Technology and the decline in the U.S. labor share of income
A well-known “stylized fact” in economics holds that the distribution of income between labor (wages and salaries) and capital (stocks and bonds) is fairly stable over time. Yet quite a number of researchers find that the share of income going to labor has been on the decline over the past several decades—the most famous, of course, being Thomas Piketty in his book “Capital in the 21st Century.”
Whether this change is as a result of changes in technology is hotly debated in economic circles. Last week, Dylan Matthews at Vox reviewed the research on the question of the labor share, noting that researchers posit a variety of hypotheses, among them the financialization of the U.S. economy, which benefits the owners of capital over wage earners, the downward pressure in U.S. wage brought about by globalization, the decline of labor unions, and the growing accumulation of capital.
But one hypothesis—what Matthews calls “robots”—now garners the most attention. Research by the University of Chicago’s Loukas Karabarbounis and Brent Neiman finds that the decline in the price of investment goods, such as computers or industrial robots, has led to the decline in the labor share of income. Firms have responded to this price decline by substituting capital for labor, meaning firms are investing more in technology and less in the size and remuneration of their workforces.
This result implies a high level of “elasticity of substitution,” or the ability to substitute one type of input for another in the making of a good or the delivery of a service. The two researchers calculate this elasticity between capital and labor at approximately 1.25, which is high compared to previous estimates of the elasticity. To put this number in context, when the elasticity is higher than 1, a decrease in the price of capital would reduce the labor share of income. And if it’s below one, the labor share would increase after a decrease in the price. The authors find that the elasticity results also hold after accounting for the depreciation of capital goods, which is important because the rate of depreciation has accelerated in recent years.
But labor’s share of income can decline without the elasticity of substitution being so high. Research by economists Ezra Oberfield of Princeton University and Devesh Raval at the Federal Trade Commission take a different approach and find a much smaller elasticity while still seeing a decline in labor’s share of income. In Oberfield and Raval’s model, there are two ways for labor’s share to decline. The first is for the price of a factor of production (labor or capital) to decline. Karabarbounis and Neiman’s research finds that the price of capital has declined and that explains the decline in the labor share.
The other option is that technology has changed so much that capital is more productive now when invested in new technology so more firms are investing in technology at the expense of labor. Oberfield and Raval’s research finds this second option is more important. Whereas Karabarbounis and Neiman look at changes in labor’s share of income across the United States to find their elasticity of 1.25, Oberfield and Raval use data from individual firms to estimate an elasticity for the entire economy. They find a far lower elasticity of about 0.7.
So Oberfield and Ravel argue that what’s causing the decline in labor’s share of income isn’t cheaper capital or more capital accumulation but rather a technological shift toward using capital more. What’s interesting is they say the cause is technology, which they broadly define to include the offshoring of jobs. For these two authors, the technology required to offshore production is a form of capital substitution for labor in the United States, yet they also point out their results don’t point toward a simple offshoring story. What’s happening is quite nuanced, they argue, involving investments in new technology and new workers overseas.
Is this new research just a distinction without a difference? Not at all. Our understanding of how technology is changing our economy, the causes of rising inequality, and the proper policy responses are all influenced by the outcomes of this kind of research. So believe it or not, whether one economic variable is larger or smaller than 1 has quite a bit of impact.