A Benchmark Nobody Has Seen

Perfect competition requires many firms, none large enough to affect the price; an identical product; free entry and exit; and full information on all sides.

Nothing looks like this. The nearest approximations are commodity markets (wheat, copper, currency) and even those involve quality grades, transport costs and information asymmetries. Perfect competition is not a description of any market that has ever existed.

It is a benchmark, and that is a legitimate and useful thing to have. It tells you what price and quantity would be if nobody had market power: price equal to marginal cost, output at the level where the last unit is worth exactly what it costs to make. Every actual market is then measured as a departure from it, which is what competition authorities spend their time doing.

The departure that matters is not the profit. A monopolist charging above marginal cost transfers money from consumers to itself, and that transfer is a distributional question. Somebody has the money either way, and reasonable people disagree about who should.

The efficiency loss is elsewhere, and it is the trades that do not happen. At the monopoly price, there are buyers who value the good above what it costs to produce and who do not buy, because the price sits between the two. Nobody gets that value; it simply does not exist. That is deadweight loss, and measuring it requires knowing the demand curve, which is why estimates of monopoly's cost to an economy vary by an order of magnitude.

The practical tool for thinking about market power is concentration, usually measured with the Herfindahl-Hirschman index: add up the squares of each firm's market share. Squaring is the trick. It weights a dominant firm far more heavily than several small ones, so a market with one firm at 50% scores very differently from five firms at 20%, which a simple count cannot distinguish.

Competition authorities set thresholds on it for screening mergers. Its weakness is not arithmetic but definitional: the index requires you to say what the market is, and that decision usually determines the answer before any calculation happens. Is the relevant market "streaming video", "home entertainment", or "things to do on a Tuesday evening"? Every merger case is substantially an argument about that question.

Which raises a distinction that gets lost in public argument about big firms. A large market share is not by itself evidence of harm. A firm can be large because it is genuinely better, arrived first, or found something cheaper, and forcing it to be smaller would cost consumers.

What matters is how the position is maintained. Superior performance is one answer. Exclusive contracts that lock up distribution, predatory pricing to make entry unprofitable, self-preferencing on a platform that competitors must use, and acquiring companies before they become rivals are others. The distinction between competing hard and excluding rivals is where essentially all antitrust litigation lives, and it is genuinely difficult, because aggressive competition and exclusion look the same from outside.

William Baumol added a useful correction to the habit of counting firms. A market is contestable if entry and exit are genuinely free, and a monopolist in a contestable market cannot raise prices, doing so invites a competitor who undercuts, takes the profit and leaves before the incumbent can respond.

On this view the number of firms currently present tells you less than the barriers to a new one arriving. Sunk costs, network effects, regulatory licences, control of data or distribution are what matter. In practice perfectly costless exit is rare, which makes contestability a lens rather than a complete theory, and it is the right first question: not how many competitors are there, but what would happen if someone tried.

The case where all of this is hardest is the one currently occupying competition authorities everywhere.

A dominant platform acquires a startup with a few hundred thousand users and no revenue. Traditional merger review asks about the price effect in a defined market, and the answer is zero, because the target sells nothing.

The concern is counterfactual: that this firm might have become the only credible challenger, and that the acquisition forecloses a future that will now not be observed. Facebook's purchases of Instagram and WhatsApp are the cases everyone argues about, and the argument cannot be settled by evidence, because the world in which they stayed independent does not exist to be measured.

That is a genuine epistemic problem rather than a failure of nerve by regulators. Several jurisdictions have responded by lowering thresholds specifically for acquisitions by dominant platforms, accepting that some harmless deals will be blocked, because the alternative is that harmful ones are invisible until it is far too late.