Distribution-sensitive assessment of air-quality and health benefits from clean-heating pilots in northern China

Clean-heating programs are central to winter air-quality management in northern China, yet most evaluations focus on average PM_2.5 reductions and provide limited evidence on severe pollution episodes that are especially relevant to short-term health risks. This study evaluates 88 clean-heating pilot cities in northern China from 2013 to 2023 using city-level daily PM_2.5 concentrations spatially aggregated from a 1-km gridded product, machine-learning-based weather normalization, and the augmented synthetic control method (ASCM). We estimate city-specific no-policy counterfactuals and quantify policy effects during the heating season for mean PM_2.5, upper-tail concentrations, and exceedance days. We then quantify the short-term mortality implications of the estimated daily exposure contrasts using a PM_2.5–mortality exposure–response model. Using weather-normalized PM_2.5, the ASCM-estimated clean-heating policy effects were −3.01 μg/m³ for the heating-season mean, −5.76 μg/m³ for P95, -5.96 μg/m³ for P99, and -7.92 exceedance days per heating season. Under standardized meteorological conditions, the health impact assessment (HIA) yielded a model-estimated mortality difference of 12,454 fewer PM_2.5-attributable all-cause deaths on post-treatment improvement days (95% CI, 8564–16,222), with 88.9% of the estimated total occurring in the 2017 and 2018 rollout cohorts and 56.2% concentrated in the top 10% of cities. High-pollution improvement days accounted for 27.8% of improvement days but contributed 48.8% of the model-estimated mortality difference. Under a stylized fixed-capacity sequencing scenario, prioritizing cities with higher projected health returns increased the projected difference by 939 deaths (7.5%). These findings show that distribution-sensitive assessment can better support health-oriented air-quality management and rollout prioritization for residential energy-transition policies.