Archives For platforms

The Amazon investigation and Europe’s “Big Tech” Crusade

Dirk Auer —  21 October 2018

The dust has barely settled on the European Commission’s record-breaking €4.3 Billion Google Android fine, but already the European Commission is gearing up for its next high-profile case. Last month, Margrethe Vestager dropped a competition bombshell: the European watchdog is looking into the behavior of Amazon. Should the Commission decide to move further with the investigation, Amazon will likely join other US tech firms such as Microsoft, Intel, Qualcomm and, of course, Google, who have all been on the receiving end of European competition enforcement.

The Commission’s move – though informal at this stage – is not surprising. Over the last couples of years, Amazon has become one of the world’s largest and most controversial companies. The animosity against it is exemplified in a paper by Lina Khan, which uses the example of Amazon to highlight the numerous ills that allegedly plague modern antitrust law. The paper is widely regarded as the starting point of the so-called “hipster antitrust” movement.

But is there anything particularly noxious about Amazon’s behavior, or is it just the latest victim of a European crusade against American tech companies?

Where things stand so far

As is often the case in such matters, publicly available information regarding the Commission’s “probe” (the European watchdog is yet to open a formal investigation) is particularly thin. What we know so far comes from a number of declarations made by Margrethe Vestager (here and here) and a leaked questionnaire that was sent to Amazon’s rivals. Going on this limited information, it appears that the Commission is preoccupied about the manner in which Amazon uses the data that it gathers from its online merchants. In Vestager’s own words:

The question here is about the data, because if you as Amazon get the data from the smaller merchants that you host […] do you then also use this data to do your own calculations? What is the new big thing, what is it that people want, what kind of offers do they like to receive, what makes them buy things.

These concerns relate to the fact that Amazon acts as both a retailer in its own right and a platform for other retailers, which allegedly constitutes a “conflict of interest”. As a retailer, Amazon sells a wide range of goods directly to consumers. Meanwhile, its marketplace platform enables third party merchants to offer their goods in exchange for referral fees when items are sold (these fees typically range from 8% to 15%, depending on the type of good). Merchants can either execute theses orders themselves or opt for fulfilment by Amazon, in which case it handles storage and shipping. In addition to its role as a platform operator,  As of 2017, more than 50% of units sold on the Amazon marketplace where fulfilled by third-party sellers, although Amazon derived three times more revenue from its own sales than from those of third parties (note that Amazon Web Services is still by far its largest source of profits).

Mirroring concerns raised by Khan, the Commission worries that Amazon uses the data it gathers from third party retailers on its platform to outcompete them. More specifically, the concern is that Amazon might use this data to identify and enter the most profitable segments of its online platform, excluding other retailers in the process (or deterring them from joining the platform in the first place). Although there is some empirical evidence to support such claims, it is far from clear that this is in any way harmful to competition or consumers. Indeed, the authors of the paper that found evidence in support of the claims note:

Amazon is less likely to enter product spaces that require greater seller efforts to grow, suggesting that complementors’ platform‐specific investments influence platform owners’ entry decisions. While Amazon’s entry discourages affected third‐party sellers from subsequently pursuing growth on the platform, it increases product demand and reduces shipping costs for consumers.

Thou shalt not punish efficient behavior

The question is whether Amazon using data on rivals’ sales to outcompete them should raise competition concerns? After all, this is a standard practice in the brick-and-mortar industry, where most large retailers use house brands to go after successful, high-margin third-party brands. Some, such as Costco, even eliminate some third-party products from their shelves once they have a successful own-brand product. Granted, as Khan observes, Amazon may be doing this more effectively because it has access to vastly superior data. But does that somehow make Amazon’s practice harmful to social social welfare? Absent further evidence, I believe not.

The basic problem is the following. Assume that Amazon does indeed have a monopoly in the market for online retail platforms (or, in other words, that the Amazon marketplace is a bottleneck for online retailers). Why would it move into direct retail competition against its third party sellers if it is less efficient than them? Amazon would either have to sell at a loss or hope that consumers saw something in its products that warrants a higher price. A more profitable alternative would be to stay put and increase its fees. It could thereby capture all the profits of its independent retailers. Not that Amazon would necessarily want to do so, as this could potentially deter other retailers from joining its platform. The upshot is that Amazon has little incentive to exclude more efficient retailers.

Astute readers, will have observed that this is simply a restatement of the Chicago school’s Single Monopoly Theory, which broadly holds that, absent efficiencies, a monopolist in one line of commerce cannot increase its profits by entering the competitive market for a complementary good. Although the theory has drawn some criticism, it remains a crucial starting point with which enforcers must contend before they conclude that a monopolist’s behavior is anticompetitive.

