Principal investigator: Viviane Pires Ribeiro
Paper Title: Robots and Jobs: Evidence from US Labor Markets
Authors: Daron Acemoglu and Pascual Restrepo
Location of the Intervention: United States
Sample Size: Industrial workers in the United States
Main theme: Labor market
Main Variable of Interest: JOBS
Type of InterventionAdoption of robots in industries
Methodology: Econometric model
The proliferation of robots, artificial intelligence, and other automation technologies has raised concerns about the future of jobs and wages. In this context, Acemoglu and Restrepo (2020) sought to analyze the effect of the increased use of industrial robots between 1990 and 2007 on the United States labor market. Using a model in which robots compete against workers, the study's results show that advances in robotics technology can reduce employment and wages. According to estimates, one additional robot for every thousand workers reduces the employment-to-population ratio by about 0,2 percentage points and wages by 0,42%.
Evaluation Context
Robotics technology advanced significantly in the 1990s and 2000s, leading to a fourfold increase in the stock of industrial robots in the United States and Western Europe between 1993 and 2007. In the United States, the increase was one new robot per thousand workers, and in Western Europe it was 1,6 new robots per thousand workers. The automotive industry employs 38% of existing robots, followed by the electronics industry (15%), plastics and chemicals (10%), and metal products (7%).
As robots and automation technologies take over tasks previously performed by human labor, there is growing concern about the future of jobs and wages. Even so, there has been relatively little research on the balancing effects of new automation technologies and, in particular, robots.
Intervention Details
Despite concerns about the spread of robots, artificial intelligence, and other automation technologies, there is little systematic evidence on the equilibrium impact of automation technologies, and especially robots, on employment and wages. Aiming to corroborate the literature addressing this topic, Acemoglu and Restrepo (2020) investigate the effects of industrial robots on local labor markets in the United States. According to the authors, robots and automation technologies, in general, displace workers from the tasks they previously performed and, therefore, should have very different effects on the labor market than capital deepening (savings used to increase the capital-labor ratio) and other types of technological change.
In the study, the authors consider industrial robots as fully autonomous machines that do not require a human operator and can be programmed to perform various manual tasks such as welding, painting, assembly, material handling, and packaging. Textile looms, elevators, cranes, or conveyor belts are not robots because they have a single purpose, cannot be reprogrammed to perform other tasks, and/or require a human operator. Thus, it excludes other types of equipment and allows for 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.
To conduct the analysis, data on industrial employment, total wages, value added, labor force participation, and capital were used. Employment and total wage data are from County Business Patterns (CBP) and NBER-CES. Data on value added and labor force participation are from the Bureau of Economic Analysis, and data on information technology capital and overall capital stock are from the Bureau of Labor Statistics (BLS).
Methodology Details
The empirical approach was based on a model in which robots and workers compete in the production of different tasks. The model is based on other studies, but allows for the proportion of tasks performed by robots to vary between sectors and for trade between labor markets specialized in different sectors. As specified by the model, greater adoption of robots negatively affects wages and employment due to a displacement effect (since they directly displace workers from tasks they previously performed), but there is also a positive productivity effect (as other industries and/or tasks increase their demand for labor). Due to the displacement effect, robots may have very different implications for labor demand than capital deepening or factor-augmentation technologies. Furthermore, the effects of robots on employment and wages can be estimated by regressing the change in these variables on robot exposure. Robot exposure is constructed from the interaction between the actions of the baseline industry in a local labor market and the technological possibilities for the introduction of robots across industries.
First, the authors document that there is considerable variation in robot adoption across sectors and show that the same sectors are rapidly adopting robots in both the United States and Europe. Furthermore, they show that, at the industry level, there is no strong positive correlation between robot adoption and other important trends affecting local US labor markets, such as import competition from China, competition from Mexico, information technology capital investment, and capital deepening.
Results
The results of the estimates indicate a negative relationship between a commuting zone's exposure to robots and its labor market outcomes after 1990. Between 1990 and 2007, the increase in the robot stock (approximately one additional robot per thousand workers from 1993 to 2007) reduced the average employment-to-population ratio in a commuting zone by 0,39 percentage points and average wages by 0,77% (compared to a commuting zone without robot exposure). These figures are substantial, but not implausible. For example, they imply that one more robot in a commuting zone reduces employment by about 6 workers; this estimate includes both direct and indirect effects, the latter caused by a drop in demand for non-tradable goods due to reduced employment and wages in the local economy.
The increased use of robots in a displacement zone generates benefits for the rest of the U.S. economy, reducing the prices of tradable goods produced with robots and creating shared capital gains. Estimates imply that one additional robot for every thousand workers reduces the aggregate employment-to-population rate by about 0,2 percentage points and wages by about 0,42% (compared to their larger local effects of 0,39 percentage points and 0,77%, respectively). Or, equivalently, one new robot reduces employment by about 3,3 workers.
The authors verified whether the measure of exposure to robots is unrelated to previous trends in employment and wages from 1970 to 1990, the period preceding the onset of rapid advances in robotics technology. Several robustness checks reinforce the results obtained. First, the results are robust to include differential trends by various baseline characteristics, linear switching zone trends, and controls for other changes affecting demand or productivity in various sectors. Second, the tests show that the automotive industry, which is the most robot-intensive sector, is not driving the results. Third, consistent with the theoretical emphasis that robots (and, more generally, automation technologies) have very different effects on the labor market than other types of machinery and capital deepening, no similar negative impact of capital, other information technology measures, or overall productivity increases was found.
The effects on employment caused by the use of robots are most pronounced in manufacturing and, in particular, in industries most exposed to robots. They are also concentrated in routine manual activities, assembly, and related occupations. Negative effects have also been estimated in construction, retail, and personal services.
Lessons in Public Policy
Acemoglu and Restrepo (2020) argue that there are relatively few robots in the US economy, so the number of jobs lost due to robots has been limited so far (a 0,2 percentage point drop in the employment-to-population ratio, or about 400.000 jobs). However, if robotic technology continues as expected by experts over the next two decades, the future aggregate implications of robots could be greater. Thus, it is crucial that any extrapolation about the future effects of robots should recognize not only the usual uncertainty associated with their use, but also the possibility that some of the general equilibrium effects that operate through the technology may emerge only slowly, and that the employment and wage response may be different once robots become sufficiently widespread.
In this sense, the authors highlight that the conceptual framework adopted in the study indicates that, in contrast to the prevailing assumption in economic discussions, automation and non-automation technologies have distinct and different impacts, and therefore, technological waves of automation can have different impacts depending on the balance between displacement and productivity effects. Thus, the next decade will likely witness major advances in artificial intelligence, machine learning, communication technologies, and new manufacturing technologies, including augmented reality and... Design modular.
References
ACEMOGLU, Daron; RESTREPO, Pascual. Robots and jobs: Evidence from US labor markets. Journal of Political Economy, v. 128, no. 6, p. 2188-2244, 2020.
