Time-Varying Volatility of Bank Betas

Battle Count: 65

The paper identifies a previously unmeasured source of unhedgeable risk in bank stocks (interest expense beta uncertainty) that is negatively priced by the market. A one standard deviation increase in this uncertainty reduces large bank stock values by 2.34% (~$47 billion for top 10 banks). This finding is relevant for: (1) factor models for financial sector equities, (2) risk management of bank stock portfolios, (3) event-driven strategies around monetary policy changes, (4) understanding the transmission of interest rate shocks to bank equity. However, the paper is more academic/descriptive than directly actionable for trading strategies, and the signal is at quarterly frequency.

The core methodology (Kalman filter with time-varying parameters) is well-established in econometrics and available in standard packages (e.g., R's dlm, Python's filterpy, MATLAB). The main complexity lies in: (1) constructing the FDIC Call Report dataset across 10 deciles with proper variable processing, (2) specifying the state-space model correctly with random walk dynamics, (3) extracting conditional variance from the Kalman filter, (4) running multiple Granger causality tests across deciles, and (5) the market pricing regression. The paper provides clear equations but no code. Estimated implementation time: 2-4 weeks for an experienced econometrician.

Data sources are publicly available (FDIC Call Reports via BankFind, FRED database for Federal Funds rate). Methods are standard econometric techniques (Kalman filter, Granger causality, OLS). The paper provides detailed equations and specifies the sample period (Oct 1992 - Jun 2024) and decile structure. However, no code or repository is provided, and the specific FDIC variable codes and processing steps (e.g., differencing cumulative annual interest expense) would need to be replicated manually.