Website CV
Fields
Labor, Economic History, Innovation
Job market paper
Occupational Reinvention: Technology, Tasks and Jobs over Eight Decades
I examine the impact of robot and computing technologies on the labor market, using a novel approach that allows the tasks within occupations to change over time. I construct a panel of U.S. occupations spanning eight decades (1939 to 2020), assembling more than 300,000 task descriptions from historical occupational records and matching them to capabilities described in robot and computing patents available at each date. I distinguish worker tasks that technology executes (“automation”) from those it assists (“augmentation”). Automation technologies tend to depress jobs and wages, whilst augmentation technologies improve worker outcomes. For example, looking at robots, a one-standard-deviation shift from augmentation towards automation, holding overall technological relatedness fixed, predicts 10 percent lower occupational employment and 3.5 percent lower real weekly pay over five decades. I show that using 2020 occupational descriptions instead of contemporaneous ones magnifies the estimated employment effects of both technologies but attenuates the negative pay effects of automation. IV estimates exploiting long-dated historical tasks yield results consistent with OLS. Holding technological capabilities fixed, occupations initially tilted towards robot automation subsequently reorganise their work, with roughly one quarter of the initial difference disappearing by 2020. Occupational work is therefore both an input into measuring technological change and an outcome of adjustment to it.