HMDA Mortgage Data 🏦

Loan-level US mortgage records from the CFPB's public HMDA data — straight from the government API. No API key, no proxy.

The Home Mortgage Disclosure Act (HMDA) makes nearly every US mortgage application public. This actor pulls those loan-level records — amounts, interest rates, lender, geography, outcome (approved/denied), property value, and applicant attributes — with the HMDA numeric codes decoded into readable labels.

Perfect for real-estate & lending analytics, fintech/mortgage market research, fair-lending & compliance analysis, and economic research.

Uses the CFPB HMDA Data Browser — public government data, no key. Filter by year, state/county, outcome, loan type and purpose.

---

Features

• ✅ No API key, no proxy — official CFPB HMDA endpoint.
• ✅ Loan-level — every reported application, not just aggregates.
• ✅ Decoded — action, loan type, purpose, occupancy turned into labels.
• ✅ Lender + geography — LEI, state, county FIPS, census tract.
• ✅ Rich fields — amount, rate, rate spread, term, property value, income, demographics.
• ✅ Filters — year, state/county, action taken, loan type & purpose.

---

Input

Field	Description
Years	HMDA years (2018 onward), e.g. ["2023"].
States	Two-letter codes, e.g. ["CA", "TX"].
Counties (FIPS)	5-digit FIPS, e.g. ["06037"] (Los Angeles).
Actions taken	1=Originated, 3=Denied, … (see below).
Loan types	1=Conventional, 2=FHA, 3=VA, 4=USDA/RHS.
Loan purposes	1=Purchase, 31=Refi, 32=Cash-out refi, 2=Improvement.
Max results	Cap (a state-year can hold millions).

A state or county filter is required (with a year) — a nationwide unfiltered pull is far too large.

Example — LA County home-purchase originations, 2023

{
  "years": ["2023"],
  "counties": ["06037"],
  "actionsTaken": ["1"],
  "loanPurposes": ["1"],
  "maxResults": 50000
}

Example — denied applications in Texas

{
  "years": ["2023"],
  "states": ["TX"],
  "actionsTaken": ["3"]
}

Output

{
  "activity_year": 2023,
  "lei": "B4TYDEB6GKMZO031MB27",
  "state": "CA",
  "county_fips": "06037",
  "census_tract": "06037...",
  "action_taken": "Loan originated",
  "loan_type": "Conventional",
  "loan_purpose": "Home purchase",
  "occupancy": "Principal residence",
  "loan_amount": 805000,
  "property_value": 1375000,
  "interest_rate": 5.75,
  "rate_spread": 0.35,
  "loan_term_months": 360,
  "applicant_income": 421000,
  "applicant_race": "White",
  "applicant_sex": "Joint",
  "applicant_age": "45-54"
}

Run it on a schedule

HMDA publishes annually (with quarterly updates for large filers) — refresh your dataset when new data drops, then pipe to a warehouse or BI tool.

Notes & limitations

• Source: CFPB HMDA public data. Loan amounts and property values are disclosed as midpoints/ranges per HMDA privacy rules; income is reported in thousands (converted to dollars here).
• Applicant demographics are self-reported and often "Not Available."
• Requires a year + a state/county filter (nationwide is too large to stream).
• Public regulatory data — no personal identifiers (no names/addresses), only geography to census-tract level.

---

Keywords

HMDA, mortgage data, CFPB, home loans, lending, mortgage rates, real estate finance, loan originations, denials, fair lending, redlining analysis, LEI, lender data, refinancing, FHA, conventional loans, housing finance, fintech, credit, mortgage market.