The Fannie Mae Office of Housing Research has confirmed the findings of the 1992 Boston Federal Reserve Bank study that found widespread lending discrimination in the Boston area.

The office set out to determine the validity of the data to improve the effectiveness of community lending initiatives and to address recent attempts to discredit the study.

The Boston Fed was criticized for inaccurately coded data and omitted variables. Although these shortcomings have been of legitimate concern, they do not represent fatal flaws in the research. In fact, closer examination of the data reinforces the Boston Fed's original findings.

A resurgence of interest in lending discrimination culminated in the 1989 overhaul of the Home Mortgage Disclosure Act data requirements, resulting in the availability of extensive information on individual loan applications.

Dramatic Finding

Although some observers suspected discrimination, the HMDA measures of applicant creditworthiness are not enough to gauge discrimination with certainty.

The Boston Fed enhanced the 1991 Boston-area HMDA data with additional applicant information and reported that it reviewed 3,062 loan applications from throughout the Boston area. Even after controlling for all objective indicators of applicant risk, it discovered that lenders rejected minorities 56% more often than identical white applicants.

Each step in the Boston Fed study has been criticized as improperly influencing the results.

In a study summarized in The Wall Street Journal, Stan Liebowitz and Ted Day alleged that typographical errors in the Boston Fed data set accounted for the negative race effect. The authors claimed that when they excluded records with inconsistent entries or extreme observations such as multimillion-dollar loans and net worths, the estimated effect of race disappeared.

The Fannie Mae analysis showed that the Boston Fed data did indeed contain a large number of inaccurately coded or atypical observations. Fannie Mae reestimated the study with clean data only, and found that the race effect persisted.

A study by Mark Zandi, summarized in American Banker, typifies another kind of criticism. He notes that while the Boston Fed collected a wide array of risk-related variables, it used only 12 of them in the final analysis.

Mr. Zandi found that one excluded variable in particular - the lender's subjective assessment of the application vis-a-vis institution credit guidelines - was an important determinant of rejection. When this variable was included as a control for legitimate application risk, the race effect largely disappeared.

In other words, Mr. Zandi claimed that the Boston Fed study is plagued by bias resulting from omitted variables, the primary problem it intended to overcome.

Mr. Zandi reestimated the Boston Fed study and contended that omission of a variable assessing applicant credit risk was responsible for the race effect.

The Fannie Mae analysis confirmed that this credit variable was in fact an important determinant of the race effect. But Mr. Zandi failed to note that the variable is itself affected by discrimination. For example, a white applicant with slow-pay credit accounts is considered creditworthy, while a similar African-American applicant is not.

The Boston Fed study clearly demonstrated disparate treatment in the application phase.

Interpretation Challenged

Other criticisms have questioned the interpretation of the Boston Fed results.

The study's conclusion on racial discrimination in mortgage lending was based on a technique that holds that for any given set of risk controls, it is possible to evaluate the expected probability of rejection of a white or a minority applicant.

Using the average white applicant in the Boston area as the reference applicant, the Boston Fed set the expected rejection rate for a white applicant at 11%, and for an African-American or Hispanic applicant at 17%.

Mr. Zandi asserted that the more appropriate reference applicant in the Boston area is the average minority applicant. Mr. Zandi claimed that using the average white applicant substantially altered the results. However, from both a theoretical and practical standpoint, the Boston Fed was correct in using the white applicant as the reference.

Consequently, the treatment of white applicants should be the benchmark for measuring the extent of discrimination in mortgage lending. The accompanying table describes the three broad types of discrimination that occur in mortgage lending: blatant, disparate treatment, and adverse impact.

Scope Misrepresented

The criticisms of Gary Becker in Business Week, Peter Brimelow in National Review, and Mr. Brimelow and Leslie Spencer in Forbes fundamentally misrepresented the scope of the Boston Fed report.

Using the average-default methodology, which many researchers have demonstrated to be defective Mr. Becker, Mr. Brimelow, and Mr. Spencer maintained that discrimination against minorities met a legitimate business purpose.

Because Boston-area default rates among whites and minorities were equal, the authors reasoned that lenders were properly evaluating the true default risk of whites and minorities.

This argument does not contradict the Boston Fed findings, which made no claims about adverse impact. The Boston Fed examined only application processing and focused only on disparate treatment.

Mr. Becker, Mr. Brimelow, and Mr. Spencer criticized the Boston Fed study for providing questionable evidence of adverse-impact discrimination, but this type of discrimination was not considered by the study.

With the validity of the Boston Fed data firmly established, the information can now be used to move beyond the question of whether discrimination exists to improving products and services for low-income and minority households and communities.

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.