Answers are grounded in this paper and cite their sources.
Disruptions propagate through inter-firm dependency networks, but when such propagation becomes self-sustaining is unclear. We model disruption and recovery as a Susceptible–Infected–Susceptible process on a directed network, a supplier’s failure traveling downstream and a customer’s upstream at a smaller rate. The spectral disruption number scales exactly as the square root of the backward-to-forward ratio on a layered multi-echelon network, and on any directed network that square-root value is a lower bound, so the symmetric convention overstates persistence by at most the inverse square root of the ratio. A pure propagation lag leaves the endemic state stable for every delay, whereas a delayed collective response destabilizes it through a Hopf bifurcation, whose reduced boundary we check against the exact characteristic equation in 800 settings. Budget-constrained controls satisfying Pontryagin’s necessary conditions cut total cost by 22 to 89 percent on density-matched topologies, and a static allocation aimed at the spectral threshold captures most of that gain. On a 402-industry U.S. input–output network, 2 to 56 percent of material dependencies, by threshold, lie on directed cycles. The other tests use communication, co-purchasing and generated graphs, so they establish internal consistency and structural sensitivity rather than field performance.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: