There\’s a well-worn conversation about the relationship between new technology and possible job displacement which goes something like this:
Concerned person: \”New developments in information technology and artificial intelligence are going to threaten lots of jobs.\”
Skeptical person: \”Economies in developed countries have been experiencing extraordinary developments and shifts in new technology for literally a couple of centuries. But as old jobs have been dislocated, new jobs have been created.\”
Concerned person: \”This time seems different.\”
Skeptical person: \”Every time is different in the specific details. But there\’s certainly no downward pattern in the number of jobs in the last two centuries, or the last few decades.\”
Concerned person: \”Still, the way in which information technology and artificial intelligence replace workers seems different than the way in which, say, assembly lines replaced skilled artisan workers or combine harvesters replaced farm workers. \”
Skeptical person: \”Maybe this time will be different. After all, it\’s logically impossible to prove that something in the future will NOT be different. But based on the long-run historical pattern, the evidence that new technology leads to shifts in the labor market is clear-cut, while the evidence that it leads to permanent job loss for the population as a whole is nonexistent.\”
Concerned person: \”Still, this current wave of technology seems different.\”
Skeptical person: \”I guess we\’ll see how it unfolds in the next decade or two.\”
The most recent Spring 2019 issue of the Journal of Economic Perspectives has a symposium on \”Automation and Employment.\” Two of the articles in particular offer a concrete arguments about how something is different with how the current new technologies are interacting with labor markets.
Daron Acemoglu and Pascual Restrepo discuss \”Automation and New Tasks: How Technology Displaces and Reinstates Labor.\” They suggest a framework in which automation can have three possible effects on the tasks that are involved in doing a job: a displacement effect, when automation replaces a task previously done by a worker; a productivity effect in which the higher productivity from automation taking over certain tasks leads to more buying power in the economy, creating jobs in other sectors; and a reinstatement effect, when new technology reshuffles the production process in a way that leads to new tasks that will be done by labor.
In this approach, the effect of automation on labor is not predestined to be good, bad, or neutral. It depends on how these three factors interact. Acemoglu and Restrepo attempt to calculate the size of these three factors for the US economy in two time periods: 1947-1987 and 1987-2017. There is of course considerable technological change through all of this 60-year period. For example, I\’ve written on this blog about \”Automation and Job Loss: The Fears of 1964\” (December 1, 2014) and \”Automation and Job Loss: Leontief in 1982\” (August 22, 2016). But the later period can be associated more closely with the rise of computers and information technology.
Their calculations suggest that in the 1987-2017 period, the effects of automation have involved a larger displacement effect, lower productivity growth, and a lower reinstatement effect. The lower demand for labor can be seen in stagnant wage growth over this period for lower- and medium-skilled workers. They argue that the real issue isn\’t whether automation displaces tasks and alters jobs–or course it does–but rather how those displacement effects compare to how automation leads to greater productivity the possibility of new job-related tasks that reinstate labor. They argue that public policy has some power to affect how the forward movement of technology will affect demand for labor: for example, they argue that public policy has tended to favor investment in new equipment and machinery over investment in human capital, like on-the-job training by employers.
Another angle on new technology and labor markets in the same issue of JEP comes from Jeremy Atack, Robert A. Margo, and Paul W. Rhode in \”`Automation\’ of Manufacturing in the Late Nineteenth Century: The Hand and Machine Labor Study.\” The focus of their paper is on a remarkably detailed US government study done in the 1890s of how machines were replacing the tasks involved in specific jobs.
The new assembly line machines of that time clearly displaced large number of tasks previously done by workers. However, the productivity effects of this wave of automation were very large. In addition, the new automation technology of that time had a powerful reinstatement effect of creating new tasks to be done by workers. They write:
[T]he net effect of the introduction of new tasks on labor demand appears to have been positive. This is because the share of time taken up by new tasks in machine labor was larger than the share of time associated with hand tasks that were abandoned—indeed, five times larger. Among other activities, these new tasks included maintenance of steam engines, a foreman supervising large numbers of workers (discussed further below), and workers packaging products for distant markets.
Atack, Margo, and Rhode also offer a broader point about technology and labor that seems to me worth considering. They point out that back in the 1890s, with a much heavier use of machines in the production process, there was a shift toward a broader division of labor: that is, the study counted more overall tasks to be done when machines were used, as compared to before the machines were used. One implication for workers of that time is that the path to a steady and well-paid job was to focus on a very particular niche of the production process. Indeed, one broad description of labor markets at this time is that there is a shift away from artisan workers (say, blacksmiths) who carried out many tasks, and toward workers who focused on a smaller set of tasks.
The massive division of labor documented front and center in the Hand and Machine Labor study dramatically affected the nature of the human capital investment decision facing successive cohorts of American workers contemplating whether to enter the manufacturing sector. Earlier in the nineteenth century, the human capital investment problem such workers faced was mastering the diverse set of skills associated with most or all of the tasks involved in making a product, along with managing the affairs of a (very) small business, an artisan shop. The human capital investment problem facing the prospective manufacturing worker in the 1890s was quite different. There was little or no need to learn how to fashion a product from start to finish; mastery of one or two tasks
would do, and such mastery might be gained quickly on the job. The more able or ambitious might gravitate to learning new skills, such as designing, maintaining, or repairing steam engines, or clerical/managerial tasks, the demand for which had grown sharply as average establishment size increased over the century.
For many decades in the twentieth century, specialization was economically beneficial to workers—the costs of learning skills were relatively modest and the return on the investment—a relatively secure, highly paid job in manufacturing—made that investment worthwhile. The prospect of widespread automation has arguably changed this calculus. No single “job” is safe and the optimal investment strategy may be very different—a suite of diverse, relatively uncorrelated skills as insurance against displacement by robotics and artificial intelligence. This is perhaps the sense in which the history of how technology affects jobs is not repeating itself, and “this time” really is different.
In watching the cohort that includes my own children move from high school into young adulthood, this observation seems to me to contain a lot of truth. When it comes to training for a future job, many of us are still mentally in the 1890s, looking for one or a few particular focused skills that will guarantee a \”good job.\” But modern technologies are likely to disrupt what tasks are actually done in a very wide array of jobs, which will put a premium on workers with the ability to shift flexibly as the job situation is reshaped.