Beyond the Conditional Mean: Digital Economy Development and Logistics Carbon Intensity—Evidence from Chinese Provincial Panel Quantile Analysis

Whether the relationship between digital economy development and logistics carbon intensity varies across the conditional distribution cannot be determined from average estimates alone. Using panel data for 30 Chinese provincial-level regions from 2013 to 2022, we compare a conventional mean two-way fixed effects estimate with benchmark Canay two-step panel quantile estimates and examine sensitivity to potential endogeneity, patent-related associations, and regional heterogeneity. The mean TWFE estimate is negative but insignificant, whereas the benchmark Canay estimates are negative and significant across quantiles 0.10–0.90. Conventional robustness checks retain this pattern, although an alternative quantile-specific fixed effects specification indicates lower-tail sensitivity. Weak-identification-robust IVQR inference supports a negative association only at the 0.10 and 0.25 quantiles. Digitalization is positively associated with green utility model patent grants at most mediator quantiles, but no stable contemporaneous negative indirect association is found; significant positive lagged associations are concentrated at lower and middle quantiles during the first four years. The negative conditional association is stronger in non-eastern regions, with significant differences at the 0.25 and 0.50 quantiles. These findings concern carbon intensity rather than absolute emissions and emphasize estimator sensitivity and inferential uncertainty.