The Complicated Relationship Between Robots And Workers

Last week I attended the American Economic Association’s (AEA) annual conference in San Diego. The AEA event is the largest annual conference for economists, in which cutting edge research is presented and where Nobel laureates are congratulated for their achievements. I was the discussant for two papers on the use of robots in firms, and these papers highlight that the relationship between robots and workers is complicated.

Robots Create and Destroy Jobs in Manufacturing

A paper by Acemoglu, Lelarge and Restrepo studied the effects of robots on employment in French manufacturing firms. The authors found that firms that adopt robots experience an increase in employment. This finding is notable for a couple reasons. First, it runs counter to the popular narrative that “robots are coming for your job.” Second, the positive relationship between robot adoption and jobs at the firm level has been documented by researchers in other countries including Canada, Denmark and Spain. In other words, the positive relationship between robot adoption and jobs at the firm level appears to be quite robust, at least in advanced economies.

Yet the authors also find that manufacturing firms are likely to experience decreases in employment when their competitors adopt robots. Moreover, they find that, in aggregate, the negative effects on employment dominate the positive effects at the firm level: even as some manufacturing firms are growing and adding jobs (the firms adopting robots), even more manufacturing firms are shrinking and losing jobs. This same result has also been found using data from Spain, but I’d argue that more work is needed to assess its robustness.

This and other recent papers highlight that the relationship between robots and jobs is nuanced, at least in manufacturing settings in advanced economies. The paper also raises a number of questions: Why do some firms adopt robots and others do not (especially given that firms that do not adopt robots seem to suffer)? What happens to workers at firms that do not adopt robots? How many of them end up working at firms that adopt robots? When firms that adopt robots increase employment, what types of workers are hired, and what are their wages?

Robots Appear to Complement Workers in Some Service Industry Settings

Another paper by Eggleston, Lee and Iizuka studied the effects of robots on workers in Japanese nursing homes. The authors find some evidence that robots appear to complement workers in Japanese nursing homes, in particular by reducing turnover. The result makes sense—work in nursing homes can be physically demanding leading to high turnover, and the robots are intentionally designed to assist with the most physically demanding tasks (such as lifting patients out of beds).

As explained in the paper, an interesting feature of the Japanese setting is that many prefectures (subdivisions of the larger country) provided subsidies for nursing homes that adopted robots. The idea of a “robot subsidy” is the exact opposite of the idea of a “robot tax”—the subsidy aims to spur adoption of robots, whereas the tax aims to slow adoption of robots. In 2017, Bill Gates famously argued in favor of robot taxes. As illustrated by this paper, however, the experience in Japan suggests that there may be many settings where it would be counter-productive to tax the adoption of robots.

The Need for More Data and Research

Both these papers highlight that the effects of robots on firms and workers is nuanced—the effects might differ across countries, across industries, and even across firms in the same industry. Thus, the papers point to the need for more data and more studies on how firms use robots and other new technologies. Statistical agencies, in the United States and elsewhere, can play a major role in this respect.

Indeed, one of the more fascinating sessions at the conference described new data being produced by the U.S. Census Bureau on firm-level adoption of robots and other new technologies such as machine learning, computer vision, and autonomous-guided vehicles. This U.S. data will soon be available for researchers to study the impact of these technologies on workers, firms, communities and industries. It will be interesting to see if the results documented in France, Japan and other countries also hold in the United States.

More generally the results from the papers presented at the AEA annual meeting highlight the importance of funding statistical agencies to collect such data. It would be folly to try to develop policy around new technologies without having a firm grasp of how these technologies truly affect the economy.

Rob Seamans is an associate professor at the NYU Stern School of Business.