Productivity: Why It Is Everything

By Thorsten Meyer | November 12, 2026 | ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

Work After Work · Episode 10 · Arc 2: The basics everyone needs

On 21 October 1987, the Royal Swedish Academy of Sciences awarded the Nobel Prize in economics to Robert Solow of MIT “for his contributions to the theory of economic growth”. Thirty years earlier, Solow had shown how to measure where growth comes from.

Three months before the prize, in July 1987, the same man wrote a single sentence in a book review that economists still argue about. And right now, because of AI, it is back.

Yesterday, in episode 9, we looked at GDP, the value of everything with a price tag. Today: what one hour of work produces, and why that number decides how rich a country gets.

Key takeaways

• Productivity is output per hour worked. At 2% a year it doubles in about 35 years; at 1% it takes 70. That one point decides whether your children or only your grandchildren are twice as well off.
• Most of it comes from better ways of working. Solow found in 1957 that about seven eighths of the rise in American output per hour from 1909 to 1949 came from technical change, only one eighth from more machines.
• The number to remember is 5. One hour of work in American business produced about five times as much in 2025 as in 1948.
• Computers took a decade to show up. Productivity growth slowed after 1973 while computers spread. It jumped to about 3% a year from 1995 to 2004, and Fed economists credited information technology with about two thirds of the speed-up.
• AI looks the same so far. Big gains inside single firms (+15% for support agents in one study), but executives put AI’s effect on productivity growth so far at about 0.2% a year.

Products Worth Considering

DJI Mini 5 Pro Fly More Combo with DJI RC 2

AmazonView Latest Price

Amazon Kindle Scribe Colorsoft 64GB with 11-Inch Color Display and Premium Pen, Graphite

AmazonView Latest Price

Samsung Odyssey G3 G30D 27-Inch FHD Gaming Monitor

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

Almost everything

Productivity, in its simplest form, is output per hour worked: how much value one hour of work produces. Statisticians call it labor productivity.

The economist Paul Krugman wrote a sentence in a 1990 book that has been quoted ever since: “Productivity isn’t everything, but in the long run it is almost everything.”

Here’s why. Growth compounds:

• At 2% a year, output per hour doubles in about 35 years.
• At 1% a year, it takes about 70 years.

That one point decides whether your children are twice as well off as you, or only your grandchildren. The other ways to get richer run into limits: you can work longer hours only up to a point, and you can put more people to work only once. That is why, in the long run, output per hour matters so much.

1957: seven eighths

But where does productivity come from? In 1957, Solow took the American numbers from 1909 to 1949 and split the growth into two parts: more capital (machines, buildings, tools) per hour of work, and everything else.

His result, in the Review of Economics and Statistics: gross output per man-hour “doubled over the interval, with 87½ per cent of the increase attributable to technical change and the remaining 12½ per cent to increased use of capital”.

More machines explained about one eighth. Seven eighths came from something else: better technology, better methods, better organisation, better skills. Economists call this leftover total factor productivity (TFP): the growth in output that more capital and more hours cannot explain. It is a residual, measured by what is left over, which is also why economists keep arguing about what is inside it.

Productivity moves the sector clock

You have seen productivity at work in this season already. In episode 2, about three in four American workers farmed around 1800. Today it is 1.5%.

The US Department of Agriculture tracks what happened on the farms since 1948:

• Output: in 2023, US farm output was nearly three times its 1948 level (+210%, 1.52% a year).
• Inputs: all inputs together (land, labour, machinery, fertiliser, energy and more) rose only 0.12% a year, about 9% over the whole period.
• Total factor productivity: +1.40% a year on average.

Nearly three times the food from almost the same total inputs: that is productivity. And it drives the sector clock. In 1950, about one in eight American workers farmed. By 2025, the International Labour Organization estimates 1.5%. The work didn’t vanish. It moved, first into factories, then into offices and services.

The golden age and the slowdown

After the Second World War came a golden age. The Bureau of Labor Statistics measures output per hour in the nonfarm business sector (the private economy without farms, government, nonprofits and households). Using its index, our calculation of annual averages gives:

• 1948-1973: +2.8% a year.
• 1973-1995: +1.4% a year.

Growth roughly halved. And this happened just as computers were spreading through offices and factories.

The Solow paradox, 1987

In July 1987, Solow reviewed the book Manufacturing Matters by Stephen Cohen and John Zysman for the New York Times Book Review. The authors believed in a revolution in manufacturing through programmable automation. Solow pointed out that “what everyone feels to have been a technological revolution, a drastic change in our productive lives, has been accompanied everywhere, including Japan, by a slowing-down of productivity growth, not by a step up”.

And then the sentence: “You can see the computer age everywhere but in the productivity statistics.”

It became known as the Solow paradox, or the productivity paradox.

Then the computers arrived

Then came a twist. From 1995 to 2004, US productivity growth jumped to about 3.1% a year (our calculation from the BLS index).

