On September 2, 2025, in the United States, in the federal court in the District of Columbia, Judge Amit P. Mehta announced remedies following a high-profile ruling in which he found that Google had unlawfully maintained its dominance in the search market. But instead of "breaking up" the company, as some members of the public wanted, the court opted for more precise remedies. It ordered Google, among other things, to remove exclusivity clauses from agreements with device and browser manufacturers, to partially open up interfaces (APIs) and protocols that had previously been closed to competitors, and to ensure the transparency of its ranking algorithms. In other words, instead of splitting the company into parts, the court ordered it to loosen its grip – to allow others to enter the market, but not to take away its technological achievements. It was a decision that was more surgical than symbolic. Why? Because in the meantime, a new competitor had emerged: generative AI, which began to take over the functions of classic search engines. "The ruling concerns a world that no longer exists," economists commented.

A few months earlier, Judge Leonie Brinkema in a Virginia federal court ruled that Google had monopolized the display advertising market, and the European Commission imposed a €2.95 billion fine for similar abuses. Except that at the same time, the adtech industry was changing under the influence of new formats, AI, and distributed auction systems. By the time officials had settled the dispute, the reference point had shifted again.

This is the essence of the problem with antitrust law: the law lags behind. Regulators, regardless of their intentions, are chasing a market that changes before a ruling is made. And it's not just about Google. It's a universal mechanism of the digital age—the market invents itself faster than bureaucracy can understand it.

To understand why this is a problem at all, we need to start from the beginning. In any antitrust case, the first question is: "What market are we analyzing?" Because only by knowing the boundaries of the market can we assess whether a company has actual market power or is simply innovative.

Defining the relevant market is like drawing up a map before a battle. Too narrow, and every big player becomes a monopolist. Too broad, and no monopoly ever exists. That is why economists and lawyers have developed tools to bring precision to this area.

One of them is the hypothetical monopolist test (HMT) – a simple idea: imagine that one company controls the entire market for a given product. Could it raise the price by 5-10% and not lose customers to substitutes? If so, the market is defined correctly (narrow and actually monopolized). If not, it should be broadened. This method is known from US court rulings (H&R Block, Staples, Cardinal Health) and from the practice of the European Commission.

The problem is that today's digital markets do not operate like boxes with a single product. They are ecosystems – search engines powered by advertising, stores with payment systems, messengers connecting users and advertisers. Here, the price is sometimes zero, and instead of one market side, there are at least two. It's like measuring temperature with a thermometer designed for another planet.

Under such conditions, the SSNIP (Small but Significant Non-transitory Increase in Price) test only works conditionally. That is why researchers – including Filistrucchi, Geradin, Van Damme, and Affeldt – have proposed applying it to the total price of the platform, i.e., the sum of prices on both sides of the market, i.e., users and advertisers. If a platform can raise the total price without upsetting the balance between the parties, there is a risk of dominance. In practice, this means that not only is the amount paid by the advertiser analyzed, but also how a change in this price affects the number of users on the other side.

Using platforms such as Google, Facebook, TikTok, Amazon, and Uber as examples—i.e., those that connect two groups of users (e.g., customers and advertisers, passengers and drivers, buyers and sellers)—economists have tried to understand how competition really works in digital ecosystems.

Argentesi and Filistrucchi, and later Song, showed that this can be measured empirically: it is possible to "recover" from market data the so-called elasticity of demand on both sides of the platform—that is, to check how strongly users and advertisers (or drivers, sellers) react to changes in price, quality, or terms of service.

If raising rates for advertisers limits the number of ads, which in turn reduces the number of free features for users, it is possible to calculate how strong this link is. In this sense, economists "recover" elasticities—they recreate them from real market behavior, rather than from declarations or hypothetical models.

Added to this is the phenomenon of multi-homing: users and contractors use multiple platforms simultaneously—just as a passenger has both Uber and Bolt, and an advertiser promotes themselves on Google and Facebook.

If we assume that each user is “stuck” to one platform, it would be easy to mistakenly conclude that someone has a monopoly. The models developed by Argentes, Filistrucchi, and Song help to avoid this by showing that in the world of platforms, dominance is a temporary rather than a structural condition — and users move faster than regulations can react.

In the case of "zero-margin" services—such as Google search, YouTube, or social media—the classic SSNIP test ("could a hypothetical monopolist raise the price by 5-10% without losing customers?") simply does not work. There is no price to raise.

That is why economists have created variants of it – SSNIC and SSNDQ.

The first (Small but Significant Non-transitory Increase in Cost) checks whether users would remain on the platform if their non-monetary costs increased – e.g., if they had to watch more ads, spend more time, give up more data about themselves, or put up with more disruptions in using the service.

The second, SSNDQ (Small but Significant Non-transitory Decrease in Quality), examines the same thing from the quality side: whether users would stay if the quality declined – e.g., search results were less accurate, videos loaded more slowly, and maps provided poorer directions.

If users remain despite a deterioration in quality or an increase in non-monetary costs, it means that the platform has real market power – that even a reduction in comfort does not prompt them to leave.

