In a forthcoming article, Seth Benzell and Felix Chang explore how antitrust regulators can use insights from a new quantitative model of Facebook that Benzell developed with Avinash Collis to estimate social welfare under various regulatory and antitrust scenarios. Benzell and Chang find that interventions such as taxes, mandated interoperability, and user rebates would be more socially beneficial than breakups. Acknowledging the limitations of the model, the authors hope it spurs regulators and policymakers to proffer their own quantitative models, advancing the conversation around digital antitrust beyond polemic.
The specter of digital platform monopolies haunts the chambers of congressmen and parliamentarians across the world. The breadth, power, and profitability of these companies have drawn bipartisan ire. While there has been legislation and regulation in Europe, the U.S. has taken a more restrained lawmaking approach. One reason has been the difficulty in measuring the deleterious effects of concentration—particularly in zero-price markets, where trade-offs are difficult to quantify. But now, with antitrust progressives installed in key positions at U.S. enforcement agencies, Big Tech is facing an onslaught of proposals, from interoperability to taxes to breakups.
Yet lawmakers, regulators, and industry rarely quantify how much the monopoly power of large digital platforms (e.g., Amazon, Meta, and Google) reduces or redistributes welfare. Such an analysis is important under the rule of reason. Under this standard, the appropriateness of antitrust enforcement depends on how much the public suffers from the monopoly. Quantification of harms and offsetting efficiencies can therefore help fine-tune regulations, taxes, and other remedies. Yet, the lack of such analyses has left prominent antitrust figures arguing past one another, often relying on polemics. In fact, antitrust enforcers have recently fallen short in establishing even more fundamental baselines. In 2021, the FTC’s complaint against Facebook was initially dismissed for failing to plausibly allege that Facebook wields monopoly power, though the FTC responded with an amended complaint.
Coincidentally, advances in the economics of multi-sided platforms are increasingly making it possible to quantitatively answer these essential questions. The innovations of Jean Tirole, Marshall Van Alstyne, Glen Weyl, and others can help us calculate if and how much a digital platform contravenes the public interest.
Building on this body of research, Seth Benzell and Avinash Collis recently devised a model to evaluate the societal consequences of oft-mentioned—and more obscure—solutions for Facebook. They calibrated the model to Facebook using regulatory filings and Internet surveys of over 57,000 users (representative of the U.S. population), which gauged demand for the platform. Using this model, the authors can estimate the social surplus from Facebook-like social networking services, and how it changes under different market structures and regulatory scenarios.
Following the economic literature, the Benzell–Collis model calculates social welfare as the sum of four variables: (1) Facebook’s consumer welfare, itself distributed across different demographic groups; (2) its advertising revenues; (3) the tax revenues raised from Facebook; and (4) the value to Facebook of maintaining a large user-base. In a forthcoming article for the Vanderbilt Law Review, we discuss what this model can tell us about the consequences of antitrust remedies. In the case of Facebook, one key finding is that redistributive solutions such as taxes on advertisements and data-as-labor rebates fare better than heavy-handed reforms such as breakups. More broadly, we also hope that the model and our essay will help regulators think more quantitatively about antitrust generally and inspire similar analyses of digital platform monopolies. Moving the conversation into a quantifiable realm can help overcome the morass of polemical arguments.
Before proceeding to the results, it is important to note that the model keeps track of several dimensions of social welfare, rather than only consumer welfare. Legal scholars debate whether antitrust should account only for the welfare of consumers or, more holistically, the welfare of firms and even noneconomic interests. Our approach is to measure all the aspects of social welfare we think we can measure well, while allowing regulators and law scholars to decide which are legally important. These aspects include the surplus that users receive from the platform (above their willingness to pay), tax revenues to the government, and the surplus that Facebook’s owners receive from their profits and maintaining a large user base.
However, our approach falls short of an impossibly broad total welfare (which would account for a plethora of difficult-to-measure consequences). For example, if some Facebook usage is addictive (and a recent study suggests that perhaps 31% of Facebook usage is), then consumer’s willingness to pay for the service might not be a good measure of its social value. Similarly, if Facebook use causes depression (in a way unanticipated by users) or spreads fake news to the social detriment, these negative effects should theoretically factor into the platform’s social value. In this area, our model only illustrates how important it is that additional quantitative research be undertaken on these nebulous supposed social ills.
