Frequentist Ridge Regression and Hierarchical Bayesian Regression Models: An Application to Regional Banks Commercial Real Estate Exposure

Abstract: We show that Frequentist Ridge (and LASSO) regression and hierarchical Bayesian regression models are equivalent in terms of the main predictions of regional banks commercial real estate exposure. In a world of noisy datasets, the penalty parameter in either of these models is critical for regulators to monitor. In the banking policy domain and other social science domains, biased estimates may actually be better for sciences. Keywords: Bayesian Models, Frequentist Models, Commercial Real Estate JEL Classification Number: C11, C21, R33