So why does Amazon move into retail segments that are already occupied by its rivals? The most likely explanation is simply that it can source and sell these goods more efficiently than them, and that these efficiencies cannot be achieved through contracts with the said rivals. Once we accept the possibility that Amazon is simply more efficient, the picture changes dramatically. The sooner it overthrows less efficient rivals the better. Doing so creates valuable surplus that can flow to either itself or its consumers. This is true regardless of whether Amazon has a marketplace monopoly or not. Even if it does have a monopoly (which is doubtful given competition from the likes of Zalando, AliExpress, Google Search and eBay), at least some of these efficiencies will likely be passed on to consumers. Such a scenario is also perfectly compatible with increased profits for Amazon. The real test is whether output increases when Amazon enters segments that were previously occupied by rivals.

Of course, the usual critiques voiced against the “Single Monopoly Profit” theory apply here. It is plausible that, by excluding its retail rivals, Amazon is simply seeking to protect its alleged platform monopoly. However, the anecdotal evidence that has been raised thus far does not support this conclusion.

But what about innovation?

Possibly sensing the weakness of the “inefficiency” line of arguments against Amazon, critics will likely put put forward a second theory of harm. The claim is that by capturing the rents of potentially innovative retailers, Amazon may hamper their incentives to innovate and will therefore harm consumer choice. Margrethe Vestager intimated this much in a Bloomberg interview. Though this framing might seem tempting at first, it falters under close inspection.

The effects of Amazon’s behavior could first be framed in terms of appropriability — that is: the extent to which an innovator captures the social benefits of its innovation. The higher its share of those benefits, the larger its incentives to innovate. By forcing out its retail rivals, it is plausible that Amazon is reducing the returns which they earn on their potential innovations.

Another potential framing is that of holdup theory. Applied to this case, one could argue that rival retailers made sunk investments (potentially innovation-related) to join the Amazon platform, and that Amazon is behaving opportunistically by capturing their surplus. With hindsight, merchants might thus have opted to stay out of the Amazon marketplace.

Unfortunately for Amazon’s critics, there are numerous objections to these two framings. For a start, the business implication of both the appropriability and holdup theories is that firms can and should take sensible steps to protect their investments. The recent empirical paper mentioned above stresses that these actions are critical for the sake of Amazon’s retailers.

Potential solutions abound. Retailers could in principle enter into long-term exclusivity agreements with their suppliers (which would keep Amazon out of the market if there are no alternative suppliers). Alternatively, they could sign non-compete clauses with Amazon, exchange assets, or even outright merge. In fact, there is at least some evidence of this last possibility occurring, as Amazon has acquired some of its online retailers. The fact that some retailers have not opted for these safety measures (or other methods of appropriability) suggests that they either don’t perceive a threat or are unwilling to make the necessary investments. It might also be due to bad business judgement on their part).

Which brings us to the big question. Should competition law step into the breach in those cases where firms have refused to take even basic steps to protect their investments? The answer is probably no.

For a start, condoning this poor judgement encourages firms to rely on competition enforcement rather than private solutions  to solve appropriability and holdup issues. This is best understood with reference to moral hazard. By insuring firms against the capture of their profits, competition authorities disincentivize all forms of risk-mitigation on the part of those firms. This will ultimately raise enforcement costs (as firms become increasingly reliant on the antitrust system for protection).

It is also informationally much more burdensome, as authorities will systematically have to rule on the appropriate share of profits between parties to a case.

Finally, overprotecting these investments would go against the philosophy of the European Court of Justice’s Huawei ruling.  Albeit in the specific context of injunctions relating to SEPs, the Court conditioned competition liability on firms showing that they have taken a series of reasonable steps to sort out their disputes privately.

Concluding remarks

This is not to say that competition intervention should categorically be proscribed. But rather that the capture of a retailer’s investments by Amazon is an insufficient condition for enforcement actions. Instead, the Commission should question whether Amazon’s actions are truly detrimental to consumer welfare and output. Absent strong evidence that an excluded retailer offered superior products, or that Amazon’s move was merely a strategic play to prevent entry, competition authorities should let the chips fall where they may.

As things stand, there is simply no evidence to indicate that anything out of the ordinary is occurring on the Amazon marketplace. By shining the spotlight on Amazon, the Commission is putting itself under tremendous political pressure to move forward with a formal investigation (all the more so, given the looming European Parliament elections). This is regrettable, as there are surely more pressing matters for the European regulator to deal with. The Commission would thus do well to recall the words of Shakespeare in the Merchant of Venice: “All that glisters is not gold”. Applied in competition circles this translates to “all that is big is not inefficient”.

In european commission, monopolization, truth on the market , , , , ,
permalink

AOL/Time Warner merger conditions are a template for disastrous tech policy

Kristian Stout —  30 July 2018

Senator Mark Warner has proposed 20 policy prescriptions for bringing “big tech” to heel. The proposals — which run the gamut from policing foreign advertising on social networks to regulating feared competitive harms — provide much interesting material for Congress to consider.

On the positive side, Senator Warner introduces the idea that online platforms may be able to function as least-cost avoiders with respect to certain tortious behavior of their users. He advocates for platforms to implement technology that would help control the spread of content that courts have found violated certain rights of third-parties.

Yet, on other accounts — specifically the imposition of an “interoperability” mandate on platforms — his proposals risk doing more harm than good.

