Small-World Spatial Network Analysis of Global Supply Chains Using International Trade Data (1988–2022)
Advisor:
Dr. Firas Khatib, Associate Professor, Department of Computer and Information Science
Committee Members:
Dr. Debarun Das, Assistant Professor, Department of Computer and Information Science
Dr. Ashokkumar Patel, Associate Teaching Professor, Department of Computer and Information Science
Abstract:
This thesis applies complex network science to analyse 35 years of global trade data and characterise the structural properties, temporal evolution, and resilience of the international trade system. Using 613,252 bilateral export observations from the UN Comtrade database covering 188 reporting economies from 1988 to 2022, the study constructs nine 4-year interval networks and one full-period aggregate network. Each country becomes a node, each bilateral trade relationship above a one million USD threshold becomes a weighted edge, and the resulting graphs are analysed using small-world theory, scale-free network theory, geospatial community detection, and percolation-based resilience simulation.
Nine formal hypotheses are tested. Seven are confirmed, two are partially confirmed. The global trade network exhibits persistent small-world properties across all nine intervals, with small-world index σ consistently above 1.2, clustering coefficient in the range 0.82 to 0.86, and average path length between 1.38 and 1.75. The network densified substantially over the study period, from 203 nodes and 3,694 edges in 1988 to 237 nodes and over 12,000 edges from 2008 onwards. The power-law exponent evolved from α = 2.38 in 1988 to α = 2.60 in 2020, indicating gradual structural shift toward less extreme hub dominance. Louvain community detection identifies three geographically coherent trade blocs — Asia-Pacific, European, and North American — with a Pearson correlation of 0.72 between geographic proximity and bilateral clustering confirming that geography drives trade network topology.
Percolation analysis reveals asymmetric resilience. The network tolerates 60 to 75 per cent random node removal before fragmenting but collapses at only 15 to 20 per cent targeted hub removal, with the largest connected component dropping from 98 per cent to 22 per cent. Cascade simulation under a severity 0.8 shock to China produces an 80 per cent immediate trade loss and a 40-step recovery trajectory under active rerouting. These findings identify the 15 to 20 per cent hub removal threshold as a critical structural vulnerability and provide empirical grounding for supply chain policies around regional diversification, strategic inventory, and backup hub strategies for semiconductors, energy, and pharmaceuticals. The thesis establishes the largest longitudinal trade network dataset yet analyzed at this methodological depth, with all code and data publicly released for reproducibility
For further information please contact Dr. Firas Khatib at fkhatib@umassd.edu.
Zoom: https://us05web.zoom.us/j/86211339649?pwd=36pCDAlu0IGRZsXUB1zGvGGox8mvEY.1
Dr. Firas Khatib
fkhatib@umassd.edu
https://us05web.zoom.us/j/86211339649?pwd=36pCDAlu0IGRZsXUB1zGvGGox8mvEY.1