Are Intelligent Machines Coming for Your Job and Your Wages?

A specter haunts Europe, the United States, and the rest of the world: the specter of massive technological unemployment and, therefore, widespread immiseration.

Or at least we keep hearing it does, and we’ve been hearing it for centuries.

We keep hearing. And we keep waiting. And the specter never shows up.

I doubt that it ever will. It’s certainly possible, and improving technology will continue to creatively destroy certain sectors. To use one example, the days of long-haul truck drivers might be numbered. In a recent paper in the European Review of Economic History, Robert Allen showed how “the ‘golden age of the hand loom weaver’ gave way to poverty and employment” ($0 version here).

I seriously doubt, though, that technology and automation are going to lead us into a future of large-scale unemployment and falling wages. Some industries will shrink or disappear altogether. Other industries will expand, and new industries will crop up to replace them.

The employment picture looks grim for some sectors, but it looks pretty bright for workers in computer and information technology. The Bureau of Labor Statistics, for example, estimates that the economy will add 302,500 jobs for software developers between 2016 and 2026.

In an article for Bloomberg last month, Noah Smith notes that the hubbub about technological unemployment and falling wages is largely a lot of scary hand-wringing over what is possible rather than clear-headed analysis of what is probable. He summarizes recent research by Francesco Caselli and Alan Manning that appears as the first article in the first issue of the brand-new journal American Economic Review: Insights (you can find a $0 version here). Caselli and Manning build a theoretical model in which falling prices of investment goods relative to consumer goods, lead to higher average wages.

As Deirdre McCloskey argues, if technology were going to lead to widespread immiseration, it probably would have done it in the nineteenth and twentieth centuries when mechanization transformed agriculture. Furthermore, major episodes in technological history have been about replacing our mental muscles with artificial intelligence. The adding machine and the cash register are types of artificial intelligence, albeit crude ones. Search engines like Google and Bing are manifestations of artificial intelligence that unambiguously make commentators and scholars more productive by making it easier to locate sources and data.

But this time is different, right? Perhaps, but as Joel Mokyr, Chris Vickers, and Nicolas L. Ziebarth put it, however, “If someone as brilliant as David Ricardo could be so terribly wrong in how machinery would reduce the overall demand for labor, modern economists should be cautious in making pronouncements about the end of work.” Maybe this time isn’t as different as we think.

How should we react? At this point, people should be thinking about how they can develop skills that complement robotics and artificial intelligence. If they do, they can expect robots and algorithms to be useful, wage-raising co-workers rather than wage-reducing competitors.

Note: This article is based in part on my article A Curse on (Intelligent) Machines? in Technology, Automation, and Employment: Will this Time be Different?, published by the Fraser Institute, edited by Steven Globerman, and available here.