For decades, the constancy of the labor share stood as one of Kaldor's stylized facts—a bedrock assumption underlying much of modern growth theory. The Cobb-Douglas production function, with its convenient unitary elasticity of substitution, encoded this stability into countless macroeconomic models. Yet the empirical record since the 1980s has fundamentally undermined this foundation.

Across advanced economies, the share of national income accruing to labor has declined by roughly four to six percentage points, with corresponding increases in capital's share. This is not a minor statistical curiosity. It reshapes our understanding of distributional dynamics, alters the transmission channels of monetary and fiscal policy, and challenges the microfoundations we impose on aggregate models.

The implications extend well beyond distributional concerns. If factor shares are variable, then the elasticity of substitution between capital and labor cannot be unity, which cascades through every prediction our growth models make about long-run dynamics, investment behavior, and the response to technological shocks. For policymakers, this raises uncomfortable questions about whether standard prescriptions—derived from models assuming stable shares—remain appropriate for economies where the fundamental structure of factor payments is evolving in ways we are only beginning to understand.

Before we can theorize about why the labor share is falling, we must confront a more prosaic difficulty: measuring it accurately is genuinely hard. National accounts do not cleanly bifurcate income into labor and capital. A substantial share of income flows to ambiguous categories—proprietor's income, mixed income from self-employment, and imputed rents on owner-occupied housing—each of which contains elements of both factor returns.

Consider proprietor's income. A self-employed physician earns a return on human capital, on physical capital invested in the practice, and on entrepreneurial risk-bearing. How the statistician allocates this mixed income between labor and capital shares can swing the aggregate measurement by two or more percentage points. Different allocation conventions across countries and time periods introduce comparability problems that plague international studies.

Housing presents an even thornier case. Imputed rents on owner-occupied housing are conventionally classified as capital income, but they represent consumption flows to households. Recent work by Gutiérrez and Piton, and by Rognlie, demonstrates that once housing capital is separated from productive capital, the decline in the non-housing labor share becomes considerably smaller—perhaps a third of the headline figure in some economies.

Intangible capital compounds the measurement problem. Software, patents, brand equity, and organizational capital are increasingly important, yet their treatment in national accounts remains uneven. Reclassifying research and development as investment, as the BEA did in 2013, mechanically shifted measured factor shares without any underlying economic change.

None of this negates the phenomenon. The decline is robust across most measurement conventions. But it does suggest that our theoretical explanations must engage with heterogeneity across sectors and asset types rather than treating the aggregate as a monolithic object requiring a single unified narrative.

Takeaway

Aggregate statistics are theoretical constructs, not raw observations. The choices embedded in measurement often predetermine which explanations appear plausible.

The literature has generated at least four serious contenders for explaining the labor share decline, each with distinct theoretical foundations and policy implications. Automation and capital-biased technical change, most rigorously modeled by Karabarbounis and Neiman, attributes the decline to falling relative prices of investment goods. In their framework, cheaper capital induces substitution away from labor when the elasticity of substitution exceeds unity.

Globalization provides a second explanation. Offshoring and the integration of low-wage labor pools into global production networks weakens the bargaining position of workers in advanced economies. Elsby, Hobijn, and Şahin find that industries with greater import exposure experienced larger labor share declines, suggesting a real trade channel operating alongside technological forces.

The third and increasingly influential explanation centers on rising market concentration and the emergence of superstar firms. Autor and colleagues document that within industries, the largest firms have both lower labor shares and growing market share. As economic activity reallocates toward these firms, the aggregate labor share falls even without changes at the firm level. This mechanism connects the labor share decline to broader concerns about market power and productivity dispersion.

Rents and markups constitute a related but distinct channel. If firms increasingly earn economic profits above competitive returns, then what national accounts label as capital income partially reflects pure rents rather than payments to productive capital. De Loecker and Eeckhout's estimates of rising markups suggest this component may be substantial, with implications for how we interpret the capital-labor split entirely.

These explanations are not mutually exclusive, and empirical work suggests each contributes something. The relative weights matter enormously for policy: automation implies redistributive responses, globalization suggests trade adjustment, and concentration points toward competition policy.

Takeaway

When multiple mechanisms plausibly explain the same phenomenon, the question is not which is correct but how they interact—and which policy levers each mechanism responds to.

The theoretical implications of variable labor shares are profound. Cobb-Douglas production functions, with their unitary elasticity of substitution, generate constant factor shares by construction. If shares are variable, we require CES production functions or more general specifications, and the value of the substitution elasticity becomes a first-order empirical parameter.

Estimates of the aggregate elasticity of substitution between capital and labor now cluster somewhat above unity in most careful studies, though considerable disagreement remains. This seemingly technical parameter has cascading consequences. It determines whether capital accumulation raises or lowers the labor share, whether skill-biased technical change increases or decreases wage inequality, and whether higher taxation of capital is borne by capital or shifted onto labor.

For monetary policy, non-unitary elasticities complicate the New Keynesian workhorse. The natural rate of interest depends on production structure, and misspecified factor substitution can bias estimates of the output gap and equilibrium real rates. Central bank communication that references potential output implicitly relies on production function assumptions that variable factor shares call into question.

Heterogeneous agent New Keynesian models are particularly sensitive to these issues. The marginal propensity to consume differs sharply across the wealth distribution, so a shift from labor income to capital income has aggregate demand implications that representative agent models cannot capture. Auclert, Kaplan, and others have shown that factor share dynamics interact with household heterogeneity in ways that amplify or dampen policy transmission.

The policy toolkit must accordingly expand. Redistribution, competition policy, and worker bargaining institutions become macroeconomically relevant, not merely distributional concerns. This challenges the traditional separation between stabilization policy and structural policy that has organized much of modern macroeconomics.

Takeaway

Parameters that seem technical often carry enormous policy weight. The elasticity of substitution is not just a modeling choice—it determines who bears the incidence of nearly every macroeconomic shock and policy response.

The decline in the labor share is not merely a distributional curiosity but a signal that foundational assumptions in our growth models require revision. Kaldor's stylized fact of constant factor shares was always an approximation; we now understand it was an approximation that concealed important structural dynamics.

The path forward requires integrating insights from industrial organization, trade theory, and household finance into macroeconomic frameworks that have traditionally abstracted from these dimensions. Heterogeneous agent models, richer production structures, and explicit modeling of market power all point toward a more complex but more accurate representation of how modern economies actually function.

For policymakers, the implication is not that existing tools have failed but that their calibration must reflect a changing economic structure. Monetary policy alone cannot address distributional shifts driven by technology and market structure. A mature policy framework acknowledges this and expands its analytical scope accordingly.