Tail Sensitivity of US Bank Net Interest Margins: A Bayesian Penalized Quantile Regression Approach

Bank net interest margins (NIM) have been historically stable in the US on average, but this stability deteriorated in the post-2020 period, particularly in the tails of the distribution. Recent literature disagrees on the extent to which banks hedge interest rate risk, and past literature shows that credit risk and persistence are also important considerations for bank NIM. I use a novel approach to Bayesian dynamic panel quantile regression to document heterogeneity in US bank NIM estimated sensitivities to interest rates, credit risk, and own persistence. I find increased sensitivity to interest rates in the tails of the conditional NIM distribution during the post-2020 period, driven by increased interest rate sensitivities of bank loans and deposits. Density forecast evaluation shows that the model forecasts outperform frequentist benchmark models, and standard tail risk measures show that risks to bank NIM have material implications for bottom-line measures of bank profitability.