Lender Register

Methodology

Every figure on this site is computed from public files: the HMDA loan/application register (LAR) and transmittal sheets published by the FFIEC and CFPB, the HMDA Data Browser aggregation API, the VA lender statistics workbooks, and Census Bureau population and household estimates. This build processed 12,259,199 LAR records for 2024 on 30 Sep 2026.

Applications, originations and purchased loans

HMDA's action-taken code says what happened to each record. This site counts codes 1 to 5 as applications: loan originated (1), application approved but not accepted (2), application denied (3), application withdrawn by the applicant (4) and file closed for incompleteness (5). Purchased loans (6) are loans a lender bought from another institution; they are reported and shown separately, never counted as applications. Preapproval requests denied or approved but not accepted (7, 8) are shown in the action-taken table only.

Denial rate

Denial rate is denied applications (code 3) divided by decided applications (codes 1, 2 and 3): the share of applications on which the lender made a credit decision that ended in a denial. Withdrawn applications and incomplete files are excluded from the denominator because the lender did not decide them. Where a table says "denials ÷ applications" the denominator is all applications of that purpose or type (codes 1 to 5), because the source series does not break decisions out at that level; those figures are lower than the denial rate and are labelled accordingly.

The CFPB's caveat applies to every denial rate on this site: HMDA data alone cannot be used to determine whether a lender is complying with fair lending laws. The data do not include some legitimate credit risk considerations for loan approval and loan pricing decisions. A higher or lower rate reflects the applicants a lender receives, the products it offers and its underwriting, none of which the file describes.

Rate spread

Rate spread is reported by the lender as the difference between the loan's annual percentage rate and the average prime offer rate (APOR) for a comparable transaction, in percentage points. It is reported on originated loans and on approved-but-not-accepted applications; this site uses originated loans with a numeric spread. The median and quartiles are computed from a histogram in 0.05-point bins. "1.5 pp or more" is the share of those loans whose spread is at or above 1.5 points, the Regulation C threshold for a higher-priced first-lien loan; it is applied to all originations here without distinguishing lien status, so it is a distribution measure, not a count of higher-priced loans.

Suppression

Any cell with fewer than 25 applications is withheld: rates, medians, loan mix and denial reasons are replaced by a dash and the note "fewer than 25 applications". Denial-reason breakdowns need at least 25 denials. Counts of applications and originations are shown for all cells because they are the published record.

Comparisons and normalisation

Lender pages compare the lender with its largest state (by applications) and the United States; lender-in-state pages add the lender's own national figures. County and state pages show originations per 1,000 households, using the Census Bureau's American Community Survey 2024 five-year household estimate for the county or state. Population is the Vintage 2024 estimate. Loan mix shares are shares of originated loans.

Trend series

Lender, lender-in-state and county trends come from the public LAR for 2022, 2023, 2024, processed the same way each year. State and national trends for 2018–2024 come from the HMDA Data Browser aggregation API, queried by state and year for action taken by loan type and by loan purpose. Each lender's page also charts the number of LAR records on its transmittal sheet for every year since 2018; that count includes purchased loans and preapproval requests, so it is larger than the applications figure.

Refresh cadence

HMDA is annual. Each year's LAR is streamed once, and the resulting aggregates (never loan-level records) are kept as a snapshot that later builds reuse; the site checks the FFIEC file server on every build and processes a new year as soon as it is published, typically the modified LAR in spring and the national snapshot in early summer. VA monthly workbooks, the Data Browser series and Census estimates are refreshed on every build, so the "data through" date in the footer follows the latest VA month.

Lender identity and names

A lender is its Legal Entity Identifier (LEI). The name shown is the respondent name on its most recent transmittal sheet, with all-capitals names converted to title case; the LEI is printed on every lender page. Names that read as a natural person rather than an institution are dropped at ingest. Mergers and renamings mean an LEI can carry a new name while earlier years still show the old filer; the site does not merge LEIs.

VA lender statistics

VA publishes loans guaranteed per lender per month, nationally, split into purchase, cash-out refinance and IRRRL. Lender names in those workbooks are matched to HMDA filers by normalising both names (case, punctuation and common abbreviations such as MTG and FCU); 1225 of 4406 VA names match, covering 68% of loans in the series. Unmatched names appear on VA monthly pages only. VA counts loans by guarantee date, HMDA by application year, so the two series are not expected to agree.

Thresholds and indexing

Lender pages are built for lenders with at least 25 applications in 2024 and indexed when they have at least 100. Lender-in-state pages need 100 applications in that state; county rows on lender pages need 50. County and state pages are built for every county and state in the file. FHA monthly figures from HUD's Single Family Portfolio Snapshot are not included because HUD's website blocks automated downloads; FHA context is taken from HMDA loan type 2 instead.

Why figures can differ from the Data Browser