The Solow paradox
Better tools. Slower productivity growth.
Total factor productivity, US business sector – annual growth, rolling 20-year average
Solow, 1987
Total factor productivity measures growth from working smarter, once extra hours and equipment are counted.
Data: John Fernald, Federal Reserve Bank of San Francisco – quarterly TFP series, US business sector
In the 20 years to 1967, it grew by about 2.2% a year. Electricity, the combustion engine and their spin-offs were still spreading.
Robert Gordon, NBER Working Paper 18315 (2012); data: John Fernald, Federal Reserve Bank of San Francisco
By the early 1970s those inventions had run their course. Growth slowed, and the 20-year average fell to about 0.6% by 1994.
Robert Gordon, NBER Working Paper 18315 (2012); data: John Fernald, Federal Reserve Bank of San Francisco
“You can see the computer age everywhere but in the productivity statistics.”
Robert Solow, New York Times Book Review, 12 July 1987, p. 36
Electricity took about 40 years to lift factory productivity. Plants first had to be redesigned around the new motors.
Paul David, American Economic Review 80(2), 1990
Computers paid off in the late 1990s, as firms reorganised work around them. Firms with stronger management gained most.
Brynjolfsson & Hitt, J. Econ. Perspectives (2000); Bloom, Sadun & Van Reenen, Am. Econ. Review (2012)
That boost faded in the mid-2000s, before the financial crisis. Leading firms kept improving; the rest fell further behind.
John Fernald, NBER Macroeconomics Annual (2015); Andrews, Criscuolo & Gal, OECD Productivity Working Paper 5 (2016)
Today it averages about 0.6% a year. AI may follow the same path: heavy investment in new ways of working first, measured gains later.
Brynjolfsson, Rock & Syverson, Am. Econ. Journal: Macroeconomics (2021); data: John Fernald, FRBSF
Each wave paid off once work was redesigned around it.
That takes years, and the gains don’t spread on their own.
Paul David (1990); Brynjolfsson & Hitt (2000); Andrews, Criscuolo & Gal (2016)