Global Networks in Transformation: Trade, Institutions, and Regionalization, 2000–2014
Description
This database accompanies a research note examining the topological transformation of the international system between 2000 and 2014. Using network analysis, it combines bilateral trade and intergovernmental organization (IGO) co-membership ties to measure connectivity, brokerage, clustering, regionalization, and systemic vulnerability across the periods surrounding the 2008 financial crisis. The material is organized into six folders containing raw dyadic data, combined multiplex networks, node-level metrics, clustering and modularity calculations, Geographic Homophily Index (GHI) and Regional Intensity Index (RII) calculations, and robustness checks using alternative community-detection algorithms. Together, these materials allow replication of the study’s findings on densification, brokerage erosion, regional clustering, and differentiated vulnerability within the international system. A standalone document containing the complete Online Appendices accompanies the six replication folders. It provides supplementary methodological details, calculations, results, and robustness checks and is intended for consultation alongside the Research Note. FOLDERS: 1-Raw Data: Two files make up the original raw data, based on the trade and intergovernmental organization (IGO) datasets from the Correlates of War (COW) project. Merging these two files creates the set of edges and nodes for the three networks (2000, 2007, 2014). 2-Network Nodes and Edges: Contains files with nodes (states and IGOs) comprising up to 201 countries and 346 IGOs (maximum 547 nodes). These global actors form the basis of networks containing more than 24,000 edges (unique connections). 3-Gini Betweenness: Contains a file showing that the Gini coefficient for betweenness centrality declined from 0.7603 in 2000 to 0.7327 in 2014, indicating a shift from a highly centralized brokerage structure—where power was concentrated in a few hub nodes like France—toward a more diffuse, mesh‑like regional topology. 4-Basic Metrics: The basic metrics are density, betweenness centrality, and modularity, measured for the 2000, 2007, and 2014 networks. This folder has four files: three with node‑level metrics for each year, and one document explaining the calculation formulas. 5-Regional Integration Index (RII): This folder contains eight files that calculate the RII for each regional bloc (EU, ASEAN, MERCOSUR, SADC). 6-Robustness Check: This folder contains a file that reports the robustness checks for community detection conducted on the global multiplex network used in the article.
Files
Steps to reproduce
NETWORK METRICS 1. Network Construction For each benchmark year (2000, 2007, and 2014), combine bilateral trade and IGO co-membership dyads into an undirected, unweighted multiplex graph with no self-loops. Standardize node names across the source datasets. 2. Metrics Calculate the following measures: - Degree – number of neighboring nodes. - Normalized betweenness centrality – Freeman-normalized proportion of shortest paths passing through each node. - Communities – detect communities using Clauset–Newman–Moore Greedy Modularity Maximization and assign community IDs based on the resulting partition. - Network-level metrics – calculate system density and modularity (Q) and append these values to each node record. Validation (2000): approximately 553 nodes, density = 0.149, modularity (Q) = 0.124, with France as the highest-betweenness node (normalized betweenness ≈ 0.026). Gini Betweenness Centrality The file Gini_Betweenness_Calculation_Steps.csv contains node-level data for 2000 and 2014. For each sovereign state and IGO, it reports normalized betweenness centrality, cumulative node share, and cumulative betweenness share after ranking nodes from lowest to highest betweenness centrality. Construct the Lorenz curve from the cumulative distribution of betweenness centrality and calculate the Gini coefficient from its deviation from the line of equality. The resulting Gini coefficient declines from 0.7603 in 2000 to 0.7326 in 2014, documenting a decline in the concentration of brokerage across the global network. Regional Intensity Index (RII) Using the dyadic trade datasets and bloc-membership definitions provided in the repository, reproduce the RII as follows: - Identify bloc membership for each observation year (2000, 2007, and 2014) for the EU, ASEAN, Mercosur, and SADC. - Calculate intra-bloc trade by summing bilateral trade flows for which both trading states belong to the same bloc. - Calculate total bloc trade by summing trade flows involving bloc member states with both intra-bloc and extra-bloc partners. - Calculate the RII as: RII(t) = Intra-bloc Trade(t) / Total Bloc Trade(t). The resulting values reproduce the RII scores reported in the calculation files. Robustness Check To reproduce the community-detection robustness check: - Construct the undirected, unweighted multiplex network for each benchmark year (2000, 2007, and 2014) by combining IGO co-membership and bilateral trade ties. - Apply three community-detection algorithms separately: Clauset–Newman–Moore Greedy Modularity Maximization, Louvain, and Leiden. For Louvain and Leiden, set resolution γ = 1.0 and use fixed random seeds for reproducibility. - Record the modularity score (Q) and number of detected communities for each algorithm and year. - Compare the resulting modularity scores and community counts across algorithms and observation years.