Hold-Capacity and Financial Crisis Survival: Evidence from Three Banking Crises Across 90 Years

This paper introduces Hold-Capacity (H), a composite metric capturing an institution's structural ability to absorb shocks without forced deleveraging, and tests its predictive power across three major banking crises spanning 90 years. We operationalize H using four balance sheet characteristics: capital adequacy, liquidity coverage, inverse leverage, and funding duration. Applying this metric to 38 financial institutions across the 1929 Great Depression (N=10), 2008 Global Financial Crisis (N=15), and 2023 Regional Banking Crisis (N=13), we find that H strongly predicts binary survival outcomes with pooled Spearman ρ = 0.86 (p < 0.001) and classification accuracy of 100%. A critical threshold emerges at H* ≈ 0.43–0.46 across all three epochs with remarkable stability (CV = 6.7%). The 2023 crisis provides striking out-of-sample validation: Silicon Valley Bank's Tier 1 capital ratio of 15.4% exceeded JPMorgan's 15.0%, yet SVB failed while JPMorgan survived—a divergence perfectly captured by their H-scores (0.27 vs. 0.88). The 1929 analysis reveals what occurs when system-wide H approaches zero: without a lender of last resort, even structurally sound institutions cannot survive coordinated runs. These findings suggest Hold-Capacity as a robust crisis resilience indicator with implications for stress testing, capital regulation, and early warning systems.