An Interaction-Aware Decision Framework for FDIC-Assisted Distressed Bank Acquisitions: Empirical Evidence from the FDIC Failed-Bank Panel and the SVB-First Citizens Case (arXiv (Cornell University))

Summary

• The study evaluates distressed bank acquisition decisions across 55 FDIC-assisted acquisitions conducted between 2008 and 2024 by U.S.-listed acquiring institutions.p.1
• The proposed Spider Network Filtering Framework formalizes four explicit interaction rules among five core dimensions to generate interpretable features for decision models.p.2
• Under nested cross-validation, the Spider-enhanced logistic regression achieves a mean area under the curve of 0.594, compared to 0.615 for basic logistic regression and 0.632 for ridge regression.p.1
• In the case of Silicon Valley Bank's resolution, approximately $72 billion in assets were acquired at a discount of about 23%, alongside $56 billion in deposits assumed by First Citizens.p.2
• Out of the 55 primary panel transactions, 35 generated positive cumulative abnormal returns to acquiring shareholders around the acquisition announcement date.p.9

Key insights

1. Empirical evaluation does not support the hypothesis that domain-derived interaction features improve ranking performance over standard regularized regression benchmarks.p.2
2. Direct use of the deterministic Spider deal score achieves a mean AUC of 0.423, ranking distressed acquisitions no better than chance.p.10
3. Discrimination performance across models depends heavily on isolated clusters, such as the negative outcome shared by four transactions acquired by a single bank.p.10
4. The apparent positive correlation between target financial distress and higher acquirer announcement returns diminishes entirely when measured against a bank-sector benchmark.p.1
5. Standard metric optimization such as maximizing F1 scores leads to excessive false positives and negative utility when false acquisitions carry substantial asymmetric downside costs.p.6
6. Accounting reports available prior to failure failed to capture the sudden liquidity runoff preceding the collapse of Silicon Valley Bank.p.10