AI technology shocks account for 25 percent of GDP swings
A Bank of Italy study finds that artificial intelligence technology shocks act as positive supply shocks, boosting total factor productivity and employment while reducing consumer prices. However, AI innovations lower the labor share and concentrate wealth gains at the top.
Higher productivity across four decades
Using U.S. patent data from 1980 to 2019, Bank of Italy economists Andrea Gazzani and Filippo Natoli constructed a monthly measure of artificial intelligence intensity in innovation.
AI-intensive patents exhibit systematically higher quality, attracting more citations and market value than broader ICT or automation patents.
Over a five-year horizon, AI technology shocks account for roughly 25 percent of output and employment fluctuations.
The shocks generate delayed but persistent increases in total factor productivity, real output, employment, and wages, alongside persistent declines in consumer prices.
Similar macroeconomic supply-side effects are observed for generative AI patents.
Capital gains over labor returns
Unlike general technology shocks, which are typically labor-neutral, AI shocks reduce the labor share of income and shift overall returns toward capital.
While job creation outweighs displacement across most sectors and educational levels, workers without a high school diploma do not experience employment gains.
Moreover, expansionary AI shocks significantly alter wealth distribution by raising the real wealth share of the top 10 percent while reducing that of the bottom 50 percent through asset valuation channels.
A rising tide that leaves the bottom behind
The study provides clear macro evidence that artificial intelligence acts as a broad supply booster rather than a simple job killer.
Yet central bankers and policy makers must not ignore the structural warning on capital bias and rising wealth concentration.
Without targeted policy responses, macroeconomic gains from AI risk widening social inequality.