In this second part of our two-part interview with Maurice Stucke and Ariel Ezrachi, the two discuss how competition authorities can mitigate tacit collusion in the age of big data and artificial intelligence and how antitrust could adapt to the new dynamic of the market.
Last week, we published the first installment of a two-part interview with Maurice Stucke from the University of Tennessee Knoxville and Ariel Ezrachi of Oxford University, whose research and upcoming book, Virtual Competition, explore the changing nature of competition in the age of big data and algorithms.
While the digital economy is often touted by politicians and regulators as the epitome of a competitive and dynamic market, Stucke and Ezrachi point to a number of developments and market dynamics that make the online marketplace much less free than it appears to be in the eyes of consumers.
In the previous installment, we discussed how the dynamics of competition are changing due to network effects and the increased reliance on big data and big analytics. We considered three main theories of harm—collusion between algorithms, behavioral discrimination, and the frenemy dynamic which operates in favor of “super-platforms”—that may lead to exclusion of some companies, to the detriment of consumers.
In the second part of our interview, Stucke and Ezrachi discuss the changes that competition authorities and antitrust enforcers need to make in order to protect both the consumers and innovation.
Q: When we look at the role of antitrust in the age of big data, what are the challenges and new threats to competition that competition authorities should be aware of?
Maurice Stucke: We note how Big Data and Big Analytics—in increasing the speed of communicating price changes, detecting any cheating or deviations, and punishing such deviations—can provide new and enhanced means to foster collusion. The danger here is not express collusion where computers limit competition through “agreement,” but more elusive forms of collusion, achieved through subtler means, which do not amount to a hard-core cartel, and are beyond the reach of the law. Our book and paper consider four scenarios in which computer algorithms may promote collusion:
The first scenario—Messenger—concerns humans’ agreeing to collude and using computers to execute their will. Under this scenario, humans collude. They use computers to assist in implementing, monitoring, and policing the cartel or to facilitate information exchange and signaling. In the United States and elsewhere, they go to jail if caught.
Our second scenario—Hub and Spoke—is more challenging. As we mentioned in last week’s interview, here we consider the use of a single pricing algorithm to determine the market price charged by numerous users. In this framework, a cluster of similar vertical agreements with many of the industries’ competitors may give rise to a classic hub-and-spoke conspiracy, whereby the algorithm developer, as the hub, helps orchestrate industry-wide collusion, leading to higher prices.
The third scenario—The Predictable Agent—is even more challenging. It explores how we are shifting from a world where executives expressly collude in smoke-filled hotel rooms, to a world where pricing algorithms act as predictable agents and continually monitor and adjust to each other’s prices and market data. In this new world, there is no collusive agreement among executives. Each firm unilaterally adopts its own pricing algorithm, which sets its own price. The result is algorithm-enhanced conscious parallelism—or as we call it, Tacit Collusion on Steroids.
Finally, we consider the most challenging collusion scenario—Digital Eye. The computers, in learning by doing, determine independently the means to optimize profit. Artificial intelligence operating in enhanced market transparency leads to an anticompetitive outcome, with no evidence of any anticompetitive agreement or intent. In this scenario we may not even know when something is amiss. In the end, we may think the markets, driven by these technologies, are competitive. And yet, we’re not benefitting from this virtual competition.
Q: What can regulators and enforcement agencies do about these collusion dynamics?
MS: Antitrust law can readily address the first scenario. The law can also reach the second scenario. The challenge is determining which legal standard applies: the more permissive rule of reason (if the agreements are characterized as vertical) or the per se illegal standard (if the agreements are characterized as horizontal).
The third and fourth scenarios are more problematic. Here, humans unilaterally design the algorithm to deliver predictable outcomes and react in a given way to changing market conditions. In the latter two categories, there is insufficient evidence of any agreement (either vertical or horizontal). Without agreement as such, the market may nonetheless exhibit the conditions for tacit collusion/conscious parallelism. As tacit collusion is not in itself illegal, proof of intent to change market dynamics is central in this scenario. The agencies have far fewer tools when there’s only evidence of anti-competitive intent. And, as in the last scenario, when there isn’t any evidence of an anti-competitive agreement or intent, the agencies often have no tools.
So when one considers the increasing enforcement challenges as we move from the first to fourth scenario, it is possible to identify the point where enforcement, under the current law, ebbs. This concern is universal within the antitrust community, namely that computers and particularly artificial intelligence might help facilitate collusion in ways that the current antitrust tools can’t readily address.
