Path Dependence, Governance, and the Limits of AI-Led Green Growth: A Dynamic Panel Analysis of 36 Economies

This paper asks whether the development of national artificial intelligence ecosystems contributes to greener economic performance, and whether public governance shapes that relationship. The analysis covers a balanced panel of 36 advanced and emerging economies from 2017 to 2023. We capture general national artificial intelligence ecosystem development through a multidimensional index built on five pillars (innovation, economic diffusion, skills, policy, computing infrastructure) aggregated by within-pillar principal component analysis, and estimate the model by two-step System-GMM, with instrumentation anchored in Wooldridge endogeneity tests robust to heteroscedasticity. Green growth is highly path-dependent, with an autoregressive coefficient close to 0.96 that corresponds to an annual convergence speed of 4.5 percent. Government effectiveness contributes positively and significantly. The artificial intelligence ecosystem index displays no detectable independent effect once persistence and endogeneity are addressed, and its interaction with government effectiveness is similarly indistinguishable from zero, a result that calls for caution in narratives expecting artificial intelligence to deliver sustainability gains on its own.