Antitrust: The Test Was Built for Prices That Never Rose
Competition law developed around a specific and testable idea of harm: a dominant firm raises prices above what competition would allow, and customers pay more than they should.
That test has real virtues. It is measurable, it is objective, and it worked for a century on railways, oil, steel, and telephony.
It fits the largest firms of the current era poorly, and the reason is simple. Many of them charge consumers nothing. Search, maps, social networks, and much else are free at the point of use. On the classic test, there is no harm to find — prices have not risen, and in several cases the services have improved.
Yet the power is real, and it shows up somewhere the old tests do not look. That mismatch, more than any failure of will, is why cases against large platforms run for a decade and frequently end inconclusively.
Where the power actually sits
If not in consumer prices, then where? Several places, each harder to measure than a price increase.
In the terms offered to suppliers. A business that must reach customers through a platform has limited ability to negotiate the commission, the ranking rules, or the data it does or does not receive. This is the aggregation position: suppliers compete against each other for access to demand the platform controls.
In self-preferencing. A firm that both runs a marketplace and sells within it has an obvious incentive to favour its own products in its own ranking. Whether this is harmful competition or ordinary vertical integration is genuinely contested, and the answer probably depends on how essential the marketplace is.
In acquisitions of potential competitors. Buying a small company that might one day have become a rival is difficult to challenge under a framework that asks about current market share. The harm is to a future that did not happen, which is close to unprovable.
In what does not get built. If founders and investors avoid categories adjacent to a dominant platform, the cost is products that never existed. Real, and impossible to enter into evidence.
In the diffuse price consumers do pay. Attention, data, and the terms of a relationship. The difficulty is that translating these into the language of consumer harm has proven hard, and attempts have been contested.
Two positions, both with something to them
Public debate tends to collapse into caricature here, so it is worth stating both cases in their reasonable forms.
The case for restraint. Some of what makes these products good is inseparable from what makes them hard to compete with. Integration produces things that work well together. Scale funds services that are free. Intervening clumsily can produce a worse product without producing more competition — and regulators trying to engineer market structure have a mixed historical record. Dominance in technology has also proven less permanent than it looked, with several apparently unassailable positions eroded by platform shifts rather than by enforcement.
The case for action. Waiting for a platform shift is not a policy. Some positions have persisted through several shifts. The absence of consumer price increases does not mean the absence of harm — it means the framework measures the wrong thing. And the acquisition of nascent competitors specifically forecloses the mechanism the restraint argument relies on.
Both are serious. The honest position is that this is a genuinely hard problem where the analytical tools are mismatched to the situation, not a case where one side is obviously right and obstructed by bad faith.
The remedies, and what is known about them
Four broad approaches, with quite different track records.
Structural separation — breaking a firm apart. Slow, blunt, and durable when it works, because it removes the incentive rather than policing it. Historically used sparingly, and difficult where the value comes from integration rather than from separable business lines.
Conduct rules — prohibiting specific behaviours like self-preferencing or bundling. Faster to implement and requires continuous enforcement, since a firm with strong incentives will find approaches the rule did not anticipate. Prone to compliance theatre, which is Goodhart's Law in legal form.
Interoperability mandates — requiring platforms to work with competitors. Attacks the network effect directly, which is where the durable power actually lives, and is technically difficult and easy to comply with badly.
Data portability — letting users leave with what they brought. Attacks switching costs, is conceptually clean, and has so far produced less movement than hoped, because portability without an equally useful destination changes little.
The pattern across attempts is that remedies aimed at the symptom — a specific behaviour — get worked around, while remedies aimed at the structural source — switching costs, network effects, control of the interface — are harder to design and more likely to matter. That is not a solved problem in any jurisdiction, and anyone claiming a clear answer is ahead of the evidence.
