Investigating the effect of industry-specific economic distance on the prediction of intercity population movement

Intercity population movement has been extensively studied since it is closely related to human society. Currently, city industry structures play dominant roles in the direction of population movement. Yet, the extent to which different kinds of industry proximity influence human mobility remains unclear. In this study, we introduce the concept of intercity industry proximity, regarded as economic distances, to forecast intercity population movement using a relational graph convolutional network. Our findings demonstrate the effectiveness of this framework in learning information from 18 industry proximity networks. Using this framework, we investigate the impact of distinct industries on population movement by traversing each industry as input separately. Results show that while all industries exhibit favorable predictive performance, slight differences exist. Specifically, the primary industry emerges as the most influential predictor of population movement, followed by secondary industries, whereas certain tertiary industries exert comparatively minimal effects. We also examine the influence of proximity thresholds for graph-generating on model performance. Theoretical explanations concerning face-to-face interactions for the diffusion of tacit knowledge are discussed, and policy implications are provided to enrich the current understanding of population movement.