Evaluating Statistical Models of Mortgage Lending Discrimination: A Bank-Specific Analysis

We present our efforts to develop bank-specific models to test for the presence of mortgage lending discrimination. We discuss the potential for selection and simultaneity biases and delineate the conditions under which a single-equation model is appropriate. The results from three national banks demonstrate that, by incorporating the specific underwriting guidelines of each bank, our alternative approach significantly improves the ability of the model to explain the outcomes of the mortgage lending decision process when compared to a single generic specification applied across all banks. Our results also demonstrate the difficulties encountered in attempting to incorporate the specifics of a bank's underwriting criteria and the remaining potential for omitted-variables problems.