Man vs Machine

By Daron Acemoglu

Posted on June 29, 2017


The decline in the share of labor in national income, the slow growth of US employment and stagnant real wages have increased the concerns that the wave of new technologies, particularly automation technologies, artificial intelligence and robotics, are making labor increasingly redundant.


Image: Protesters outside automated factorySuch concerns are not new, however. Keynes articulated similar fears in 1930, introducing the notion of technological unemployment as an ill that would be a byproduct of future economic growth. Another giant of early 20th-century economics, Wassily Leontief, was equally pessimistic about the implications of new machines. By drawing an analogy with the technologies of the early 20th century that made horses redundant, he speculated in 1952 that “Labor will become less and less important...More and more workers will be replaced by machines. I do not see that new industries can employ everybody who wants a job.”


Yet, these fears did not come to pass. Could this time be different? Perhaps. But in order to understand whether it is or not, and what can be done about these developments, we need a conceptual framework to shed light on why past episodes of technologies replacing labor in a range of tasks did not lead to technological unemployment, and on the contrary, were based on typically accompanied by wage and employment growth.


One obvious reason for this is that during many episodes of major technological changes, we witness not only the replacement of labor by capital in certain tasks, but also the creation of new industries, occupations and tasks. This is illustrated by technological and organizational changes during the Second Industrial Revolution, which brought about the replacement of the stagecoach by the railroad, sailboats by steamboats, and of manual dock workers by cranes, but simultaneously, there was also another process of technological changes which led to the creation of new labor-intensive tasks. These new tasks generated jobs for a new class of engineers, machinists, repairmen and conductors, as well as modern managers and financiers involved with the introduction and operation of new technologies.


Today, industrial robots, digital technologies and computer-controlled machines are indeed replacing labor and leading to lower wages and employment. Yet simultaneously we are also witnessing the emergence of new tasks ranging from engineering and programming functions to those performed by audio-visual specialists, executive assistants, data administrators and analysts, meeting planners or computer support specialists. Indeed, during the last 30 years, new tasks and new job titles account for a large fraction of US employment growth.


To document this fact, consider data from Lin (2011) that measures the share of new job titles in which workers perform newer tasks than those employed in more traditional jobs within each occupation. In 2000, about 70% of the workers employed as computer software developers (an occupation employing one million people at the time) held new job titles. Figure 2 shows that for each decade since 1980, employment growth has been greater in occupations with more new job titles. The regression line shows that occupations with 10 percentage points more new job titles at the beginning of each decade grow 5.05% faster over the next 10 years. Similarly, in 1990 a radiology technician and in 1980 a management analyst were new job titles. From 1980 to 2007, total employment in the US grew by 17.5%. About half (8.84%) of this growth is explained by the additional employment growth in occupations with new job titles relative to a benchmark category with no new job titles.

Image: Chart of new job titles

A new working paper by Acemoglu and Restrepo (2016), titled ‘The Race Between Machine and Man: Implications of Technology for Growth, Factor Shares and Employment’, develops a conceptual framework for understanding how different types of technological change coexist and impact the fortunes of labor. Some, like robotics and automation, replace labor in existing tasks, reducing employment, labor share and potentially also wages. Others, like the creation of new tasks and occupations, increase the demand for labor, raising employment, wages and labor share. It is the balance between these two types of technologies that determines the future prospects for labor.


More formally, in the model economy, there are two types of technological changes: the automation of existing tasks and the creation of new complex tasks in which labor has a comparative advantage. Our static model provides a rich but tractable framework to study how automation and the creation of new complex tasks impact factor prices, factor shares in national income and employment. Automation allows firms to produce tasks previously performed by labor with capital, while the creation of new complex tasks allow firms to replace old tasks by new variants in which labor has a higher productivity. In contrast to the more commonly-used models featuring factor-augmenting technologies, here automation always reduces the share of labor in national income and employment, and may even reduce wages. Conversely, the creation of new complex tasks always increases wages, employment and the share of labor, and may even reduce the rate of return to capital. Critically, this framework implies that when the creation of new complex tasks keeps up with (or is even faster than) the process of automation, employment and wages will increase even as some workers are being replaced by machinery and new technology. In contrast, when automation runs ahead of the process of creation of new, labor-intensive tasks, technological change will bring lower employment, lower share of labor in national income and also potentially lower wages.


But what determines the pace of these two different types of technological changes? Does the fact that we are seeing more rapid replacement of labor in existing tasks imply that the future is bleak for labor? Or are there powerful self-correcting forces in the economy that could restore some of the lost ground for labor?


To provide a theoretical perspective on these questions, the basic framework outlined above is then embedded in a dynamic setting in which the direction of technological change is endogenous. This, in particular, implies that depending on the profitability of different types of technologies, firms are the ones that invest and develop these technologies. In this setting, Acemoglu and Restrepo (2016) identify a potentially important theoretical effect: If automation runs ahead of the creation of new complex tasks, market forces induce a slowdown of subsequent automation and countervailing advances in the creation of new complex tasks. As a result, there are new economic forces that may restore, in the long run, the share of labor in national income and employment back to their initial levels. The economics of these self-correcting forces are instructive and highlight a crucial new force: a wave of automation pushes down the effective cost of producing with labor. When technology is endogenous, this discourages further efforts to automate additional tasks and pushes the economy to redirect its research efforts towards the creation of new (labor-intensive) tasks.


This stability of balanced growth path implies that periods in which automation runs ahead of the creation of new complex tasks tend to self-correct. Contrary to the increasingly widespread concerns discussed above, this framework thus raises the (theoretical) possibility that rapid automation need not signal the demise of labor, but might simply be a prelude to a phase of new technologies favoring labor. In addition, it clarifies the long-run implications of different types of technological shocks. For example, if a wave of automation is triggered by a change in the innovation possibilities frontier (that is, in the technology for creating new technologies) that makes it easier to automate tasks, the economy will settle in a new balanced growth path with a greater share of tasks performed by capital, lower employment and lower labor share.


Overall, the central new insight of this framework is the presence of self-correcting forces, which help restore some of the lost ground for labor because labor becomes cheaper as a result of automation, making the creation of new labor-intensive tasks more profitable. These forces, do not, however, imply that the future is necessarily bright for labor. First, as already noted, these forces might take the economy to a new balanced growth equilibrium (rather than the one we started with) if new technologies also make the creation of further new technologies replacing labor cheaper. In this case, labor will not permanently disappear as a major factor of production, but the future level of employment and labor share may be lower in the future than the past. Second, these economic forces do not imply that the balance between the two types of technologies is efficient. In particular, to the extent that labor gets paid above its opportunity cost (eg, the value of leisure), firms will have a stronger incentive to adopt automation technologies than what a social planner wishing to maximize output would do. This suggests that policies that affect the composition of new technologies might be welfare-improving.


Ultimately, some of the questions posed in this white paper are not just conceptual but also empirical. Only future empirical work can fully inform us about the extent to which new technologies can create sufficient employment opportunities to make up for those lost to automation and robotics. Nevertheless, the conceptual framework outlined here is an important input in understanding how different economic forces might interplay in the future and what types of evidence we should look for.

Image: Robots


The preceding is republished on TAP with permission by its author, Professor Daron Acemoglu and the Toulouse Network for Information Technology (TNIT). “Man vs Machine” was originally published in TNIT’s May 2017 newsletter.


Read more from Professor Daron Acemoglu on Artificial Intelligence and the Labor Market: