The paper tackles a fundamental challenge in economics: understanding how shocks and risks ripple through large, complex networks of interconnected firms, banks, or other economic actors. Think of supply chains where companies depend on each other, or financial systems where banks are exposed to one another's debts. When these networks are large, computing exactly what happens when something goes wrong becomes extremely difficult. The authors develop a mathematical framework that makes this tractable by showing that, for large "sparse" networks (where each participant connects to only a relatively small number of others), you can approximate the behavior of the whole system by zooming in on local neighborhoods around each node. The key mathematical tool is something called local weak convergence, which formalizes the idea that the network looks increasingly tree-like when you examine small patches of it.
The main theoretical results show that, as the network grows large, the distribution of outcomes across all participants converges to something you can calculate using only these local, tree-structured approximations. This holds under two sets of conditions covering different economic situations: one where responses to shocks shrink predictably, and a more subtle case where the system might have multiple possible equilibria but the local approximation still pins down the right answer regardless of which equilibrium actually forms. This second result is particularly striking because it means the local method is robust even in ambiguous situations where economic theory does not give a unique prediction.
The authors then extend these results to risk measurement, specifically showing that standard financial risk metrics like conditional value-at-risk (which measures expected losses in bad scenarios) can also be accurately approximated through local calculations. They test their approach on two real-world inspired examples: production networks, where firms buy inputs from each other, and payment clearing systems, where financial institutions settle mutual obligations. In both cases, the local approximation matches full-network calculations closely while being far cheaper to compute, suggesting this framework could be practically useful for regulators and analysts trying to assess systemic risk in large financial or economic systems.