Why It Matters
HMDA is why we know investor-share statistics exist, why fair-lending discrimination research is possible, and why CFPB can monitor lending patterns. Every year, U.S. mortgage lenders above certain size thresholds report every single loan application to the CFPB: borrower race, income, loan amount, property location (at census-tract level), denial reason if applicable, and the final disposition (approved, denied, withdrawn, incomplete). That's the HMDA Loan Application Register (LAR). CFPB publishes the LAR annually. Researchers, journalists, and investor-platform providers use it for everything from denial-rate analysis by race/income to investor-purchase-share by ZIP to identification of lending deserts. For investors, HMDA gives real-time data on financing availability, race/income lending patterns in target markets, and competitive intelligence on which lenders dominate which markets.
At a Glance
- What it is: Federal law (1975) requiring mortgage lenders to report every application to federal regulators annually.
- Who reports: All U.S. mortgage lenders above ~$50M in assets (most bank and non-bank lenders).
- Key data points collected: Borrower race/ethnicity/sex, income, loan amount, property type/location (census tract), denial reason, final disposition.
- Publication: CFPB publishes annual LAR dataset in Q2 of each year covering the prior calendar year. 2024 data publishes in Q2 2025, for example.
- Access: Free, public. Downloadable at ffiec.cfpb.gov.
- Scale: ~20 million loan application records per year across ~5,000 institutions.
How It Works
What lenders report. Every HMDA-reportable institution must submit details on every mortgage application received during the year: loan type (conventional, FHA, VA, USDA), purpose (purchase, refinance, home improvement), loan amount, property type (1-4 family, multifamily, manufactured), property location (state, county, census tract), borrower demographics (race, ethnicity, sex, age), borrower income, AUS (automated underwriting system) results, denial reasons (if denied), interest rate spread, whether the loan was originated or purchased, and the final action (originated, approved but not accepted, denied, withdrawn by applicant, file closed for incompleteness). All of that gets consolidated into the annual Loan Application Register.
The investor-share statistic. When NAR or CoreLogic reports "investors bought 28% of homes in Q4," that number is partially derived from HMDA data plus supplementary sources. HMDA indicates loan purpose — a purchase loan's "occupancy" field shows owner-occupant vs non-owner-occupant (investor or second home). Researchers can identify investor-purchased properties by combining the loan application data with property records. Not all investor purchases show up in HMDA (cash purchases don't have loans), but the mortgage-financed investor subset is captured.
How CFPB, researchers, and journalists use HMDA. Fair-lending analysis: the CFPB watches for denial-rate disparities by race that can't be explained by credit factors — HMDA's demographic + denial-reason fields are the raw material for that. Academic research: university researchers use HMDA to study redlining, gentrification, access to credit. Journalists: outlets like Reveal, ProPublica have used HMDA to expose lending bias in specific metros. For real estate investors, HMDA provides two useful reads: (1) competitive lender analysis — which banks dominate which ZIPs, useful for identifying local lenders for your next deal; (2) denial-rate context — metros with high denial rates signal lending constraint, often correlated with rate-sensitive buyer behavior.
HMDA's public vs private data. The full HMDA LAR contains the complete application detail per lender. A version for the public replaces some sensitive fields (exact loan amount, exact income, exact date) with ranges to protect borrower privacy. The FFIEC-published public file is what researchers and journalists typically access. Regulators and CFPB have access to the full unmasked file for enforcement.
Real-World Example
Ana Castillo uses HMDA to pick a lender for a Memphis investment property.
Ana is buying her third Memphis rental and wants to find a local lender that specializes in investor loans in that market. She downloads HMDA LAR data for Shelby County TN for the most recent year and filters:
- Loan purpose: purchase
- Occupancy: non-owner-occupant (investor)
- Property type: 1-4 family
- Action taken: originated (loan funded)
The filter shows 2,400 investor loans originated in Shelby County last year. She groups by lender. Top 10 lenders originated 65% of all investor loans. She picks the top 3 by volume and adds the two that specialize in smaller dollar amounts (her target is $180K). She reaches out to all five, gets pre-approval letters, and picks the one offering the best combination of rate, closing speed, and investor-friendly underwriting.
This kind of competitive lender analysis would be impossible without HMDA. The dataset revealed which lenders ACTUALLY fund investor loans in her target county, not who SAYS they do.
She also checks the HMDA denial data. Memphis overall denial rate for conventional investor loans is 18% — higher than the national average of 14%. That tells her to prepare stronger documentation (proof of rent rolls on existing properties, bank statements showing reserves) before applying. The HMDA denial-reason field shows the top reasons: debt-to-income (33% of denials), collateral (24%), credit history (19%). She focuses on her DTI specifically, paying down other debts before applying.
Pros & Cons
- Completely free, publicly available at ffiec.cfpb.gov
- Covers ~20 million loan applications annually
- 30+ years of historical data (modernized post-2018)
- Granular to census-tract level for geographic analysis
- Required by law — no sampling bias, full population coverage
- Annual publication lag — Q2 release covers prior calendar year (12-16 month lag)
- Complex data dictionary — not casual-user friendly
- Cash purchases invisible — only financed transactions captured
- Some fields masked in public version for privacy
- Application-level not transaction-level — doesn't tie to sale price directly
Watch Out
- Denial rates aren't discrimination proof: A high denial rate in a metro could reflect borrower credit quality rather than lender discrimination. Fair-lending analysis controls for credit factors; casual denial-rate reading doesn't.
- Action taken codes matter: "Originated" means the loan funded. "Approved but not accepted" means the borrower got approval and chose not to take it (often because they got better rates elsewhere). Separate these in analysis.
- Small lender coverage is incomplete: Lenders below the HMDA size threshold (~$50M assets for banks; lower for non-banks) don't report. Small community banks and local credit unions may not appear.
- Non-owner-occupant classification: Investor flag is based on borrower statement, not verified ownership. Some properties flagged as non-owner-occupant are second homes, not investments.
- Census tract geography: HMDA uses Census tract boundaries. Tract boundaries change with each decennial Census — when comparing 2010-based and 2020-based data, some tract IDs changed.
Ask an Investor
The Takeaway
HMDA is the federal mortgage-application disclosure law that produces the annual Loan Application Register — the largest publicly available source of mortgage lending data in the U.S. Used by CFPB for fair-lending monitoring, by researchers for access-to-credit analysis, and by journalists for lending-discrimination reporting. For real estate investors, HMDA provides competitive lender analysis (who funds what in which ZIP), denial-rate context for target markets, and intelligence on financing availability. Free access at ffiec.cfpb.gov. Annual publication in Q2 covering the prior year. Complements CFPB's TRID and QM rules as part of the post-2008 regulatory framework.