Home > Mortgage Lending > Methodology
Mortgage Lending Methodology & Technical Specification
Document Scope: Vertical 1 — Residential Mortgage Lending Transparency
Primary Reporting Year: 2025
Comparison Reporting Year: 2024
Last Updated: September 2026
1. Introduction & Methodology Philosophy
OpenFileData.com is committed to complete technical transparency and absolute reproducibility. Every rate, count, dollar total, ranking, and distribution published across our mortgage lending pages is computed from official federal government records using documented, deterministic algorithms.
This methodology document explains: 1. The exact federal datasets and freeze dates used. 2. The analytical filters used to isolate consumer residential mortgages. 3. The mathematical formulas used to derive our metrics. 4. The statistical rules and disclosures governing small samples and data limitations.
2. Official Data Sources & Annual Datasets
All mortgage data on OpenFileData.com originates from public records collected under the Home Mortgage Disclosure Act (HMDA) of 1975 (as amended) and published pursuant to Regulation C (12 CFR Part 1003) by the Federal Financial Institutions Examination Council (FFIEC) and the Consumer Financial Protection Bureau (CFPB).
A. Primary Dataset (2025)
- Official Publication Name: 2025 Snapshot National Loan-Level Dataset
- Publishing Agency: FFIEC / CFPB
- Snapshot Freeze Date: June 2, 2026
- Official Endpoint / Distribution URL:
https://ffiec.cfpb.gov/data-publication/snapshot-national-loan-level-dataset/2025
- Included Public Files:
2025_public_lar_csv.zip: Loan Application Register (LAR) containing individual loan-level records with official CFPB-derived fields.2025_public_ts_csv.zip: Transmittal Sheet (TS) containing legal institution names, addresses, supervisory agency codes, and total loan counts for all reporting institutions.2025_public_msamd_csv.zip: Reference table of Metropolitan Statistical Area / Metropolitan Division (MSA/MD) numeric codes and names.- Dataset Characteristics & Timing:
The 2025 Snapshot provides full-calendar-year 2025 loan-level records available at the June 2, 2026 snapshot freeze. It represents the most current complete-calendar-year HMDA dataset available as of September 2026. The Snapshot may differ from the eventual One-Year Dataset because institutions may submit revisions after the Snapshot freeze date. Later revisions will be incorporated into the eventual 2025 One-Year Dataset when released by federal regulators.
B. Comparison Dataset (2024 Year-over-Year Baseline)
- Official Publication Name: 2024 One-Year National Loan-Level Dataset
- Publishing Agency: FFIEC / CFPB
- Freeze Date: June 2, 2026
- Official Endpoint / Distribution URL:
https://ffiec.cfpb.gov/data-publication/one-year-national-loan-level-dataset/2024
- Role: Serves as the historical baseline for multi-year trend analysis. The One-Year dataset incorporates revisions, corrections, and late submissions made by reporting institutions during the 12 months following initial submission.
C. Policy on Current-Year (2026) Data
- Full-calendar-year 2026 HMDA data is currently in collection by federal regulators and is scheduled for initial snapshot release in mid-2027.
- Certain large depository institutions submit quarterly HMDA data during 2026. However, quarterly data covers only a small minority of total mortgage lenders and excludes independent mortgage banks that originate a substantial portion of consumer loans. Therefore, quarterly data is not used as an annual baseline on OpenFileData.
3. Scope: Defined Application Universes
The raw national HMDA dataset contains tens of millions of records spanning diverse credit activities, including commercial apartment financing, farm loans, subordinate home equity lines, second homes, and secondary market portfolio trades.
