Tuesday, June 25, 2019

Lina M. Khan On Structural Separation

In It's The Enforcement, Stupid! I argued that anti-trust enforcement was viable only if there were "bright lines". I even went further and, following Kim Stanley Robinson's Pacific Edge, suggested a hard cap on corporate revenue, as a way of making anti-trust self-executing.

Much of the recent wave of attention to anti-trust was sparked by Lina Khan's masterful January 2017 Yale Law Journal article Amazon's Antitrust Paradox (a must-read, even at 24,000 words). Now Cory Doctorow writes:
Khan (who is now a Columbia Law fellow) is back with The Separation of Platforms and Commerce -- clocking in at 61,000 words with footnotes! -- that describes the one-two punch of contemporary monopolism, in which Reagan-era deregulation enthusiasts took the brakes off of corporate conduct but said it would be OK because antitrust law would keep things from getting out of control, while Reagan-era antitrust "reformers" (led by Robert Bork and the Chicago School) dismantled antitrust). 
You should definitely read Khan's latest magnum opus. OK, maybe you can skip the footnotes, I admit I did. Below the fold I examine two threads among many in the article.

Structural vs. Conduct Remedies

Khan's abstract summarizes her argument for reviving structural separation as an anti-trust mechanism (my emphasis):
Structural separations have been a mainstay element of American economic regulation. Traditionally applied to critical networks and essential infrastructure, structural separations prohibited entry in certain markets and prevented dominant intermediaries from directly competing with the businesses reliant on their services. In recent decades, structural separations have been largely abandoned. At the same time that lawmakers have weakened or eliminated sector- specific regulatory regimes, judicial interpretation of antitrust law has drastically narrowed the forms of vertical conduct and structures that register as anticompetitive. And when antitrust enforcers have targeted these forms of conduct and structures, they have applied remedies that generally (1) fail to target the underlying source of the problem and (2) overwhelm the institutional capacities of the actors assigned to oversee them.
Tracing the history of separations reveals that they have been motivated by a host of functional goals, ranging from fair competition and system resiliency to media diversity and administrability. Recalling this broader set of concerns brings into focus the range of factors at stake when dealing with dominant intermediaries and invites consideration of the degree to which separations in platform markets would also respond to a diverse set of problems.
In other words, Khan agrees with me that the key to effective anti-trust is enforcement triggered by "bright lines". Current conduct-based anti-trust enforcement lacks "bright lines" and as a result simply isn't effective (Page 1031 in the Columbia Law Review PDF):
In practice, the Obama Administration proved reluctant to issue strong structural remedies in vertical cases; it approved two major vertical deals—both described by critics as raising significant anticompetitive concerns—by issuing primarily conduct remedies. These conduct remedies—in the Ticketmaster–Live Nation and Comcast–NBC mergers—have proved difficult to oversee and enforce. Concerns that Live Nation has failed to abide by the remedies in any meaningful sense have prompted the Justice Department to open a Section 2 investigation, examining whether Live Nation is indeed using its control over concert facilities to pressure customers to also use its ticketing service and retaliating against those who decline its ticket service but still seek access to the concert facility. Comcast, too, has violated the conduct remedies that enforcers imposed when permitting the merger.

These incidents raise broader questions about the relative efficacy and administrative costs of imposing conduct remedies over structural ones. As Professor Spencer Weber Waller has noted, the retreat from structural remedies has led the antitrust agencies to adopt highly complex remedies that typically “exceed the resources and strengths” of the Justice Department and FTC.
Large companies feel confident that their legal teams can either intimidate the regulators, or tie them up in court while they reap the benefits of ignoring the conduct remedies. What's the worst that can happen? In practice, a cost-of-doing-business fine and more conduct remedies to be ignored.

The difference between structural separations and behavioral or conduct remedies is summarized on Page 980:
This Article argues that these combined problems of discrimination and information appropriation invite recovering common carriage’s forgotten cousin: structural separations. Structural separations place clear limits on the lines of business in which a firm can engage. Rather than prohibit particular business practices, separations proscribe certain organizational structures. In antitrust, structural remedies are contrasted with behavioral ones: Whereas behavioral remedies seek to prevent firms from engaging in specific types of conduct, structural remedies seek to eliminate the incentives that would make that conduct possible or likely in the first place.
Khan reviews the history of anti-trust, showing that structural remedies used to be applied in many areas, for example the railroads, the phone system and banking. She identifies six justifications (Page 1049):
Although policymakers applied structural limits in a variety of sectors, six justifications recur: (1) eliminating conflicts of interest, (2) preventing cross-financing that would extend existing dominance, (3) preserving system resiliency, (4) promoting diversity, (5) preventing excessive concentration of power, and (6) prioritizing administrability.
The last of these is the one I focus on here. Khan discusses it on Page 1059:
A final functional justification for structural separations is that they are highly administrable. Issuing outright bans obviates the need to engage in lengthy rule-of-reason type analysis; structural limits prescribe rules instead of standards. Structural separations are sometimes criticized for being far-reaching, crude, and overly broad, prohibiting benign as well as pernicious activity. This criticism is fair, given that rules are “by nature both over- and under-inclusive.” They accept some degree of error in return for clarity and predictability.

