Deep learning: Spatiotemporal impact of digital economy on energy productivity

Abstract

Can digital economy in developing countries always promote energy productivity? Using a panel data of thirty provinces in China, this study utilizes artificial neural network and SHapley Additive exPlanations for a sample-by-sample analysis, which effectively mine the nonlinear correlation and spatiotemporal evolution between digital economy and energy productivity. The findings show: First, the influence of digital economy and various factors on energy productivity demonstrates a U-shaped pattern. Particularly, digital inclusive finance is steadily emerging as the predominant factor. Second, the lagging effect of digital infrastructure and digital innovation environment stand as crucial drivers without exhibiting a cumulative trend. As a nascent motivator, digital economy effectively disrupts development inertia, significantly contributing to the enhancement of energy productivity. Third, the effect unfolds as a gradual intensification, also characterized by a non-equilibrium distribution pattern spatially, with varying dominant factors across distinct regions. Fourth, forecast indicates that vigorous promotion of the digital economy will bolster the realization of China's 2025 energy productivity target (2.64 > 2.53). Furthermore, the digital economy exerts a marginal increasing effect on energy productivity. This research not only supplements the analysis from a methodological standpoint, but also provides reliable academic support and policy inspiration for integrating digital economy development with energy productivity advancement across countries.