“Tasks, Automation, and the Rise in US Wage Inequality,” by Daron Acemoglu and Pascual Restrepo, was circulated as NBER Working Paper 28920 in 2021 and published in Econometrica in 2022. It tackles a long-standing puzzle in labor economics: why did wage inequality in the United States widen so sharply over the last four decades, and how much of that is the fault of automation rather than other forces?
Using a task-based framework, the authors argue that automation displaces workers from the tasks where they held a comparative advantage, pushing their wages down relative to workers whose tasks were not automated. Their headline empirical claim is striking: between 50 and 70 percent of the changes in the US wage structure between 1980 and 2016 can be accounted for by the relative wage declines of worker groups concentrated in routine tasks within industries that automated quickly. Crucially, they find that other commonly blamed factors, such as rising market power, markups, and the decline of unions, play only a minor role in their decomposition.
The paper is significant because it reframes inequality as a direct, measurable consequence of which tasks machines take over, not just a vague side effect of “skill-biased technical change.” For anyone trying to understand whether AI will widen or narrow the income gap, it provides the analytical machinery and a sobering historical baseline: past automation mostly widened it.