Eigenvector Centrality Reveals Structural Undervaluation in U.S. Interstate Commerce Networks

GDP is the default measure of a state's economic importance. But GDP counts value produced locally and says nothing about a state's structural role in enabling production elsewhere. We construct a weighted, directed network of interstate commodity flows (51 nodes, ~2,500 edges) from the 2017 Commodity Flow Survey and apply eigenvector, betweenness, and out-degree centrality to ask: which states are structurally important to the national trade system in ways GDP does not capture?

Eigenvector centrality produces the clearest signal. The GDP-eigenvector correlation is strong (Spearman rho = 0.93) but 39% of states diverge by five or more rank positions. Physical-economy states are systematically undervalued: Kentucky ranks 28th in GDP but 14th in eigenvector centrality (+14 positions); Mississippi, Indiana, Louisiana, Tennessee, South Carolina, Michigan, and Montana show similar patterns. Their commodity flows sustain production in higher-GDP states, but this enabling role is invisible in output-based accounting.

Service-economy states show the opposite pattern. Massachusetts, Colorado, Minnesota, and Washington, D.C. rank substantially higher in GDP than eigenvector centrality. GDP captures software, finance, and consulting value that does not flow through the physical goods network.

An out-of-sample correlation sharpens the finding: seven of the eight eigenvector overperformers are now attracting over $170 billion in AI data center investment (2024-2026). The same physical infrastructure factors that generate high commodity flow centrality (cheap power, water access, industrial grid capacity, logistics corridors) are driving site selection for the next generation of compute infrastructure. The centrality framework, applied to 2017 data, identifies structural assets that the market has only recently begun to price.

We test boundary sensitivity by comparing the 51-node domestic network against a 52-node network that includes an aggregate international trade node. Eigenvector rankings are robust (rho = 0.982), with gateway states (Washington, Hawaii) gaining and interior bridge states (Missouri, Ohio) declining. Out-degree is nearly invariant (rho = 0.994). Betweenness centrality, however, is substantially sensitive: 96% of states change rank, and interpretation requires care. Because the underlying data records origin-destination flows rather than shipment itineraries, betweenness captures structural bottleneck position in the weighted topology, not literal intermediary routing. Disentangling the mechanical insertion effect of a dominant international node from genuine structural reshuffling remains an open methodological question.

The results suggest that eigenvector centrality applied to commodity flow networks offers a complementary lens to GDP for understanding structural economic importance, with implications for infrastructure policy, fiscal federalism, and the emerging geography of AI compute.