The interoperability mandate was included by Senator Warner in order to “blunt [tech platforms’] ability to leverage their dominance over one market or feature into complementary or adjacent markets or products.” According to Senator Warner, such a measure would enable startups to offset the advantages that arise from network effects on large tech platforms by building their services more easily on the backs of successful incumbents.

Whatever you think of the moats created by network effects, the example of “successful” previous regulation on this issue that Senator Warner relies upon is perplexing:

A prominent template for [imposing interoperability requirements] was in the AOL/Time Warner merger, where the FCC identified instant messaging as the ‘killer app’ – the app so popular and dominant that it would drive consumers to continue to pay for AOL service despite the existence of more innovative and efficient email and internet connectivity services. To address this, the FCC required AOL to make its instant messaging service (AIM, which also included a social graph) interoperable with at least one rival immediately and with two other rivals within 6 months.

But the AOL/Time Warner merger and the FCC’s conditions provide an example that demonstrates the exact opposite of what Senator Warner suggests. The much-feared 2001 megamerger prompted, as the Senator notes, fears that the new company would be able to leverage its dominance in the nascent instant messaging market to extend its influence into adjacent product markets.

Except, by 2003, despite it being unclear that AOL had developed interoperable systems, two large competitors had arisen that did not run interoperable IM networks (Yahoo! and Microsoft). In that same period, AOL’s previously 100% IM market share had declined by about half. By 2009, after eight years of heavy losses, Time Warner shed AOL, and by last year AIM was completely dead.

Not only was it not clear that AOL was able to make AIM interoperable, AIM was never able to catch up once better, rival services launched. What the conditions did do, however, was prevent AOL from launching competitive video chat services as it flailed about in the wake of the deal, thus forcing it to miss out on a market opportunity available to unencumbered competitors like Microsoft and Yahoo!

And all of this of course ignores the practical impossibility entailed in interfering in highly integrated technology platforms.

The AOL/Time Warner merger conditions are no template for successful tech regulation. Congress would be ill-advised to rely upon such templates for crafting policy around tech and innovation.

In technology, truth on the market , , , ,
permalink

Amazon and the Unwisdom of the Populist Crowd

Kristian Stout —  1 May 2018

There are some who view a host of claimed negative social ills allegedly related to the large size of firms like Amazon as an occasion to call for the company’s break up. And, unfortunately, these critics find an unlikely ally in President Trump, whose tweet storms claim that tech platforms are too big and extract unfair rents at the expense of small businesses. But these critics are wrong: Amazon is not a dangerous monopoly, and it certainly should not be broken up.  

Of course, no one really spells out what it means for these companies to be “too big.” Even Barry Lynn, a champion of the neo-Brandeisian antitrust movement, has shied away from specifics. The best that emerges when probing his writings is that he favors something like a return to Joe Bain’s “Structure-Conduct-Performance” paradigm (but even here, the details are fuzzy).

The reality of Amazon’s impact on the market is quite different than that asserted by its critics. Amazon has had decades to fulfill a nefarious scheme to suddenly raise prices and reap the benefits of anticompetive behavior. Yet it keeps putting downward pressure on prices in a way that seems to be commoditizing goods instead of building anticompetitive moats.

Amazon Does Not Anticompetitively Exercise Market Power

Twitter rants aside, more serious attempts to attack Amazon on antitrust grounds argue that it is engaging in pricing that is “predatory.” But “predatory pricing” requires a specific demonstration of factors — which, to date, have not been demonstrated — in order to justify legal action. Absent a showing of these factors, it has long been understood that seemingly “predatory” conduct is unlikely to harm consumers and often actually benefits consumers.

One important requirement that has gone unsatisfied is that a firm engaging in predatory pricing must have market power. Contrary to common characterizations of Amazon as a retail monopolist, its market power is less than it seems. By no means does it control retail in general. Rather, less than half of all online commerce (44%) takes place on its platform (and that number represents only 4% of total US retail commerce). Of that 44 percent, a significant portion is attributable to the merchants who use Amazon as a platform for their own online retail sales. Rather than abusing a monopoly market position to predatorily harm its retail competitors, at worst Amazon has created a retail business model that puts pressure on other firms to offer more convenience and lower prices to their customers. This is what we want and expect of competitive markets.

The claims leveled at Amazon are the intellectual kin of the ones made against Walmart during its ascendancy that it was destroying main street throughout the nation. In 1993, it was feared that Walmart’s quest to vertically integrate its offerings through Sam’s Club warehouse operations meant that “[r]etailers could simply bypass their distributors in favor of Sam’s — and Sam’s could take revenues from local merchants on two levels: as a supplier at the wholesale level, and as a competitor at retail.” This is a strikingly similar accusation to those leveled against Amazon’s use of its Seller Marketplace to aggregate smaller retailers on its platform.

But, just as in 1993 with Walmart, and now with Amazon, the basic fact remains that consumer preferences shift. Firms need to alter their behavior to satisfy their customers, not pretend they can change consumer preferences to suit their own needs. Preferring small, local retailers to Amazon or Walmart is a decision for individual consumers interacting in their communities, not for federal officials figuring out how best to pattern the economy.

