Integrating granular data into a multilayer network: an interbank model of the euro area for systemic risk assessment

Micro-structural models of contagion and systemic risk emphasize that shock propagation is inherently multi-channel, spanning counterparty exposures, short-term funding and roll-over risk, securities cross-holdings, and common-asset (fire-sale) spillovers. Empirical implementations, however, often rely on stylized or simulated networks, or focus on a single exposure dimension, reflecting the practical difficulty of reconciling heterogeneous granular collections into a coherent representation with consistent identifiers and consolidation rules. We close part of this gap by constructing an empirically grounded multilayer network for euro area significant banking groups that integrates several supervisory and statistical datasets into layer-consistent exposure matrices defined on a common node set. Each layer corresponds to a distinct transmission channel, long- and short-term credit, securities cross-holdings, short-term secured funding, and overlapping external portfolios, and nodes are enriched with balance-sheet information to support model calibration. We document pronounced cross-layer heterogeneity in connectivity and centrality, and show that an aggregated (flattened) representation can mask economically relevant structure and misidentify the institutions that are systemically important in specific markets. We then illustrate how the resulting network disciplines standard systemic-risk analytics by implementing a centrality-based propagation measure and a micro-structural agent-based framework on real exposures. The approach provides a data-grounded basis for layer-aware systemic-risk assessment and stress testing across multiple dimensions of the banking network.

💡 Research Summary

The paper addresses a critical gap between theoretical multi‑channel contagion models and the empirical measurement of systemic risk in the euro‑area banking sector. By leveraging a suite of granular supervisory and statistical datasets—including AnaCredit (loan‑by‑loan data), EMIR (derivatives contracts), SFTDS (securities financing transactions), MMSR, FINREP/COREP (financial statements), as well as the master register of institutions (RIAD) and the list of significant banking groups (ROSSI) – the authors construct a multilayer interbank network for 114 significant banking groups.

The construction proceeds in four methodological steps. First, a common node set is defined at the banking‑group level, using RIAD identifiers and consolidation rules that aggregate subsidiaries and affiliates. Second, five distinct layers are created, each representing a separate transmission channel: (1) long‑term credit (exposures from AnaCredit), (2) short‑term credit (short‑term loans and EMIR contracts), (3) securities cross‑holdings (CSDB issuance and SHSG holdings), (4) short‑term secured funding (repo and reverse‑repo transactions from SFTDS), and (5) overlapping external portfolios (common‑asset holdings derived from FINREP/COREP). For each layer, directed weighted adjacency matrices are built, preserving exposure amounts, maturities, and collateral status, and all matrices share the same node ordering to enable cross‑layer comparison.

Topological diagnostics reveal pronounced heterogeneity across layers. The short‑term secured‑funding layer exhibits a highly concentrated core‑periphery structure, with a few large banks dominating the network, whereas the securities‑cross‑holding layer is more evenly distributed. Centrality measures (PageRank‑type and DebtRank‑type) differ markedly between layers, indicating that institutions deemed systemically important in one market may be peripheral in another. When the five layers are aggregated into a single “flattened” network, these distinctions disappear, leading to misidentification of systemic importance and under‑estimation of channel‑specific vulnerabilities.

To demonstrate the practical implications of layer‑aware measurement, the authors implement two contagion frameworks on the real‑exposure network. The first is a centrality‑based propagation model: an exogenous shock is applied to a node, and distress is transmitted through weighted edges, with layer‑specific attenuation factors. Simulations show that shocks originating in the long‑term credit layer propagate broadly across all layers, while shocks in the short‑term funding layer tend to be absorbed locally. The second framework is a micro‑structural agent‑based model in which each bank is an agent holding balance‑sheet items, engaging in interbank lending, securities holding, and fire‑sale dynamics. The agent‑based simulations confirm that multilayer interconnections generate complex cascade patterns; a failure in one layer can trigger secondary failures in other layers, amplifying systemic impact.

The paper makes three principal contributions: (1) a reproducible pipeline for integrating heterogeneous supervisory data into a coherent multilayer network, complete with documentation of identifier reconciliation and consolidation rules; (2) empirical evidence of cross‑layer structural heterogeneity and the pitfalls of using aggregated networks for systemic‑risk diagnostics; (3) an illustration of how layer‑aware contagion models alter the identification and magnitude of systemic impact, thereby providing a more nuanced basis for stress‑testing and macro‑prudential supervision. The authors argue that policymakers should incorporate multilayer exposures into stress‑test scenarios and that future research should extend the static snapshot to dynamic, time‑varying multilayer networks and explore cross‑border extensions. Overall, the study offers a data‑grounded, layer‑sensitive framework that bridges the gap between theoretical contagion mechanisms and real‑world systemic‑risk assessment.