From self-checkout kiosks to self-driving cars, the rapid acceleration of technology and simultaneous growth of economic inequality over the last several decades have sparked new fears that artificial intelligence (AI) could soon eliminate a number of jobs. But while commentators have tended to focus on the question of whether automation could cause mass unemployment over the long term, what are some of the shorter-term effects of automation on the labor market?
In a new report for the Oxford Review of Economic Policy, RSF grantee Frank Levy (MIT) projects how artificial intelligence (AI) will affect the economy in the next five to seven years, and the subsequent effects that these economic changes could have on politics. Drawing from the Bureau of Labor Statistics, Levy shows that the jobs most vulnerable to near-term automation are not low-skilled jobs, as many assume. As he points out, many low-skilled jobs—such as janitorial jobs and home health aide jobs—required unstructured conversation and unstructured physical movement, making them extremely difficult to automate. On the other end of the spectrum, high-skilled occupations that require unstructured cognitive work are similarly unlikely to be subject to AI. Instead, it is middle-skilled occupations—jobs like medical transcription and assembly line work—that are most vulnerable. As Levy writes, “on balance, near-term AI will have the greatest effects on blue-collar work, clerical work, and other mid-skilled occupations.”
The figure above displays the 2000 and 2016 distribution of all employed people in the U.S. across occupational categories. Category bars in black—which include transportation and material moving, office and administrative support, production, and installation and repair—are occupations that score high in routine cognitive content and/or routine physical content, meaning that they are potentially vulnerable to automation and outsourcing. As Levy notes, the percentage of people employed in these categories declined from 39.2 percent to 33.3 percent between 2000 and 2016. On the other hand, the percentage of people in both higher and lower wage categories increased during the same time period, suggesting a pattern of occupational polarization.
When middle-skilled jobs decline, most of the displaced workers will move to lower-wage jobs. Levy cautions that this downward mobility could create the conditions for a populist backlash like the one that animated the 2016 presidential election. “Since automation anxiety is already high,” Levy writes, “it is likely like AI-induced job losses will eventually create their own political reaction.”