Abstract

I analyze the effect of artificial intelligence exposure on wages, employment, and the emergence of new work in the United States between 2010 and 2021. I rely on novel indices at the occupational level to distinguish between AI that automatizes tasks and AI complementing output. Using IV estimators, I find a negative effect of automation AI exposure on hourly wages. Heterogeneity analysis shows that this effect is concentrated in low-skill occupations. AI exposure complementing occupational output positively affects the emergence of new work, especially in high-skill occupations. I do not find discernible effects of automation AI and complementing AI on employment size.