures the FTC followed in adopting the rule. In brief, the panel decision authored by Judge Patrick E. Higginbotham held that the FTC promulgated the CARS rule “in violation of its own regulations.”
The details are a bit complicated, as they involve the interplay between Section 18 of the FTC Act (regarding consumer-protection rulemaking); FTC procedural rules for adopting federal regulations under Section 18; and the provisions of the Dodd-Frank Act that provide for FTC regulation and permit such rulemaking under Administrative Procedure Act (APA) rulemaking procedures. At the statutory level, the FTC Act sets out certain requirements for consumer-protection rulemaking by the FTC, while the Dodd-Frank Act says, roughly, that with regard to certain auto-dealer regulations, the FTC might follow the relatively (if not terribly) streamlined requirements for rulemaking under the APA.
The problem, in the court’s view, is that the FTC’s own regulations regarding how it may adopt consumer-protection rules require, among other things, publication of an advance notice of proposed rulemaking (ANPR) before it publishes a notice of proposed rulemaking (NPRM). But there were no other agency rules of process for auto-dealer regulations and there was no such ANPR for the CARS rule. That is, the panel argued that Dodd-Frank provides a statutory basis on which the FTC could have adopted different internal rules with regard to how it might promulgate consumer-protection rules in one specific industry, but the FTC did not do so.
Still with me? The dissent issued by 5th Circuit’s Judge Stephen A. Higginson is not incredible; on a different panel, it might well have been the majority opinion. The majority put a good deal of stress on the proper construction of “authorized” in the pertinent part of the Dodd-Frank Act.
Still, the majority opinion is a careful one with a real issue at its core: was Dodd-Frank a new and independent grant of rulemaking authority? The majority says “no” because of the plain language of Dodd-Frank, which says that the FTC may adopt its auto dealer rules “under sections 45 and 57a(a)(1)(B) of title 15” (that is, under Sections 5 and 18(a)(1)(B) of the FTC Act).
This may have been an own goal by the FTC’s 2024 majority and its outgoing chair. Readers may recall that the FTC revised its process rules for consumer-protection rulemaking in July 2021. According to the commission, the revisions were adopted to streamline and “modernize” the way the FTC issues consumer-protection rules under Section 18 of the FTC Act. But as I noted in a prior post, it was a controversial bit of streamlining:
Those revisions were adopted by a 3-2 vote, with Commissioners Christine Wilson and Noah Phillips dissenting.… Wilson and Phillips pointed out that the changes diminished both the transparency of the rulemaking process and the opportunity for independent input, and eliminated the requirement of an expert staff report (and for Bureau of Economics review of a preliminary staff report), as we noted in ICLE’s comments on the commission’s Advance Notice of Proposed Rulemaking on Commercial Surveillance and Data Security.
Just as it did in later amendments to its rules of practice regarding Part 3 proceedings, the FTC adopted rule changes that amplified the authority and autonomy of the FTC chair. That is, under then-Chair Khan, the FTC considered a raft of rulemakings on both competition and consumer-protection matters. It revised its process rules for the adoption of Section 18 consumer-protection regulations and, subsequently, it issued ANPRs under its revised process rules.
There was the “junk fees” ANPR (formally titled the Unfair or Deceptive Fees Federal Trade Regulation Rule), which was published Nov. 8, 2022, with an NPRM published Nov. 8, 2023 and a final rule published Dec. 17, 2024. And there was the controversial “commercial surveillance” ANPR, which was published Aug. 22, 2024, with no subsequent NPRM or final rule.
The FTC could have amended its rules further to account for the Dodd-Frank carveout. As far as I can tell, it still can. But it hasn’t yet, and it did not follow even its streamlined process rules in adopting the CARS rule. In its rush to regulate, it didn’t think it had to. But a federal court of appeals thinks otherwise. So, for now at least, there goes the rule and a good deal of staff work.
What’s a Meta For?
Old metaphors never die, exactly, but the hoary, overused, and prosaic at best are sometimes called “dead metaphors.” That’s what happens to metaphors when their ship has sailed. (See what I did there?)
And that brings me to Lina Khan’s “guest essay” (op-ed) in The New York Times this week: “Stop Worshipping the American Tech Giants.” Her title reminds me that I know so little about religious diversity in today’s America.
Taking a stylistic cue from Tim Wu’s recent column in The Atlantic, she opens by comparing one moribund metaphor with another. Marc Andreessen, she notes, has called recent news about DeepSeek “A.I.’s Sputnik moment.” (Sputnik? Really?) But “[a]s an antitrust enforcer” (she presumably means former antitrust enforcer), Khan sees:
…a different metaphor. DeepSeek is the canary in the coal mine. It’s warning us that when there isn’t enough competition, our tech industry grows vulnerable to its Chinese rivals, threatening U.S. geopolitical power in the 21st century.
“Canary in a coal mine” is a dead metaphor—if, once upon a time, a viable carbon-monoxide detector. The canaries signaled danger not by song but by silence: when the sensitive birds died, they stopped singing. Ideally, the miners noticed.
