Robots and Work

Robots, capable of performing both manual and cognitive tasks autonomously, have been integrated in various industries. The increasing adoption of robots has generated concern about the loss of jobs and skills. Our paper addresses these concerns by studying the introduction of robots in US manufacturing plants during the period 2010-2022. 

Robots transform the workplace through three main channels: substitution, complementarity, and productivity. Robots substitute for certain tasks, especially routine, precision-intensive, or hazardous ones, thereby reducing the demand for labor. At the same time, their introduction generates new tasks, such as designing robotized workplaces, installing and programming robots, and supervising, maintaining, or “babysitting” them. These tasks create demand for additional engineers, technicians, and operators, or require existing workers to upgrade their skills. This is the channel of complementarity. Finally, robot adoption raises productivity: Adopting plants produce not only more but also higher-quality goods, which boosts sales at the expense of outcompeted non-adopters. Furthermore, robots are commonly introduced only in some stages of the production process in a plant, where increased productivity also increases production and employment in other stages of the production process.

A novel approach: plant-level analysis

Early research on industrial robots relied mainly on industry-level variation in exposure to automation and generally found negative effects on manufacturing employment. More recent research shifts the focus inside firms and shows a more nuanced picture: when firms adopt robots, productivity can increase, tasks are reallocated, and robots often complement human labor. As a result, employment can grow within adopting firms even when overall employment in the industry or economy is falling.

However, most manufacturing firms operate multiple plants, and robot adoption typically occurs in only a subset of them. Firm-level studies therefore combine the effects of adoption in some plants with any spillovers to non-adoption plants within the same firm, potentially offsetting positive and negative effects. Plant-level analysis is crucial to isolate the direct effects of robot adoption where it actually occurs.

In our paper, we identify robot adoption at the establishment level in U.S. manufacturing from the late 1990s to the early 2020s and show that adopting plants increase both employment and production. The increase in production exceeds that in employment, consistent with productivity gains. Employment rises across skill groups in production as well as in support functions.

Robot adoption increases employment in adopting plants across all occupations and skill-levels

The gains in job postings extend across all occupations—production and support roles alike—and span the entire skill spectrum, from low- to high-skilled work. Even occupations directly affected by automation, such as welders, painters, and packagers, experience increases in job postings. For example, the General Electric plant in Norwich, NY, advertised 169 job openings between 2014 and 2022. Of these, five were specifically for robotic welding positions (SOC code 51-4122), beginning in 2017. The first posting sought a welder capable of setting up a welding robot. The trend continued, with 33 additional welder positions posted from 2017 onward. Before adoption, only 20% of non-robotic welding job postings required troubleshooting skills; after adoption, more than half did. This example illustrates both the growth of job postings in directly affected occupations and how robot adoption reshapes skill requirements, driving demand for complementary capabilities and supporting overall productivity growth. Expanding plant production also raises demand for support functions, though the relative increase is smaller than that observed for production workers. 

Positive spillovers of robots from adopting to non-adopting plants

Robots enhance competitiveness through productivity and quality enhancement, resulting in greater output that increases demand for employees in the non-robotic parts of a plant and in some upstream and downstream plants that belong to the same firm. In multi-plant firms in which some plants adopt robots but others do not, demand for labor in non-adopting plants rises but much less than in the adopting plants. This is evidence for positive spillover effects at the firm level, with the spillover effect being much smaller than the direct effect at the plant level.

Co-Authors:

• Adrianto, Ministry of Finance, Government of Indonesia
• Avner Ben-Ner, Professor, Carlson School of Management, University of Minnesota