Spatial Spillover Effects and Nonlinear Impact Intensity of Artificial Intelligence on Green Total Factor Productivity

Abstract

This study focuses on the spatial evolution logic and heterogeneous impact intensity of artificial intelligence (AI) technology in driving green total factor productivity (GTFP). First, using spatial econometric models, the study confirms through global and local Moran's I indices that significant positive spatial autocorrelation and agglomeration characteristics exist in AI development levels across Chinese provinces. Empirical results show that AI not only significantly enhances local green TFP but also generates notable spatial spillovers through technology diffusion and diffusion effects, with positive coefficients for direct, spillover, and total effects. Subsequently, to more precisely measure variations in impact intensity, the study employed K-Means clustering to classify provincial samples into high, medium, and low development tiers. Dynamic GMM estimation was applied to address model endogeneity. Results indicate that AI exerts a sustained, positive influence on GTFP and its sub-dimension of green technology efficiency, with significant historical effects. However, at the sub-dimensional level, AI's direct impact on green technological progress has not yet exhibited statistically significant characteristics. Through comparisons of multiple regression models and lagged term treatments, this study confirms the robustness of the conclusions, providing quantitative support for analyzing AI's enabling mechanisms in the spatial dimension.