Partitioning for “Common but Differentiated” Precise Air Pollution Governance: A Combined Machine Learning and Spatial Econometric Approach

Abstract

Effective governance of air pollution requires precise identification of its influencing factors. Most existing studies attempt to identify the socioeconomic factors but lack consideration of multidimensional heterogeneous characteristics. This paper fills this long-ignored research gap by differentiating governance regions with regard to multidimensional heterogeneity characteristics. Decision tree recursive analysis combined with a spatial autoregressive model is used to identify governance factors in China. Empirical results show several interesting findings. First, geographic location, administrative level, economic zones and regional planning are the main heterogeneous features of accurate air pollution governance in Chinese cities. Second, significant influencing factors of air pollution in different delineated regions are identified, especially significant differences between coastal and non-coastal cities. Third, the trends of heterogeneity in urban air governance in China are to some extent consistent with national policies. The approach identifies factors influencing air pollution, thus providing a basis for accurate air pollution governance that has wider applicability.