Topological features for the robustness of global supply chain networks

The topological structure of modern global supply chain networks (SCNs) has become increasingly complex. In recent years, major disruption risk events haveemerged, disrupting complex SCNs. Since a global SCN is the aggregation of several sub-SCNs in terms of industries and countries (regions) and is temporally changing in its topological characteristics, robustness toward risk should be analysed by the community in consideration of its temporality. In this study, we aim (i) to propose a method to generate temporal SCNs of multiple communities, (ii) to evaluate the robustness of each SCN against error and attack risks, and (iii) to identify the topological features that influence the robustness of SCNs using real transaction data between firms. As a result, eight SCNs weredetected based on industries and countries, and the size of these SCNsincreased over time. The average shortest path length and degree distribution have similar impacts on each SCN, while the cluster structure diverges among SCNs. Regarding robustness against error and attack risk, unlike in existing studies, the SCNs are significantly partitioned at the initial firm removal rate. Only for attack risk, percolation transition was found at approximately 10% removal of firms. The relationship between robustness indicators and topological features is identified by panel data analysis, and we find that the significant topological features affecting robustness differ by type of risk.