Ariel Ezrachi: When we consider the antitrust enforcement toolbox, most often we focus on ex-post tools. That is–the use of remedies to address an antitrust violation which already took place. So, for example, following a finding of abuse of market power or monopolization, the company will be ordered to chan
ge the way it operates on the market and will be subjected to financial sanctions. Or, following a finding of collusion between companies, they will be ordered to terminate their illegal activity and will be subjected to administrative or criminal sanctions.
As Maurice indicated, these tools, which dominate our enforcement landscape, can deal with some issues like clear collusion or abuse of market power, but they are not likely to be triggered by some of the other scenarios we discuss.
In our research we venture beyond ex-post intervention and also consider possible ex-ante instruments. Pre-emptive intervention, when carefully executed, may offer us a fine tuned mechanism adequate for fast moving technology markets. It may provide competition agencies with a range of tools to change the dynamics of market power, the transparency of the market, or the way online companies are able or allowed to present the goods they are selling.
Ex-ante means, can enable the competition agencies and other regulators to focus on the competition dynamics and better align the interests of the super-platforms and consumers.
Q: Some might say that we’ve only just begun exploiting the efficiencies, the innovation, and the possibilities of online platforms, and already you want to have more regulation, more restrictions.
AE: It is important to note that the current online markets are far from the ideal of perfect competition that only exists in economic casebooks. In a perfectly competitive market, both buyers and sellers are fully informed. In the online world, the buyer’s rational decision-making is often undermined by a controlled ecosystem in which transparency and choice can be distorted. Perhaps most striking, the targets of these practices—the buyers—are often unaware of the extent of the manipulation.
Of course, antitrust enforcement should not be taken lightly and only deployed where harm is likely or established. One needs to be mindful of the risk of over intervention—the risk that competition enforcement might chill innovation and competition. At the same time, one needs to be mindful of the cost of under-enforcement.
We are certainly not in favor of intervention for the sake of intervention. What we do think, is that possible intervention should not be categorically set aside as irrelevant for new dynamic markets. The enforcement toolbox offers many mechanisms which could be deployed—to address some of the new dynamics while allowing competition and innovation to flourish.
The opponents of intervention often argue that antitrust law is ill-suited for data-driven markets and that existing case law does not support intervention. But, antitrust law is only one tool in a bigger enforcement toolbox. Harm and market failure cannot be defined by the enforcement tools. Rather, once they are identified, a range of enforcement actions—some ex-ante, some ex-post—should be considered.
It is true that the digital economy is booming. But our concerns are not with technological advances or successful online businesses. Our concerns go deeper to the core of the new market dynamics: where entry is possible, but expansion will likely be controlled by super-platforms; where choice is ample but competition may be limited; and where wealth is transferred through discriminatory pricing while retaining a competitive façade.
In line with this, an argument that “no previous case law supports intervention” fails to look at the full enforcement toolbox and to acknowledge the change in market and competitive dynamics.
Q: In recent years we’ve seen a stark difference between the approaches taken by the competition authorities in Europe and in the U.S. toward digital giants. Are Europeans better equipped to deal with the challenges presented by data-driven markets?
MS: Yes and no. On one level, it will depend on the jurisdiction, because different jurisdictions have different tools in their toolkit. The U.K.’s Competition and Markets Authority, for example, has the power to undertake market investigations that the U.S. Department of Justice lacks. So a particular agency might be able to reach a novel anticompetitive practice that others cannot.
But on a basic level, both the U.S. and E.U. agencies have the tools to target a broad range of abuses by dominant firms. So it comes down to intellectual leadership and political will.
As we discuss in Big Data and Competition Policy, the harm from data-driven monopolies can be greater and longer lasting. Big Data can provide some dominant firms with a “Nowcasting Radar,” which makes them more dangerous than past monopolies. So one sign of a healthy antitrust program is when the competition authority, to protect society from these harms, is proactively identifying and challenging anticompetitive strategies that monopolies use to maintain or attain their power.
In the U.S., monopolization pays. The DOJ criminally prosecuted more persons in one year under the Migratory Bird Treaty Act (227 in 2012) than it has civilly and criminally prosecuted monopolies over the past 35 years (13 since 1980). Between 2005 and 2014, the DOJ opened only nineteen monopolization investigations, and brought only one case.