Even if these effects can be better measured, there will still be much to argue about. Antitrust has traditionally focused on price and output without regard to whether the underlying products (be they cancer-causing cigarettes or carbon-intensive oil) are actually good for the public. Paradoxically, then, antitrust enforcement may have the foreseeable consequence of lowering welfare. For Facebook in particular, neo-Brandeisians have certainly pointed to a host of negative externalities attendant to the “bigness” of the dominant digital platforms. Some of these (e.g., subverting democracy) are plausible but extraordinarily difficult to pin down numerically. Nonetheless, these theoretical and practical difficulties do not detract from the imperative of measuring what we can.
Those caveats aside, one important conclusion is that we should proceed with caution if Facebook were to be broken up. Many terrific papers have been written about breakups, including how they might affect labor markets, but botched breakups can destroy network effects without enhancing competition. For example, a horizontal breakup, which did not preserve network effects across the platforms, might lead to two Baby Facebooks, each monopolizing a segment of the market (as did the “Baby Bells,” progeny of the Bell Telephone Company breakup). Under these circumstances, the Benzell-Collis model predicts consumer welfare would plummet by 33% and user participation by 21.8%. Similarly, a vertical breakup, such as divestiture of Instagram or WhatsApp, might lead to a 5.3% drop in consumer welfare.
That said, if regulators could leverage a breakup of Facebook into creating a more competitive social media industry, significant benefits could flow. A key component of such a plan would be “mandated interoperability.” This would entail users of competitors offering Facebook-like services being able to interact with each other and port content across platforms. While a botched breakup would be disastrous, a breakup that created a competitive market and preserved the value of connections across platforms would boost social welfare by 4.8%.
Certain types of taxes also might help. A 3% tax on advertising revenues increases social welfare by 1.1%. This tax would also entail desirable distributional consequences—the tax would tend to reduce the profits of the owners of Facebook, while raising consumer surplus and revenue for the government. A tax on the number of users generating the same amount of revenue would slightly decrease social welfare, however, as it would incentivize the platform to more intensely monetize a smaller user base.
An interesting option we explored was data coalitions, which adhere to a data-as-labor approach that puts users in control over the monetization of their data. If collective bargaining by users leads to a recurring rebate, consumer welfare grows from both the rebate and by attracting a larger user-base (which would boost both platform value and network effects to other users). Meta’s profits would decrease, but the gain in welfare for users would more than offset this. If we plausibly assume that Meta shareholders are richer on average than Facebook users, this reform would have desirable distributional consequences as well.
Altogether, taxes, interoperability, and user rebates—which we group loosely as redistributive proposals—seem to augment social welfare better than more severe structural remedies. Of course, at this stage, we acknowledge that many of these possibilities are beset by practical limitations (e.g., how to organize a data union).
We also note that this version of the model is calibrated to Facebook and therefore reflects the platform’s idiosyncrasies. Although well-functioning platforms share general traits, such as positive economies of scale and scope, every platform exhibits a distribution of consumer demand that is nonetheless unique. Benzell and Collis approximated demand by conducting a massive “willingness to accept” experiment that measured demand for Facebook components—for example, certain friend groups— as well as the overall platform, by asking versions of “Would you give up Facebook for 1 month in exchange for $[X]?” Digital platforms and regulators could measure these essential parameters more precisely by running experiments directly on the platform. Given the unique demand heterogeneity characteristics of every digital platform, any model needs to be tuned to the idiosyncrasies of each platform—and each model’s developers should be transparent about that process. And if a new digital regulator were created in the future, the ability to compel platforms to perform these measurements and create models of participation based on them might be essential to the regulator’s success.
Our essays and the Benzell-Collis model will surely generate conversations as much about the model’s assumptions as its findings. The model does not attempt to account for negative externalities such as Internet addiction and subversion of democracy—noneconomic consequences that are typically omitted in the consumer welfare paradigm. There might be positive non-economic consequences from the spread of social media as well. All of these would have an impact on social welfare, but they are also hard to quantify.
Over time, we hope new tools will be devised to gauge more of Big Tech’s externalities so that they can be folded into a more comprehensive model of social welfare. In the interim, we hope that ours is a first step in prompting scholars, policymakers, and industry to publicize their own models. In doing so, the discourse around remedies will be driven toward precision—and broader social gains.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