All of this is not to say that Amazon is not large, or important, or that, as a consequence of its success it does not exert influence over the markets it operates in. But having influence through success is not the same as anticompetitively asserting market power.

Other criticisms of Amazon focus on its conduct in specific vertical markets in which it does have more significant market share. For instance, a UK Liberal Democratic leader recently claimed that “[j]ust as Standard Oil once cornered 85% of the refined oil market, today… Amazon accounts for 75% of ebook sales … .”

The problem with this concern is that Amazon’s conduct in the ebook market has had, on net, pro-competitive, not anti-competitive, effects. Amazon’s behavior in the ebook market has actually increased demand for books overall (and expanded output), increased the amount that consumers read, and decreased the price of theses books. Amazon is now even opening physical bookstores. Lina Khan made much hay in her widely cited article last year that this was all part of a grand strategy to predatorily push competitors out of the market:

The fact that Amazon has been willing to forego profits for growth undercuts a central premise of contemporary predatory pricing doctrine, which assumes that predation is irrational precisely because firms prioritize profits over growth. In this way, Amazon’s strategy has enabled it to use predatory pricing tactics without triggering the scrutiny of predatory pricing laws.

But it’s hard to allege predation in a market when over the past twenty years Amazon has consistently expanded output and lowered overall prices in the book market. Courts and lawmakers have sought to craft laws that encourage firms to provide consumers with more choices at lower prices — a feat that Amazon repeatedly accomplishes. To describe this conduct as anticompetitive is asking for a legal requirement that is at odds with the goal of benefiting consumers. It is to claim that Amazon has a contradictory duty to both benefit consumers and its shareholders, while also making sure that all of its less successful competitors also stay in business.

But far from creating a monopoly, the empirical reality appears to be that Amazon is driving categories of goods, like books, closer to the textbook model of commodities in a perfectly competitive market. Hardly an antitrust violation.

Amazon Should Not Be Broken Up

“Big is bad” may roll off the tongue, but, as a guiding ethic, it makes for terrible public policy. Amazon’s size and success are a direct result of its ability to enter relevant markets and to innovate. To break up Amazon, or any other large firm, is to punish it for serving the needs of its consumers.

None of this is to say that large firms are incapable of causing harm or acting anticompetitively. But we should accept calls for dramatic regulatory intervention  — especially from those in a position to influence regulatory or market reactions to such calls — to be supported by substantial factual evidence and legal and economic theory.

This tendency to go after large players is nothing new. As noted above, Walmart triggered many similar concerns thirty years ago. Thinking about Walmart then, pundits feared that direct competition with Walmart was fruitless:

In the spring of 1992 Ken Stone came to Maine to address merchant groups from towns in the path of the Wal-Mart advance. His advice was simple and direct: don’t compete directly with Wal-Mart; specialize and carry harder-to-get and better-quality products; emphasize customer service; extend your hours; advertise more — not just your products but your business — and perhaps most pertinent of all to this group of Yankee individualists, work together.

And today, some think it would be similarly pointless to compete with Amazon:

Concentration means it is much harder for someone to start a new business that might, for example, try to take advantage of the cheap housing in Minneapolis. Why bother when you know that if you challenge Amazon, they will simply dump your product below cost and drive you out of business?

The interesting thing to note, of course, is that Walmart is now desperately trying to compete with Amazon. But despite being very successful in its own right, and having strong revenues, Walmart doesn’t seem able to keep up.

Some small businesses will close as new business models emerge and consumer preferences shift. This is to be expected in a market driven by creative destruction. Once upon a time Walmart changed retail and improved the lives of many Americans. If our lawmakers can resist the urge to intervene without real evidence of harm, Amazon just might do the same.

In antitrust, e-books, truth on the market , , , , , ,
permalink

Soylent Analytica: The Graph is too Damn Open

Gus Hurwitz —  21 March 2018

The world discovered something this past weekend that the world had already known: that what you say on the Internet stays on the Internet, spread intractably and untraceably through the tendrils of social media. I refer, of course, to the Cambridge Analytica/Facebook SNAFU (or just Situation Normal): the disclosure that Cambridge Analytica, a company used for election analytics by the Trump campaign, breached a contract with Facebook in order to unauthorizedly collect information on 50 million Facebook users. Since the news broke, Facebook’s stock is off by about 10 percent, Cambridge Analytica is almost certainly a doomed company, the FTC has started investigating both, private suits against Facebook are already being filed, the Europeans are investigating as well, and Cambridge Analytica is now being blamed for Brexit.

That is all fine and well, and we will be discussing this situation and its fallout for years to come. I want to write about a couple of other aspects of the story: the culpability of 270,000 Facebook users in disclosing the data of 50 million of their peers, and what this situation tells us about evergreen proposals to “open up the social graph” by making users’ social media content portable.

