A Research Note on Discrimination in Mortgage Lending

This paper examines the hypothesis that mortgage lenders rank applications from better to worst and encourage the better ones to apply. A second ranking occurs when the application is ranked by the loan committee and funds are approved from the top of the list until exhausted. A theoretically correct procedure for analyzing the resulting multivariate ordinal data is the little known rank multiple discriminant analysis. Preliminary results have revealed that this technique produces a “best” model with fewer variables and a higher classification rate than the commonly known multiple discriminant analysis, logit, or probit.