National big data pilot zones and urban carbon emission efficiency: Causal evidence from China via double machine learning.

Based on panel data from 282 Chinese prefecture-level cities (2008-2021), this study employs double machine learning and causal forest methods to evaluate the National Big Data Comprehensive Pilot Zone policy's causal effects on urban carbon emission efficiency. The policy increases carbon efficiency by 9.82% on average through four synergistic pathways: green technology innovation (28.4% mediation), industrial structure upgrading (14.1%), resource allocation efficiency (12.3%), and energy structure transformation (6.9%), collectively explaining 61.7% of total effects. Heterogeneity analysis reveals significant spatial differentiation, with policy effects 61% higher in central-western cities than eastern cities, 25% higher in resource-based cities, and 23% higher in low-digitalization cities, demonstrating "catch-up effects" and a "digital paradox." Conditional average treatment effect estimation identifies digital infrastructure level, baseline efficiency, and industrial structure as key moderators, while spatial spillover analysis shows positive externalities on neighboring cities (13.8% indirect effects). Nineteen robustness tests confirm high reliability (4.3% coefficient variation). This study contributes by constructing a "big data → four mechanisms → carbon efficiency" theoretical framework, pioneering DML and causal forest application in environmental policy evaluation, and identifying optimal policy profiles for central-western, resource-based, low-digitalization, low-baseline-efficiency cities, providing empirical evidence for optimizing pilot layouts and formulating differentiated support policies.