I Have Seen the Enemy and the Enemy is Us

Most discussion of Cambridge Analytica’s use of Facebook data has focused on the large number of user records Cambridge Analytica obtained access to – 50 million – and the fact that it obtained these records through some problematic means (and Cambridge Analytica pretty clearly breached contracts and acted deceptively to obtain these records). But one needs to dig a deeper to understand the mechanics of what actually happened. Once one does this, the story becomes both less remarkable and more interesting.

(For purposes of this discussion, I refer to Cambridge Analytica as the actor that obtained the records. It’s actually a little more complicated: Cambridge Analytica worked with an academic researcher to obtain these records. That researcher was given permission by Facebook to work with and obtain data on users for purposes relating to his research. But he exceeded that scope of authority, sharing the data that he collected with CA.)

The 50 million users’ records that Cambridge Analytica obtained access to were given to Cambridge Analytica by about 200,000 individual Facebook users. Those 270,000 users become involved with Cambridge Analytica by participating in an online quiz – one of those fun little throwaway quizzes that periodically get some attention on Facebook and other platforms. As part of taking that quiz, those 270,000 users agreed to grant Cambridge Analytica access to their profile information, including information available through their profile about their friends.

This general practice is reasonably well known. Any time a quiz or game like this has its moment on Facebook it is also accompanied by discussion of how the quiz or game is likely being used to harvest data about users. The terms of use of these quizzes and games almost always disclose that such information is being collected. More telling, any time a user posts a link to one of these quizzes or games, some friend will will invariably leave a comment warning about these terms of service and of these data harvesting practices.

There are two remarkable things about this. The first remarkable thing is that there is almost nothing remarkable about the fact that Cambridge Analytica obtained this information. A hundred such data harvesting efforts have preceded Cambridge Analytica; and a hundred more will follow it. The only remarkable things about the present story is that Cambridge Analytica was an election analytics firm working for Donald Trump – never mind that by all accounts the data collected proved to be of limited use generally in elections or that when Cambridge Analytica started working for the Trump campaign they were tasked with more mundane work that didn’t make use of this data.

More remarkable is that Cambridge Analytica didn’t really obtain data about 50 million individuals from Facebook, or from a Facebook quiz. Cambridge Analytica obtained this data from those 50 million individuals’ friends.

There are unquestionably important questions to be asked about the role of Facebook in giving users better control over, or ability to track uses of, their information. And there are questions about the use of contracts such as that between Facebook and Cambridge Analytica to control how data like this is handled. But this discussion will not be complete unless and until we also understand the roles and responsibilities of individual users in managing and respecting the privacy of their friends.

Fundamentally, we lack a clear and easy way to delineate privacy rights. If I share with my friends that I participated in a political rally, that I attended a concert, that I like certain activities, that I engage in certain illegal activities, what rights do I have to control how they subsequently share that information? The answer in the physical world, in the American tradition, is none – at least, unless I take affirmative steps to establish such a right prior to disclosing that information.

The answer is the same in the online world, as well – though platforms have substantial ability to alter this if they so desire. For instance, Facebook could change the design of its system to prohibit users from sharing information about their friends with third parties. (Indeed, this is something that most privacy advocates think social media platforms should do.) But such a “solution” to the delineation problem has its own problems. It assumes that the platform is the appropriate arbiter of privacy rights – a perhaps questionable assumption given platforms’ history of getting things wrong when it comes to privacy. More trenchant, it raises questions about users’ ability to delineate or allocate their privacy differently than allowed by the platforms, particularly where a given platform may not allow the delineation or allocation of rights that users prefer.

The Badness of the Open Graph Idea

One of the standard responses to concerns about how platforms may delineate and allow users to allocate their privacy interests is, on the one hand, that competition among platforms would promote desirable outcomes and that, on the other hand, the relatively limited and monopolistic competition that we see among firms like Facebook is one of the reasons that consumers today have relatively poor control over their information.

The nature of competition in markets such as these, including whether and how to promote more of it, is a perennial and difficult topic. The network effects inherent in markets like these suggest that promoting competition may in fact not improve consumer outcomes, for instance. Competition could push firms to less consumer-friendly privacy positions if that allows better monetization and competitive advantages. And the simple fact that Facebook has lost 10% of its value following the Cambridge Analytica news suggests that there are real market constraints on how Facebook operates.

But placing those issues to the side for now, the situation with Cambridge Analytica offers an important cautionary tale about one of the perennial proposals for how to promote competition between social media platforms: “opening up the social graph.” The basic idea of these proposals is to make it easier for users of these platforms to migrate between platforms or to use the features of different platforms through data portability and interoperability. Specific proposals have taken various forms over the years, but generally they would require firms like Facebook to either make users’ data exportable in a standardized form so that users could easily migrate it to other platforms or to adopt a standardized API that would allow other platforms to interoperate with data stored on the Facebook platform.

In other words, proposals to “open the social graph” are proposals to make it easier to export massive volumes of Facebook user data to third parties at efficient scale.

If there is one lesson from the past decade that is more trenchant than that delineation privacy rights is difficult it is that data security is even harder.

