Methodology

This information is also contained in a documentation file along with field definitions. Methods and field definitions for 1990–2017 data are stored in a legacy documentation file.

The HMDA Lender File provides information for all lenders that filed a Home Mortgage Disclosure Act (HMDA) report at any time from 2018 through 2024. Each lender’s HMDA ID is matched to information in the Federal Reserve System’s National Information Center (NIC) system, Federal Financial Institutions Examination Council (FFIEC) and National Credit Union Administration (NCUA) Call Reports, and the Federal Home Loan Bank (FHLB) membership directory for each filing year. The HMDA ID is also matched to the FFIEC’s list of Community Reinvestment Act (CRA) wholesale, limited purpose and strategic plan institutions, CRA exam results, and the Ginnie Mae list of approved issuers.

A key objective of the Lender File is to attempt to reconcile the name of each HMDA-filing institution in order to connect institutions under the same parent company and to link the HMDA records to other lender-level data sets. The transmittal sheet accompanying each HMDA filing contains the year, name, address, legal entity identifier (LEI), and tax ID of each filer. In most — but not all — instances, this is sufficient to uniquely identify the filer. There are 10–12 filers each year whose tax ID and/or LEI appear to be inconsistent with their name. For example, the name given might be that of a mortgage company subsidiary, but the tax ID is that of the parent institution. We try to reconcile these inconsistencies with rules that generally prioritize the name. Reconciliation often relies on how the filer reported in other years. Once reconciled, we provide the NIC Research, Statistics, Supervision, and Discount (RSSD) identifier for the filer, their direct parent (if any), and organization “high holder.”

Matches and associations among depositories are derived from the Federal Reserve NIC system. Such matches are more difficult with nondepositories (generally independent mortgage companies). In years past, the NIC system assigned an RSSD identifier to all independent mortgage companies filing HMDA. This has not been the case in recent years, though, resulting in an increasing number of filers that are not in the NIC system. Even when they were assigned a NIC identifier, the NIC system generally did not track associations among nondepositories. We attempt to identify such associations (e.g., joint ownership) using information available on the Internet, but this process is incomplete and surely misses some.

A second feature of the lender file is that we identify the CRA-covered institution (ENTITY) that is likely associated with each filer. If the HMDA filer is a depository, the match is straightforward (ENTITY=RSSD). If the filer is a subsidiary of a bank or thrift, the HMDA filer is matched to the parent institution (the one likely to be subject to a CRA exam). If the filer is a subsidiary of a bank holding company, the filer is matched to the lead (largest) bank of the holding company. HMDA filers are supposed to be in existence at the end of the calendar year. In some cases, however, institutions merge into another institution during the year but still file a separate HMDA report. In this case, the CRA assignment (ENTITY) of the acquiring institution is assigned. Thus, all ENTITY matches that we show are to institutions in existence on December 31 of the filing year (according to the NIC system). Once matched, the institution ID enables us to merge data in from the FFIEC public record of CRA exams. (Credit unions are not subject to CRA exams; nevertheless, we assign them an ENTITY ID using the same criteria as other depositories.)

This HMDA Lender file 2018–2024 reflects the most recent information downloadable for each filer/year as of June 24, 2025, from the FFIEC public HMDA website. For 2025, this is the “snapshot” file as of June 12, 2025 (4,908 filers); for 2023, it reflects the “one year national loan-level” data set as of June 12, 2025 (5,129 filers); for 2022 it reflects the “one-year national loan-level” data set as of May 1, 2024 (4,480 filers); for 2021, it reflects the “three-year national loan-level” data set as of June 12, 2025 (4,376 filers); for 2020, it reflects the “three-year national loan-level” data set as of December 31, 2023 (4,527 filers); for 2019, it reflects the “three-year national loan-level” data set as of December 31, 2022 (5,569 filers); and for 2018, it reflects the “three-year national loan-level” data set as of December 31, 2021 (5,732 filers). A filer will only appear in the data set in a particular year if it filed HMDA data in that year.

Some adjustments were made to the downloaded files. In 2021, a “fake” record with LEI “B90YWS6AFX2LGWOXJ111” was inadvertently left in both the LAR file and the Transmittal files on the FFIEC website. It has been removed here (leaving 4,367 filers for 2021). Also, in filing the transmittal sheets used to determine the filer’s name and address, some filers mixed up the columns (e.g., putting their institution name in the city location field). These have been corrected here.

Some users may be using different versions of the files (e.g., the first-posted “snapshot” files for the years 2018–2023, rather than the more recently updated versions described above). We caution that the updated postings correct errors in the original postings and are thus the most advisable to use. Nevertheless, we also include in the lender file records for any filer that was in the original snapshot file for the years 2018–2023 and give the record counts for the original snapshot filings (LARSNAP). There are five filers in 2018 that were included in the original snapshot but not the most up-to-date files. These filers will show zero entries in the data count variables (i.e., LAR=0 but LARSNAP>0). Filers added after the snapshot posting will have LARSNAP=0 with positive LAR.

Users may want to adjust records for 2018 and 2019 when using the updated files for LEIs “894500SMOMUFH0UZXT46” and “984500DA79C1B97ACF16,” which are duplicated in those LARs (but not the snapshot). The latter appears to be the correct lender code as it is used for 2020–2024 and corresponds to NIC data. For the 2023 one-year file, LEI “254900FR7KYE3QFAIX50” appears to have been replaced by “254900QI30VK43QFSL75,” but records for “254900FR7KYE3QFAIX50” were not removed from either the LAR or TS, resulting in 23 exactly duplicated records. A solution to this would be to remove the records for “254900FR7KYE3QFAIX50.”

HMDA data expanded dramatically in 2018 to cover more fields and more kinds of loans. In addition, the identification fields for reporters changed to LEI, whereas in prior years, the unique key for a reporter in a given year was respondent ID and agency code. For lender information before 2018, see the Legacy HMDA Lender File, also available through the Federal Reserve Bank of Philadelphia’s Consumer Finance Institute.

Finally, we point out that this file is not an “official” record. It is our estimate of lender characteristics derived from the publicly posted official HMDA filings. These are our best estimates, but may, in some cases, be wrong. Please use the file in that spirit.