A new National Bureau of Economic Research working paper finds automation has been the primary driver of wage inequality in the US over the past four decades, more so than other factors, such as the offshoring of jobs, decline of unions, and market concentration of firms.
Authors Daron Acemoglu and Pascual Restrepo find that wage decreases for workers specializing in routine tasks with high exposure to automation account for 50 to 70 percent of the changes in the US wage structure between 1980 and 2016. Now, in 2021, many consider the COVID-19 pandemic an “automation forcing” event (PDF), with companies replacing workers in some low-wage jobs with machines, a trend likely to continue after the economy fully reopens and more workers return to the labor market.
Acemoglu, institute professor of economics at the Massachusetts Institute of Technology, and Restrepo, assistant professor of economics at Boston University, observe through an analysis of government and private data that automation displaces workers without college degrees who perform specific tasks at a much higher rate. Workers in the top quintile of task displacement saw their real wages decline by 12 percent, compared with workers least exposed to automation, who experienced wage growth of about 26 percent.
The authors also contend that automation affects different demographic groups differently. Men without college degrees have experienced high levels of both task displacement and wage declines, and men and women with postgraduate degrees and college-educated women have experienced negligible task displacement and strong wage growth.
For Black and Hispanic workers, the evidence of automation’s effect on their job security and economic mobility is concerning. They account for 13 percent and 18 percent of the US labor force, respectively, but are overrepresented in jobs at high risk of being eliminated or significantly changed by automation, according to a recent Hamilton Project study.
Acemoglu and Restrepo use the decline in the labor share of income—the share of economic output going to workers’ wages and compensation rather than to employers and owners—in a specific industry to infer how much automation and task displacement is occurring within that industry. Workers who specialized in industries that experienced labor share declines saw their relative wages decrease between 1980 and 2016.
The paper also provides a more nuanced explanation for wage inequality than one offered by a commonly accepted framework known as skills-biased technical change (SBTC). SBTC proposes that wage inequality is the result of technological progress creating higher demand for higher-skilled workers and lower demand for workers with limited skills. This framework, however, doesn’t account for how technology plays out differently in different industries, with workers performing routine tasks being vulnerable to both task displacement by automation and wage declines.
Because task displacement by automation is the primary lever of wage decline, even workers with more education and skills are susceptible to wage decline if their tasks, too, become routinized and automated, according to Acemoglu and Restrepo.
SBTC also assumes technological progress will lead to productivity gains. But the paper’s authors find rapid automation has led to sluggish productivity, documenting that automation only increased productivity by 3.8 percent between 1980 and 2016. Automation, they conclude, “can explain a sizeable fraction of changes in wage structure and wage declines in the data, while having a tiny impact on productivity growth.”
In media interviews, Acemoglu notes that “excessive” automation has resulted in the adoption of “so-so technologies”—such as self-check-out kiosks at the grocery store—that replace workers and jobs without saving costs or adding value. Yet experts say there is nothing inevitable about automation’s encroachment upon work and wages; policymakers can advance policies to harness the power of artificial intelligence and other technologies to complement and augment human capital (PDF) and create new tasks, skills, and competencies for workers.
Race and Jobs at Risk of Being Automated in the Age of COVID-19 (PDF), Hamilton Project and Brookings