To produce meaningful consumer lending statistics, OpenFileData separates raw records into two explicitly defined analytical tiers:
[ All Raw HMDA Records ]
│
▼ (Apply 8 Structural Consumer Filters)
[ CONSUMER MORTGAGE UNIVERSE ] (Actions 1, 2, 3, 4, 5)
│
▼ (Exclude Withdrawn [4] & Incomplete [5])
[ ACTIONABLE APPLICATION UNIVERSE ] (Actions 1, 2, 3)
├── Approvals: Action 1 (Originated) + Action 2 (Approved, Not Accepted)
└── Denials: Action 3 (Denied)A. The Consumer Mortgage Universe (Actions 1–5)
The Consumer Mortgage Universe consists of all loan application records that satisfy our 8 structural consumer filters: 1. Dwelling Category: 1-to-4 family residential properties (derived_dwelling_category LIKE 'Single Family (1-4 Units)%' OR total_units IN ('1', '2', '3', '4')). Excludes 5+ unit commercial multifamily complexes. 2. Construction Method: Site-built or manufactured residential housing (construction_method IN (1, 2)). 3. Occupancy: Principal residence only (occupancy_type = 1). Excludes second homes and investor rental properties. 4. Lien Priority: First lien only (lien_status = 1). Excludes subordinate/junior liens. 5. Loan Purpose: Home purchase (loan_purpose = 1), Refinancing (loan_purpose = 31), or Cash-out refinancing (loan_purpose = 32). 6. Consumer Purpose: Non-commercial purpose (business_or_commercial_purpose = 2). Excludes commercial LLC/entity business loans. 7. Credit Product: Closed-end mortgages (reverse_mortgage != 1 AND open-end_line_of_credit != 1). Excludes reverse mortgages and open-end lines of credit (HELOCs). 8. Exclusion of Non-Retail Applications: Excludes secondary market loan purchases (action_taken = 6) and preapproval commitment requests (action_taken IN (7, 8)).
Records Retained: Applications resulting in Action 1, Action 2, Action 3, Action 4 (withdrawn by applicant), and Action 5 (closed for incompleteness) are retained in this universe. Preserving Actions 4 and 5 ensures complete transparency regarding total consumer application volume, applicant attrition, and application paperwork completion rates.
B. The Actionable Application Universe (Actions 1, 2, and 3 Only)
The Actionable Application Universe is the subset of the Consumer Mortgage Universe where the lending institution reached a formal, definitive credit underwriting decision:
- Action 1: Loan originated.
- Action 2: Application approved but not accepted by borrower.
- Action 3: Application denied by lender.
Actions 4 and 5 are strictly excluded from this universe.
4. Treatment of Official HMDA Action Taken Codes
Under Regulation C, reporting institutions assign one of eight numeric codes to the action_taken field:
| Code | Official HMDA Description | OpenFileData Treatment | Analytical Justification |
|---|---|---|---|
| 1 | Loan originated | Numerator & Denominator | Application approved by lender, accepted by borrower, and funds disbursed. |
| 2 | Application approved but not accepted | Numerator & Denominator | Lender completed credit evaluation and approved the loan, but the applicant chose not to proceed. Under fair-lending analytical conventions, this represents a formal credit approval. |
| 3 | Application denied | Denominator (Numerator for Denial Rate) | Lender evaluated creditworthiness and issued a formal adverse action/denial notice. |
| 4 | Application withdrawn by applicant | Excluded from Credit Decision Denominator | Applicant withdrew the application before a formal credit decision was made. The lender neither approved nor denied credit. |
| 5 | File closed for incompleteness | Excluded from Credit Decision Denominator | File closed because the applicant failed to submit required verification materials after written notice. No credit decision was made. |
| 6 | Purchased loan | Excluded from Retail Applications | Represents previously originated loans purchased on the secondary market from other lenders. Does not reflect an application submitted to or evaluated by the reporting lender. |
| 7 | Preapproval request denied | Excluded from Closed-End Application Rates | Preapproval requests are optional requests for a preliminary credit commitment prior to property identification. |
| 8 | Preapproval approved but not accepted | Excluded from Closed-End Application Rates | Preliminary preapproval commitment approved by lender but not accepted by applicant. |
5. Derived Metrics & Mathematical Formulas
### ⚠️ Notice Regarding Official Government Terminology
OpenFileData’s Approval Rate and Denial Rate are our derived analytical metrics.
The federal government (CFPB/FFIEC) and Regulation C collect individualaction_takencodes, but do not publish or define an official government “approval rate.”
Our formulas reflect standard public-interest and fair-lending analysis conventions, documented below.
A. Approval Rate (%)
Measures the proportion of actionable applications where the lender reached a positive credit decision:
$$\text{Approval Rate (\%)} = \left( \frac{\text{Count}(\text{action\_taken} \in \{1, 2\})}{\text{Count}(\text{action\_taken} \in \{1, 2, 3\})} \right) \times 100$$
B. Denial Rate (%)
Measures the proportion of actionable applications where the lender issued a formal credit denial:
$$\text{Denial Rate (\%)} = \left( \frac{\text{Count}(\text{action\_taken} = 3)}{\text{Count}(\text{action\_taken} \in \{1, 2, 3\})} \right) \times 100$$
- Mathematical Complement: Under OpenFileData’s analytical definition, $\text{Approval Rate (\%)} + \text{Denial Rate (\%)} = 100.00\%$ exactly.
- Why Actions 4 and 5 Are Excluded:
Including withdrawn applications (Action 4) or incomplete files (Action 5) in the denominator would distort approval and denial rates by conflating borrower-initiated departures or missing paperwork with lender credit decisions. Actions 4 and 5 are tracked separately as procedural metrics within the broader Consumer Mortgage Universe.
C. Origination Rate (%)
Measures the proportion of actionable applications that culminated in an active, funded loan:
$$\text{Origination Rate (\%)} = \left( \frac{\text{Count}(\text{action\_taken} = 1)}{\text{Count}(\text{action\_taken} \in \{1, 2, 3\})} \right) \times 100$$
D. Loan Amounts & Dollar Volume
To prevent distorted averages caused by denied or hypothetical loan amounts, all loan size statistics are calculated strictly from originated loans (action_taken = 1):
- Average Loan Amount: Arithmetic mean of
loan_amountfor originated loans. - Median Loan Amount: Middle value of
loan_amountfor originated loans. Median is highlighted prominently on OpenFileData because it is resistant to luxury outlier skew. - Total Originated Dollar Volume: Sum of
loan_amountfor all originated loans in the geography or institution.
6. Demographic & Income Analysis Methodology
A. Household Income Brackets
In HMDA records, applicant income is reported in thousands of dollars (income), representing the gross annual income relied upon by the lender during underwriting.
OpenFileData groups applications into five consumer-facing income tiers: 1. Under \$50,000: income < 50 2. \$50,000 – \$99,999: income >= 50 AND income < 100 3. \$100,000 – \$149,999: income >= 100 AND income < 150 4. \$150,000 or more: income >= 150 5. Income Not Available / Entity: Applications where income is reported as 'NA', missing (NULL), or negative (business operating loss).
- Analytical Clarification: These fixed dollar bins are OpenFileData analytical categories designed for intuitive consumer comprehension. They are not official statutory HMDA income categories (which often rely on local Area Median Income ratios).
B. Race and Ethnicity Classifications
To evaluate fair lending patterns, OpenFileData utilizes the CFPB's official derived demographic fields (derived_race and derived_ethnicity), which synthesize primary applicant and co-applicant disclosures according to standard federal civil rights categories: 1. Hispanic or Latino: Any applicant identifying as Hispanic or Latino, regardless of race. 2. White (Non-Hispanic): Applicant identified as White and not Hispanic. 3. Black or African American (Non-Hispanic): Applicant identified as Black or African American and not Hispanic. 4. Asian (Non-Hispanic): Applicant identified as Asian and not Hispanic. 5. American Indian or Alaska Native (Non-Hispanic): Applicant identified as American Indian or Alaska Native and not Hispanic. 6. Native Hawaiian or Other Pacific Islander (Non-Hispanic): Applicant identified as Native Hawaiian or Other Pacific Islander and not Hispanic. 7. 2 or More Minority Races (Non-Hispanic): Applicant identified with more than one minority racial group and not Hispanic. 8. Joint: One applicant reported as White Non-Hispanic and the co-applicant reported as a minority race or Hispanic. 9. Race/Ethnicity Not Available: Applications where the borrower opted not to disclose demographic information or the application was completed entirely via mail/internet without visual observation.
