Analyzing the heterogeneous impacts of high-speed rail entry on air travel in China: A hierarchical panel regression approach

It has been widely recognized that the entry of high-speed rail (HSR) has significant impacts on air travel. Most, if not all, studies quantify such impacts using HSR entry dummy variables, namely, dummy variables that indicate the entry of HSR, in various econometric models. Although existing studies have attempted to explain the heterogeneity of the impacts of HSR entry, their approaches are often based on simple segmentation rules based on the distance between the origin and destination cities or whether the city pair involves a hub city. In this research, we use a systematic approach to analyze the heterogeneous impacts of HSR entry on air travel, which is new to the literature. Specifically, we take a hierarchical panel regression approach to analyze the Chinese domestic air travel data, which records, for each city pair, its monthly passenger volume and flight frequency from January 2011 to December 2017. We conduct data filtering and aggregation in the data preprocessing. At the lower level of our hierarchical model, we explore the heterogeneity of the coefficients for the HSR entry dummy variables and allow them to vary across city pairs; while at the upper level, we explain this heterogeneity using a structural equation that considers the difference of in-vehicle travel time between HSR and air travel, and the geo-economical characteristics of the origin and destination cities. Our main results include: (1) Consistent with existing literature, the entry of HSR reduces air passenger volumes and flight frequencies by 17.88% and 15.80%, respectively; (2) Strong heterogeneity exists in the impacts of HSR entry on both passenger volumes and flight frequencies; (3) For air passenger volumes, the impact of HSR entry intensifies at an accelerated rate when the difference of in-vehicle travel time between HSR and air decreases; (4) For air passenger volumes, the impact of HSR entry is stronger for city pairs that contain a hub city; (5) For air passenger volumes, the impact of HSR entry is smaller among city pairs that contain a western region city; (6) For flight frequencies, we find similar results, except that the factor of whether the city pair contains a hub city is non-significant; and (7) The determinants explain about 70% of heterogeneity of impacts of HSR entry on both the passenger volumes and flight frequencies.