“Robots and Jobs: Evidence from US Labor Markets” is a paper by MIT economist Daron Acemoglu and Pascual Restrepo, circulated as NBER Working Paper 23285 in 2017 and published in the Journal of Political Economy in 2020. It is one of the first studies to put a credible causal number on what industrial robots did to American workers, using local labor markets (commuting zones) as the unit of analysis over the 1990 to 2007 period.
The authors build a model in which robots compete directly with human labor at the task level, then estimate the effect of robot adoption on employment and wages by exploiting differences in how exposed each local market was to robots based on its mix of industries. Their central estimate is that one more robot per thousand workers reduces the employment-to-population ratio by roughly 0.18 to 0.34 percentage points and wages by about 0.25 to 0.5 percent. Expressed differently, the arrival of one new industrial robot in a local labor market is associated with the loss of about 5.6 jobs. The effects are robust to controlling for Chinese import competition, broader IT capital, offshoring, and other trends.
The finding mattered because it cut against the comforting view that automation only ever creates as many jobs as it destroys. Robots, in this data, produced net local losses that were not offset within the affected communities. The paper became a cornerstone of the task-based approach to automation economics and a frequent reference point in debates over what generative AI might do next.