A wage tells us what one worker receives. The labor share asks what labor receives from the whole. The distinction changes the scale of the question without making the individual bargain disappear. A person may earn more and still belong to a group whose part of the income is shrinking; a country's average pay may rise while the worker in the middle gains little. The old question about the product and its division returns as a fraction: compensation paid to labor divided by the income out of which it is paid. What appears above the line depends on the bargains already examined. What appears below it depends on production, prices and the boundaries of the account. Before interpreting a movement in the fraction, we have to know which fraction moved.
The question at the head of this plate carries a historical clock. Its forty years belong to the debate about the decades after 1973, not to a claim that the same trend has continued unaltered up to the day of reading. In a 2015 analysis of the United States, Josh Bivens and Lawrence Mishel compared productivity and typical compensation over 1973–2014. Their account separated a median worker from the average and separated changes within labor income from changes in labor's share. Karabarbounis and Neiman's international study, published in 2014, examined a corporate labor-share decline in data spanning 1975–2012. The questions overlap, but the country coverage, periods and measures differ. Neither date should be allowed to roll forward silently.
Begin with an economy small enough to count. Suppose ten hours produce a thousand in value added, and compensation for those hours is six hundred. The labor share is sixty percent; average compensation is sixty an hour; value added per hour is a hundred. If value added rises to twelve hundred while compensation rises to six hundred and sixty, the share falls to fifty-five percent. Average pay has risen ten percent and productivity twenty percent. There is no contradiction. Had compensation risen to seven hundred and twenty, both would have grown twenty percent and the share would have remained sixty. This is an illustrative account with unchanged prices, hours and coverage, not a description of a historical year. It makes the relation visible before a theory is asked to explain it.
With consistent hours and output prices, average compensation divided by productivity is exactly the labor share. But take-home wages are not all compensation. Employer-paid benefits belong in a compensation account, even when they cannot be spent on this month's rent. Nor is sales revenue the same as value added: purchased intermediate inputs must be treated consistently. The income of a self-employed person may reward both labor and ownership, requiring an allocation rather than a label. A gross account also includes depreciation, while a net account deducts the capital used up. Elsby, Hobijn and Şahin showed why such choices materially affect a historical labor-share series. Care with the measurement does not dissolve the distributional question. It establishes what the evidence can answer.
The median introduces a second division, within the wage bill itself. Imagine three people working equal hours and each receiving forty. The total is a hundred and twenty, and the median is forty. Now pay them thirty, thirty and sixty. The total is still a hundred and twenty, but the median has fallen to thirty. Holding value added fixed, labor's aggregate share is unchanged. A widening gap between typical and average compensation can therefore worsen the position of the worker in the middle without moving income from labor to capital at all. High salaries are still labor compensation in this account. The distinction is uncomfortable only if the categories are expected to tell a complete story about social position. Their purpose here is to distinguish two movements that require different explanations.
Prices create another difference. Productivity describes real output per hour, while real compensation often describes what pay can purchase for the worker. The price index for output and the price index for consumption need not move together. If the goods workers buy become more expensive relative to the output being measured, purchasing-power compensation can lag productivity even when compensation valued in output prices keeps pace. Neither measure is fraudulent; they answer different questions. The Bureau of Labor Statistics made this distinction explicit in its 2017 discussion of the productivity-compensation gap. The historical decomposition by Bivens and Mishel likewise separated the price wedge from the labor-share wedge and inequality within compensation. A pair of lines cannot identify all three by itself.
Once the account is consistent, the explanation remains open. There is no general law of competition that makes labor's fraction constant. A familiar competitive model with a fixed-exponent Cobb–Douglas technology gives that result under its assumptions, but changing the technology or the relevant substitution possibilities can change it. Karabarbounis and Neiman examined a route through cheaper investment goods: if capital and labor are sufficiently substitutable, cheaper capital can induce a large enough shift to reduce labor's share. The condition matters. The claim is not that every cheaper machine reduces the fraction, or that a lower fraction necessarily means lower real wages. Their own analysis distinguishes capital-driven changes from changes in markups with different welfare implications.
The machine of the preceding plate supplies a related but distinct route. Moving tasks from workers to capital directly changes where labor is used. Lower costs can expand output and new labor-intensive tasks can emerge; their strength determines how far displacement is offset. Trade and offshoring can also change the activities performed within a measured economy and the alternatives available to employers and workers. Elsby and his coauthors treated exposure to imports as a potentially important part of the historical United States explanation, while being explicit about measurement and competing hypotheses. These mechanisms have to be tested against the relevant pattern. Naming globalization or technology does not show how much of a particular change each caused, much less establish that the worker whose pay lagged became less productive.
The fraction can change through the firms that grow as well as through what happens inside each firm. Suppose two firms have labor shares of seventy and forty percent and initially account for equal amounts of value added. The aggregate share is fifty-five percent. If the first firm's weight falls to a quarter and the second's rises to three-quarters, the aggregate becomes forty-seven and a half percent, though neither firm's own fraction changed. Autor and his coauthors developed and investigated this kind of reallocation in their 2020 study of superstar firms. Their model permits competitive factor markets alongside product-market markups. A falling aggregate labor share can therefore arise without assuming that each firm has newly cut its workers' share through monopsony.
This leaves room for bargaining power without turning the aggregate account into its unique measure. An employer's alternatives, a worker's ability to leave, collective organization and the terms on which technology is introduced can all affect a wage bargain. Yet the same aggregate fraction can be reached through different combinations of bargaining, production and composition. A labor-share decline does not, on its own, tell us the size of a markdown at a particular workplace. Nor does competitive hiring prove that the division must remain fixed: product-market markups and the distribution of activity can still matter. The earlier plates identify mechanisms through which a worker can receive less than another feasible arrangement would permit. This plate asks which of those mechanisms survive aggregation, and what additional evidence distinguishes them.
There is a practical consequence to keeping the distinctions. If the gap lies mainly between median and average pay, restoring labor's aggregate fraction would not by itself determine how the larger wage bill is distributed. If depreciation or self-employment treatment drives a measured change, a policy directed at an imagined transfer may miss the object. If low-share firms gain weight by producing more efficiently, preventing their growth raises a different question from preventing exclusion or suppressing workers' alternatives. An account of the division must ask both how much income is created and who receives it. Those inquiries can support one another, but neither can be substituted for the other merely because one produces a cleaner chart.
A larger product creates room for higher pay; it does not settle who receives the room.
The volume began with a bargain that neither party could strike from nowhere. It ends with an account that cannot be understood by pretending all bargains were identical. Skill, treatment, the floor, the employer's alternatives, the collective voice, the capacity to wait and the machine all affect what particular workers can receive. The labor share records one dimension of the result, while typical pay records another. Productivity enlarges the possibilities; it does not write every term of their division. To ask where the gain went is therefore to return to the wage with a larger ledger: to follow the income through firms and households, retain the distinctions the numbers require, and determine which arrangements would let more of the people who produce it participate in its increase.