BOARD OF GOVERNORS
OF THE
FEDERAL RESERVE SYSTEM
WASHINGTON, D.C. 20551
DIVISION OF CONSUMER
AND COMMUNITY AFFAIRS
CA 05-1
January 31, 2005
TO THE OFFICERS AND MANAGERS IN CHARGE OF CONSUMER AFFAIRS SECTIONS:
SUBJECT: Provisional HMDA Data Sampling Procedures
A recent review of preliminary 2004 HMDA data found numerous instances of errors in some of the new key data fields; therefore we are temporarily revising the sampling procedures for HMDA data. In particular, the review identified issues related to the collection and reporting of race, ethnicity, sex, lien status, Home Ownership and Equity Protection Act ("HOEPA") status, and loan pricing data. To ensure that all data items are being collected and reported correctly, increased sample sizes appear to be necessary. These provisional sampling procedures are to be used until January 31, 2006.
I. Changes to Sample Size
Using the sample sizes outlined in CA 04-4 as a starting point, samples should be reviewed and possibly increased to ensure that loan applications originated by the bank (HMDA action code 1) comprise at least 50 percent of the items in the sample. If the original randomly selected sample includes less than 50 percent of applications originated by the bank, continue randomly selecting applications with action code 1until the number of originations reaches at least 50 percent of the original number of items in the sample. For example:
| HMDA universe: | 100 | |
| Sample size using CA 04-04 guidelines | 39 | |
| Random sample selected: | ||
| Action Code 1 (Originations) | 11 (28 percent) | |
| Action Code 2 (Approved not accepted) | 4 | |
| Action Code 3 (Denied) | 7 | |
| Action Code 4 (Withdrawn) | 4 | |
| Action Code 5 (Incomplete) | 2 | |
| Action Code 6 (Purchased) | 5 | |
| Action Code 7 (Preapproval Denied) | 4 | |
| Action Code 8 (Preapproval Not Accepted) | 2 | |
| Additional originations required for sample: | 9 | |
| Revised sample size: 48 | ||
Please refer to Appendix A to this letter for the revised sampling schedule.
II. Samples for HOEPA Loan Originations
This letter also establishes additional sampling procedures to review HMDA LARs for loans that are subject to HOEPA. This sampling method is designed to determine if the bank's procedures for calculating APR spreads and identifying HOEPA loans are accurate and to ensure that those reported as HOEPA loans, as well as those that are not, have been identified correctly. If the random sample selected for HMDA data verification, as outlined in Section I, does not include enough loans to fulfill the sampling requirements described below, a targeted sample of loans should be selected to meet the minimum requirements. The loan sample developed for these purposes should only be reviewed to determine if the rate spread has been accurately computed and the HOEPA status correctly reported.
For banks that originate less than 10 percent of their applications with APRs above the established thresholds set by HOEPA, a sample size of six loans each of first and subordinate lien loan originations should be reviewed, if possible, for a total of 12 loans (see section 226.32 of Regulation Z for a discussion of the appropriate thresholds). If possible, the sample should include three high cost non-HOEPA loans with APRs within one point below the HOEPA trigger and three HOEPA loans with APRs within one point above the HOEPA trigger for both first and subordinate liens. If the bank does not have enough loans within the one point margin above or below the HOEPA trigger to fulfill the sample requirements, loans with APRs beyond the one-point margin should be selected to achieve the requisite 12. This methodology has been selected because looking at close cases is most likely to reveal whether the creditor is correctly designating HOEPA loans. For both first and subordinate liens, if the bank originated fewer than 3 high cost non-HOEPA loans with APRs below the HOEPA threshold or fewer than 3 loans with APRs above the HOEPA threshold, 100 percent of the loans in that category should be reviewed.
If the bank originates more than 10 percent of its loans with APRs above the HOEPA thresholds, a minimum sample of 10 loans each of first and subordinate lien loan originations should be reviewed, if possible, for a total of 20 loans. If possible, the sample should include five high cost non-HOEPA loans with APRs within one point below the HOEPA trigger and five HOEPA loans with APRs within one point above the HOEPA trigger for both first and subordinate liens. If the bank does not have enough loans within the one point margin above or below the HOEPA trigger to fulfill the sample requirements, loans with APRs beyond the one-point margin should be selected. For both first and subordinate liens, if the bank originated fewer than 5 high cost non-HOEPA loans with APRs below the HOEPA thresholds or fewer than 5 loans with APRs above the HOEPA threshold (but nonetheless meets the 10 percent criterion), 100 percent of the loans in that category should be reviewed.
Questions regarding the revised sampling procedures should be referred to the review examiner assigned to your district.
Sincerely,
(signed)
Suzanne G. Killian
Assistant Director