Explainable Graph Neural Networks for Interbank Contagion Surveillance: A Regulatory-Aligned Framework for the U.S. Banking Sector

2026-04-14

AI Summary

This study addresses contagion risk in the U.S. banking system by proposing an interpretable spatiotemporal graph attention network (ST-GAT) framework for early warning and macroprudential supervision. Leveraging quarterly data from 8,103 FDIC-insured institutions between 2010Q1 and 2024Q2, the authors reconstruct a dynamic directed weighted graph via maximum entropy estimation and integrate bidirectional LSTM with graph attention mechanisms, embedding regulatory-logic-aligned interpretability. The model achieves state-of-the-art performance among graph neural networks with an AUPRC of 0.939 ± 0.010, approaching XGBoost’s accuracy. It identifies return on assets (ROA) and non-performing loan ratio as key predictive factors and successfully provides ex post interpretability for the 2023 regional banking crisis, revealing a persistent association between temporal attention weights and structural vulnerabilities.

Technology Category

• Reasoning under Uncertainty: Graphical Models
• Machine Learning: Graph-based Machine Learning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

The Spatial-Temporal Graph Attention Network (ST-GAT) framework was created to serve as an explainable GNN-based solution for detecting bank distress early warning signs and for conducting macro-prudential surveillance of the interbank system in the United States. The ST-GAT framework models 8,103 FDIC insured institutions across 58 quarterly snapshots (2010Q1-2024Q2). Bilateral exposures were reconstructed from publicly available FDIC Call Reports using maximum entropy estimation to produce a dynamic directed weighted graph. The framework achieves the highest AUPRC among all GNN architectures (0.939 +/- 0.010), trailing only XGBoost (0.944). Ablation analysis confirms the BiLSTM temporal component contributes +0.020 AUPRC; temporal attention weights exhibit a monotonically decreasing pattern consistent with long-run structural vulnerability weighting. Permutation importance identifies ROA (0.309) and NPL Ratio (0.252) as dominant predictors, consistent with post-mortem analyses of the 2023 regional banking crisis. All data are publicly available FDIC Call Reports and FRED series; all code and results are released.

Problem

• Interbank Contagion
• Explainable AI
• Macro-prudential Surveillance
• Bank Distress
• Graph Neural Networks

Innovation

• Explainable GNN
• ST-GAT
• Interbank Contagion
• Macro-prudential Surveillance
• Maximum Entropy Estimation