In banks selected as part of the OCC’s random sample of banks to receive fair lending examinations, select a sample size within the appropriate range based on risk. For banks and focal points selected through the risk-based screening process, use the maximum sample size for the range unless the Compliance Management Review resolves concerns about the specific indications of risk that caused the bank to be selected for examination.
NOTE: Do not use these tables to evaluate focal points that involve credit scoring systems or the results of self-evaluations or of self-tests. Instead, see
“Considering Automated Underwriting and Credit Scoring Risk Factors”
(appendix B) and “Using Self-Tests and Self-Evaluations to Streamline the Examination” (appendix H). Do not use these tables when conducting a pricing examination. See note #1 for sample sizes for pricing examinations.
Table A: Underwriting (Accept/Deny) Comparisons
Sample 1 Sample 2
Prohibited Basis Group Denials Control Group Approvals Number of
Table B: Terms and Conditions Comparisons
Sample 1 Sample 2
Prohibited Basis Group Approvals Control Group Approvals Number of See explanatory notes on the following pages.
Explanatory Notes to Sample Size Tables
1. When performing a pricing examination, conduct a full file review over a specific time range when the pricing criteria were constant. Do not just review loans that received a rate-spread, but all pricing decisions for the specific product being reviewed.
2. When performing both underwriting and terms and conditions comparisons (NOTE: OCC examinations typically should include only one of the
comparisons), use the same control group approval sample for both tasks.
3. If there are fewer than five prohibited basis denials or 20 control group approvals, refer to “Sample Size” instructions in the procedures.
4. “Minimum” and “maximum” sample sizes: select a sample size between the minimum and maximum identified above. Base the size for the sample on the level of risk identified during scoping and the outcome of the compliance management system review. Once the sample size has been determined, select individual transactions judgmentally (refer to procedures). If the
minimum number of approved files called for in a sample-size table exceeds the maximum (as calculated using the table), select the smaller number of files for the approved sample.
5. If two prohibited basis groups (e.g., Black and Native American) are being compared against one control group, select a control group that is five times greater than the larger prohibited basis group sample, up to the maximum.
6. If the bank’s discrimination risk profile identifies significant discrepancies in withdrawal/incomplete activity between the control group and prohibited basis group, or if the number of marginal prohibited basis group files
available for sampling is small, consider supplementing samples by applying the following rules:
• If prohibited basis group withdrawals/incompletes occur after the
applicant has received an offer of credit that includes pricing terms, this is a reporting error under Regulation C (the bank should have reported the application as approved but not accepted), and, therefore, these
applications should be included as prohibited basis group approvals in a terms and conditions comparative file analysis.
• If prohibited basis group incompletes occur due to lack of an applicant response with respect to an item that would give rise to a denial reason, then include these incompletes as denials for that reason when conducting an underwriting comparative file analysis.
Whenever possible, select the sample from the 12-month period immediately preceding the examination, not from an earlier period. In addition,
transactions or classes of transactions of particular interest may be identified to include in the sample. For banks and mortgage companies listed on the final fair lending screening lists each year, use the appropriate HMDA data as follows:
• For banks and mortgage companies listed on the final fair lending screens for Mortgage Lending Underwriting, Terms and Conditions, Rate-Spread Mortgages, and Fed Output Reports, use the HMDA data for the year used to develop the screening lists.
• For banks and mortgage companies listed on the final Redlining and Marketing screen, use the HMDA data and any other data useful for
conducting a redlining and marketing analysis for the year used to develop the screening list.
• For banks on the Random Sample and Credit Card Banks final screens, use the most current data for the focal points identified.
For banks selected in the random sample of banks to be examined, set the sample size based on the estimated risk of discrimination. The more risk factors identified during examination scoping and the weaker the compliance management process, the larger the sample should be within the range.
If no HMDA-LAR for the product exists and the bank is not subject to the Fair Housing Home Loan Data System requirements, request that the bank
estimate or count the numbers of racial and national origin group applications for home purchase, or refinance loans. Alternatively, examiners themselves may count them. (This is feasible because Regulation B requires monitoring information for home purchase and refinance applications.)
Note: Regardless of application volume or sample size, any clear instance of potential disparate treatment – even if the comparison consists of only two files – must be treated as an apparent violation.