Artificial intelligence and national productivity: Evidence of distributional heterogeneity

Artificial intelligence (AI) is widely regarded as a transformative general-purpose technology with the potential to reshape production and long-run economic growth. Yet evidence linking national AI development to aggregate productivity remains scarce, reflecting a modern manifestation of the Solow paradox. Existing studies have predominantly examined firm- or task-level outcomes or average cross-country effects, leaving unclear whether the AI-productivity relationship varies across the productivity-growth distribution. This study investigates the relationship between national AI development, measured by the Stanford Global AI Vibrancy Index, and total factor productivity (TFP) using an unbalanced panel of 35 countries over 2017–2023. Results from conventional panel and quantile regression models indicate that lagged AI development is positively related to TFP growth, with the association strengthening toward the upper end of the TFP-growth distribution but remaining weak or insignificant at the lower end. Robustness checks, including a placebo permutation test, further indicate that these findings are unlikely to be driven by random variation. The findings contribute to the emerging macroeconomic literature on AI by providing evidence of distributional heterogeneity in the AI-TFP relationship, with implications for the design and evaluation of AI-related policy interventions.