For twenty years, one statistic has fueled debates on contemporary capitalism: the share of national income going to workers has fallen. This statistic, documented on nearly every continent since the early 1980s, became the empirical foundation for a generation of analyses, policy proposals, and social movements. It justified calls to tax capital more heavily, to establish a universal basic income, to reduce working hours. It nourished the theses of economists as different as Thomas Piketty and left-wing critics of the global market.

An analysis published in June 2026 by Bruegel and based on OECD data complicates this narrative. Not to say that the decline did not occur—it did—but to show that the usual measurement instruments have inflated it, sometimes massively. In the United States, approximately one-third of the measured decline would be a statistical artifact. In most advanced economies, if one replaces gross GDP with net income after deducting capital depreciation, the trend weakens very strongly, or even disappears entirely.

The Essentials

  • Labor’s share of gross GDP has declined in most advanced economies since the 1980s, but the magnitude of this decline depends heavily on the numerator and denominator chosen to calculate it.
  • From the perspective of net income after deducting capital depreciation, the measured decline almost disappears in several major countries (Bruegel / OECD, June 2026).
  • In the United States, approximately one-third of the measured decline would be attributable to accounting conventions regarding self-employed workers’ income, not an actual wage compression.
  • The increase in capital depreciation—linked to the acceleration of obsolescence of digital equipment—explains a large fraction of the apparent decline, without real wages falling proportionally.
  • The diagnosis remains real but narrower than claimed: income inequalities exist, but their causes and appropriate remedies deserve to be reconsidered.

We must begin by understanding how this labor share is measured, because the problem lies there, in the construction of the indicator itself.

Labor’s Share Is Measured Poorly, and That Is Where It All Begins

Labor’s share of GDP is in principle simple: take the wage bill, divide it by national income, and you obtain the fraction going to workers rather than capital holders. Since the 1970s-1980s, this fraction has been declining in most advanced economies. The OECD has documented it. The IMF has documented it. Dozens of academic economists have documented it.

The problem is that this calculation rests on two conventions that have evolved in the same direction unfavorable to the measurement of labor, without the underlying economic phenomenon necessarily having changed.

The first convention concerns the denominator: gross GDP. When we say “labor’s share of GDP,” this GDP includes capital depreciation—that is, the value of equipment consumed during the year, machines that age, software that becomes obsolete. This depreciation is income for no one: neither for shareholders nor for employees. It is wealth destroyed in the production process. Yet since the 1980s, the digital economy has accelerated the rate of equipment obsolescence. A server depreciates much faster than a steel assembly line. Capital wears out faster, so depreciation rises as a proportion of GDP, so the net share available to be divided between labor and capital decreases—even if nothing has changed in the power balance between employees and employers.

If we recalculate labor’s share not in gross GDP but in net national income (GDP minus depreciation), the picture changes. In several European countries, the observed decline weakens considerably. In some cases, it nearly disappears entirely.

One-Third of the American Decline Would Be an Accounting Artifact

The second convention poses an even more specific problem, and it affects the United States particularly acutely. It concerns the treatment of self-employed workers’ income.

A self-employed worker—a freelance physician, an individual entrepreneur, a craftsperson—receives income that mingles both the compensation for their work and the return on their own capital. They are simultaneously an employee and an investor in themselves. In national accounts, this mixed income must be divided between “labor share” and “capital share,” but this division is a convention, not a direct measurement. National accountants make an assumption about what the self-employed person would earn if they were a salaried employee, and attribute the remainder to capital.

Now since the 1970s, the share of self-employed workers in the American workforce has evolved, and the attribution conventions have ended up allocating an increasing fraction of their incomes to capital rather than labor. According to Bruegel’s analysis, this accounting convention explains approximately one-third of the measured decline in labor’s share in the United States. It is not that workers have lost ground to shareholders of large corporations. It is that the way of classifying self-employed workers’ incomes changed the statistic without necessarily reflecting a change in the actual distribution of value.

This is not a minor marginal correction. One-third is the order of magnitude that shifts us from a diagnosis of “systemic retreat of labor against capital” to a diagnosis of “moderate distortion partly linked to labor market structures.”

What the Real Decline Actually Says About Inequality

We must be precise about what this reassessment implies and does not imply. It does not say that inequality has decreased. It does not say that workers are living better. It says that the macroeconomic statistic of labor’s share is too crude an instrument to capture the transformations that really matter.