Contrast the U.S. with the E.U., where several agencies have been holding hearings about the implications of Big Data on competition. The French and German competition authorities are studying (and investigating) conduct by data-opolies. They recognize that reductions in privacy could be a matter of abuse, if for example, the monopoly collects data by clearly breaching the data protection law and there is a strong interplay between the data collection and the undertaking’s market position.
The U.K. Competition and Markets Authority in its 2015 report, considered how dominant firms could weaken competitors by denying them access to data. Likewise, the U.K. House of Lords, in an informative report examined how large online platforms can act as gateways to markets. They also observed a lack of competition between several dominant platforms on privacy standards, and suggested that online platforms could potentially abuse a dominant position by downgrading their privacy standards.
And besides its active, ongoing investigations, the European Commission released its Preliminary Report on its e-commerce sector inquiry which addresses many market dynamics and the way businesses have adjusted and reacted to onli
The subject of monopolies seems to be gaining greater momentum on both sides of the Atlantic. With that in mind, the Europeans seem to be more willing to take a hard look on the effects these dynamics may have on consumers and to follow up with enforcement actions.
Q: Would you say that since this industry is developing so fast, there is an acute problem of asymmetric information between regulators and the industry?
MS: It’s hard to make any broad generalization about the competition authorities without looking at what particular tools are in each competition authority’s toolkit, and to what extent can those tools address the anticompetitive scenarios we identify.
In Virtual Competition, we identify several looming enforcement challenges presented by our scenarios. The competition agency may not see any problem. This may be due to informational asymmetries. Or it may be due to ideological beliefs in self-correcting markets. Even if the agency does see the problem, it may not have any enforcement tools to fix the problem. The issues may transcend antitrust, given the difficult legal and ethical issues of humans’ accountability for a computer’s behavior. Finally, even if the competition agency has the tools, when should it intervene?
So the enforcement challenges presented by the new market dynamics may be divided into several groups. The first is ideology around how markets operate and the goals of competition law. The second is political will. The effects of intellectual and regulatory capture cannot be underestimated. The third is ingenuity.
The lack of tools may represent a policy failure—namely, the agency’s pursuit of what is readily quantifiable over what is important. It is all the more startling that in the age of Big Data and Big Analytics, the government is the entity that claims it lacks the data or analytical tools to assess a restraint’s impact. For the areas where their current tools may be ill-suited, the agencies should study how they can refine them or develop alternatives.
Again, it comes down to intellectual leadership and political will.
Q: Maurice, in the book you wrote with Allen Grunes, Big Data and Competition Policy, you discuss what you call data-driven mergers, and their implications for consumer welfare and competition law. In your submission to the House of Lords, you quote an estimate that said big data-related mergers doubled between 2008 and 2013.((European Data Protection Supervisor, Report of Workshop on Privacy, Consumers, Competition and Big Data (2014).)) What can competition authorities do about this, and should they even intervene?
MS: Big Data exposes several weaknesses in today’s antitrust framework for evaluating mergers.
First, antitrust has gravitated to assessing mergers’ short-term price effects in narrowly-defined markets. What is readily measurable has become increasingly important. Thus the agencies have pretty good tools for assessing a merger’s likely price effects in some markets. But as Hayek noted, what is measurable in economics is not always what is important. Parameters of competition that are important but harder to quantify (such as quality when the product is free) are often marginalized or ignored.
Competition agencies have very crude tools for assessing a merger’s impact on non-price parameters of competition such as quality and privacy protections. Basically, their price-centric approach doesn’t work very well when the product or service is free. Thus antitrust’s price-centric approach fails in data-driven markets, where products or services are often offered for free, many consumers are accustomed to not paying for the product or service, and the potential harms, while significant, are harder to quantify.
Second, data-driven mergers can defy the agencies’ conventions that work well for other mergers. For example, data-driven mergers at times defy the existing way competition authorities categorize mergers as either horizontal, vertical, or conglomerate. Suppose Google sought to acquire Twitter. Google does not currently compete with Twitter in this space. Nor is Google a key buyer or supplier to Twitter. Nor is Google a perceived potential entrant in Twitter’s market. As a Google-Twitter merger illustrates, data-driven mergers can differ from traditional mergers in that key data at times are neither sold nor bought. The acquirer buys the target not for its product per se but the data the acquiring company collects.
