Arts & Sciences Magazine

By Mara Sassoon

When economist Pascual Restrepo began his PhD at MIT in 2011, he was hearing a lot of buzz about the impacts of technology on the job market. “People were talking about robots. And people were just starting to talk about artificial intelligence. At that time, it sounded like this cryptic thing of the future,” he says. “Fast forward more than 10 years, and AI’s everywhere.”

Restrepo, an associate professor of economics who studies the interaction of automation and labor, was inspired to conduct some of his earlier research after he observed that a lot of people had positive outlooks on technology’s impact on jobs. He felt that view lacked nuance. “One thing that puzzled me was that everyone was making analogies to past technological revolutions—for example, the Industrial Revolution. And there were many people saying things like, ‘Oh, it’s the usual thing where technology changes the nature of jobs, and everyone benefits,’” he says. “I was skeptical about that narrative.”

For one, that outlook didn’t mesh with macro trends he was observing, including increasing wage inequality and deteriorating employment opportunities for young men with lower education levels. Today, Restrepo continues to look at how automation technologies introduced since the 1980s have contributed to these trends. He has also begun to focus on the impact of AI on the labor market and is intrigued to see how it plays out in the next few decades.

The Robots Are Coming

Restrepo says that the people who had positive outlooks on the impact of new technologies on jobs weren’t considering the vastly different kinds of technology and industries out there. “There are technologies that are tools that can augment the capabilities of humans,” he says. “Then there are automation technologies that work by substituting humans in things that humans were doing.” More sophisticated computer systems might help office workers collaborate or be more efficient, but advanced robotics can replace an entire factory of laborers.

To better understand the impact of automation technologies across regions and industries, Restrepo worked with Daron Acemoglu, an economist at MIT, on a 2019 study published in the Journal of Political Economy, which looked at changes in employment and wage growth in US industries from 1980 to 2007. They found less wage growth in regions where more automating industrial robots were used for manufacturing.

“There are some industries where you can use robots very effectively,” Restrepo says. The automobile, petrochemical, and metallurgical industries are all examples. With car manufacturing, he points out, welding jobs were hit especially hard. “You used to have people doing the welding and now you have a robot.”

However, Restrepo and Acemoglu found that other industries, like apparel manufacturing, have proven more resilient against the impacts of automation on human labor. “It’s very hard to come up with a robot that can do the whole process of sewing, because fabrics come in uneven shapes and sizes. It’s not the same as welding a car which always comes in the same shape,” says Restrepo. Still, that could change. “We’re starting to develop robots that are coupled with computer vision and can automate sewing,” he says.

Restrepo and Acemoglu also examined the differences in wage trends across demographic groups in a 2022 study published in Econometrica. They found that since the 1980s, young men without college degrees have seen their wages decline or stagnate. “Despite all of the technological progress we’ve had, the real earning power hasn’t gone up over time,” Restrepo says. He and Acemoglu estimated that half of that wage stagnation was due to the automation of labor. On the other hand, they found that highly educated people haven’t lost many jobs to automation. “This is an uneven phenomenon,” he says. “It’s hitting some people in some particular industries and jobs, and many people remain completely shielded from this process.”

“It might seem like I would be very negative about automation,” Restrepo says. “Of course, it also generates benefits in the form of cheaper products for everyone.” But those benefits come at a cost to somebody—and that’s where he tends to focus his research. “It’s more a story of winners and losers,” he says. “And this is relevant for thinking about AI.”

Enter AI

Restrepo says he is keeping an eye on whether or not AI has an uneven impact on the job market. “If AI creates pockets of automation in the economy everywhere, and it’s more about automating small components of everyone’s jobs, then I wouldn’t be too worried about it, at least in the short term,” he says.

Overall, Restrepo envisions AI’s impacts will be felt across a diverse set of industries and have more widespread impacts than automation has had over the last four decades. “I think that parts of our jobs will be automated, but as a whole, we are going to benefit because the costs and benefits will be shared evenly. That’s the way in which AI might be different from previous waves of automation.”

Restrepo also acknowledges that there is a chance for AI to follow a different trajectory and affect certain segments of the economy much more than others. If that is the case, “content creators, writers, financiers, even researchers are going to start being more at risk of automation,” he says. “The winners and losers from this wave of AI could be very different from the winners and losers from previous waves of automation.”

Restrepo is now interested in researching how generative AI—systems that can create content—will impact specific jobs. “Are these algorithms going to augment and help workers become more productive? Or are these algorithms going to lead to more automation like we’ve seen in the past? I think that the jury is still out on that. That’s what keeps me awake at night.”