Can Developing Big Data Improve Urban Energy Efficiency? A Double Machine Learning Based Approach

This paper examines the causal impact of big data development on urban energy efficiency using China Big Data Integrated Pilot Zone policy as a quasi natural experiment. Through double machine learning and spatial econometric models applied to panel data from 238 Chinese cities during 2008 to 2019, we identify a critical paradox. Implementation of the policy significantly reduces local energy efficiency by 10 percent after controlling for variables and fixed effects while generating positive spatial spillover effects that improve efficiency in neighboring areas by 8.4 percent. Mechanism analysis demonstrates that the policy induces clustering of computer and software industries, increasing local energy consumption and lowering local energy efficiency. Simultaneously, it attracts enterprise relocation from peripheral regions to pilot zones, reducing energy burdens in surrounding areas and improving their efficiency. Heterogeneity analysis reveals stronger negative local effects in eastern and central regions compared to western regions, and more pronounced impacts in non-ordinary prefecture level cities. Robustness tests including extended SFA modeling confirm these findings. These results challenge conventional assumptions about uniform benefits of big data and highlight the necessity for spatially coordinated policies to balance efficiency gains across regions.