Simulating Endogenous Global Automation

Abstract

Comparing our baseline transition to a future without automation identifies winners and losers. Western Europe, Japan, Canada, the UK, and the U.S. experience the largest percentage increases in GDP. Compared to no-automation, output in 2050 is, respectively, 12.9, 12.7, 12.1, 10.0, and 5.1 percent larger. China defers automating until 2027. Automation makes slow-or non-automating regions, including India, Brazil, and Sub-Saharan Africa, worse off. This reflects the flow of capital from non-automating to automating regions. Mexico, for example, produces 3.6 percent less output in 2050 due to a 9.2 percent decline in the amount of capital that can profitably operate in that region.We also consider alternate scenarios where access to new technologies is restricted to the U.S., where automation technologies reduce labor input-shares of all skill groups proportionally, and where technological change is more rapid than historical experience. In virtually all cases, automation benefits the U.S. and raises the welfare of middle-and high-skilled American workers. As indicated, relative to no automation, U.S. GDP in 2050 increases by 5.1 percent in our baseline scenario. This figure is 8.3 percent if the new technology is exclusive to the U.S., 6.1 percent if the new technology is not skill-biased, and 11.9 percent if global technological change occurs at twice the historical rate.As suggested by stylized models (e.g. Benzell et al. 2016;Sachs et al. 2015a;Acemoglu and Restrepo 2018a;Peretto and Seater 2013), automation increases the world interest rate. However, this effect is relatively modest in our model and does not significantly deter countries from automating. Absent automation, the global interest rate falls from 6.0 percent in 2017 to 2.8 in 2050. In our baseline scenario, the 2050 interest rate decreases by less -to 3.2 percent. Two fundamental trends contribute to a global savings glut (Eichengreen, 2015). The first is population aging in developed and many emerging regions that raises global wealth per worker. The second is the high saving rate of Chinese and Indian households coupled with their rising share of world GDP.Wage inequality increases dramatically in our baseline scenario. In the U.S, for example, the high-skilled wage increases by 28.8 percent between 2017 and 2050. The low-skilled wage falls by 22.8 percent. In 2050, 92.5 percent of the world's high-skilled workers are better off due to automation. But 85.7 percent of low-skilled workers and 71.5 percent of all workers are worse off. In the same year and relative to no-automation, the world's average worker is 0.9 percent worse off in terms of lifetime welfare, measured as a compensating differential. Our paper's final section considers policy responses. One example is mandating automation in regions that would not otherwise automate. Such 'forced industrialization' policies (see, for example, Preobrazhensky 1926; Erlich 1950) have been pursued historically by the Soviet Union and other countries. In India, but no other region, forced adoption ra...