The study that makes people uncomfortable

In February 2026, Erik Brynjolfsson (Stanford) and a team of economists published NBER Working Paper 32789, titled Minimum Wages and Automation. They analyzed 20 years of data (2005-2025) from manufacturing establishments in the United States, using variations in the minimum wage between neighboring counties (the “border pairs” method).

The central finding: a 10% increase in the minimum wage is associated with an 8.4% increase in the probability of adopting industrial robots over the following two years.

This is not a weak correlation. The study controlled for productivity, firm size, sector and the business cycle. The effect held.

How a higher minimum wage speeds up robot adoption

Why does it happen? The mechanism, step by step

  1. The minimum wage goes up (for example, California went from $15 to $18 an hour in 2024).
  2. For an employer who uses human labor in repetitive tasks (packaging, basic welding, assembly), the marginal cost of an hour of work rises by 20%.
  3. At the same time, the cost of a collaborative robot (cobot) has fallen: a basic Universal Robots arm cost $35,000 in 2020 and $19,000 in 2025.
  4. The employer calculates the break-even point: if the robot pays for itself in 18 months versus the previous 30 months, the decision to automate speeds up.
  5. The result is not immediate mass layoffs, but not replacing workers who retire or quit. Automation happens through attrition, which makes it politically invisible.

The international evidence confirming the pattern

Brynjolfsson's study is not an isolated case. Three recent investigations found the same phenomenon:

Turkey (2023-2025) : A 33.5% increase in the real minimum wage between 2022 and 2023 (due to inflation and a government decision) led manufacturing companies to increase their robot imports by 41% within 18 months, according to data from the Turkish Machinery Exporters' Association.

China (2008-2012) : A study by Fan and colleagues (2024) in the Journal of Development Economics showed that the new Labor Contract Law of 2008 (which raised labor costs by 15-20%) explained 22% of the increase in robot adoption in China's coastal provinces during that period.

Germany (2015-2025): Unlike the cases above, Germany introduced a late and low minimum wage (8.50 euros in 2015). The effect on automation was nil. Why? Because the German minimum wage was set below the equilibrium wage in most manufacturing sectors. The lesson: the effect appears only when the minimum wage exceeds the market wage for the tasks that can be automated.

Expectations: will this process speed up?

Most of the specialists consulted by the Journal of Economic Perspectives (Spring 2026 edition) agree on three trends:

  1. Short term (2026-2028): The effect will intensify because the new generation of robots (with built-in AI) no longer requires specialized programming. A robot that “learns by watching” an operator costs half as much as a traditional one. This lowers the barrier to entry for small businesses.
  2. Medium term (2029-2032): The U.S. Congress is expected to debate an “automation tax” or a “tax credit for hiring humans.” Several economists (among them Lawrence Summers) have proposed taxing the use of robots in tasks previously done by humans, something that already exists in South Korea (since 2022, a 2% tax on automation investment, though a symbolic one).
  3. Long term (2033-2040) : If robots become so cheap that the marginal cost is almost zero, the relationship between the minimum wage and automation will disappear because all automatable jobs will already be automated. In that scenario, the debate will shift to basic income or a job guarantee.

What governments and agencies propose in response to automation

What governments and institutions have proposed (concrete cases)

City of Seattle (2024): Faced with evidence that its minimum wage of $19.97 an hour (the highest in the U.S.) was accelerating the replacement of cashiers with self-service kiosks, the city launched the Seattle Automation Offset Fund: a 1% sales tax on companies that use more than 10 self-checkout terminals, intended to subsidize the hiring of human cashiers during high-demand hours.

South Korea (Industrial Automation Promotion Act, revised in 2025) : Introduced the concept of a “negative robot tax” : companies that keep the same number of human employees after introducing robots receive a refund of 50% of the robot's cost. Those that lay off workers pay an additional 5% on the labor savings.

European Commission (Proposal for a Directive on Fair Work in Automation, December 2025): Proposes that any company that automates more than 10% of an occupational category in a year must:

  • Submit an internal retraining plan
  • Offer displaced workers a transfer to another plant or a severance package equivalent to 3 months per year worked (versus 1 month under current rules)

ILO (International Labour Organization, World Employment and Social Outlook 2025 report) : Recommends not raising the minimum wage in isolation, but accompanying it with:

  • Reductions in social security contributions for low-automation jobs (to narrow the cost gap with robots)
  • Public investment in collaborative robotics (designed to work with humans, not in place of them)

Dube’s objections and Brynjolfsson’s reply

The critical voice: are we misreading the study?

Not everyone accepts Brynjolfsson's conclusion without qualification. Arindrajit Dube (UMass Amherst), one of the most respected economists in minimum wage research, published a rebuttal in May 2026 in the NBER Digest noting that the 8.4% effect is statistically significant but economically modest: a 30% increase in the minimum wage (like the one that occurred in several U.S. cities) would explain only 2.5% more robot adoption. “The fear that raising the minimum wage will set off a wave of automation is exaggerated,” Dube wrote. “The main driver remains the falling price of robots, not the rise in labor costs.”

Brynjolfsson responded in the same issue that Dube ignores the threshold effect: companies do not automate gradually, but when labor cost passes a critical point. That point is reached with small increases if the price of the robot is already close to the break-even level.