Why robots may not replace humans after all

Megan E. Doherty | Jul 07, 2017

Sections Economics

Economist John Maynard Keynes conjectured in 1930 that all human workers would be replaced by machines, leading to “technological unemployment.” This and similar prophesies have inspired a cottage industry of research examining what might happen in a robot-led future. Previous studies, including the 2016 World Bank’s World Development Report, have estimated that roughly half of all workers will be at risk of losing their jobs to automation over the next 20 years.

But MIT’s Daron Acemoglu and Boston University’s Pascual Restrepo analyze the effect of industrial robots and find a more modest effect on job loss.

Industrial robots (IRs) are fully autonomous, reprogrammable, multipurpose machines, such as those used in the automotive industry. Acemoglu and Restrepo analyzed IRs’ effect on 19 industries between 1990 and 2007. They used data from nine European economies as a proxy for the United States prior to 2004, which is when data on the degree of exposure to IRs within specific industries, beyond the aggregate total, became available in the US.

Their findings indicate that introducing one new robot per 1,000 workers reduced the overall employment‐to‐population ratio by between 0.18 and 0.34 percentage points and wages by 0.25 to 0.5 percent. This is equivalent to each robot causing between 3 and 5.6 workers to lose their jobs.

The findings suggest that the impact of robots is distinct from, and only weakly correlated with, potentially confounding variables.

The researchers caution that it can be hard to estimate these aggregate results because, while automation may displace workers, directly lowering employment opportunities and wages, other industries might hire those laborers. Furthermore, there could be productivity gains that may wind up expanding employment opportunities in the very industries in which IRs are used.

The findings suggest that the impact of robots is distinct from, and only weakly correlated with, potentially confounding variables such as imports, offshoring, the decline of routine jobs (such as information processing), and other computer technology. The researchers also determined that their results, in areas that saw great increases in robot usage, were not due to a preexisting downward trend in employment or wages prior to 1990.

The effects they observe were more pronounced among men, people with less than a college degree, workers in manual and blue-collar occupations, and those in industries that have adopted a greater share of robots.

With relatively few robots (the total is estimated to be between 1.5 and 1.75 million) in the US economy, the number of jobs lost due to robots so far is between 360,000 and 670,000, the researchers estimate. Even if the number of IRs across the world were to quadruple by 2025, it would lead to only a 0.94–1.76 percentage-point drop in the employment‐to‐population ratio, and to between 1.3 and 2.6 percent lower wage growth.