- Essential Methodology Guardrail:
These categories report empirical differences in historical outcomes. They do not establish causation or explain why approval or denial differences occur. HMDA records do not contain borrower credit scores, credit tier histories, down payment percentages, or asset reserve levels. Differences in rates between demographic groups reflect gross outcomes across the reporting population and must not be interpreted as definitive proof of lender bias or creditworthiness.
7. Institutional Mapping & Lender Rankings
A. Institution Legal Identification
- Every reporting lender is uniquely identified in HMDA by a 20-character alphanumeric Legal Entity Identifier (LEI) pursuant to ISO 17442.
- OpenFileData maps each LEI to its official legal corporate name by joining records to the FFIEC's 2025 Public Transmittal Sheet (
2025_public_ts_csv). - If an institution submitted loan records but lacks a name in the Transmittal Sheet, it is transparently designated as
"Unlisted Institution [LEI]".
B. Ranking Criteria
- State and metropolitan lender tables rank institutions by Originated Loan Count (
action_taken = 1) descending. - Rankings reflect total retail consumer mortgage volume. They do not constitute an endorsement, recommendation, or qualitative rating.
C. Multi-Year Organizational Shifts
Lending institutions frequently undergo mergers, acquisitions, charter changes, or corporate restructuring across calendar years. When comparing 2025 and 2024 institutional activity, OpenFileData matches records strictly by LEI. Readers should note that corporate acquisitions or portfolio transfers can cause substantial year-over-year volume changes for specific entities.
8. Year-over-Year (YoY) Comparison Standards
When comparing the primary year (2025) with the baseline year (2024):
- Volume Metrics (Applications, Originated Loans, Dollar Volume):
Reported as percentage change: $$\text{Volume Change (\%)} = \left( \frac{\text{Volume}_{2025} - \text{Volume}_{2024}}{\text{Volume}_{2024}} \right) \times 100$$
- Rate Metrics (Approval Rate, Denial Rate):
Reported as arithmetic percentage point (pp) deltas: $$\text{Rate Delta (pp)} = \text{Rate}_{2025} - \text{Rate}_{2024}$$
- Dataset Asymmetry Limitation:
The 2025 baseline represents a Snapshot dataset frozen on June 2, 2026, whereas the 2024 baseline represents a One-Year dataset that incorporates 12 months of institutional revisions. While comparable for broad market analysis, minor differences may reflect post-snapshot resubmissions.
9. Statistical Reliability & Small Sample Handling
Calculating percentages on very small sample sizes creates mathematical volatility. In a rural county or for a small demographic group with only 5 applications, a single denial shifts the denial rate by 20.0 percentage points.
To protect readers from misinterpreting high-variance statistics:
- The 30-Application Threshold: Any table, cell, or group with fewer than 30 actionable applications is explicitly flagged with the notation:
"Fewer than 30 applications; rate subject to high sample variance."
- Analytical Restraint: OpenFileData does not rank, highlight, or draw comparative conclusions from samples below this threshold.
10. Summary of Data Limitations & Disclosures
Users of OpenFileData must keep the following structural boundaries in mind:
- Absence of Credit Scores:
Under federal privacy laws, applicant credit scores (FICO, VantageScore) are redacted from public HMDA files. Because credit scores are a fundamental underwriting factor, gross denial rates cannot be adjusted for credit score differences. 2. Missing & Exempt Data Fields: Under the Economic Growth, Regulatory Relief, and Consumer Protection Act (EGRRCPA), smaller banks and credit unions originating fewer than 500 closed-end loans annually are exempt from reporting certain detailed data fields, including specific denial reasons, credit scores, and debt-to-income ratios. 3. No Causal Inference: Differences in denial rates across lenders, geographies, income brackets, or racial groups reflect historical reporting patterns. They do not demonstrate unlawful discrimination, nor do they prove lender favoritism. 4. No Individual Recommendations: OpenFileData is a public data analysis platform, not a financial advisor. Past aggregate approval rates of an institution do not predict whether an individual borrower will be approved or offered favorable terms.