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

TL;DR

This paper develops an empirically grounded multilayer interbank network for euro area significant banking groups by harmonizing granular supervisory datasets into layer-consistent exposures across long-term credit, short-term credit, securities cross-holdings, short-term funding, and external portfolios. Nodes are enriched with balance-sheet data, enabling layer-aware systemic-risk analytics, and the authors demonstrate that aggregation can obscure channel-specific structure. By extending DebtRank and a micro-structural agent-based contagion framework to operate on real multi-layer exposures, the study shows channel-specific systemic importance and strong sensitivity of short-term liquidity contagion to liquidity buffers. The results highlight the value of layer-aware stress testing for macro- and micro-prudential policy design and provide a data-grounded basis for evaluating multi-channel contagion mechanisms in resilience planning.

Abstract

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.

Figures (9)

• Figure 1: 1. Using ROSSI, RIAD and COREP data a list of banking groups, including their group structure is constructed. 2. A multitude of granular datasets are integrated with banking groups. 3. A multi-layer network is constructed, with nodes depicting banking groups, layers depicting financial markets and edges representing financial relationships between nodes. 4. Further enrichment of the nodes is carried out to represent nodes by their balance sheet.
• Figure 2: Empirical degree distributions across layers (June 2021 snapshot). Left panel: in-degree distributions; right panel: out-degree distributions (x-axes depict the number of edges). Layers are color-coded: short-term credit (dark blue), long-term credit (yellow), cross-securities (red), short-term funding / repo market (light green), external securities (light blue, dubbed overlapping portfolio), and the flattened (aggregated) layer (dark green). Density estimates are obtained via a Gaussian kernel density estimator with bandwidth selected according to Scott's rule scott1992multivariate.
• Figure 3: Centrality distributions across layers (June 2021 snapshot). Rows depict (normalized) centrality measures: PageRank $C^{PR}(i)$, betweenness $C^{b}(i)$, and closeness $C^{c}(i)$. Columns represent layers; x-axes depict centrality values normalized to $[0,1]$. Layers are color-coded: short-term credit (dark blue), long-term credit (yellow), cross-securities (red), short-term funding / repo market (light green), external securities (light blue), and the flattened (aggregated) layer (dark green). The black vertical dashed line in each subplot denotes the median for the corresponding layer--measure pair. Density estimates are obtained via a Gaussian kernel density estimator with bandwidth selected according to Scott's rule scott1992multivariate.
• Figure 4: Top panel: densities of DebtRank values for the long-term credit layer (left in yellow) and cross-securities layer (right in red); x-axes depict DebtRank values in (%). Bottom panel: cumulative counts of the same DebtRank values for the two layers
• Figure 5: DebtRank values for the two layers that are concerned with funding risk, the short-term credit market (top panel, color-coded in blue shades) and the short-term funding market (bottom-panel, color-coded in green shades). First three columns represent DebtRank values for respectively $\beta = 0.05$, $\beta = 0.1$ and $\beta = 0.2$, with $\beta \in [0,1]$ a liquidity buffer scaler with higher $\beta$ values aligned with a higher net cash outflow. The last column shows the cumulative counts of DebtRank (darker shades indicate higher $\beta$ values). x-axes depict DebtRank values in (%) .