The most stubborn empirical fact in US labor economics is that median real wages have grown far more slowly than aggregate productivity since roughly 1979. Productivity, measured as real output per hour, has roughly doubled. The median male wage has risen modestly, and for some sub-populations has barely moved at all. This is the decoupling puzzle, and a coherent account of it has been the central project of US labor economics for forty years.

What the Data Actually Show

The headline number — productivity up ~70%, median compensation up ~12% from 1979 to 2019 — comes from the Economic Policy Institute's sustained work on this question. The Brookings Hamilton Project, working with overlapping data, reaches similar conclusions. The number is sensitive to methodology: how you measure inflation, whether you include health benefits as compensation, whether you compare hourly or weekly earnings, whether you look at median or mean wages. Different choices yield estimates ranging from "significant decoupling" to "modest decoupling", but every credible methodology finds some decoupling.

The Composition Effects

Some of the gap is mechanical. Health insurance costs grew faster than wages, so "total compensation" rose faster than "wages". The mix of jobs shifted toward services with lower marginal productivity, so aggregate productivity gains overstate what a typical worker experienced. Inflation indexes for consumer goods and producer goods diverged, which inflates the apparent productivity gain when measured the standard way.

Adjusting for these effects narrows the gap but does not close it. The most rigorous accounting — Lawrence Mishel's work at EPI and Larry Summers' work with Anna Stansbury — agrees that real residual decoupling is around 30-50% of the headline number even after all adjustments. That residual is what needs explaining.

The Falling Labor Share

The aggregate labor share of national income has fallen from roughly 65% in 1979 to about 58% by the late 2010s. This is the macro-level fact that corresponds to the micro-level wage stagnation: the share of output going to workers as a whole has shrunk, and the share going to capital owners (interest, dividends, retained earnings) has grown. Across the OECD the trend is similar, though the US declines are sharper than European declines.

The competing explanations for the falling labor share are technology, globalization, market power, and policy. Technology arguments — that automation and AI substitute for labor more than they complement it — are plausible but the empirical case is weaker than many advocates suggest. Globalization arguments — China shock, offshoring — are strong for tradable manufacturing but cannot explain decoupling in non-tradable sectors. Market-power arguments (monopsony in local labor markets, declining unionization, declining outside options) are the most active research area.

The Union-Density Story

US union density has fallen from roughly 25% of the workforce in 1979 to under 11% by 2024. The cross-country and cross-industry correlations between union density and wage growth are strong. The causal direction is debated — does deunionization cause wage stagnation, or do both respond to a common underlying shift in bargaining power? — but the empirical association is robust enough that no serious analysis of decoupling ignores it.

The 2022-2024 labor revival — UAW Stand-Up Strike, Starbucks Workers United, Amazon Labor Union — has produced real wage gains in the unionized sectors. The aggregate effect on national wage statistics is small because unionized sectors are now small. The political-economy test is whether the revival can extend into the non-union sectors and generate broader bargaining pressure, or whether it remains a niche phenomenon.

The Monopsony Frame

Local labor markets typically have fewer employers than they should under perfect competition, which means workers face limited outside options and employers can suppress wages below productivity. The work of Ioana Marinescu, Suresh Naidu, Alan Manning, and others has documented this systematically. The Department of Justice's 2010s guidance against no-poach agreements (Silicon Valley wage-fixing litigation, restaurant chain agreements, hospital agreements) is the policy response. The FTC's 2024 non-compete rule, if it survives litigation, would be a much broader response.

The China Shock

Autor, Dorn, and Hanson's work on the China shock (the 2001-2010 displacement of US manufacturing by Chinese imports following WTO accession) documented that the labor-market damage from a single trade-policy choice was much larger and longer-lived than mainstream trade economics had predicted. The China shock is causally distinct from the broader decoupling story — it is concentrated in particular geographies and sectors — but it interacts with the bargaining-power story by reducing the outside options for displaced workers.

The Honest Reading

Wage stagnation is not a measurement artifact. It is a real phenomenon with multiple causes, all of which have some explanatory weight: globalization shocks, declining unionization, monopsony power, technological displacement of specific job categories, and policy choices on tax, trade, and labor regulation. Different researchers weight the causes differently, but the policy implication across most of the credible accounts is similar: restoring wage growth requires changes on multiple fronts, no single fix is going to unwind forty years of compounding distributional shifts. The 2020s labor revival is the first sustained pushback in decades; whether it broadens or stalls will shape the next phase of the puzzle.

The Coalition Problem

Reversing wage stagnation requires changes on multiple fronts simultaneously — labor law, tax policy, trade policy, antitrust enforcement, education and training policy, healthcare financing. No single intervention is large enough to move the aggregate numbers. The political coalition required to coordinate across all of these fronts has not existed in the US since the late 1960s. Whether it can be assembled again — and whether the underlying constituency exists for the assembly — is the long-run political-economy question that the 2020s have begun to test. The 2022-2025 labor revival is one input; the remaining inputs have not yet materialized.