Carlos Lastres
Carlos Lastres1,393 words

Japan's Manufacturing Was the Most Productive in the OECD. Now It Is 20th of 35.

In 2000, Japanese manufacturing was the most productive in the OECD. Not top five. First.

In the 2024 figures, it is 20th out of the 35 countries the measurement covers.

That is from the Japan Productivity Center's International Comparison of Labour Productivity 2025, published in December 2025. Japanese manufacturing added $80,411 per worker in 2024. The report traces the fall precisely: first in 2000, seventh by 2005, seventh again in 2010, and somewhere between 15th and 20th every year since 2015.

I keep coming back to that timeline because of what else was being measured in the same window.

The wider picture, which is worse

Manufacturing is the part of the Japanese economy that is supposed to be good at this. The rest of the numbers are harder to read.

Japan's labour productivity per hour worked in 2024 was $60.10, or ¥5,720 at purchasing power parity. That puts it 28th of the 38 OECD members, between Portugal at $60.70 and New Zealand at $59.60. The OECD average is $79.40. The United States is $116.50.

Per worker rather than per hour, Japan produced $98,344, which is 29th of 38 and the lowest in the G7. It is about 54% of the American figure. Even against the United Kingdom, the next lowest in the G7, it is just under 80%.

The trend line is not encouraging either. Japan ranked 21st in 2018, fell to 28th by 2020, recovered somewhat, and was back at 28th in 2024. Real productivity growth in 2024 was negative 0.6%, which is 33rd of 38 and the first negative reading since 2020. Employment kept rising because of the labour shortage while real GDP fell 0.2%, and productivity is a ratio, so it went down.

There is one genuinely good number in the report and it deserves saying. Measured against 2019, Japan's real hourly productivity is at 102.7%, which is second best in the G7 behind the United States at 109.7% and ahead of Britain, Germany, France and Italy. Japan is improving faster than most of its peers. It is improving faster from much further back.

And the improvement has not reached anyone's pay. Comparing real average annual wages in 2024 against 2019, the only G7 country below Japan is Italy.

What a productivity number actually measures

Value added divided by hours worked. That is the whole formula.

Which means it is not a measure of effort, and I want to be clear about that because the conversation in Japan usually turns into one about working harder or working longer. Nothing about Japanese effort is the problem. If effort were the input, this country would be first.

The numerator is value added. What the market paid, minus what you bought in. So productivity is a question about pricing power and about what you chose to make. The denominator is hours, which is a question about how much of the work was necessary.

Both halves are decisions. Neither is exertion.

The number I cannot stop putting next to it

I have written before about a footnote in METI's 2018 Design Management Declaration. It cites a survey sent to every Japanese manufacturer with more than 100 employees, twenty five thousand companies, asking what made them successful at opening a market.

For the domestic market, 38.3% said quality and function. Design got 0.8%. For the American market, 63.6% said quality and function, and design got 0.0%.

Those two datasets describe the same period and, I think, the same habit. Between 2000 and 2010, Japanese manufacturing lost its productivity lead while its manufacturers reported, in writing, that the discipline responsible for deciding what to build and what it should cost had nothing to do with why they won.

I am not claiming one caused the other. Currency moved, China industrialised, the measurement basis changed, and anyone who tells you a single input explains a decade of macroeconomics is selling something. What I am claiming is narrower and I think harder to argue with: a company that files design under appearance has no function that owns the numerator. It has excellent functions for the denominator. Manufacturing engineering, quality control, kaizen, supply chain, all world class, all pointed at making the same thing more efficiently.

Ask a Japanese manufacturer to build something cheaper and you will get a genuinely astonishing answer. Ask what it should be worth and to whom, and in a lot of companies you will find that no single team is accountable for that question. It gets split between sales, who want volume, and engineering, who want specification, and it resolves as a compromise nobody chose.

Where the hours actually go

The denominator is the half I see up close, because software is where I work.

METI's own IT talent supply and demand study has the line nobody quotes. Everyone cites the projected shortfall of 450,000 IT workers by 2030. The same document says that at 3.54% annual productivity growth, supply and demand balance and the shortfall disappears entirely.

The reason nobody quotes it is probably that the baseline in the same model is 0.7% a year, which is what Japan's information and communications sector actually managed through the 2010s. So the ask is five times the historical trend, which sounds impossible right up until you look at where the hours go.

3.54% a year in software is not a heroic figure. It is roughly what you get from not building the wrong thing.

Almost every hour I have watched get wasted in a product organisation was spent building something correctly that should not have existed. Not slow work. Precise, careful, well tested work on a feature that one stakeholder wanted, that shipped, and that nobody used. Those hours land in the denominator at full weight and contribute nothing to the numerator.

Deleting a feature from a plan is the highest leverage productivity intervention available to a software team, and it costs nothing. It is also the one most organisations have no mechanism for, because saying no requires someone whose job is the whole rather than a part.

The part that is genuinely mine to argue

I have run into a specific pattern often enough here to trust it. A company brings me in to improve a product's interface. Somewhere in the second or third conversation it becomes clear that the interface is not the constraint. The constraint is that nobody in the room can say which of the eleven things on the roadmap matter, because eleven different people each own one.

Fixing the screens in that situation raises the numerator slightly and leaves the denominator untouched. The productive intervention is upstream and it is unpopular, because it means telling eight of those eleven people that their thing is not happening this year.

METI's declaration named the two conditions for calling something design management. One is a design leader on the management team. The other is that design is involved from the most upstream point of business strategy rather than after it. I would throw away the rest of the paper and keep the second condition, because it is the only one that touches either half of the ratio.

Japan is 20th in manufacturing productivity and 29th per worker, and it is improving faster than most of the G7 from a long way behind. Both of those are true at once, and the second one is the reason to be interested rather than gloomy.

What closes the remaining distance is not more effort. It is somebody senior who is accountable for what gets built and what it is worth, and who has the standing to cancel things. That role has a name in most of the companies that pulled ahead. In a lot of Japanese companies it still does not exist, which means the answer to the productivity problem is a hiring decision that nobody has framed as one.

Sources

← All posts