A paper estimating AI’s macro impact via software engineering

daveholtz · x · 2026-07-21

A paper presented at NBER asks how AI affects software engineering productivity and, through that channel, GDP.

The authors infer the effect from stock prices: firms more exposed to software engineering should benefit more from AI, so their returns should co-move more strongly with an AI index. They map that cross-sectional relationship into an estimate of productivity gains. Under the baseline calibration, AI news from Nov 2022 to Dec 2025 implies a permanent 30.5% gain in software engineering productivity, translating into a one-time GDP increase of 3.3% from production alone, or 6.0% if higher software productivity also boosts R&D. Updating the model with early-2026 returns — amid rapid progress in AI coding agents — suggests even larger implied gains.

The paper’s broader point is that market prices can be used as a real-time signal for the macroeconomic effect of AI.

Related event: NBER Paper Estimates AI's Macroeconomic Impact via Stock Prices(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →