Robots are job creators, for better or worse
A comprehensive analysis of a decade's worth of robot investments finds that robots create more jobs than they eliminate — but that may be the bad news
One of the most important and contentious issues facing business and social leaders today is the impact that new technologies will have on labor and employment. As AI and robotics merge in novel ways to take on ever-more complex human tasks, there have been widespread reports of the devastation that these new technologies will have on human labor. For example, a 2017 study predicted that up to 47% of all jobs in the United States could be displaced by robots. OECD researchers went even further in 2016, predicting that up to 70% of tasks performed by labor could be automated. Studies consistently predict that somewhere around 50% of current jobs could be eliminated by the arrival of new robots with higher AI-enabled capabilities.
For the most part, this bleak outlook is accepted in the general business press, where stories of robot-driven unemployment and income inequality are impossible to miss. These reports typically end with a dire warning to policymakers to start thinking about what to do with all those displaced workers. They have even led to efforts such as the Universal Basic Income movement to protect those displaced by these new technologies. On the surface, all of these books and articles make a pretty compelling case, but recent research by Jay Dixon (Statistics Canada), Bryan Hong (NYU), and Lynn Wu (Wharton) provide a somewhat less pessimistic and more nuanced view of this issue.
Their report is based on analysis of over ten years of detailed investment and labor data found in various indices maintained by Statistics Canada. As the authors note, "this study provides the most comprehensive evidence at the level of individual businesses on employment and organizational effects of robot investments." The study is important research not just for its breadth and depth but also for its consideration of a key issue: previous wide-scale general-purpose technology adoptions have tended to increase overall job availability. This was the case with the arrival of computers and the internet, for example. Thus, robots would have to be a different kind of technology, if they negatively affect employment. It's this new and interesting question, typically ignored in discussions about robots, that the researchers wanted to examine.
After looking at all the data, the authors' first surprising conclusion is that most firms did not invest in robots to simply shed workers or cut costs. They made these investments to increase quality and productivity. Two features of robots drove this trend. First, robots can, of course, "follow complex multi-step protocols with a high degree of precision, which can greatly increase consistency and efficiency in production." In warehouses, for example, "robots have automated a wide range of warehousing logistics activities by effectively transporting objects between locations without human intervention." Second, robots "have much greater manual dexterity in performing certain tasks relative to humans" and "can be trained to perform specific types of tasks effectively, and at times surpass human abilities." Thus, robots "can substantially reduce variance in production quality by minimizing human errors across a wide range of tasks."
The effect of deploying these kinds of machines was not typically to eliminate the jobs that had previously been done by humans. What really happened in Canada is a multi-dimensional shifting of labor and management priorities and roles. The most important of these impacts are summarized below.
1: Robots increased jobs that are complementary to robots
Returning to the warehouse example, the authors note that "redesign of warehousing logistics still requires humans to pack boxes before shipping since robot capabilities remain too limited to effectively manipulate the range of objects of various sizes and shapes." As a consequence, "the demand for low-skilled workers doing residual tasks such as those engaged in packing has increased." This finding is consistent with the anecdotal evidence coming out of companies such as Amazon. A major adopter of robots, the company is also the fastest-growing hirer of human labor (~1,400 per day in 2020). In other words, the more productive the robots get, the more people are needed alongside them to do the things robots cannot. Of course, this is not a judgment on the desirability of those generally low-skilled jobs, just a finding about their growth in parallel with the adoption of robots. Indeed, recent evidence suggests that humans can find working alongside robots challenging at best and dangerous at worst.
2: Robots decreased the need for managers
Because robots tend to increase productivity and quality, there is generally a lower need for managers whose traditional role was increasing productivity and ensuring quality. "As monitoring and supervision constitute a significant portion of managerial activities," note the authors, "the demand for managerial labor is likely to decrease after robot adoption as robots replace routine and predictable segments of the production process." This finding surprised the researchers, who expected high-skill managers to be more immune from displacement. They theorize, however, that "this reduction may be the consequence of both a decrease in the need for certain types of supervisory work from robot adoption as well as an indirect effect from the changing composition of non-managerial employees."
3: Managerial span of control increased, but authority decreased
Consistent with the findings of more low-skilled workers and fewer managers, the managers who did remain faced an increased span of control, both in terms of worker count and process reach. Put simply, a smaller number of managers must now oversee more workers, more robots, and more activities. Paradoxically, while managers oversaw more workers and work, their decision-making power decreased. This outcome, again, is to be expected since robots can handle more complex tasks that are typically not changed at the spur of the moment. In other words, managers are needed not to make complex decisions but to make sure that the robots are doing (along with their human support) their jobs.
4: Traning is more decentralized while technology selection is more centralized
Because managers are no longer the primary definers of tasks, workers receive training that is localized to and defined by, their robot-support role, which removes much of the training component of the managerial role. Conversely, because robot technologies are major investments, these decisions become centralized and taken as strategic initiatives by a firm's senior management. In this latter aspect, robots are different from past general-purpose technologies, the authors note — at least for the time being. Perhaps in the future, as happened with computers, robot costs will decrease, and individuals will have more autonomy over robot selection and purchase. That outcome may be decades away, however.
5: Employment and pay are linked more closely to individual performance
Consistent with the decrease in employee monitoring costs, the authors write, "we observe an increase in the adoption of performance pay linked to individual employee performance after robot adoption." Again, field evidence supports this finding, with robot-connected workers' employment and compensation more linked to their individual human actions and daily productivity than to team, or firm, metrics.
In presenting all of their conclusions, the authors are careful to add an important caveat:
...with robot adoption rapidly increasing in prevalence and capability, we expect that the allocation of decision authority, incentives and other complementary work practices are likely to continue to evolve. Those firms that can best match their capabilities and work practices to productive opportunities can benefit substantially from robot investments and develop potential competitive advantages, highlighting the need to understand the different types of complements to robots as a new technology.
In summary, a comprehensive analysis of the Canadian economy spanning over a decade of investment and hiring indicates that robots are associated with "increases in total employment in robot-adopting firms." Thus, they are "similar to past generations of general-purpose technologies that ultimately increased labor demand and provide a stark contrast to labor declines found in earlier empirical studies using data at industry or geographic region levels." That said, robots do change the composition of the increased workforce because of "decreases in employment of middle-skilled workers and increases in employment of low- and high-skilled labor."
As a new general-purpose technology, moreover, robots also "lead to significant changes in how firms organize production activities and manage their human capital." Moreover, firms find that "new sets of organizational practices will be necessary to effectively utilize the new technology." Indeed, the authors believe that just as AI led to a hiring boom in highly-skilled programmers and engineers, so too they "expect employment of high-skilled workers to also increase after robot adoption."
As with all good studies, this report raises as many questions as it answers. How were manager wages impacted, for example? What is the relationship between robot task complexity and specific worker profiles? Interestingly, the authors speculate that robots may reduce employment in firms that do not invest in these new technologies and thus find themselves unable to compete in productivity or quality — yet another area for further research.
Overall, the authors conclude that "the effect of robots on labor is more nuanced than earlier work has predicted." Thus, "this issue requires a deeper examination beyond more aggregated levels of analysis to understand how they are used to complement and substitute labor, as well as how organizational practices need to evolve around the changing nature of work." For executives, managers and policymakers, this important research suggests the need to go beyond the dramatic headlines to fully understand the complex ways in which the next generation of robots are changing human labor and management.
Dixon, Jay and Hong, Bryan and Wu, Lynn. The Robot Revolution: Managerial and Employment Consequences for Firms (June 3, 2020). NYU Stern School of Business. http://dx.doi.org/10.2139/ssrn.3422581