Real-world complex systems, including modern economies and financial markets, feature a highly heterogeneous network structure. When a network of firms or banks is subject to supply-demand shocks or financial stresses, the consequences for societal stability crucially depend on initial conditions (which nodes are shocked first) and network topology (who is connected to whom). However, empirical information about these networks is privacy-protected and typically available only at a resolution level that is much coarser (e.g. industry-to-industry) than the actual microscopic level (e.g. firm-to-firm) at which processes of concern take place. Consequently, the outcomes of stress-tests or risk calculations obtained at a coarse-grained level become misleading and uninformative about the actual system-wide response. In this interdisciplinary project, we will establish an international cooperation to develop a statistical-physics renormalization-group approach to consistently represent and model networks at multiple levels of resolution, thus providing rigorous tools to coherently coarse-grain and fine-grain the description of network structure and dynamics. Moreover, we will collaborate with Dutch societal partners (the central bank, the statistical office and the largest private bank) to reconstruct and model, in a consistent yet privacy-preserving way, the Dutch economic network at multiple resolutions and improve tools for risk estimation, bank supervision, and policymaking.