These last two points do not sum together well. The easier that Facebook makes it for its users’ data to be exported at scale, the easier Facebook makes it for its users’ data to be exfiltrated at scale. Despite its myriad problems, Cambridge Analytica at least was operating within a contractual framework with Facebook – it was a known party. Creating external API for exporting Facebook data makes it easier for unknown third-parties to anonymously obtain user information. Indeed, even if the API only works to allow trusted third parties to to obtain such information, the problem of keeping that data secured against subsequent exfiltration multiplies with each third party that is allowed access to that data.

In consumer protection, federal trade commission, regulation, technology, truth on the market , , , , ,
permalink

Pushing Ad Networks Out of Business: Yershov v. Gannett and the War Against Online Platforms

Kristian Stout —  10 May 2016

The lifecycle of a law is a curious one; born to fanfare, a great solution to a great problem, but ultimately doomed to age badly as lawyers seek to shoehorn wholly inappropriate technologies and circumstances into its ambit. The latest chapter in the book of badly aging laws comes to us courtesy of yet another dysfunctional feature of our political system: the Supreme Court nomination and confirmation process.

In 1988, President Reagan nominated Judge Bork for a spot on the US Supreme Court. During the confirmation process following his nomination, a reporter was able to obtain a list of videos he and his family had rented from local video rental stores (You remember those, right?). In response to this invasion of privacy — by a reporter whose intention was to publicize and thereby (in some fashion) embarrass or “expose” Judge Bork — Congress enacted the Video Privacy Protection Act (“VPPA”).

In short, the VPPA makes it illegal for a “video tape service provider” to knowingly disclose to third parties any “personally identifiable information” in connection with the viewing habits of a “consumer” who uses its services. Left as written and confined to the scope originally intended for it, the Act seems more or less fine. However, over the last few years, plaintiffs have begun to use the Act as a weapon with which to attack common Internet business models in a manner wholly out of keeping with drafters’ intent.

And with a decision that promises to be a windfall for hungry plaintiff’s attorneys everywhere, the First Circuit recently allowed a plaintiff, Alexander Yershov, to make it past a 12(b)(6) motion on a claim that Gannett violated the VPPA with its  USA Today Android mobile app.

What’s in a name (or Android ID) ?

The app in question allowed Mr. Yershov to view videos without creating an account, providing his personal details, or otherwise subscribing (in the generally accepted sense of the term) to USA Today’s content. What Gannett did do, however, was to provide to Adobe Systems the Android ID and GPS location data associated with Mr. Yershov’s use of the app’s video content.

In interpreting the VPPA in a post-Blockbuster world, the First Circuit panel (which, apropos of nothing, included retired Justice Souter) had to wrestle with whether Mr. Yershov counts as a “subscriber,” and to what extent an Android ID and location information count as “personally identifying information” under the Act. Relying on the possibility that Adobe might be able to infer the identity of the plaintiff given its access to data from other web properties, and given the court’s rather gut-level instinct that an app user is a “subscriber,” the court allowed the plaintiff to survive the 12(b)(6) motion.

The PII point is the more arguable of the two, as the statutory language is somewhat vague. Under the Act, PIII “includes information which identifies a person as having requested or obtained specific video materials or services from a video tape service provider.” On this score the court decided that GPS data plus an Android ID (or each alone — it wasn’t completely clear) could constitute information protected under the Act (at least for purposes of a 12(b)(6) motion):

The statutory term “personally identifiable information” is awkward and unclear. The definition of that term… adds little clarity beyond training our focus on the question whether the information identifies the person who obtained the video…. Nevertheless, the language reasonably conveys the point that PII is not limited to information that explicitly names a person.

OK (maybe). But where the court goes off the rails is in its determination that an Android ID, GPS data, or a list of videos is, in itself, enough to identify anyone.

It might be reasonable to conclude that Adobe could use that information in combination with other information it collects from yet other third parties (fourth parties?) in order to build up a reliable, personally identifiable profile. But the statute’s language doesn’t hang on such a combination. Instead, the court’s reasoning finds potential liability by reading this exact sort of prohibition into the statute:

Adobe takes this and other information culled from a variety of sources to create user profiles comprised of a given user’s personal information, online behavioral data, and device identifiers… These digital dossiers provide Adobe and its clients with “an intimate look at the different types of materials consumed by the individual” … While there is certainly a point at which the linkage of information to identity becomes too uncertain, or too dependent on too much yet-to-be-done, or unforeseeable detective work, here the linkage, as plausibly alleged, is both firm and readily foreseeable to Gannett.

Despite its hedging about uncertain linkages, the court’s reasoning remains contingent on an awful lot of other moving parts — something not found in either the text of the law, nor the legislative history of the Act.

The information sharing identified by the court is in no way the sort of simple disclosure of PII that easily identifies a particular person in the way that, say, Blockbuster Video would have been able to do in 1988 with disclosure of its viewing lists.  Yet the court purports to find a basis for its holding in the abstract nature of the language in the VPPA:

Had Congress intended such a narrow and simple construction [as specifying a precise definition for PII], it would have had no reason to fashion the more abstract formulation contained in the statute.