So for these data-driven mergers, the agencies should broaden their inquiry and examine (i) the merger’s impact on each side of the multi-sided platform (both on advertisers and Twitter users), (ii) whether the merger increases the likelihood of the firm degrading quality on the free side (including the privacy protections on the data collected and its uses), (iii) whether the data from the acquired or acquiring entity helps the firm attain or maintain its power in any market, and (iv) whether the merger increases entry barriers in the social network market or other markets.
Q: A lot has been written about the wide-ranging influence internet giants have over lawmakers and regulators, particularly in the U.S. What is your opinion?
AE: On-going efforts by pressure groups, chambers of commerce, corporations and other interested parties, have always dominated the antitrust landscape. Indeed, since its early days, competition policy has been shaped by social, political and economic considerations. This susceptibility, the sponginess of competition law, certainly opens the gate to intellectual capture.
We have become accustomed to lobbying which is used to affect the enforcer’s views as to the appropriate scope of competition enforcement, the merit of pursuing a case or investigating an industry. This may be case-specific or more generally used to affect public discussions through conferences, think tanks, publications and other means. Naturally, large successful companies, such as the online giants, have greater capacity to invest in such efforts and translate their economic power into influence on law, policy and politics.
What is interesting and different, is that in the new online environment, the big platforms benefit not only from deep pockets but from greater control over our communication interfaces, our news feeds and our data. One could see how this may provide some of them with the power to capture the public debate through media and online channels.
So, as we increasingly rely on a specific platform to access the web, and as we move toward increased use of voice-activated digital assistants, we will likely have less control over our online eco-system, over our purchasing options, news feeds, and over our view of the world.
Q: Should the focus of antitrust policy shift from consumer welfare to market power and political power?
MS: First, there’s no consensus among competition authorities on what precisely the term “consumer welfare” means. “Consumer welfare” is one of those open-ended terms that conceivably could host various political, social, and economic objectives.
One point in Virtual Competition and Big Data and Competition Policy is to inquire “What do we mean by consumer welfare? And what does that term mean in the world of big data?” Historically, competition authorities looked at consumer surplus, but how is that relevant when it comes to free products? And because it’s free, does that necessarily mean that consumers always benefit?
Even though a product may be free, you can’t assume that consumers automatically benefit. Free is simply one point along a price spectrum. You can have negative prices whereby the service pays for your data.
Moreover, other non-price dimensions of competition such as quality and the level of privacy protections may take on greater importance in these markets. So our conception of consumer welfare has to move beyond consumer surplus, to account more generally for consumer well-being.
Q: Should we be concerned about the political influence of these super-platforms?
MS: Yes. What we’re dealing with are not ordinary commodities like bolts or washers. We are dealing with platforms that people are increasingly relying upon for many aspects of their day-to-day lives. These platforms aren’t just providing emails and texting. They are also gateways to getting news, information, and the like.
Federal Trade Commission Chairman Robert Pitofsky discussed how excessive concentration of economic power will breed antidemocratic political pressures. In an interview about the AOL-Time Warner transaction, Chairman Pitofsky observed:
“Antitrust is more than economics . . . . And I do believe if you have issues in the newspaper business, in book publishing, news generally, entertainment, I think you want to be more careful and thorough in your investigation than if the very same problems arose in cosmetics, or lumber, or coal mining. I mean, if somebody monopolizes the cosmetics fields, they’re going to take money out of consumers’ pockets, but the implications for democratic values are zero. On the other hand, if they monopolize books, you’re talking about implications that go way beyond what the wholesale price of the books might be.”((Alec Klein, A Hard Look at Media Mergers, Wash. Post, Nov. 29. 2000, at E1, available at 2000 WL 29918451 (quoting Robert Pitofsky, FTC Chairman).))
We raised a similar concern with digital personal assistants. If one or two personal assistants dominate the market, they can have power to reduce consumer surplus. Importantly, however, the power does not stop there. The control over the key interface provides the platform with the ultimate power – to affect not only what we buy, but to affect our views and the public debate.
The increasing reliance on this gatekeeper could enable its operator to intellectually capture users, and subsequently decision makers, in an attempt to ultimately ensure that public opinion and government policies align with the corporate agenda.
So we are at a critical crossroad in the U.S. Not only over the future of antitrust policy and enforcement, but how antitrust can ensure that we get the benefits of a data-driven economy, while mitigating the risks.