When we hear about automation and artificial intelligence replacing jobs, it may seem like a tsunami of technology is going to largely wipe out workers in the name of greater efficiency. But a study co-authored by an MIT economist shows markedly different dynamics in the US since 1980.
Rather than implementing automation in the pursuit of maximum productivity, companies have often used automation to replace workers who typically receive a “wage premium”, earning higher wages than other comparable workers. In practice, this means that automation has often eroded the earnings of non-college-educated workers, who earned better wages than most workers with similar qualifications.
This discovery has at least two major implications. For one thing, automation has driven the increase in American income inequality to a greater extent than many observers realize. At the same time, automation has led to increases in average productivity, due to companies’ focus on controlling wages rather than finding more technology-driven ways to increase efficiency and long-term growth.
“There has been an inefficient targeting of automation,” says MIT’s Daron Acemoglu, co-author of a published paper detailing the study’s results. “The higher the employee salary in a particular industry or occupation or function, the more attractive automation becomes for companies.” In theory, he notes, companies could automate efficiently. But they have not done so by emphasizing it as a tool for pay cuts that helps their own internal short-term figures without creating an optimal path for growth.
The study estimates that automation is responsible for 52 percent of the increase in income inequality from 1980 to 2016, and about 10 percentage points come specifically from firms replacing workers who were earning a wage premium. This inefficient targeting of certain employees has reduced the productivity gains from automation by 60-90 percent over a period of time.
“This is one of the possible reasons why productivity improvements have been relatively slow in the U.S., despite the fact that we have an amazing number of new patents and an amazing number of new technologies,” Acemoglu says. “Then you look at the productivity figures, and they’re quite pathetic.”
The paper, “Automation and Rent Dispersion: Implications for Wages, Inequality, and Productivity,” appears in the May print issue. Quarterly Journal of Economics. The authors are Acemoglu, institute professor at MIT; and Pascual Restrepo, associate professor of economics at Yale University.
implications of inequality
Since the 2010s, Acemoglu and Restrepo have together conducted several studies about automation and its effects on employment, wages, productivity, and firm growth. In general, their findings suggest that the effects of automation on the workforce after 1980 are more significant than many other scholars have assumed.
To conduct the current study, researchers used data from multiple sources, including U.S. Census Bureau statistics, data from the bureau’s American Community Survey, industry numbers, and more. Acemoglu and Restrepo analyzed 500 broad demographic groups, sorted by five levels of education, as well as gender, age, and ethnic background. The study combines this information with an analysis of changes in 49 US industries to take a detailed look at how automation has affected the workforce.
Ultimately, the analysis allowed scholars to estimate not only the total amount of jobs lost due to automation, but how many of them were caused by companies specifically trying to remove the wage premium they received for some of their workers.
Among other findings, the study shows that within the groups of workers affected by automation, the impacts are greatest for workers in the 70th–95th percentile of the salary range, indicating that higher-income workers bear the greater brunt of the process.
And as the analysis indicates, this single factor accounts for almost one-fifth of the total increase in income inequality.
“I think it’s a big number,” says Acemoglu, who shared the 2024 Nobel Prize in Economic Sciences with his longtime colleagues Simon Johnson of MIT and James Robinson of the University of Chicago.
He says: “Sure, automation is an engine of economic growth and we’re going to use it, but it creates huge inequalities between capital and labor and between different worker groups, and so it may be a huge contributor to the increase in inequality in the United States over the last several decades.”
productivity puzzle
The study also highlights a fundamental but often overlooked choice for firm managers. Imagine a type of automation—for example call-center technology—that could actually be inefficient for a business. Nevertheless, firm managers have an incentive to adopt it, reducing salaries and overseeing a less productive business with increased net profits.
On a larger scale, it seems that some version of this has been happening in the US economy since the 1980s: greater profitability does not equate to increased productivity.
“Those two things are different,” Acemoglu says. “You can reduce costs while increasing productivity.”
Indeed, Acemoglu and Restrepo’s current study recalls an observation by the late MIT economist Robert M. Solow, who wrote in 1987, “You can see the computer age everywhere except in productivity statistics.”
In that vein, Acemoglu believes, “If managers can reduce productivity by 1 percent but increase profits, many of them may be happy with that. It depends on their priorities and values. So the other important implication of our paper is that good automation at the margin is being coupled with very good automation.”
To be clear, the study does not imply that less automation is always better. Some types of automation can boost productivity and promote a virtuous cycle in which a firm makes more money and hires more workers.
But currently, Acemoglu believes, the complexities of automation have not yet been recognized clearly enough. Perhaps looking at the broader historical patterns of US automation since 1980 would help people better understand the tradeoffs involved – and not just economists, but also firm managers, workers, and technologists.
“The important thing is whether it gets into people’s thinking and where we get to in terms of a better holistic assessment of automation in terms of inequality, productivity and labor market impacts,” Acemoglu says. “So we hope this study moves the dial there.”
Or, as they concluded, “We could miss out on potentially better productivity gains by calibrating the type and extent of automation more carefully and in a more productivity-enhancing way. It’s all a choice, 100 percent.”
