Journal of Real Estate Finance and Economics, 24:3, 207±237, 2002
# 2002 Kluwer Academic Publishers. Manufactured in The Netherlands.
Anatomy of a Fair Lending Exam: The Uses and
Limitations of Statistics*
PAUL S. CALEM
Board of Governors of the Federal Reserve System, Division of Research and Statistics, 20th and C Streets,
Washington, D.C. 20551
STANLEY D. LONGHOFER
Wichita State University, Barton School of Business, 1845 N. Fairmount, Wichita, KS 67260-0077
In this paper, we examine how statistical analysis is used to help conduct fair lending compliance examinations.
We present a case study of an actual fair lending examination of a large mortgage lender, illustrating how
statistical techniques are used to focus examiner efforts. Our case also highlights the limitations inherent in
statistical analysis of discrimination. The study suggests that statistical analysis and the more-traditional
comparative ®le reviews complement one another in the overall examination process, offsetting some of the
limitations inherent in each.
Key Words: mortgage discrimination, bank examinations, fair lending, logistic regression
In recent years, statistical analysis has played an increasingly important role in the
enforcement of the nation's fair lending laws. Both the Federal Reserve and the Of®ce of
the Comptroller of the Currency (OCC) have incorporated statistical techniques into
regularly scheduled compliance examinations of the depository institutions they supervise
to help identify possible discriminatory patterns in decisions to grant credit and in the
pricing of credit. In addition, statistical analysis has ®gured prominently in recent fair
lending cases pursued by the Justice Department.
Whereas the traditional judgmental examination process tends to focus attention on
individual instances in which minority applicants appear to have been treated differently
than comparable white applicants (which may re¯ect purely random outcomes), statistical
testing in principle can provide evidence of a pattern of discriminatory treatment. Despite
this straightforward rationale, the practice of applying statistical techniques to test for
lending discrimination has been somewhat controversial.
Most pointedly, logistic regression models suchas those used by compliance examiners
have been criticized as being inadequate to represent the complex array of factors
underlying lenders' decisions to accept or reject mortgage loan applications. Indeed, these