In May 2000, two Federal Reserve economists, Stephen Oliner and Daniel Sichel, measured how much of the late-1990s rebound came from computers. Their estimate: the use of information technology and the production of computers “accounted for about two-thirds of the 1 percentage point step-up in productivity growth between the first and second halves of the decade”. Their answer to the question in their title: “information technology largely is the story”.

The computer age had arrived in the statistics, about a decade after Solow’s sentence. Then growth fell back: about 1.5% a year from 2004 to 2019.

Goldman Sachs economists later described a similar pattern for earlier general-purpose technologies. In their reading, the productivity booms after the electric motor and the personal computer started “about 20 years after the technological breakthrough, at a point when roughly half of US businesses had adopted the technology”, and then raised labour productivity growth by around 1.5 percentage points a year for a decade.

Economists Erik Brynjolfsson, Daniel Rock and Chad Syverson offer one explanation, which they call the productivity J-curve: a new general-purpose technology first needs large, hard-to-measure investments in new processes, skills and business models. These costs make measured productivity look weak early on, and too strong later, when the benefits are harvested.

The number to remember: 5

Put the decades together. In 1948, the BLS index of output per hour in US nonfarm business averaged 23.1 (2017 = 100). In 2025, it averaged 118.6.

That’s the number to remember: 5. One hour of work in American business today produces about five times as much as in 1948. Same hour, five times the output.

In principle, this is what pays for higher wages, shorter weeks and cheaper goods. How much of it actually reached workers is the topic of the next episode.

Today: the US, Germany and Asia

• United States: since 2019, output per hour has grown about 2.3% a year (our calculation, 2019 to 2025), faster than the 1.5% from 2007 to 2019. In 2025 it rose by about 2.4%. Whether AI already plays a part is debated.
• Germany: output per hour worked fell by 1.2% in 2023 and 0.3% in 2024, then rose by 1.1% in 2025, according to the federal statistics office (price-adjusted, as of 25 August 2026).
• Levels: in 2024, one hour of work produced about $116.5 of value added in the United States, $98.9 in Germany, $60.1 in Japan and $57.5 in South Korea, converted at purchasing power parity, according to the Japan Productivity Center’s comparison of OECD data. The OECD average was $79.4. Japan ranked 28th of 38 OECD countries; its real labour productivity fell by 0.6% in 2024 as employment grew faster than output.

Germany is a useful reminder that the number can also go down. Japan is a reminder that a rich, highly automated country can still be in the lower half of the OECD table per hour of work.

Will AI repeat the Solow paradox?

Inside single firms, the gains can be large. In a study published in the Quarterly Journal of Economics in 2025, Erik Brynjolfsson, Danielle Li and Lindsey Raymond followed 5,172 customer support agents as an AI assistant was introduced. Access to the assistant raised issues resolved per hour by 15% on average. Less experienced and lower-skilled agents gained about 30%. The most experienced and highest-skilled agents saw small gains in speed and small declines in quality.

Across whole firms, the effect is still small. In a survey of nearly 9,000 chief financial officers, chief executives and senior executives in the US, UK, Germany, Australia, Canada, Mexico and South Africa, run in summer 2026, executives reported that AI adoption has raised annual labour productivity growth by 0.2% over the last three years. They expect 0.6% a year over the next three. 78% of the firms use AI. These are the executives’ own estimates, published as NBER Working Paper 34836 (revised October 2026).

The forecasts are far apart:

• Goldman Sachs (March 2023): generative AI could raise US labour productivity growth by about 1.5 percentage points a year over a ten-year period after widespread adoption. The authors called the estimate “highly uncertain” and gave a range of 0.3 to 3.0 points.
• Daron Acemoglu (MIT, April 2024): total factor productivity gains of “no more than a 0.71% increase” over ten years in total, and less than 0.55% once harder tasks are taken into account.

Both are estimates, not facts. They differ mainly in how many tasks AI will take over and how much cheaper each task becomes.

Myth check: “Being more productive means working longer hours”

Verdict: false.

• Productivity is output per hour. Longer hours raise total output, not output per hour.
• Most of the rise came from better methods. Solow attributed seven eighths of the rise in output per hour from 1909 to 1949 to technical change, not to more machines per worker.
• Long hours, low output per hour. By the Japan Productivity Center’s figures for 2024, the average South Korean worker put in roughly 400 more hours a year than the average German worker (our calculation from output per worker and per hour), yet produced $57.5 per hour against Germany’s $98.9.

Productivity is about how you work, not how long.

What it means for businesses

• Buying AI is like buying the first computers. The tool alone changes little. The gains come when you change how the work is done: processes, training, roles.
• Expect a J-curve. The first year often costs more than it saves, because the complementary investments come first.
• Measure output per hour, not licences. Track what a team produces per hour, and quality, before and after.

What it means for you

• Your productivity is your strongest argument for a raise. It is what makes an hour of your work worth more.
• Learn the tools that let you do more in the same hour. The support-agent study suggests the biggest gains go to people still learning the job.
• More output per hour is not automatically more pay. Whether productivity gains reach paychecks is the topic of the next episode.

