Research Highlights Article
March 6, 2025
The uneven labor market impact of industrial robots
How are the effects of automation spread across the population?
Source: Phonlamaiphoto
Over the past 25 years, industrial robots have revolutionized manufacturing, reshaping factory floors and displacing human workers. However, little is known about how these effects are distributed across demographic groups.
In a paper in the American Economic Journal: Macroeconomics, Benjamin Lerch shows that automation simultaneously narrowed gender employment gaps while widening racial and ethnic disparities, revealing the uneven effects of technological change.
"I was interested in the negative effects of industrial robots in the United States and I wanted to dig a bit deeper into what is actually going on," Lerch told the AEA in an interview. "Not everybody is affected in the same way by automation technologies."
To understand these uneven impacts, Lerch combined several key datasets. He paired industry-level data from the International Federation of Robotics, which tracked robot adoption across sectors since the early 1990s, with US Census and American Community Survey data on employment patterns across local labor markets, allowing him to compare outcomes among areas with varying degrees of exposure to robots depending on their industrial makeup.
Brains vs brawn
The chart below illustrates how industries adopted industrial robots between 1993 and 2014, measured as the logarithmic growth in robots per thousand workers. Manufacturing industries with high physical labor requirements (brawn task content) showed greater robot adoption, while industries requiring more cognitive skills (brain task content) incorporated fewer robots.
The results paint a complex picture of the effects of automation. Between 1993 and 2014, robots reduced employment by 3.7 percentage points for men compared to 1.6 percentage points for women. This helped narrow the gender employment gap—albeit through job losses rather than gains. Meanwhile, robots cut employment for non-White workers by 4.5 percentage points versus 1.8 points for White workers, widening racial and ethnic employment disparities.
These divergent effects are, in part, explained by occupational segregation—men and racial and ethnic minorities are more concentrated in manufacturing jobs, which are most susceptible to automation. But the biggest impacts often came through indirect "spillover" effects on service sector jobs. When robots displace manufacturing workers, local consumer spending might decrease, reducing demand for services, such as hospitality and retail. These spillover effects hit minority workers particularly hard.
“I find that most of the effects, especially the differential effect across gender and race and ethnicity, happen outside of the manufacturing sector,” Lerch said.
The wage effects were also nuanced. Men's wages fell more than women's, and Lerch estimates that each additional robot per thousand workers decreased the gender wage gap by 0.348 percent. For racial gaps, White workers' wages fell while non-White workers' wages remained stable—primarily because displaced White workers took lower-paying service jobs while displaced non-White workers were more likely to exit the workforce entirely.
If we can develop targeted skills for people so that they can make good use of technologies, they will actually be very beneficial to society.
Benjamin Lerch
These findings may have important implications as new technologies like artificial intelligence (AI) transform a greater number of industries. While industrial robots primarily threatened manufacturing jobs, AI could especially affect white-collar work, where women are more heavily represented.
"Women are employed more often in white-collar jobs and AI seems particularly good at executing some of these tasks," notes Lerch. "Therefore, with respect to AI, women might be exposed to a greater extent than men."
Lerch’s research underscores a crucial point about technological change: its effects depend heavily on existing patterns of inequality and occupational segregation. As long as certain demographic groups are concentrated in specific types of work, even seemingly neutral technologies can have disparate impacts. Moreover, policymakers may need to consider not just the direct effects of automation on particular jobs, but also the broader ripple effects on local economies.
To mitigate unwanted impacts, targeted retraining programs and stronger safety nets could help vulnerable workers adapt rather than leaving the workforce. "If we can develop targeted skills for people so that they can make good use of technologies, they will actually be very beneficial to society," Lerch said.
♦
“From Blue- to Steel-Collar Jobs: The Decline in Employment Gaps?” appears in the January 2024 issue of the American Economic Journal: Macroeconomics.