Unraveling the dynamics of China railway express (CRE) in China: A multi-method analysis

Abstract

As a flagship project of the Belt and Road Initiative (BRI), the China Railway Express (CRE) has significantly reshaped Eurasian trade dynamics through transcontinental rail connectivity. However, systematic quantitative analysis of its spatiotemporal evolution and the heterogeneous drivers behind its development remain understudied. This study utilizes a unique monthly panel dataset from 2013 to 2021, covering involved provinces, and employs an integrated methodology to examine the CRE's evolving patterns and drivers. The key findings include: (1) Spatial econometric analysis reveals distinct cargo flow patterns—outgoing shipments diffuse from northeastern and coastal regions toward central and western hubs, while incoming flows shift from central and eastern China toward southwestern and western regions, establishing the central and western provinces as key nodes in bidirectional logistics networks. (2) Regression analysis identifies a set of significant driving factors, this set of factors is further excavated through DeepAR forecasting and 50 iterations of Permutation Feature Importance (PFI), which uncovers region-specific drivers: outgoing flows are predominantly influenced by policy interventions and supply-side factors (e.g., infrastructure and government attention), while incoming flows are driven by demand-side forces (e.g., consumer markets and informatization level). Coastal areas exhibit a substitution effect with sea-rail transport. Based on these PFI results, targeted recommendations for regional policy differentiation are proposed. (3) Wavelet coherence analysis reveals a dynamic evolution in the relationship between government policy attention and cargo flow volumes, signifying a shift from active governmental engagement towards more passive facilitation. Methodologically, this study introduces a novel analytical framework integrating spatiotemporal pattern analysis, machine learning-driven explainable artificial intelligence (XAI) for driver decomposition, and policy response assessment. Practically, it provides actionable recommendations for tailored regional strategies and offers a replicable methodological blueprint for optimizing multi-regional rail freight systems.