Productivity growth proves additive rather than exponential
A study by Callum Jones, David López-Salido and Thomas Philippon finds that total factor productivity growth is additive rather than exponential. Evaluating US data since 1890 alongside international series, Bayesian model selection assigns near-total probability weight to linear expansion.
Constant steps, declining percentages
Standard macroeconomic models assume that total factor productivity expands exponentially.
Evaluating US data from 1890 to 2022, the authors find that productivity increments are arithmetic, adding a constant level each year.
Since 1947, US productivity normalized to one has increased by roughly 250 basis points annually, causing measured percentage growth rates to halve naturally as the base doubles.
Bayesian model comparison shifts posterior probability weights to nearly one for the additive specification over the exponential benchmark.
In ten-year point forecasts, the additive model yields a mean squared error of 0.068 compared with 0.158 for the geometric alternative, cutting errors by over half.
Dismantling the slowdown puzzle
The linear pattern holds across 22 of 23 advanced economies and emerging markets such as Taiwan and South Korea.
Projections from the Survey of Professional Forecasters and the Council of Economic Advisers also mirror additive trajectories rather than compounding.
This framework reframes the post-1970s productivity slowdown as an artifact of geometric misspecification.
Furthermore, Robert Solow adopted exponential growth in 1957 because a structural shift around 1930 obscured linearity in his shorter sample.
Challenging growth orthodoxy
Replacing exponential benchmarks with additive dynamics fundamentally upends long-term fiscal and climate modeling.
The paper delivers a rigorous empirical corrective to decades of unverified theoretical convenience.
Policymakers must stop misdiagnosing natural arithmetic deceleration as structural stagnation.
Source: Working Papers. Is growth additive?
IN: