Decoding air passenger flows: Identifying the role of network autocorrelation in air travel

Abstract

With the rapid expansion of the global aviation industry, especially in the Eurasian continent, understanding the factors driving interregional air passenger flow is of increasing importance. While most existing studies emphasize node-level and edge-level influencing factors such as economic scale, population size, and geographic distance, they often neglect the pivotal variable of network autocorrelation. This research is the first to introduce network autocorrelation within the Eurasian context and systematically analyze it using the Eigenvector Spatial Filtering Negative Binomial Gravity Model. Our findings highlight: (1) A significant network autocorrelation in the Eurasian continental aviation network. The eigenvector spatial filtering negative binomial regression model effectively captures this autocorrelation, considerably reducing model estimation bias. Specifically, the leading 3.28% of eigenvectors capture a high degree of this network autocorrelation. (2) The presence of network autocorrelation introduces estimation biases in related variables, resulting in underestimations of economic size, population size, visa restriction, and international trade, while overestimating cultural and institutional differences, geographical distance, colonial relationship. (3) Various factors affect the Eurasian continental sub-region's air passenger flows differently, indicating regional variations. This study takes a step towards improving our understanding of network autocorrelation in air passenger flows research.