Bank Run Exposure in a Paycheck-to-Paycheck Economy with Loss-Averse Depositors

Battle Count: 87

The paper is primarily focused on bank run theory, liquidity risk management, and financial regulation rather than quantitative trading strategies. However, the behavioral model of loss-averse depositors, the half-Cauchy stochastic process for loss aversion, and the martingale representation of bank run exposure could inform risk management components of trading systems. The Post-SVB event analysis and small-bank vulnerability findings may be relevant for financial sector risk assessment in portfolio management. The paper's focus on endogenous liquidity demand and behavioral amplification of withdrawal pressure is more relevant to banking supervision and systemic risk than to algorithmic trading.

The theoretical model involves advanced mathematics including Radon measures, prospect theory value functions, KKT conditions for constrained optimization, martingale representations, Ornstein-Uhlenbeck diffusions, probability-integral transforms, inverse half-Cauchy quantile constructions, and Itô's formula for SDEs. The empirical implementation requires panel data econometrics with two-way clustered standard errors, bank and quarter fixed effects, interaction terms, event-window analysis, and composite index construction. The full model implementation would require account-level banking data that is typically proprietary. The mathematical derivations span multiple sections with proofs collected in an appendix, and the stochastic process modeling (half-Cauchy SDE) adds significant complexity.

The empirical component uses publicly available FFIEC Call Report data, which enhances reproducibility. However, the theoretical model involves complex mathematical derivations (Radon measures, martingale representations, half-Cauchy SDEs, probability-integral transforms) that require significant mathematical expertise to verify. The paper references an Internet Appendix for technical derivations, simulation details, and diagnostic output, but this is not included in the extract. No code repository is mentioned. The proof-of-concept nature of the empirical work limits definitive replication of the behavioral mechanism.