Differences in treatment of applicants that may constitute a violation of fair lending laws can be detected through the comparative analysis of a sample of loan files. The sample should include a target group of applicants from a protected group most at risk of discrimination on a prohibited basis. These should be compared against a control group of others not at risk. It is necessary to inspect and compare individual loan files for both accepted and rejected applicants to determine differences in the actual treatment of applicants and possible illegal discrimination.
The following guidelines present a step-by-step approach to a comparative analysis of loan files. The focus here is on residential mortgage loans because residential mortgage loan applications provide most of the monitoring information necessary to test for illegal discrimination. However, the procedures also can be used to analyze other types of loan applications, such as consumer loans, to the extent that information contained in the applications would allow. 1 The basic steps suggested provide some examples of how loan product features, underwriting standards, or instances of lender assistance to borrowers can be compared to discern differences in treatment. The more detailed the loan file comparison becomes, the better one will be able to test for discrimination and judge an institution's fair lending performance.
First, decide which group from a prohibited basis category to test. The selection process should begin with a review of the demographics of the institution's community, an analysis of the institution's HMDA data, and discussions with community representatives, including minority real estate brokers and small business owners who can provide insight into the credit needs of minority communities. A "hypothesis" can then be formed about which types of applicants might be discriminated against, in which loan products, and on what prohibited bases.
Focus on groups in each prohibited basis category that have traditionally been disadvantaged and are most at risk of illegal discrimination. For example, when testing for gender discrimination, target women applicants; or when testing for discrimination based on race or national origin, separately target Black, Native American, Asian, or Hispanic applicants. Be careful not to group Non-White applicants as a "minority target group"; rather, compare each ethnic or racial group individually to White applicants. The loan product to consider might be the institution's most popular product in the community or a product line with a high number of denials to minority or other protected class applicants.
The control group should be a sample of applicants who seem least likely to have experienced discrimination on a prohibited basis. This group will usually be white males or white couples but may vary depending on local situations. Compare the treatment of applicants in the target group against this group to determine if the target group received less favorable treatment.
It is important that all applications in the sample be drawn from the same time period to avoid any mitigating effect that changes in product features, underwriting standards or lending personnel would likely have on the test. For institutions subject to the Home Mortgage Disclosure Act (HMDA) the sample can be selected from entries on the institution's HMDA Loan Application Register (LAR).
Include only approved applications in the control group sample 2, as this will provide evidence of the actual standards and practices used to approve loans.
For the target group sample, the population should include rejected, approved, and withdrawn applications. These should be compared against applications in the control group to identify common factors that would upon evaluation indicate disparate or less favorable treatment. Include approved applications and withdrawn applications from members of the target group, in addition to rejected applications, because these may also evidence disparate treatment. For example, members of protected groups may have been approved for loans but on more onerous terms, or applicants from protected groups may have been encouraged to withdraw applications.
Where there are several branches or decision centers, the office locations may affect the type of applicant likely to be attracted to them. For example, if testing for race or ethnic origin discrimination, applications from a branch in an integrated or minority area should be included in the sample with applications from a branch in a predominantly white area to assure an adequate comparison of the treatment of minority and white applicants.
Unless the total number of rejected and withdrawn application files is unusually large, all such files should be included in the target group sample. All rejected and withdrawn applicants of the targeted protected group should be presented on a sample spreadsheet. Once the target group sample is selected, you can calculate the size of the control group. Generally, the control group sample should be at least as large as the target group sample; but if the target group is small, the control group may need to be significantly larger (e.g., four times the size or more) to ensure that there is an adequate basis for comparison.
The size of the sample will vary depending on the size of the institution, volume of loans, and the number of offices. If there is an insufficient loan volume from which to draw a valid sample, the period covered by the sample should be expanded.