In at least two instances, public officials introduced structural regimes by citing their administrability, noting the limits of the government’s capacity to consistently detect discrete acts of wrongdoing. The FCC, for example, stressed its inability to “monitor carefully” the types of activities it had prohibited, “since even the injured party may not be aware of them.” The Commission observed that “subtle forms of favoritism” are “numerous and difficult to detect,” and that it was unlikely that the agency would “be prompt in cracking down on discovered abuses.” Relying on the agency to track individual acts of injury would risk extensive harm to competition. Structural bans, the agency explained, could also aid “the deterrence of foreseeable abuse.”
In contrast, behavioral or conduct remedies have fundamental problems (Page 1070):
Unlike structural remedies, behavioral remedies seek to change the firm’s conduct, while leaving the underlying incentives untouched. In effect these remedies constitute “attempts to require” a merged firm to “operate in a manner inconsistent with its own profit-maximizing incentives”—an effort that proves both “paradoxical” and “likely difficult to achieve.”

Behavioral remedies carry at least four substantial costs. First, there are the direct costs of monitoring the merged firm’s activity to ensure compliance with the decree. Second, there are costs of evasion associated with the merged firm sidestepping the spirit of the decree.  Third, there are costs of restraining potentially procompetitive behavior. And fourth, a behavioral remedy may hamper the firm’s ability to adapt effectively to changing market conditions. Stating that “a structural remedy can in principle avoid” these costs, the Justice Department has historically “strongly preferred” structural merger remedies to behavioral ones.
In practice, the regulators do not monitor conduct remedies. They delegate that to the monopolist's smaller competitors, waiting for a complaint. In the case of platforms such as the FAANGs this is unlikely to work. Kahn explains on Page 1084:
the European Commission’s remedy in the Google Shopping case—which required Google to implement a nondiscrimination approach—has not changed the underlying market dynamic, prompting content producers to describe it as “neither compliant nor effective.”

A remedy that was more attuned to the significant asymmetry in leverage would not rely entirely on third parties to contest the very intermediary on which their business often depends. Imposing a structural separation—that targets the underlying incentive to discriminate—would mitigate these shortcomings.
Oligopolies dominate most of the US economy, and especially the provision of Internet connectivity (the cable companies) and the services that ride on it (the FAANGs). This will continue until effective anti-trust enforcement is in place. Conduct remedies cannot be effective, because they lack "bright lines". Structural separation, for example banning Amazon from selling its own products, is an essential way to lay down "bright lines".

