Contemporaneous Causal Analysis of Housing Prices Across Guangdong’s Major Cities: Employing Vector Error-Correction Modeling and Directed Acyclic Graphs

Abstract

This research adopts a cohesive analytical strategy, combining vector error-correction modeling (VECM) with directed acyclic graphs (DAGs), to rigorously assess evolving connections in monthly home prices across six key Guangdong cities. These cities include Dongguan, Zhongshan, Foshan, Guangzhou, Huizhou and Zhuhai. The analysis spans a significant period of nearly thirteen years (November 2011–July 2024). Supporting computational procedures determine the fundamental DAG structure: Initially, the PC algorithm identifies a preliminary set of potential causal relationships; subsequently, the LiNGAM (Linear Non-Gaussian Acyclic Model) approach leverages non-Gaussian data characteristics to eliminate directional ambiguity, establishing a conclusive causal sequence. Leveraging this DAG-based causal order, detailed innovation accounting — including impulse response functions — quantifies dynamics and shock effects within the price network. Findings reveal intricate transmission channels and diverse adjustment speeds within the provincial housing market following external disturbances. Empirical results show that policies designed to increase real estate values in Dongguan, Foshan, Huizhou and Zhuhai can substantially boost wider market recovery throughout Guangdong, attributable to their central network roles. For the remaining two cities exhibiting lower systemic influence, however, insights recommend localized revitalization approaches as more direct and efficient than provincial stimulus or measures targeting core areas, highlighting the necessity for policy differentiation grounded in spatial network functions.

The authors' abstract, as published at the source. Journal of Uncertain Systems, 2026 · DOI ↗