Vectorium/Shutterstock
Economies are always adjusting to transformative technologies—from windmills, steam engines, and electricity to computers, the internet, and now artificial intelligence. Can lessons from those earlier technological revolutions help us anticipate how AI will affect inequality and the labor market?
Chicago Booth’s Rodrigo Adão, MIT’s Martin Beraja, and University of Texas’s Nitya Pandalai-Nayar examined two specific periods of major technological change, the manufacturing revolution of the early 20th century and the transition to computers and the internet in the late 20th century, and find a stark difference in how they played out in the workforce. Their findings may tell us just how pronounced the changes from AI will be.
The researchers used US census data from the late 19th century to 2019, focusing on male workers aged 16–64, to measure the employment and wage response in occupations most exposed to the new technologies, labor-supply adjustments across worker generations, and the ability of current workers to apply the new technologies. They then constructed a model to analyze how the transferability of worker skills affects the economy’s adjustment to new technologies.
They find that the manufacturing revolution tapped into existing skills—from agriculture and manual labor—that were easily transferable to factories, making for a rapid transition without widening inequality. Older and younger workers both contributed similarly to the new manufacturing economy; the relative wage in the new manufacturing jobs rose less, meaning less financial inequality on the basis of specific skill sets; and the broad economic benefits appeared relatively quickly.
“If a technology comes in, and it can be used by anyone, it’ll be quickly adopted, will have big aggregate effects quickly, positive effects in general, and very little inequality or displacement effects, simply because we all have the skills to use it,” Adão says. Slow-adopting technologies may also have big aggregate effects, but they are delayed.
“You could, for example, come from a farm and be doing labor work on the farm, come to a factory and spend a few weeks or months learning how to do the labor work in the factory,” he says. “But fundamentally it was using similar skills that people already had that could complement those technologies.”
The decisions of companies, governments, and educators will help to shape the ultimate outcomes of the A.I. revolution.
A.I. Is Going to Disrupt the Labor Market. It Doesn’t Have to Destroy It.Adjustments to innovations in information and communications technologies, on the other hand, were slower and more unequal, dramatically slowing the transition, the researchers find. That revolution created jobs requiring more cognitive skills and higher education, necessitating the training of a whole new generation of workers and disproportionately benefiting younger people.
The slower the labor-market transition to a new technology, as in the case with ICT, the more the gains will accrue for people in the future, the researchers find. Those most negatively affected will be workers today who may be displaced because they cannot redeploy their skills effectively to participate in the sectors expanding as a result of the new technologies.
“When there is more inequality, it means that the aggregate effects such as GDP growth and overall production are going to take longer to accrue,” Adão says.
He suggests that policymakers might consider compensating workers who are losing today through the gains of those who are most likely to win in the future. Such policies could include transfers—or direct payments—to displaced workers that would slow over time as younger workers entered the system, or taxes on new technologies that could fund the social safety net, or programs to help workers learn new skills. (For more on policies to help displaced workers, read “What’s the best way to retrain jobless workers?”)
The researchers don’t focus on AI but do note that their findings raise the question of whether the adjustment to the new technology will more closely resemble that of the manufacturing or of the ICT revolution.
The key question, Adão says, is whether AI is more likely to complement skills that already widely exist, making many workers more efficient, or to tap into the skills of only a small part of the labor force. The researchers think it’s too early to tell.
What is clear, Adão says, is that if the labor market ultimately adjusts to AI as it did to the manufacturing revolution, it will be a much faster transition. But if it adjusts as it did to the ICT revolution, the transition will be slower, even as the technology’s diffusion accelerates.
Rodrigo Adão, Martin Beraja, and Nitya Pandalai-Nayar, “Fast and Slow Technological Transitions,” Journal of Political Economy Macroeconomics, forthcoming.
If research using 2021 and 2022 data is any indication, the odds are good.
Could Stablecoins Destabilize Other Markets?
MIT’s Daron Acemoğlu joins hosts Bethany McLean and Luigi Zingales to discuss the key challenges of ensuring that new technology creates broadly shared benefits.
Capitalisn’t: Is Technological Progress Good for Everyone?
Chicago Booth’s Alexander Todorov and University of Chicago’s Wilma A. Bainbridge and Ben Zhao discuss biases relating to faces and the implications of facial-recognition technology.
Is it Ethical to Use Facial Imaging in Decision-Making?Your Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.