Thorsten’s thesis: what productivity says about it

My thesis for this season is that automation could become its own sector, up to 90% of the economy within a decade or a few decades. That is my thesis, not a fact, and this season tests it with evidence for and against.

Productivity is its engine. Evidence for it: if machines can do an hour’s work in seconds, output per human hour can rise very fast, the way farm output tripled while farm jobs nearly disappeared. Evidence against it, at least so far: AI shows up strongly inside single companies, but only weakly in the national statistics. And the last big technologies, the electric motor and the computer, took decades to change the numbers. Episode 47 will test the thesis in full.

Limits of this episode

• Our calculations. Period growth rates and the “five times” come from annual averages of the BLS index (FRED series OPHNFB, data to the third quarter of 2026); BLS’s own period tables use slightly different endpoints.
• Measurement. Output per hour depends on price indices and on how services and free digital goods are measured (see episode 9). Total factor productivity is a residual, not something measured directly.
• International levels are converted at purchasing power parities and get revised; rankings can change between editions.
• Hours comparison. The Korea-Germany difference in hours is derived from the Japan Productivity Center’s per-worker and per-hour values, not taken from an official hours table.
• Surveys and estimates. The executives’ figures are self-reported. The Goldman Sachs and Acemoglu figures are model estimates from 2023 and 2024.
• Quotes. Krugman’s sentence is quoted via William Nordhaus (NBER, 2001); Solow’s 1987 review via its citation in NBER Working Paper 19837 and a transcript.
• Interpretations. The lessons for businesses and for you, and the reading of my thesis, are my interpretation.

Series navigation

• Previous episode: Episode 9, What GDP Measures and What It Doesn’t.
• Next episode (tomorrow): Episode 11, How Wages Are Set. One hour of work produces five times what it did in 1948. Did pay rise five times too, and who decides what an hour is worth?

Sources

• Nobel Prize, press release, “The Prize in Economics 1987” (21 October 1987). nobelprize.org
• Robert M. Solow, “Technical Change and the Aggregate Production Function”, Review of Economics and Statistics 39(3), 312-320 (1957). doi.org
• Robert M. Solow, “We’d Better Watch Out”, New York Times Book Review, 12 July 1987, p. 36, as cited in Acemoglu, Autor, Dorn, Hanson and Price, NBER Working Paper 19837 (2014). nber.org (PDF)
• Paul Krugman, The Age of Diminished Expectations (MIT Press, 1990), p. 9, as quoted in William D. Nordhaus, “Alternative Methods for Measuring Productivity Growth”, NBER Working Paper 8095 (2001). nber.org (PDF)
• US Bureau of Labor Statistics, nonfarm business sector: labor productivity (output per hour of all persons), index 2017 = 100, via FRED series OPHNFB (retrieved 9 November 2026). fred.stlouisfed.org
• USDA Economic Research Service, “Agricultural Productivity in the United States: Summary of Recent Findings” (updated 25 August 2026). ers.usda.gov
• US employment shares: Stanley Lebergott, in US Census Bureau, Historical Statistics of the United States (1975), own calculation (1950); ILO modelled estimates via World Bank (2025), see episode 2. data.worldbank.org
• Stephen D. Oliner and Daniel E. Sichel, “The Resurgence of Growth in the Late 1990s: Is Information Technology the Story?”, Federal Reserve Board, FEDS 2000-20 (May 2000). federalreserve.gov (PDF)
• Erik Brynjolfsson, Daniel Rock and Chad Syverson, “The Productivity J-Curve: How Intangibles Complement General Purpose Technologies”, NBER Working Paper 25148 (revised January 2020). nber.org
• Destatis, “Arbeitsproduktivität, Durchschnittslöhne und Lohnstückkosten nach Wirtschaftsbereichen” (hours concept, as of 25 August 2026). destatis.de
• Japan Productivity Center, “International Comparison of Labor Productivity 2025” (22 December 2025), based on OECD data. jpc-net.jp · summary (PDF, Japanese)
• Erik Brynjolfsson, Danielle Li and Lindsey Raymond, “Generative AI at Work”, Quarterly Journal of Economics 140(2), 889-942 (2025). doi.org · arXiv
• Ivan Yotzov, Jose Maria Barrero, Nicholas Bloom, Steven J. Davis et al., “Firm Data on AI”, NBER Working Paper 34836 (February 2026, revised October 2026). nber.org
• Joseph Briggs and Devesh Kodnani, “The Potentially Large Effects of Artificial Intelligence on Economic Growth”, Goldman Sachs Global Economics Analyst (26 March 2023). gspublishing.com (PDF)
• Daron Acemoglu, “The Simple Macroeconomics of AI”, paper prepared for Economic Policy (5 April 2024). economics.mit.edu (PDF)

This article is for information only. It is not career, legal, financial or investment advice.

Researched and drafted with AI assistance, including Anthropic’s Claude.