In other words, the "price" in such markets does not disappear, but changes form – we pay with time, privacy, or quality of experience.

This methodology is no longer purely academic. The European Commission's 2023 Market Definition Notice recognized the SSNIC and SSNDQ tests as fully-fledged analytical tools in the digital economy. Thanks to them, it is now possible to demonstrate a monopoly even where the service is formally "free" and competition is for the user's attention, not their wallet.

On the other hand, although the SSNIC and SSNDQ tests have already become part of practice, they have an inevitable element of arbitrariness.

The classic SSNIP assumes that we examine the market's reaction to a 5-10% price increase – but this threshold is purely a convention adopted by officials and economists decades ago. In the digital economy, where prices change dynamically and value for the user has many dimensions, such an experiment becomes more of an illustration than proof.

The same applies to the SSNDQ test. What exactly does "a slight but significant decline in quality" mean?

Are search results that are 5% worse still the same service? Are more ads on the screen a reduction in quality or just a change in the business model?

Regulators are trying to translate these concepts into numbers: they examine system response time, search accuracy, number of ads, privacy level, and data portability. But these are only approximations — proxy indicators that are supposed to describe the user experience. Market information is scattered, incomplete, and subjective, so any attempt to encapsulate it in a single formula carries the risk of oversimplifying reality.

From this perspective, SSNIP, SSNDQ, and SSNIC tests are more indicative tools than hard evidence. They help regulators understand how the market reacts "here and now," but they do not describe what is most important — the discovery process. Because innovation does not happen in 5% increments. For some users, a change of a millisecond in page loading time may be insignificant, for others – the beginning of migration to a competitor.

Digital markets are not equations with a single variable; they are living organisms in which "quality" is often a subjective feeling, not an indicator.

This is best seen in the ride-hailing market. In the US and Europe, disputes over the position of Uber or Lyft have become a testing ground for economists and regulators. It was necessary to determine: are Uber and taxis the same market?

On the surface, yes: both products offer transportation from point A to point B. But economic research has shown otherwise. Analyses of demand in American cities have shown that when Uber raises its prices (e.g., during surge pricing), only about a quarter of passengers opt out of the service or return to taxis. Cross-elasticity is about 0.26, which means that when Uber prices increase by 10%, demand for taxis increases by 2.6%. The result: taxis are only a partial substitute. The market is related, but not identical.

This shows two things. First, that market boundaries are not intuitive – they cannot be drawn on a napkin. Second, that precise measurement protects against regulatory errors in the current legal framework. If Uber and taxis were considered identical services, it could be argued that neither company has dominance. On the other hand, narrowing the market down to transportation apps alone would make Uber a monopolist. The truth lies somewhere in between – and the market is changing faster than transportation laws.

It is worth adding that the same study shows that the benefits of ride-hailing services to consumers are enormous – estimated at billions of dollars a year. This means that users value time, reliability, and ease of use more than they lose on prices. As in any system of creative destruction, innovation breaks old models but creates more value for the user.

Competition policy is necessary—but like any medicine, it can be harmful if the dose is too high or administered too late.

The history of the GDPR (General Data Protection Regulation) clearly shows how a regulation intended to protect users can inadvertently strengthen the position of those it was meant to restrict. The cost of complying with the new requirements was so high that only giants — Google, Meta, and Amazon — could afford entire departments of lawyers and engineers dedicated to regulatory compliance. For small businesses, what was an "investment in compliance" for the big players became an insurmountable barrier. The result? The biggest players survived and strengthened their position, while thousands of smaller companies disappeared from the advertising market. The GDPR created what economists call a regulatory moat—a barrier not to abuse, but to competition.

A similar paradox can be seen today with the Digital Markets Act (DMA). It is not the act itself that is the problem, but its implementation — where good intentions meet bureaucratic practice and resource asymmetry. In 2025, the European Commission fined Apple (€500 million) for restricting developers from directing users outside the App Store (anti-steering), and Meta (€200 million) for its "pay-or-consent" model, which did not offer a real choice between privacy and payment. These decisions were intended to show that Brussels can strike even the biggest players.

But the reality turned out to be less black and white. In response to the DMA, Apple introduced a "Core Technology Fee" of €0.50 for each app installation after one million downloads. For companies such as Proton and RevenueCat, it is this fee, rather than the "anti-steering" ban itself, that has become a new barrier: the market has formally opened up, but its cost has increased. A startup that wants to enter the game now has to pay for every step—for integration, auditing, reporting, and new interfaces. For large companies, this is a budget item. For small ones, it often means the end of the project.

This creates another regulatory moat, this time built not of paragraphs, but of compliance bills. The regulation that was supposed to open the door is turning into a complex system of border crossings. Each of them theoretically keeps order, but in practice slows down traffic. And although the DMA's intention was to break down barriers, it is increasingly difficult to say who is really building bridges today and who is just building new walls.

