Economists Warn: Don't Read "AI Hurts Wages" Papers Without the Treatment/Control Averages

paulnovosad · x · 2026-09-16

Economist paulnovosad cautions that "negative wage/employment effects of AI" papers must be read very carefully: the same result can be presented two ways, and dynamic DiD plots alone can suggest AI-exposed workers' wages are declining when he would not conclude they're worse off.

EmilVerner adds the methodological rule: always show treatment/control averages, not just dynamic DiD estimates — example #104.

Related event: Economists Debate GenAI Exposure and Wages: Relative Decline or Rising for All(12 posts)→

Original post →

More from AGI Musings

AGI Musings channel →