Working Paper
2026
Leila Aghabarari · IFC, World Bank Group
GAIA integrates three previously disconnected lenses on AI's labor-market reach — observed platform usage, theoretical task exposure, and pre-GPT adoption baselines — into a single harmonized index. It covers 923 O*NET occupations and 138 countries on a common crosswalk, enabling comparable exposure and adoption measures across the world.
Key finding
A unified index across 923 occupations and 138 countries shows theoretical exposure and observed usage diverge sharply — the two are far from interchangeable.
AI EconomicsMeasurementLabor Markets
Preliminary
May 2026
Leila Aghabarari · IFC, World Bank Group
Theoretical AI-exposure scores are widely used as proxies for where AI will bite, yet they are rarely validated against measured capability. Using observed AI task performance across U.S. occupations, this paper tests whether exposure predicts performance — and finds a robust negative relationship that survives a horse race against alternative measures.
Key finding
Higher theoretical exposure predicts lower measured AI performance (β = −17.47, p < 0.001); the SML alternative is insignificant (p = 0.467).
AI EconomicsLabor MarketsMeasurementUSA
Forthcoming
2026
Leila Aghabarari · IFC, WBG · Bernardus Van Doornik · Banco Central do Brasil
This paper traces how occupational AI exposure transmits through the banking system into local labor markets in Brazil. Linking the central bank's credit registry to matched employer–employee records and the GAIA crosswalk, it studies whether AI-exposed sectors face tighter credit and adjusting employment after the generative-AI shock.
Key finding
A transmission chain — AI exposure → credit tightening → employment and wages — links the generative-AI shock to Brazilian labor markets via bank lending.
AI EconomicsCredit MarketsLabor MarketsBrazil
No papers match your filters.