A new National Bureau of Economic Research working paper adds simplification to automation and augmentation, and projects a narrowing pay gap rather than a widening one.
The working paper argues that artificial intelligence will change how jobs are done rather than erase whole occupations, and that pay across the workforce will eventually rise — with the largest increases going to workers in lower-skilled roles.
The mechanism the authors describe is simplification. By lowering the skill threshold a given task demands, AI opens higher-paying roles to workers who could not previously perform them. Over the span of a career, the analysis estimates, lower-skilled workers could earn at least 15 to 45 percent more than they would in a scenario without AI — and more than that if the technology develops quickly.
Every worker gains something from tasks being done more efficiently, but lower-skilled workers gain most because their pay rises faster than everyone else's. If the projection holds, the net effect would be to shrink the gap between top and bottom earners — a gap that has been widening.
What is new is the modelling. Earlier work treated technological change as either automation, where machines take over work from people, or augmentation, where machines make workers more productive. The framework here — built by Lukas Althoff, an assistant professor of economics in Stanford's School of Humanities and Sciences and a faculty fellow at the Stanford Institute for Economic Policy Research, with Hugo Reichardt of the Barcelona School of Economics — treats simplification, making a task easier to do, as a third, distinct outcome.
Althoff told SIEPR that the paper departs from most existing research on AI and employment, which has concentrated on identifying which occupations are most exposed to disruption. His work, he said, translates those exposure metrics into wage and employment outcomes for each individual worker, factoring in a person's particular strengths and limitations, their capacity to learn new skills and move between occupations, and economy-wide shifts such as changing demand for goods and services.
The model qualifies its own result. Lower-skilled workers have picked up AI more slowly than other workers — though, by this account, not slowly enough to cancel out the equalizing effect on wages.
The same framework is far less encouraging about skill in general. It projects that skills become less valuable overall and that fewer people pursue degrees beyond a bachelor's. Among those who do enrol, the analysis expects majors built on applied, hands-on capability to produce better job-market outcomes than those built on communication and interpersonal skill. In the first group: engineering, mechanics, physics, architecture, transportation, culinary arts, cosmetology. In the second: law, religion, philosophy, sociology, political science and women's studies. Rising pay at the bottom, on this account, arrives alongside a falling premium on skill itself.