ROBOTS AND JOBS: NATIONAL BUREAU OF ECONOMIC …

1 NBER WORKING PAPER SERIES. ROBOTS AND JOBS: EVIDENCE FROM US LABOR MARKETS. Daron Acemoglu Pascual Restrepo Working Paper 23285. NATIONAL BUREAU OF ECONOMIC RESEARCH. 1050 Massachusetts Avenue Cambridge, MA 02138. March 2017. We thank David Autor, Lorenzo Caliendo, Amy Finkelstein, Matthew Gentzkow and participants at various seminars and conferences for comments and suggestions; Joonas Tuhkuri for outstanding research assistance; and the Institute for Digital Economics and the Toulouse Network of Information Technology for financial support. The views expressed herein are those of the authors and do not necessarily reflect the views of the NATIONAL BUREAU of ECONOMIC Research.

2 NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications. 2017 by Daron Acemoglu and Pascual Restrepo. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including notice, is given to the source. ROBOTS and Jobs: Evidence from US Labor Markets Daron Acemoglu and Pascual Restrepo NBER Working Paper No. 23285. March 2017. JEL No. J23,J24. ABSTRACT. As ROBOTS and other computer-assisted technologies take over tasks previously performed by labor, there is increasing concern about the future of jobs and wages.

3 We analyze the effect of the increase in industrial robot usage between 1990 and 2007 on US local labor markets. Using a model in which ROBOTS compete against human labor in the production of different tasks, we show that ROBOTS may reduce employment and wages, and that the local labor market effects of ROBOTS can be estimated by regressing the change in employment and wages on the exposure to ROBOTS in each local labor market defined from the NATIONAL penetration of ROBOTS into each industry and the local distribution of employment across industries. Using this approach, we estimate large and robust negative effects of ROBOTS on employment and wages across commuting zones.

4 We bolster this evidence by showing that the commuting zones most exposed to ROBOTS in the post-1990 era do not exhibit any differential trends before 1990. The impact of ROBOTS is distinct from the impact of imports from China and Mexico, the decline of routine jobs, offshoring, other types of IT capital, and the total capital stock (in fact, exposure to ROBOTS is only weakly correlated with these other variables). According to our estimates, one more robot per thousand workers reduces the employment to population ratio by about percentage points and wages by percent. Daron Acemoglu Department of Economics, E52-446.

5 MIT. 77 Massachusetts Avenue Cambridge, MA 02139. and CIFAR. and also NBER. Pascual Restrepo Department of Economics Boston University 270 Bay State Rd Boston, MA 02215. and Cowles Foundation, Yale 1 Introduction In 1930, John Maynard Keynes famously predicted the rapid technological progress of the next 90 years, but also conjectured that We are being afflicted with a new disease of which some readers may not have heard the name, but of which they will hear a great deal in the years to come namely, technological unemployment (Keynes, 1930). More than two decades later, Wassily Leontief would foretell similar problems for workers writing Labor will become less and less important.

6 More and more workers will be replaced by machines. I do not see that new industries can employ everybody who wants a job (Leontief, 1952). Though these predictions did not come to pass in the decades that followed, there is renewed concern that with the striking advances in automation, robotics, and artificial intelligence, we are on the verge of seeing them realized ( , Brynjolfsson and McAfee, 2012; Ford, 2016). The mounting evidence that the automation of a range of low-skill and medium-skill occupations has contributed to wage inequality and employment polarization ( , Autor, Levy and Murnane, 2003; Goos and Manning, 2007; Michaels, Natraj and Van Reenen, 2014) adds to these worries.

7 These concerns notwithstanding, we have little systematic evidence of the equilibrium im- pact of these new technologies, and especially of ROBOTS , on employment and wages. One line of research investigates how feasible it is to automate existing jobs given current and presumed technological advances. Based on the tasks that workers perform, Frey and Osborne (2013), for instance, classify 702 occupations by how susceptible they are to automation. They conclude that over the next two decades, 47 percent of US workers are at risk of automation. Using a related methodology, McKinsey puts the same number at 45 percent, while the World Bank estimates that 57 percent of jobs in the OECD could be automated over the next two decades (World Development Report, 2016).

8 Even if these studies were on target on what is technolog- ically feasible,1 these numbers do not correspond to the equilibrium impact of automation on employment and wages. First, even if the presumed technological advances materialize, there is no guarantee that firms would choose to automate; that would depend on the costs of sub- stituting machines for labor and how much wages change in response to this threat. Second, the labor market impacts of new technologies depend not only on where they hit but also on the adjustment in other parts of the economy. For example, other sectors and occupations might expand to soak up the labor freed from the tasks that are now performed by machines, and productivity improvements due to new machines may even expand employment in affected 1.

9 Arntz, Gregory, and Zierahn (2016) argue that within an occupation, many workers specialize in tasks that cannot be automated easily, and that once this is taken into account, only about 9 percent of jobs in the OECD. are at risk. 1. industries (Acemoglu and Restrepo, 2016). In this paper we move beyond these feasibility studies and estimate the equilibrium impact of one type of automation technology, industrial ROBOTS , on local US labor markets. The Inter- NATIONAL Federation of Robotics IFR for short defines an industrial robot as an automati- cally controlled, reprogrammable, and multipurpose [machine] (IFR, 2014).

10 That is, industrial ROBOTS are fully autonomous machines that do not need a human operator and that can be programmed to perform several manual tasks such as welding, painting, assembling, handling materials, or packaging. Textile looms, elevators, cranes, transportation bands or coffee makers are not industrial ROBOTS as they have a unique purpose, cannot be reprogrammed to perform other tasks, and/or require a human Although this definition excludes other types of capital that may also replace labor (most notably software and other machines), it enables an internationally and temporally comparable measurement of a class of technologies industrial ROBOTS that are capable of replacing human labor in a range of tasks.