For the real phenomenon of the past four decades is not the decline of labor’s share in aggregate. It is the concentration of gains within the labor share itself. Wages at the top of the distribution—executives, technology engineers, business lawyers—have progressed very strongly. Wages at the bottom and middle of the distribution have stagnated in real terms in many countries. When you add all this together to calculate a “labor share,” the two effects partially cancel each other out, and you obtain a statistic that says neither the stagnation at the bottom nor the progression at the top.

Bruegel’s analysis points to exactly this problem: labor’s share is a measure of distribution between factors of production (labor versus capital), but it is blind to distribution within the labor factor itself. And it is precisely this second dimension that has moved the most. Economists working on income inequality—using tax data rather than national accounts, as the team around Piketty and Saez has done—document this phenomenon far more precisely. It is not that labor has lost against capital in general. It is that certain workers have captured a disproportionate fraction of the gains, while others stagnated.

This reversal of perspective is important: income inequalities are real, but they play out largely within the labor market, not principally between labor and capital in the macroeconomic sense of the term.

Public Policies Built on a Misdiagnosis

This is where the issue becomes political. If the decline in labor’s share has been overestimated and misattributed, the responses built on this diagnosis merit reassessment.

Let us take three concrete examples. Taxing capital has been justified, notably by Piketty in Capital in the Twenty-First Century, by the finding that the return on capital structurally exceeded economic growth, condemning societies to increasing concentration. This thesis remains defensible on its own grounds, but if a fraction of the decline in labor’s share was due to accelerated depreciation of digital capital—and not to increased appropriation of surplus by shareholders—then taxing capital would not solve the identified problem: accelerated depreciation would continue to erode the available net share regardless of the tax rate.

The basic universal income has often been presented as a response to the “expropriation” of workers by capital, notably in automation scenarios. If productivity gains go massively to capital holders, redistribution must occur through a universal transfer. But if labor’s share has not declined as strongly as measured, and if the real problem is wage dispersion within the labor market itself, then universal income addresses a poorly identified symptom. Professional training policies, strengthening wage bargaining power, or investment in sectors with strong median wage growth would seem better calibrated.

The four-day week, a third example, has sometimes relied on the argument that productivity was increasing while wages stagnated—in other words, that workers were producing more but not capturing the gains. While this argument remains partially valid, OECD data suggest a more nuanced picture: labor productivity has indeed progressed in most advanced economies, but real wages have also progressed in many countries, albeit unevenly according to skill levels. The finding of a generalized “great wage stagnation” holds up less well under examination than the finding of stagnation concentrated on certain segments of the labor market.

What Production Data Still Signal

It would be excessive to conclude that all is well and that income distribution poses no problem. The OECD data mentioned in Bruegel’s analysis indicate that from a production perspective—that is, by maintaining gross GDP in the denominator and correcting only accounting conventions on self-employed workers—the decline in labor’s share remains statistically significant, even if more moderate than announced.

This residual signal deserves to be taken seriously. It points to real structural transformations: the rise of capital-intensive enterprises in technological sectors, the concentration of markets in a limited number of industries (digital, pharma, finance), and the capacity of certain firms to generate durable rents thanks to network effects or patents. These phenomena do indeed generate increased capture of gains by shareholders of a few large companies, independent of statistical effects. Jean Tirole, in his work on market design, has shown how sectoral concentration could produce exactly this type of asymmetric distribution, visible not in macroeconomic aggregates but in individual company data.

The right question is therefore not “is labor losing against capital?” but “in which sectors, for which companies, and for which workers?” This is a question of entirely different precision, calling for different measurement tools and more targeted public policies. On this terrain, tax data and company microdata are far more informative than labor’s share in GDP.

The distinction between gross and net measurement, between aggregate and distribution, between statistical signal and underlying economic phenomenon, should henceforth be integrated into any serious discussion of income distribution. This is not a quarrel among statisticians. It is the condition for economic diagnoses to truly serve as a basis for useful policy decisions—and for remedies to be equal to real ills, not measured ills.

Sources

  1. Bruegel — Barata da Rocha, Darvas, Greppi Maturana, Lappe, “The Decline of the Labour Share of Income” (June 16, 2026): https://www.bruegel.org/blog-post/decline-labour-share-income
  2. OECD — data on labor share and productivity, National Accounts Database
  3. IMF — “Understanding the Downward Trend in Labor Income Shares”, World Economic Outlook, April 2017
  4. Thomas Piketty, Capital in the Twenty-First Century, Harvard University Press, 2014
  5. Jean Tirole, Economics for the Common Good, Princeton University Press, 2017