AI Is Creating More Jobs Than It Is Eliminating
The forecast that artificial intelligence will lead to mass unemployment took another hit this week as demand for workers rose sharply in April data.
The Bureau of Labor Statistics said that job openings surged 731,000 to 7.618 million. That beat even the rosiest forecasts from Wall Street economists.
The biggest source of the rise was in professional and business services, which accounted for 668,000 of the monthly increase, or roughly 91 percent of the total move. The sector went from 1.047 million to 1.715 million openings in a single month. This is one of the more leading and cyclically sensitive categories in the Job Openings and Labor Turnover Survey (JOLTS) because firms ramp up demand for those services early in an expansion cycle and cut them early in contraction. So the increase here suggests that growth is accelerating.
This bolsters the thesis that AI may be giving rise to a Jevons Paradox in the labor market. As AI raises the marginal product of skilled labor, it raises the return on deploying skilled labor, which increases total demand for skilled labor beyond what the direct displacement effect removes.
Construction and Manufacturing Openings on the Rise
The AI buildout boom is also having more direct effects. Construction openings rose by 25,000 to 259,000. That level of openings does not look all that impressive by the standards of the immediate post-pandemic period—openings peaked above 450,000 in March 2022—but it is healthy on a longer timeline. Much of the demand for construction workers is likely artificial intelligence-related, reflecting the construction spending data that saw data center construction up more than 28 percent year-over-year.
Openings in manufacturing rose by 24,000, with a 1,000 openings contraction in nondurables and a 25,000 expansion in durables. At 474,000, openings in manufacturing are way down from the post-pandemic hiring spree but elevated compared with the decade that preceded the pandemic. Openings in durable goods—at 321,000—are particularly impressive by historical standards and have been trending up since autumn.
Of course, Jevons Paradox does not mean there is no displacement. Openings in the information sector inched up by 4,000 to 87,000. But they are significantly below the year ago level. And for the second month in a row, layoffs and discharges in the information sector came in at 60,000. The layoff rate dipped to 2.1 percent after March’s 2.2 percent, both of which are elevated by historical standards. In data going back to 2002, the typical layoff rate for information jobs is around 1.8 percent. If we exclude the pandemic, it’s 1.5 percent. And while hires still exceeded layoffs, the gap was narrower than typical. If you add in quits, information payrolls appear to have contracted on net. For workers in the information sector, this low level of demand is rare outside of economic downturns.
Low Hires Are Evidence of Supply Constraint
Skeptics to the bullish read of the JOLTS report have pointed out that hires were down broadly. Trade and transportation were off 134,000, professional and business services itself were down 131,000 despite the opening surge. The hires rate at 3.2 percent is near the low end of the post-pandemic range. But this is likely a function of supply constraint rather than a lack of demand. Unemployment has been hovering around four percent, leaving few workers sitting available. The quits rate at 1.9 percent means that labor market circulation — the job-to-job mobility that normally recycles workers through the economy — has nearly stalled. Immigration enforcement has structurally reduced the marginal labor supply in exactly the lower-skill segments where unfilled openings are most concentrated.
Nonetheless, the net effect of AI on labor demand appears to be positive. Indeed, the western region saw openings rise by 493,000. That looks like workers being rotated out of information jobs and into business services.
A more pertinent point is that the JOLTS data tends to be volatile and has some residual seasonality. The professional and business services sector, in particular, is volatile around April, perhaps due to tax preparation. But the April growth was impressive even if we discount for a surprisingly strong seasonal boost.
The picture that emerges from JOLTS is one of efficiency gains expanding total demand for labor while displacing the tasks—like coding—that AI directly substitutes. The AI boom and the AI disruption are happening at the same time. They are the same event, observed from two different seats in the labor market.