A price is written on the shelf; the cost of making one more unit is not. The observer sees twelve dollars and has yet to learn whether the additional unit costs eleven dollars, eight, or three. The same posted figure can accompany a thin margin or a large one. Even a set of accounts, with sales on one side and expenses on the other, does not immediately supply the missing number. Those accounts record the cost of running the business over a period. The question of power concerns the cost that changes when the business supplies another unit. Measurement begins by refusing to confuse the two.
The Lerner index names the proportion of the price that lies above marginal cost. If the price is twelve and marginal cost is eight, the index is four divided by twelve: one-third. A price-to-cost ratio answers a different question and gives twelve divided by eight, or one and a half. Calling this a fifty-percent markup over cost is consistent with calling the Lerner share one-third of price. The relationship is exact: the Lerner index is one minus the reciprocal of the price-to-cost ratio. Much apparent disagreement can be manufactured simply by changing denominators while retaining the same word.
The gap still does not measure the profit left after the enterprise has paid for its existence. Suppose the business sells one hundred units at twelve, incurs eight of variable cost for each, and has four hundred in fixed economic costs. Its revenue is twelve hundred and its total cost is twelve hundred. Economic profit is zero despite the price-to-marginal-cost ratio of one and a half. A lower marginal cost might reflect a costly invention or an efficient plant; the gap may help repay the resources committed to it. That does not make every gap harmless. It makes the distinction between the price of another unit and the financing of the whole activity indispensable.
Concentration supplies a map of shares, not the missing cost. Four sellers with forty, thirty, twenty and ten percent of a defined market have a Herfindahl index of three thousand when percentage shares are squared and summed. Four equal sellers have twenty-five hundred. The first allocation places more sales in larger hands, but the index does not say how close the products are as substitutes, how capacity constrains the firms, or what a potential entrant could supply. Change the boundary of the market and the shares change. Concentration can be evidence relevant to a theory of power; it cannot replace that theory or identify marginal cost from a price.
Demand offers one route to the cost that the accounts conceal. A single-product firm choosing an interior profit maximum equates marginal revenue to marginal cost. If the elasticity of its demand is known, that condition connects the observed price to the missing marginal cost: the Lerner share equals the inverse of the absolute demand elasticity. But the elasticity must describe demand for this firm’s product under the relevant competitive response. And the pricing equation changes when a firm internalizes effects on its other products, faces a binding capacity limit, or pursues an objective the simple profit condition omits. The inference is obtained by putting a model beside the price. The price does not carry the model inside it.
The production approach follows a different margin: the firm’s purchase of a variable input. Under cost minimization, an input that can adjust freely and is bought at a price the firm takes as given has a marginal contribution tied to its expenditure. The price-to-marginal-cost ratio then equals that input’s physical-output elasticity divided by its expenditure share in revenue. An elasticity of 0.6 and a share of 0.4 imply a ratio of 1.5. This route does not require a specified demand curve or a particular form of output-market competition. It does require the production elasticity and an input choice that obeys the condition from which the ratio is derived.
Those elasticities are difficult precisely where price-setting matters. Increasing an input may raise physical output while the extra sales require a lower selling price. Revenue therefore rises less rapidly than quantity. In the simple model just used, a physical-output elasticity of 0.6 and a Lerner share of one-third imply a revenue elasticity of 0.4. Substituting that revenue elasticity into the ratio gives one, erasing the actual markup of 1.5. Bond and his coauthors made this identification problem explicit in their 2020 working paper. The algebra does not say that the firm has no power. It says that a numerator contaminated by the price response is answering a different question.
The resulting empirical debate concerns consequential estimates, not merely the names of fractions. In their published 2020 study, De Loecker, Eeckhout and Unger estimated a rise in the sales-weighted average price-to-marginal-cost ratio in their U.S. publicly traded-firm sample from 1.21 in 1980 to 1.61 in 2016. The distribution mattered: their weighted median did not rise with the average, and higher markups at the upper end were joined by a reallocation of sales toward high-markup firms. These are dated, model-based findings for a defined sample. They do not say that each firm raised its margin by the same amount, nor do they turn a historical sample of listed firms into an unqualified description of every business.
The price is observed. The marginal cost must be learned.
The accounts help only after their categories have been interpreted. A cost recorded as selling or administrative expenditure may contain activities essential to producing or distributing the product; an apparent variable expense may adjust only slowly. A change in outsourcing can move spending between categories without an equivalent change in power. Conversely, improved production can lower marginal cost faster than price, increasing a markup while making the product cheaper. A convincing measurement therefore asks whether the result survives plausible ways of treating inputs, fixed costs, output quality and the units in which production is observed. Agreement between several measures is more informative when their assumptions fail in different ways.
A measured markup becomes most useful when it is connected back to the mechanism that could sustain it. Evidence about substitution can test the demand constraint; evidence about technology can test the cost estimate; entry and conduct can help explain persistence. The object is neither to dismiss a difficult estimate nor to allow its numerical precision to do the work of an explanation. The original question asks how cost can be learned from price. It is learned through additional evidence and a stated model, with the uncertainty carried through to the conclusion. The next plate asks what the law should do with power once its existence and operation have been established. Measurement gives that inquiry an object; it does not decide the remedy.