Press Releases

News from EPI No evidence that automation leads to joblessness or inequality

In The zombie robot argument lurches on, EPI President Lawrence Mishel and Research Director Josh Bivens challenge the popular media narrative that the pace of automation is increasing, and that it will lead to overall joblessness and greater inequality.

Automation—the implementation of new technologies as capital equipment or software replace human labor in the workplace—has been ongoing for decades, and there is no evidence that this has led to greater joblessness or increasing unemployment, even though certain jobs were lost. Simply put, if automation necessarily led to higher unemployment, the United States would have experienced an ever-rising unemployment rate over the last century, or more. If an automation surge were currently underway—and if automation led to joblessness—we would not have seen the unemployment rate decline from ten percent to under five percent.

What explains the failure of overall joblessness to rise despite ongoing automation? Automation allows businesses to cut costs, which leads to lower-priced goods—giving consumers additional money to spend elsewhere and creating jobs. Automation also creates complementary jobs in new industries. While there is no question that automation eliminates jobs in particular occupations or industries, historically it has not led to increased overall joblessness. There is little reason to believe that this pattern will not continue in the future.

“What is remarkable about the media narrative around automation is how strong the desire to believe it is, despite so little evidence to support these claims,” said Mishel. “There clearly are serious problems in the labor market that have suppressed job and wage growth for far too long, but these problems have their roots in intentional policy decisions regarding globalization, collective bargaining, labor standards, and unemployment levels—not technological change.”

Mishel and Bivens critique media coverage of a widely reported paper by Daron Acemoglu of M.I.T. and Pascual Restrepo of Boston University. Acemoglu and Restrepo’s paper contained a high-quality estimate of the impact of robots on local commuting zones, but their translation of these estimates into a nationwide number of jobs lost is unconvincing. While there is no question that some workers have been displaced by robots, the computations used in Acemoglu and Restrepo’s national model are not persuasive, and do not take into account the many complementary jobs that are created by automation and the associated cost savings. Furthermore, despite reporting that characterized them as large, Acemoglu and Restrepo’s estimates of the number of jobs lost are relatively small—about 2 percent of total jobs gained in the period studied (and just a fourth of the jobs lost to trade with China in the 2000s).

“Studies that attempt to estimate the number of jobs that will be potentially lost to automation in the future never seem to take into account automation’s positive effects on employment,” said Bivens. “The invention of the automobile eliminated jobs in the horse-drawn carriage industry, but it led to new jobs in the repair, and sales of autos, as well as the construction of highways. There is no evidence that future automation will somehow be different.”

Furthermore, Mishel and Bivens argue, there is no evidence that automation has led to wage stagnation or increasing inequality. Indicators of automation increased rapidly in the late 1990s, a period that saw the best across-the-board wage growth for American workers in a generation. There are frequent claims that the polarization of occupational employment growth (where occupations in the middle shrink while those of high- and low-wage occupations expand) due to automation has hurt middle-class wages and generated inequality. In fact, as Bivens and Mishel point out, occupational employment patterns are unrelated to wage trends and moreover, occupational employment growth has not been polarized since 1999, although inequality has increased.

Lastly, Bivens and Mishel challenge the idea that there’s been a recent surge in automation ushering in an accelerating wave of joblessness. Rather, trends in productivity, capital investment, information equipment investment, and software investment all suggest that automation has decelerated in the last decade. And the rate of shifts in occupational employment patterns has been slower since 2000 than in any period since 1940.

“Rather than wringing our hands over possible problems that are more than a decade away, policymakers need to focus on addressing the decades-long crisis of wage stagnation by creating good jobs and supporting wage growth,” said Mishel. “Our experience with manufacturing in recent decades shows that education and training will not help displaced workers unless there are good jobs available. We should stop worrying about the coming robot apocalypse and start working to raise wages, strengthen collective bargaining and labor standards, and make sure that all jobs are good jobs.”