How this maps onto DeepSeek is unclear, but enough about canaries, whether in song, just resting, or having shuffled off their mortal coils. The editors at The New York Times apparently found Khan’s story compelling, or at least, somehow worth its column inches. Does it make a lick of sense?
I’m not so sure. But I want to start by pointing to a couple of threads on the platform formerly known as Twitter—one by my ICLE colleague Brian Albrecht and one by Neil Chilson—as each responds in a clear and sensible way to Khan’s column.
Now to be fair, if not kind, I don’t think her recent op-ed is worse than the one that the Times published in May 2023 in which she called for regulating artificial intelligence (AI). At the time, I wrote “Artificial Intelligence Meets Organic Folly” (a title of mine paraphrasing a better one by Drew McDermott) in response, in which I suggested that Khan had not actually told us how to regulate AI or, for that matter, what it was about AI that cried out for regulation. The whole thing seemed a muddle: a claim that “enforcers and regulators must be vigilant” (ok, nothing new there), posed with a motley of possible risks, all unanalyzed and unsubstantiated but presented as a parade of horribles and a paean to regulations undescribed.
The story, such as it was, seemed to be that innovation had run amok, and that it was bound to be disruptive. Therefore, more regs, of some sort, please. Plus, the FTC would do its part:
…to uphold America’s longstanding tradition of maintaining the open, fair, and competitive markets that have underpinned both breakthrough innovations and our nation’s economic success – without tolerating business models or practices involving the mass exploitation of their users.
That seemed like a lot, and I was not sure what, exactly. But if it just meant enforcing Section 5 of the FTC Act, then fine. That was, and remains, the FTC’s job.
That provides a connection to one clear unobjectionable claim in Khan’s new column. AI firms ought not to be excepted from the antitrust laws on some sort of “national champions” theory—not if we assume any established understanding of antitrust law and economics, I’m inclined to add.
But it also seems to highlight a tension with her earlier column that was noted in Chilson’s thread: in the summer of 2023, we needed to contemplate AI regulations (and antitrust intervention) because AI innovation had run amok. But by the end of 2024, the claim is that we need to intervene because innovation has been stifled, at least in the United States, as evidenced by innovation somewhere else. In either case—heck, in every possible case—vigorous antitrust enforcement, on some conception of antitrust, would be a key part of the answer.
How do we get from inadequate or impaired competition to threats to “U.S. geopolitical power in the 21st century”? I’m not sure. Her argument—if it’s an argument—is a bit of a hodgepodge, if not a shaggy dog tale, with a few straw men thrown in for good measure.
Both Chilson and Albrecht point out (rightly) that innovation in AI (even in the good old US of A) is not obviously (or actually) dead. They highlight, for example, extremely recent announcements from OpenAI (about Deep Research) and Google (about Gemini 2.0 flash)—as well as investments (and progress) by Meta, among others—to illustrate some of the considerable investments and progress in AI from both within and without “GAFAM.”
I think it’s a bit early to say that we know all of the implications of DeepSeek’s debut (and attendant publications), and it seems unlikely that the touted development costs remotely resemble the whole accounting story. But that’s not to say that it’s unremarkable. Khan, along with everyone else, is right that it’s remarkable. But the question is whether that implies anything at all about a failure of competition, or of U.S. competition enforcement. That is, as my ICLE colleague Lazar Radic points out, DeepSeek’s emergence:
…shows that AI development is not limited to a few American firms with deep pockets, and may be possible with far lower levels of investment than previously imagined.
But that is not, as he also notes, “a new story at all.” And it is not a showing that American firms with deep pockets, among others from the United States (and not just the United States), are failing to innovate. It’s certainly not an argument for different antitrust standards or sweeping “precautionary” antitrust regulations.
China is indeed a U.S. rival in various and serious ways. I do not at all mean to suggest that there are no policy concerns to be had there. But why on earth would some significant innovation in a nation with tremendous natural and human resources—and a population in excess of 1.4 billion persons—imply competitive or innovative failure in the United States?
Indeed, Khan’s argument about AI seems most reminiscent of the U.S. Justice Department’s (DOJ) arguments about the resources required to reach minimally efficient scale in general search in the Google Search case. There, the government argued (with a bit of hand waving) that Google’s scale advantages in data and compute over plucky “little” would-be rivals like OpenAI and (cough) Microsoft were, in fact, insurmountable. Or maybe not.
I’m just scratching the surface here. Perhaps the best overarching point is that AI is a complex and highly dynamic space. Vigilance is warranted, but panic and overreaction are unhelpful. And on that point, I’ll note, as Chilson does, some recent remarks by Commissioner Holyoak:
We must avoid slowing innovation in artificial intelligence through misguided enforcement actions. While the technology is still nascent, it is critically important to America’s economic and national security. The Commission needs to be circumspect when policing AI. Our focus should be on stopping fraudulent conduct and developing further our understanding of this industry, not seeking to stretch our statutory authorities.
That is not a call for antitrust immunity, but a brief for economically, technically, and legally grounded enforcement. That is an eminently sensible approach.
