The Productivity-Pay Gap

In the AI Story

The gap's magnitude is substantial. Between 1973 and 2020, American productivity per worker-hour roughly doubled while real median hourly compensation grew by approximately 15%. The remaining productivity gains were distributed to capital owners (through higher profits and stock returns), to top earners (through the widening wage distribution), and to consumers (through product quality improvements that national accounts imperfectly capture).

The gap's mechanisms are debated. Erik Brynjolfsson's work attributes substantial portions to skill-biased technological change, with workers whose skills complement new technologies capturing gains while those whose skills are substituted by technology lose bargaining power. Acemoglu and Restrepo emphasize the role of automation in shifting income from labor to capital. Others emphasize declining union power, globalization, and the shift from manufacturing to services. Schor's framework treats the gap as a compound outcome of multiple institutional mechanisms, all of which AI threatens to accelerate.

For AI specifically, the productivity-pay gap implies that AI-driven productivity gains will flow overwhelmingly to firms and capital owners unless institutional mechanisms are constructed to redirect them. The historical record is unambiguous: without such mechanisms, the gains do not reach workers as either income or time. They are captured, and the capture is structural rather than discretionary.

The gap's implication for Schor's framework is that the time dividend's institutional capture problem is not merely about converting income into time — it is about ensuring that any of the productivity gain reaches workers at all. If the productivity-pay gap continues its historical trajectory under AI, the question of whether workers convert their share into hours or income becomes moot, because their share will be too small to matter. The first institutional challenge is therefore ensuring that workers capture a meaningful fraction of the AI productivity surplus; the second is ensuring that the fraction they capture is allocated to time rather than consumption.

Origin

The gap was first extensively documented in the 1990s as productivity growth accelerated while wages stagnated. Lawrence Mishel and colleagues at the Economic Policy Institute have maintained the most-cited longitudinal tracking.

Theoretical explanations developed across multiple research programs, including Brynjolfsson on skill-biased technological change, Acemoglu and Restrepo on automation and capital share, and Piketty and Saez on the broader distributional dynamics.

Key Ideas

Structural redirection. Productivity gains flow systematically to capital rather than workers since approximately 1973.

Magnitude substantial. Productivity per hour doubled; real median compensation grew ~15%; the remainder flowed to capital, top earners, and unmeasured consumer surplus.

Multiple mechanisms. Skill-biased technological change, automation, declining union power, globalization, and sectoral shifts all contribute.

AI amplification risk. Without institutional mechanisms, AI productivity gains will follow the same structural pattern, flowing overwhelmingly to capital.

Compound challenge. Capturing the time dividend requires first ensuring that workers capture the income dividend at all.

Debates & Critiques

The specific mechanisms driving the gap remain actively debated. Some economists argue that the gap is overstated by measurement choices; others argue that it is understated by failure to capture informal compensation and consumer surplus. The policy implications also diverge: some analysts emphasize supply-side solutions (training, education) while others emphasize demand-side interventions (regulation, collective bargaining, tax policy).