ABSTRACT
This chapter reviews the methodology of the logistic regression model and influence statistics. This methodology is applied to the public-use sample of the data used in the Boston Fed study. The Boston Fed study has drawn its share of criticism, but its basic findings have been confirmed by research at the Federal National Mortgage Association and at the Office of the Comptroller of the Currency. Influence statistics will often identify a relatively small set of cases that materially affect the minority coefficient in a loan approval model. For a given mortgage rate, a mortgage lender would be expected to maximize profits by approving good credit risks and denying bad credit risks. The chapter concludes with a summary of findings and suggests how influence statistics could be used with pending mortgage applications to efficiently identify cases that would be highly influential in affecting the minority coefficient, possibly indicating disparate treatment.