Again… maybe. Maybe Congress meant to future-proof the provision, and didn’t want the statute construed as being confined to the simple disclosure of name, address, phone number, and so forth. I doubt, though, that it really meant to encompass the sharing of any information that might, at some point, by some unknown third parties be assembled into a profile that, just maybe if you squint at it hard enough, will identify a particular person and their viewing habits.

Passive Subscriptions?

What seems pretty clear, however, is that the court got it wrong when it declared that Mr. Yershov was a “subscriber” to USA Today by virtue of simply downloading an app from the Play Store.

The VPPA prohibits disclosure of a “consumer’s” PII — with “consumer” meaning “any renter, purchaser, or subscriber of goods or services from a video tape service provider.” In this case (as presumably will happen in most future VPPA cases involving free apps and websites), the plaintiff claims that he is a “subscriber” to a “video tape” service.

The court built its view of “subscriber” predominantly on two bases: (1) you don’t need to actually pay anything to count as a subscriber (with which I agree), and (2) that something about installing an app that can send you push notifications is different enough than frequenting a website, that a user, no matter how casual, becomes a “subscriber”:

When opened for the first time, the App presents a screen that seeks the user’s permission for it to “push” or display notifications on the device. After choosing “Yes” or “No,” the user is directed to the App’s main user interface.

The court characterized this connection between USA Today and Yershov as “seamless” — ostensibly because the app facilitates push notifications to the end user.

Thus, simply because it offers an app that can send push notifications to users, and because this app sometimes shows videos, a website or Internet service — in this case, an app portal for a newspaper company — becomes a “video tape service,” offering content to “subscribers.” And by sharing information in a manner that is nowhere mentioned in the statute and that on its own is not capable of actually identifying anyone, the company suddenly becomes subject to what will undoubtedly be an avalanche of lawsuits (at least in the first circuit).

Preposterous as this may seem on its face, it gets worse. Nothing in the court’s opinion is limited to “apps,” and the “logic” would seem to apply to the general web as well (whether the “seamless” experience is provided by push notifications or some other technology that facilitates tighter interaction with users). But, rest assured, the court believes that

[B]y installing the App on his phone, thereby establishing seamless access to an electronic version of USA Today, Yershov established a relationship with Gannett that is materially different from what would have been the case had USA Today simply remained one of millions of sites on the web that Yershov might have accessed through a web browser.

Thank goodness it’s “materially” different… although just going by the reasoning in this opinion, I don’t see how that can possibly be true.

What happens when web browsers can enable push notifications between users and servers? Well, I guess we’ll find out soon because major browsers now support this feature. Further, other technologies — like websockets — allow for continuous two-way communication between users and corporate sites. Does this change the calculus? Does it meet the court’s “test”? If so, the court’s exceedingly vague reasoning provides little guidance (and a whole lot of red meat for lawsuits).

To bolster its view that apps are qualitatively different than web sites with regard to their delivery to consumers, the court asks “[w]hy, after all, did Gannett develop and seek to induce downloading of the App?” I don’t know, because… cell phones?

And this bit of “reasoning” does nothing for the court’s opinion, in fact. Gannett undertook development of a web site in the first place because some cross-section of the public was interested in reading news online (and that was certainly the case for any electronic distribution pre-2007). No less, consumers have increasingly been moving toward using mobile devices for their online activities. Though it’s a debatable point, apps can often provide a better user experience than that provided by a mobile browser. Regardless, the line between “app” and “web site” is increasingly a blurry one, especially on mobile devices, and with the proliferation of HTML5 and frameworks like Google’s Progressive Web Apps, the line will only grow more indistinct. That Gannett was seeking to provide the public with an app has nothing to do with whether it intended to develop a more “intimate” relationship with mobile app users than it has with web users.

The 11th Circuit, at least, understands this. In Ellis v. Cartoon Network, it held that a mere user of an app — without more — could not count as a “subscriber” under the VPPA:

The dictionary definitions of the term “subscriber” we have quoted above have a common thread. And that common thread is that “subscription” involves some type of commitment, relationship, or association (financial or otherwise) between a person and an entity. As one district court succinctly put it: “Subscriptions involve some or [most] of the following [factors]: payment, registration, commitment, delivery, [expressed association,] and/or access to restricted content.”

The Eleventh Circuit’s point is crystal clear, and I’m not sure how the First Circuit failed to appreciate it (particularly since it was the district court below in the Yershov case that the Eleventh Circuit was citing). Instead, the court got tied up in asking whether or not a payment was required to constitute a “subscription.” But that’s wrong. What’s needed is some affirmative step – something more than just downloading an app, and certainly something more than merely accessing a web site.

Without that step — a “commitment, relationship, or association (financial or otherwise) between a person and an entity” — the development of technology that simply offers a different mode of interaction between users and content promises to transform the VPPA into a tremendously powerful weapon in the hands of eager attorneys, and a massive threat to the advertising-based business models that have enabled the growth of the web.

How could this possibly not apply to websites?

