Dynamic Network Determinants of Economic Growth

This study investigates how the structure of national production networks influences long-run economic growth. Using data from the EORA26 Multi-Region Input-Output Database for 187 countries (19912017), we reconstruct annual upstream and downstream sectoral networks and extract detailed topological metrics. We combine machine learning (XGBoost) with SHAP-based interpretability and two-way fixed effects panel regressions to identify and validate key structural determinants. The results show that growth is driven primarily by sector-specific network positions rather than aggregate network geometry at the country level. A pronounced bottleneck effect emerges in the transport sector, where insufficient mediation severely constrains growth. More broadly, strong outbound supply connectivity in key sectors promotes growth, whereas excessive input dependence in certain service sectors is associated with weaker performance. Our findings highlight the importance of structural positioning within production networks for understanding cross-country growth heterogeneity.