Decoding global economic dynamic: A graph-based examination of contemporary ETF markets

Abstract

Analyzing the dynamic interdependencies among global economies is critical for understanding financial risk and opportunity. In this work, we introduce a spatio-temporal attention-based graph neural network that models these complex relationships using Exchange-Traded Fund (ETF) indices from major global markets. Our framework provides a data-driven lens on economic interactions, uncovering how influence shifts between economies like the United States, China, and European nations, particularly during critical financial events such as the COVID-19 pandemic. The model not only reveals interpretable network structures—identifying the U.S. as a dominant global hub and China as a key regional one, but also significantly outperforms baseline methods in predicting future market trends. Ultimately, this study presents a powerful and transparent framework for decoding global economic dynamics and improving financial forecasting.