Money & State

For 40 years, productivity and wages have stopped rising in step. AI will widen that gap further.

When productivity gains flow into software instead of labour, they don't flow into wages either.

That wages grow with productivity was negotiated practice in Austria for two decades, and it had a name: the Benya formula. Wage growth equals inflation plus the productivity gain. Produce more, earn more. The benchmark for every collective bargaining round, the foundation of the social partnership.

Since the mid-1980s the promise has held only on paper. The economy creates far more value per hour worked than it did back then. Only part of that has shown up in the hourly wage. The gap is smallest in Austria, largest in the US, the EU sits in between. But it is growing everywhere.

What doesn’t flow into wages doesn’t vanish. It lands in share buybacks, where corporations pump profits into their own share price, and in rising property and securities prices, in wealth that concentrates at the top. Whoever owns capital shares in the productivity growth. Whoever has only their labour shares in less and less.

AI didn’t cause this decoupling. It has been running for forty years. But AI changes its character. Until now, a productivity gain meant the same workers producing more with better tools. With AI agents it increasingly means fewer workers and more software. The gain comes not from more productive work but from replaced work. For the slice of work that goes entirely to software, the Benya formula comes out at zero. Zero paid hours, zero wage claim.

Earlier waves of automation hit the factory first, then the routine office. AI reaches the work above that: writing, research, legal review, first-pass medical diagnosis, programming. Exactly the professions long held up as proof that technical progress need not lead to wage stagnation. If the gains land with capital there too, the decoupling no longer hits only those who have known it for forty years. Then it hits the middle class.

We need to talk about this

What does this topic stir up in you?

In depth

How big is the gap, exactly?

EU, Eurostat: between 2000 and 2024, real labour productivity per hour (adjusted for inflation) grew by 25.2 percent, the real hourly wage by 9.8 percent. The gap is roughly 15 percentage points.

Austria, Eurostat: productivity per hour up 26.8 percent, real hourly wage up 19.3 percent, a gap of about eight percentage points. Lower than the EU average, but the pattern is the same.

US, Economic Policy Institute (as of 2024): since 1979, net labour productivity has risen by more than 80 percent, while the typical hourly wage of full-time workers has risen by barely 30 percent. The gap is more than 50 percentage points, and it has been growing since the 1980s.

What the Benya formula was

Named after Anton Benya, the long-serving ÖGB president. The formula: annual nominal wage growth equals inflation compensation plus the medium-term rise in economy-wide labour productivity (GDP per hour worked). Introduced in the 1960s, established as the benchmark for collective bargaining in the Austrian social partnership.

The idea: whoever produces the productivity gains should share in them. Purchasing power holds, private consumption carries domestic demand, the social partnership moderates distributional conflict.

Gerhartinger, Haunschmid and Tamesberger (Arbeiterkammer Oberösterreich, 2018) date the break to the mid-1980s. Since then, wage growth and productivity have decoupled in Austria. The wage share of national income (labour's slice of all income generated) fell from 76 to 68 percent between 1995 and 2011. The authors put it down to a loss of power on the workers' side: falling union membership, globalisation, weaker institutional footing.

What does the research forecast for AI effects?

Goldman Sachs Research estimated in 2023 that generative AI could lift global GDP by roughly seven percent over ten years, about seven trillion US dollars in additional output. Around a quarter of the work currently paid for in the US and Europe would be technically automatable by today's AI models.

McKinsey estimated in June 2023 that the additional annual productivity growth would run from 0.1 to 0.6 percentage points through 2040, depending on how fast AI takes hold. The BIS (the central bank of central banks) expects a moderate effect in its 2024 annual report, not a shock.

These forecasts say nothing about who collects the gains. They only say that gains exist.

Where does the money land instead?

The World Inequality Database tracks wealth and income concentration back to 1820. Right now the richest one percent of the population holds about 25 percent of total wealth in Europe, about a third in the US. Both figures have climbed steadily since the 1980s.

At the corporate level, a growing share of US profits has gone into share buybacks since the 1980s: from practically zero to more than 60 percent at times during the 2010s and 2020s (Yardeni Research, S&P 500 data).

In Europe the buyback phenomenon is smaller, but here too property and shares have grown more expensive. The Eurostat house price index for the EU rose by more than 50 percent between 2010 and 2024. Anyone who owns assets gains regardless of their wage income.

With AI a new component joins in: the productivity gains don't flow into new jobs, they flow into software licences. Invoices from OpenAI, Anthropic, Microsoft. The money doesn't pass through a payslip, it goes straight onto the balance sheet of a US corporation.

Sources