Structural Change in the Multilayer World Trade Web: An Economic Dominance Theory Approach

Abstract

Most studies of the World Trade Web (WTW) describe its topology: degrees, strengths, centralities, communities, or the similarity between product networks and years. This paper presents Economic Dominance Theory (EDT), initiated by Lantner (1974), and positions it as an alternative theory of structural dependence in directed flow networks rather than an additional toolbox. EDT normalizes bilateral flows into a Ghosh matrix of directed propagation and measures circular interdependence through its determinant, which aggregates the feedback circuits by which influence returns to its origin. Through the log-determinant and its trace expansion, a single quantity decomposes exactly by feedback length, arc and node deletion, product layer, network partition and, in the dynamic extension developed here, time. Structural change can therefore be measured without defining a distance between graphs, and layers and regions become exact components of one object rather than separate descriptions. EDT differs from centrality and community detection, which describe positions and densities rather than feedback closure; from economic complexity, which addresses capabilities rather than dependence; and from input-output and value-chain analysis, whose algebra it shares but which locate positions along chains where EDT measures their closure. An application to CEPII BACI data (1995-2023, 96 HS2 layers) shows what this adds. Circular interdependence did not deepen with trade: the diffusion determinant, which rises when feedback weakens, goes from 0.12 in 1995 to 0.21 in 2006, falls to 0.16 in 2013 and reaches 0.19 in 2023. This net weakening comes from cross-product feedback, which carries more than 90% of order-two and about 99% of order-three feedback. Geography depends on feedback length: short reciprocal loops remain regional, while long feedback is overwhelmingly transregional.