In fact, there is no way this opinion won’t be picked up by plaintiff’s attorneys in suits against web sites that allow ad networks to collect any information on their users. Web sites may not have access to exact GPS data (for now), but they do have access to fairly accurate location data, cookies, and a host of other data about their users. And with browser-based push notifications and other technologies being developed to create what the court calls a “seamless” experience for users, any user of a web site will count as a “subscriber” under the VPPA. The potential damage to the business models that have funded the growth of the Internet is hard to overstate.

There is hope, however.

Hulu faced a similar challenge over the last few years arising out of its collection of viewer data on its platform and the sharing of that data with third-party ad services in order to provide better targeted and, importantly, more user-relevant marketing. Last year it actually won a summary judgment motion on the basis that it had no way of knowing that Facebook (the third-party with which it was sharing data) would reassemble the data in order to identify particular users and their viewing habits. Nevertheless, Huu has previously lost motions on the subscriber and PII issues.

Hulu has, however, previously raised one issue in its filings on which the district court punted, but that could hold the key to putting these abusive litigations to bed.

The VPPA provides a very narrowly written exception to the prohibition on information sharing when such sharing is “incident to the ordinary course of business” of the “video tape service provider.” “Ordinary course of business” in this context means  “debt collection activities, order fulfillment, request processing, and the transfer of ownership.” In one of its motions, Hulu argued that

the section shows that Congress took into account that providers use third parties in their business operations and “‘allows disclosure to permit video tape service providers to use mailing houses, warehouses, computer services, and similar companies for marketing to their customers. These practices are called ‘order fulfillment’ and ‘request processing.’

The district court didn’t grant Hulu summary judgment on the issue, essentially passing on the question. But in 2014 the Seventh Circuit reviewed a very similar set of circumstances in Sterk v. Redbox and found that the exception applied. In that case Redbox had a business relationship with Stream, a third party that provided Redbox with automated customer service functions. The Seventh Circuit found that information sharing in such a relationship fell within Redbox’s “ordinary course of business”, and so Redbox was entitled to summary judgment on the VPPA claims against it.

This is essentially the same argument that Hulu was making. Third-party ad networks most certainly provide a service to corporations that serve content over the web. Hulu, Gannett and every other publisher on the web surely could provide their own ad platforms on their own properties. But by doing so they would lose the economic benefits that come from specialization and economies of scale. Thus, working with a third-party ad network pretty clearly replaces the “order fulfillment” and “request processing” functions of a content platform.

The Big Picture

And, stepping back for a moment, it’s important to take in the big picture. The point of the VPPA was to prevent public disclosures that would chill speech or embarrass individuals; the reporter in 1987 set out to expose or embarrass Judge Bork.  This is the situation the VPPA’s drafters had in mind when they wrote the Act. But the VPPA was most emphatically not designed to punish Internet business models — especially of a sort that was largely unknown in 1988 — that serve the interests of consumers.

The 1988 Senate report on the bill, for instance, notes that “[t]he bill permits the disclosure of personally identifiable information under appropriate and clearly defined circumstances. For example… companies may sell mailing lists that do not disclose the actual selections of their customers.”  Moreover, the “[Act] also allows disclosure to permit video tape service providers to use mailing houses, warehouses, computer services, and similar companies for marketing to their customers. These practices are called ‘order fulfillment’ and ‘request processing.’”

Congress plainly contemplated companies being able to monetize their data. And this just as plainly includes the common practice in automated tracking systems on the web today that use customers’ viewing habits to serve them with highly personalized web experiences.

Sites that serve targeted advertising aren’t in the business of embarrassing consumers or abusing their information by revealing it publicly. And, most important, nothing in the VPPA declares that information sharing is prohibited if third party partners could theoretically construct a profile of users. The technology to construct these profiles simply didn’t exist in 1988, and there is nothing in the Act or its legislative history to support the idea that the VPPA should be employed against the content platforms that outsource marketing to ad networks.

What would make sense is to actually try to fit modern practice in with the design and intent of the VPPA. If, for instance, third-party ad networks were using the profiles they created to extort, blackmail, embarrass, or otherwise coerce individuals, the practice certainly falls outside of course of business, and should be actionable.

But as it stands, much like the TCPA, the VPPA threatens to become a costly technological anachronism. Future courts should take the lead of the Eleventh and Seventh circuits, and make the law operate in the way it was actually intended. Gannett still has the opportunity to appeal for an en banc hearing, and after that for cert before the Supreme Court. But the circuit split this presents is the least of our worries. If this issue is not resolved in a way that permits platforms to continue to outsource their marketing efforts as they do today, the effects on innovation could be drastic.

Web platforms — which includes much more than just online newspapers — depend upon targeted ads to support their efforts. This applies to mobile apps as well. The “freemium” model has eclipsed the premium model for apps — a fact that expresses the preferences of both consumers at large as well as producers. Using the VPPA as a hammer to smash these business models will hurt everyone except, of course, for plaintiff’s attorneys.

In advertising, internet, privacy, technology, truth on the market , , , ,
permalink