It is also worth remembering that the regulator is not a neutral entity. It is an institution with people who respond to stimuli in the same way as entrepreneurs. A commissioner or judge has their own goals—publicity, career, reputation as a consumer advocate. The more high-profile the case, the greater the publicity. But this is where the risk lies: when institutional incentives encourage excessive rather than necessary action.

Institutional economics reminds us that every bureaucracy is subject to the laws of supply and demand – it grows with the scope of its own powers. The more complex the regulatory system, the greater the demand for regulators, consultants, auditors, and law firms. This does not necessarily stem from ill will; it is a self-perpetuating mechanism. Regulation becomes a way of life rather than a tool for solving problems.

In practice, this leads to a paradox: the apparatus designed to defend competition may inadvertently stifle it by creating cognitive and procedural barriers that hinder innovation. Because innovators don't have legal departments – they have ideas, teams, and limited time. When the system requires filings, audits, and annual reports, creative destruction becomes trapped in a bureaucratic maze.

The point is not to reject regulation, but to design it the way the market designs products: lightly, iteratively, with the user in mind. The law should keep pace not through the speed of decisions, but through an understanding of the process it is trying to regulate. Market order does not arise from a plan, but from the interaction of individuals. The role of the regulator is therefore not to control the market, but to remove obstacles from its course. The regulator cannot replace the market in creative destruction. Its task should be to allow it to flow unhindered.

Scientific publications:

1. Filistrucchi, L., Geradin, D., Van Damme, E., Affeldt, P. (2014).

Market Definition in Two-Sided Markets: Theory and Practice.

Journal of Competition Law & Economics, Vol. 10, No. 2, pp. 293–339.

https://doi.org/10.1093/joclec/nhu007

2. Argentesi, E., & Filistrucchi, L. (2007).

Estimating Market Power in a Two-Sided Market: The Case of Newspapers.

Journal of Applied Econometrics, Vol. 22, No. 7, pp. 1247–1266.

https://doi.org/10.1002/jae.994

3. Song, Y. (2022).

Empirical Analysis of Multi-Homing in Platform Markets.

SSRN Electronic Journal, Working Paper No. 4219823.

https://ssrn.com/abstract=4219823

4. Nakamura, Y., & Ida, T. (2020).

SSNIC and SSNDQ Tests for Zero-Price Digital Services: Measuring Market Power Without Prices.

Discussion Paper Series on Industrial Organization, Kyoto University.

5. Brynjolfsson, E., & Collis, A. (2019).

GDP-B: Accounting for the Value of New and Free Goods in the Digital Economy.

NBER Working Paper No. 25695, National Bureau of Economic Research.

https://www.nber.org/papers/w25695

6. Evans, D. S., & Schmalensee, R. (2016).

Matchmakers: The New Economics of Multisided Platforms.

Harvard Business Review Press.

7. Posner, R. A. (2001).

Antitrust Law (2nd ed.).

University of Chicago Press.

8. OECD (2022).

Defining Markets in the Digital Age.

OECD Competition Committee Background Note.

https://www.oecd.org/daf/competition/defining-markets-digital-age.htm

Institutional reports and court rulings

1. European Commission (2023).

Communication from the Commission — Market Definition Notice.

Official Journal of the European Union C 106/1 (10 March 2023).

https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52023XC0310(03)

2. European Commission (2025).

Case AT.40670 – Google AdTech.

Decision of 5 September 2025 – Fine of €2.95 billion.

(Press release and summary pending publication in O.J.)

3. European Commission (2025).

Digital Markets Act (DMA) Enforcement Actions — Apple and Meta.

April 2025 — Fines of €500 million (Apple, anti-steering) and €200 million (Meta, pay-or-consent).

https://ec.europa.eu/commission/presscorner/detail/en/ip_25_1783

4. United States v. Google LLC, Civil Action No. 1:20-cv-03010 (D.D.C.).

Opinion of Judge Amit P. Mehta (2 September 2025).

United States District Court for the District of Columbia.

5. United States v. Google LLC (Ad Tech Case), Civil Action No. 1:23-cv-00108 (E.D. Va.).

Opinion of Judge Leonie M. Brinkema (17 April 2025).

United States District Court for the Eastern District of Virginia.

6. European Commission (2018–2022).

Google Search (Shopping), Google Android, and Google AdSense Cases.

Consolidated fines totaling ~€8 billion for abuse of dominance.

https://competition-policy.ec.europa.eu

7. European Commission (2016).

General Data Protection Regulation (GDPR) – Regulation (EU) 2016/679.

Official Journal L 119 (4 May 2016).

8. OECD (2024).

Competition Enforcement in Digital Markets: Lessons from Big Tech Cases.

OECD Policy Paper No. 45, Paris.

9. U.S. Department of Justice (2023).

Antitrust Division Annual Report.

Section on Technology and Platform Enforcement.

10. European Commission (2024–2025).w

Digital Markets Act – List of Gatekeepers and Enforcement Timeline.

Official Journal L 265/1 (6 September 2023); implementation reports through 2025.

https://digital-markets-act.ec.europa.eu