Robots are replacing humans in routine tasks and, more recently, in nonroutine cognitive tasks. Meanwhile, studies have documented a drop in the labor share in national income in multiple countries since the early 1980s, including one study that shows a decline of 5 percentage points, a significant shift by historical standards.1 The declining labor share is widely attributed to the rise in automation, which is being aided by advances in artificial intelligence, machine learning, and dexterous automation.

Few studies have quantified automation's effect on labor's share in income. To help fill this gap in the literature, Hong Cheng, Lukasz Drozd, Rahul Giri, Mathieu Taschereau-Dumouchel, and Junjie Xia, in their paper, "The Future of Labor: Automation and the Labor Share in the Second Machine Age," examine how the expansion of industrial robots in manufacturing firms in China has impacted the contribution of labor.

China is a useful case study because it has been one of the fastest adopters of industrial robots in recent years. Its density of robots used in production increased from only 25 units per 10,000 workers in 2013 to 187 units in 2019.2 The Chinese government may have had a role in the acceleration of robotics adoption through its large industrialization program, Made in China 2025, which, among other things, promotes high-tech labor-saving automation technologies by providing subsidies to firms that purchase industrial robots and related equipment. These sizable subsidies effectively discount the price that firms pay for automation equipment.

To quantify the relationship between the price of automation capital and the labor share in income, the authors had to disentangle other factors that can influence the usage of automation equipment (such as demand shocks and labor prices in other countries). They use a novel approach to overcome this challenge: They incorporate into an existing model of automation the observed variation in the price of automation capital that results from the nonuniform subsidies firms have received since 2015 under Made in China 2025.3 These subsidies vary by city, industry, and firm because they are implemented by local municipalities that have a large degree of discretion in setting subsidy rates. Given the varied subsidies and timing of implementation, the authors are able to identify automation's causal impact — based on the variation in the price of automation — on labor shares within firms.

Their data, which come from a large survey of manufacturing companies across China, provide detailed information on firm-level manufacturing operations from 2015 to 2017, including how much each automating firm spent on industrial robots and how much each received as government subsidies.4 The authors estimate the key parameter responsible for the relationship in the model — that is, they calculate a measure of how the share of labor used in automated manufacturing firms changes in response to the price of automation.

The authors find that a reduction in the price of automation equipment has a large negative impact on labor shares among firms that automate in China. By contrast, they find no substantial negative impact on the labor share associated with variations in the cost of capital unrelated to automation equipment. They estimate that automation capital and labor are substitutes for one another across firms that use automation technologies, and a decline in the cost of automation technologies induces firms to use more automation capital and less labor. 

If (quality-adjusted) prices for robotic equipment continue to decline, the authors argue, robots will displace many more Chinese workers in automating firms and much more labor income will be lost. They note, however, that what prevents automation from having a nearly apocalyptic impact on demand for labor and labor income is that not as many firms automate at present. But that fraction may increase with future progress in artificial intelligence.

Although they don't test their results empirically using U.S. data, their findings shed light on the relationship between workers and automated firms in the U.S. The decline in the quality-adjusted price of automation in the U.S. has been well documented, and the labor share in the manufacturing sector has fallen from about 60 percent in 1960 to 40 percent in the 2000s.5 The evidence documented in their paper establishes a connection between the decline in the labor share in U.S. manufacturing and automation.

In summary, the research by Cheng et al. broadens the literature by measuring and evaluating the impact that modern robotic technology has on labor's income share at the firm level. The effects they document are large, and if the scope of automation broadens as expected, many jobs will need to be created to offset the adverse impact of automation on labor’s share in national income and on income inequality.6

  1. L. Karabarbounis and B. Neiman, "The Global Decline of the Labor Share," Quarterly Journal of Economics, 129 (2013), pp. 61–103. Updated evidence can be found in M.C. Dao, M.M. Das, Z. Koczan, and W. Lian, "Why Is Labor Receiving a Smaller Share of Global Income? Theory and Empirical Evidence," International Monetary Fund (2017).
  2. International Federation of Robotics (World Robotics Reports), Statistica.com.
  3. The existing model used is along the lines of (i) G. Graetz and G. Michaels, "Robots at Work," Review of Economics and Statistics, 100:5 (2018), pp. 753–768, and (ii) D. Acemoglu and P. Restrepo, "The Race Between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment," American Economic Review, 108 (2018), pp. 1488–1542.
  4. The China Enterprise General Survey was undertaken by Wuhan University Institute of Quality Development Strategy and designed in conjunction with researchers from the Hong Kong University of Science and Technology, Stanford University, and the Chinese Academy of Social Sciences.
  5. M. Kehrig and N. Vincent, "The Micro-Level Anatomy of the Aggregate Labor Share Decline," National Bureau of Economic Research Working Paper 25273 (2018). 
  6. One-quarter of the U.S. labor force will be highly exposed to automation in the coming decades according to M. Muro, R. Maxim, and J. Whiton, "Automation and Artificial Intelligence: How Machines Are Affecting People and Places," Metropolitan Policy Program at Brookings (2019).