The Kill Zone

One result of the complete ineffectiveness of US anti-trust enforcement has been the growth of the FAANGs. As Khan points out the growth of the platforms has been largely due to acquisitions, not to in-house innovation. For example, Android wasn't developed in-house, Google bought it in 2005. The FAANGs have two huge advantages in the market. First, they exploit their platform to obtain early, accurate and detailed information about both possible competitors and markets worth entering (Page 1002):
Like Amazon and Google, Facebook has established a systemic informational advantage (gleaned from competitors) that it can reap to thwart rivals and strengthen its own position, either through introducing replica products or buying out nascent competitors. Strikingly, one of Facebook’s more recent acquisition—the burgeoning social network tbh—had achieved limited market penetration by the time Facebook purchased it. Analysts speculate that Facebook spotted tbh’s rapid pace of growth through Onavo and then bought it out.
Second, their monopoly rents provide capital much more cheaply than other companies, and in particular the potential acquisition targets (Page 1063):
Since dominant platforms report earnings and revenue at a highly generalized level, without breaking revenues and profits down to specific lines of business, we can mostly only speculate about the degree to which these firms are cross-financing. For example, Google’s operating margins over the last decade have hovered between 22% and 35%, margins that would qualify as supracompetitive and that derive from a market that Google dominates. Since 2004, Alphabet has purchased close to 200 companies. Several of these acquisitions strengthened Google’s position in digital advertising, its core market. But many of its purchases have established its position in new markets; indeed, Alphabet has built its strength outside of advertising almost entirely through acquisitions. Google established its home-automation business, for example, primarily through buy-ups. Most recently, the race for capturing the AI market is spurring a new flurry of acquisitions. Its pattern of acquisitions suggests that the company “will continue to push into entirely new areas, from genomics and healthcare to autonomous transport.” 
A dominant digital platform that uses its supracompetitive profits to buy its way into other markets can raise entry barriers in two ways. First, the platform can bundle its various services, such that any new firm seeking to compete in any one line of business may be unable to enter unless it could enter in multiple lines. Second, entering multiple markets positions a digital platform to combine multiple sources of data, potentially enabling a “super-platform” to control “key portals of data, which helps it attain or maintain its power across many products.” Amazon’s growing suite of acquisitions—which have picked up since Amazon Web Services (AWS) started reporting enormous profits—has also led analysts to speculate that Amazon uses AWS profits to finance entry into new markets.
Khan provides examples of the anti-competitive behavior the platforms use to either stunt competitors or reduce the cost of buying them (Page 977):
Consider Spotify’s effort to reach users through Apple’s iPhone while Apple sought to promote Apple Music. In 2016, Spotify revealed that Apple had blocked the streaming application from the App Store, “continu[ing] a troubling pattern of behavior by Apple to exclude and diminish the competitiveness of Spotify on iOS and as a rival to Apple Music.” Or take the challenge faced by Yelp, Foundem, and scores of online services to reach internet users while Google sought to build out its own competitor offerings. In Europe and India, competition authorities have found that Google ranks its own services higher than those offered by rivals, a “search bias” that means anyone competing with Google properties may effectively disappear from Google search results. Merchants that rely on Amazon to reach consumers are in a similar bind: Not only must they jostle for placement against Amazon’s
own goods, but they also face the constant risk that Amazon will spot their bestselling items and produce them itself. Facebook, equipped with technology that lets it detect which rival apps are succeeding, would often give companies a choice: Be acquired by Facebook, or watch it roll out a direct replica. Competing with one of these giants on the giant’s own turf is rife with hazards.
Given the impossibility of competing with the FAANGs, companies that might conceivably be viewed as competition can't be funded. To get funding a company needs to be designed to be acquired, severely limiting the upside (Page 978):
Venture capitalists now factor this risk into their investment decisions. Indeed, the power of these gatekeeper platforms to steer the fate of countless other firms is described by entrepreneurs and investors as “having a profound impact on innovation in Silicon Valley” and “choking off the start-up world.” Venture capitalists now discuss a “kill-zone” around digital giants—“areas not worth operating or investing in, since defeat is guaranteed.” Discussing how tech platform giants today use their integrated structure to undermine rivals, a product manager who worked for Microsoft leading up to its antitrust suit observed, “It’s what we did at Microsoft.”
Microsoft got caught doing it, and in the long term benefited greatly from it. They have now even become stalwarts of Open Source. But that requires the long-term view so lacking today.

The FAANGs can  now "use their integrated structure to undermine rivals" with impunity, and the result is a significant decline in entrepreneurship (Page 1008):
Data on investment trends do not offer a decisive answer but generally seem consistent with the story told by surveyed investors. Venture capital funding as a whole appears to be booming: In 2018, the total annual venture capital invested surpassed $100 billion for the first time since the dot-com period. The number of angel and seed investments, meanwhile, has been declining since 2015, signaling that it has become harder for startups to secure an initial round of financing. Indeed, it is late-stage deals with mature  companies that account for an “outsized proportion” of total capital today,while startups see fewer first financings, even as the deal value for startups has increased. In other words, venture capital markets seem to be following a winner-take-most model: Fewer firms receive funding, but those that do are raising more capital. These trends come against a backdrop of falling entrepreneurship: Startup formation is at a thirty-year low, contributing to a loss of business dynamism.
The Economist's American tech giants are making life tough for startups makes the same point, and graphs the decline in early stage fundings:
The behemoths’ annual conferences, held to announce new tools, features, and acquisitions, always “send shock waves of fear through entrepreneurs”, says Mike Driscoll, a partner at Data Collective, an investment firm. “Venture capitalists attend to see which of their companies are going to get killed next.” But anxiety about the tech giants on the part of startups and their investors goes much deeper than such events. Venture capitalists, such as Albert Wenger of Union Square Ventures, who was an early investor in Twitter, now talk of a “kill-zone” around the giants. Once a young firm enters, it can be extremely difficult to survive. Tech giants try to squash startups by copying them, or they pay to scoop them up early to eliminate a threat.
I'm advising a startup that, like almost all such companies, is hosting its IT operation on AWS. If it succeeds, its Amazon bills will rise rapidly. Some algorithm at Amazon will monitor both the amount and rate of change, and at some point alert someone at Amazon to look into buying the company.

Cory Doctorow makes a related point in his New York Times "op-ed from the future" entitled I Shouldn't Have to Publish This in The New York Times, that because the platforms lack effective competition they are tools for censorship:
Think back to the days when companies like Apple and Google — back when they were stand-alone companies — bought hundreds of start-ups every year. What if we’d put a halt to the practice, re-establishing the traditional antitrust rules against “mergers to monopoly” and acquiring your nascent competitors? What if we’d established an absolute legal defense for new market entrants seeking to compete with established monopolists?

Most of these new companies would have failed — if only because most new ventures fail — but the survivors would have challenged the Big Tech giants, eroding their profits and giving them less lobbying capital. They would have competed to give the best possible deals to the industries that tech was devouring, like entertainment and news. And they would have competed with the news and entertainment monopolies to offer better deals to the pixel-stained wretches who produced the “content” that was the source of all their profits.

But instead, we decided to vest the platforms with statelike duties to punish them for their domination. In doing so, we cemented that domination. Only the largest companies can afford the kinds of filters we’ve demanded of them, and that means that any would-be trustbuster who wants to break up the companies and bring them to heel first must unwind the mesh of obligations we’ve ensnared the platforms in and build new, state-based mechanisms to perform those duties.
A recent example of what Doctorow is talking about comes from Tim Cushing in Indian Gov't Uses National Security Law, Bad Information To Block Twitter Accounts All Over The World:
US social media companies are continuing to act as proxy censors for governments around the world. This is adding some bizarre twists to stories of social media content takedowns as governments target posts by non-citizens located thousands of miles away. ... an American college student's tweets were targeted by the Indian government, which claimed the student was engaging in spreading propaganda. ... Given the circumstances, it appears the Indian government is targeting Twitter accounts in the US and Britain for spreading misinformation… based on bad information.

Pushback from US tech companies has been minimal. Users no longer need to fear just their own governments, but every government in the world. As more governments enact laws directly targeting speech, the effects will continue to be felt around the globe.
The Financial Times' Rana Foroohar had an interview with Lina Khan in late March, before The Separation of Platforms and Commerce was published:
“The new Brandeis movement isn’t just about antitrust,” she says ... Rather, it is about values. “Laws reflect values,” she says. “Antitrust laws used to reflect one set of values, and then there was a change in values that led us to a very different place.”
It used to be an American value that anyone could start a company (as we did with Nvidia) and, if their idea and execution was good, build a significant company, take it public, and become a part of the US economy for the long term. Not so much any longer. Sadly, the best that's on offer is to become a small part of one of the FAANGs.


  1. In Is Amazon getting Too Big?, Wharton’s Barbara Kahn and Emory University’s Ryan Hamilton examine the question:

    "According to Hamilton, the issues provoked by Amazon require an entirely new way of thinking about antitrust. Monopolies have long been about the ability to control price and profit. Amazon could get broken up into 20 different entities, but that wouldn’t solve the problems surrounding access to data. The issues with data management could be addressed through legislation, which Hamilton thinks is more likely than the government intervening to fracture the companies.

    “People are talking about breaking up Amazon; people are talking about breaking up even Facebook, which is a free service. So, it’s no longer about just price; it’s about information, power. It’s about leverage, about control. I would be surprised if there’s kind of an appetite for that broadly, but opinion on this seems to be shifting rapidly,” he said."

  2. Matt Day's Amazon Falls After Report That the Company Prioritized Profit in Its Search Listings shows that Amazon's been doing exactly what got Google fined €2.42 billion by the EU in 2017:

    "Amazon.com Inc. shares fell following a report that the e-commerce giant adjusted its product search results to emphasize items that are more profitable to the company.

    The Wall Street Journal reported Monday that, following an internal debate, Amazon late last year changed its secret algorithm that ranks search results to lift more profitable items, a departure for a company that typically emphasized customer satisfaction and bestsellers.
    The change to the search algorithm could steer customers toward Amazon private-label products that deliver higher profit margins than competing listings, the Journal reported."

    John Herrman's Everything on Amazon Is Amazon! shows that Amazon has been making the fact that they are favoring their own brands harder to spot by stealthing them:

    "In 2009, the company started selling products under its own name. It soon moved beyond the first AmazonBasics — items including budget electronics and batteries — to a wider range of Amazon-branded products. This was followed by an explosion of company-owned brands, including dozens with Amazon-free names.

    Lark & Ro sells women’s wear, Buttoned Down sells men’s dress shirts; Pike Street sells linens; Strathwood sells furniture. These brands are intended to stand on their own, sort of. They are associated with Amazon, and listed on the site’s dozens of different contexts as “Our Brand” or “by Amazon” or “An Amazon Brand.”
    This effort is broadly understood to have been a success, generating up to $7.5 billion this year and potentially $25 billion by 2022, according to analysis by SunTrust Robinson Humphrey.

    Amazon-affiliated brands are promoted in search results on the site and inflated by reviews from Amazon’s Vine program, in which users receive items in exchange for their feedback. And, compared to better known competitors, they tend to be priced aggressively."