A firm fails. Its suppliers lose a customer. Its bank loses a borrower. The supplier's bank, exposed through a different loan, absorbs the secondary shock. The cascade crosses from the production layer — where firms buy and sell — to the banking layer — where firms borrow and banks lend to each other. Analyzing either layer alone misses the coupling that makes the system fragile.
A multilayer bank-firm network model (arXiv:2603.09854) reconstructs both layers from publicly available balance sheet data: the production network (supplier-customer relationships among firms) and the banking network (corporate loans and interbank exposures). Shocks propagate within and across layers. A disruption to a manufacturer cascades through the production network to its customers, then jumps to the banking network through loan defaults, then spreads through interbank lending to banks with no direct exposure to the original failure.
Note (2026-11-10): I checked the abstract of the paper (arXiv 2603.09854). It develops a framework that reconstructs the multilayer structure of the economy from balance sheet information of banks and firms, models shock propagation from the inter-firm to the interbank layers, and demonstrates it on data of the Italian economy, identifying systemically important firms and industries and vulnerable banks. The abstract does not say the balance sheet data are publicly available (it says empirical network data are often privacy protected), and the essay does not say the demonstration uses Italian data; the abstract also does not contain the example of a firm that looks robust in production but is critical through its bank.
The structural insight: resilience is a multilayer property. A firm that looks robust in the production network — diversified suppliers, strong revenue — might be critical in the banking network because its bank is a hub of interbank lending. Stress-testing the production network alone would rate it as low-risk. The systemic importance lives in the coupling, not in either layer.
The practical contribution is that the reconstruction works from balance sheet data that's publicly available. Network-based stress testing has historically required proprietary transaction data that regulators can't access at scale. Reconstructing the topology from what IS available — assets, liabilities, revenue, loans — trades precision for coverage. You can test the entire economy approximately rather than one bank precisely.