For Investors & Quant Researchers · Commercial Residential Mortgage Finance

Commercial & Residential Mortgage Finance Data for Investors & Quant Researchers

Commercial Residential Mortgage Finance data for investors: 10 datasets on one shelf. Every one delivered as API, files, or warehouse rows.

best alternative data sources for investing · satellite imagery data for hedge funds · point-in-time fundamentals database · how do quants use commercial residential mortgage finance data

10datasets cleared the bar for this shelf
8rated top-tier for this persona
8.8mean quality, our 10-point scoring

API, files, or your warehouse. Daily, weekly, or hourly.

How do quants use commercial & residential mortgage finance data?

Five patterns dominate. Prepayment and credit modelling: Freddie Mac pairs origination terms with monthly performance and actual loss components through March 2026. Rate-factor work: weekly MORTGAGE30US plus MBA purchase and refinance indices give rate-sensitive strategies a vintage-stamped driver. Credit-cycle timing: Z.1 debt decomposed by holder type - banks, GSEs, insurers, mortgage pools - is the leverage gauge. Regional stress screens: NMDB performance tables cover 100 metros quarterly. Execution: Ginnie Mae's Disclosure Plus dashboards expose CPR, RPB and forbearance pool by pool.

One structural caveat: price signals skew agency, because FHFA HPI measures repeat sales on mortgages Fannie Mae or Freddie Mac bought or securitized; broader-market exposure arrives through the expanded-data flavor adding FHA loans below limits and county recorder deeds.

For the industry view beyond investor workflows, see our commercial-residential-mortgage-finance data hub.

How much backtest history do these mortgage series carry?

The long runs are why this slice screens well for time-series work: Z.1 successors reach back to 1945, MORTGAGE30US starts April 1971, HPI begins January 1975 (its master file carries 90,000+ observations), AHS microdata run biennially since 1973, and Freddie Mac's loan-level era opens January 1999. NMDB originations start in 1998, performance tables in 2002 Q1 and outstanding balances in 2013 Q1; PUDB covers acquisition years 2008-2024. The binding constraint is Ginnie Mae, which retains monthly archives per file prefix only back to January 2020 - that cap limits any government-collateral backtest.

Where is the loan-level detail for MBS prepayment and credit work?

The tier splits by collateral type. Conforming: Freddie Mac ships one origination file and one monthly performance file per quarter as pipe-delimited text with no header row, split into a Standard dataset of CRT-like loans and a Non-Standard dataset for the rest, plus a 50,000-loan-per-vintage-year sample. At disposition, records carry actual loss components - net sales proceeds, MI and non-MI recoveries, expenses - which is what makes loss-severity modelling possible.

Government collateral: Ginnie Mae publishes the only public loan-level view of FHA/VA/USDA-insured mortgages - daily new issues every business morning, July 2026 portfolio zips near 430 MB, quarterly Loan Performance files around 782 MB. Its MBS Loan Level Disclosure File hit Version 2.0 in August 2026, the first pipe-delimited release, adding Re-Performing Loan Indicator (LL-55), ARM Adjustment Effective Date (LL-53) and Scheduled UPB (LL-54).

What survivorship and revision caveats belong in the research spec?

Four, handled explicitly. Index selection: HPI publication thresholds require roughly 1,000 total transactions and at least 10 per quarter per metro, so thin markets show no value rather than a stale one. Tape selection: Freddie Mac's Standard dataset keeps CRT-like loans, so a model trained on it never sees the Non-Standard tail. Sample weight: NMDB aggregates are weighted up from a 5% sample with suppressed cells flagged rather than zero-filled.

Series breaks: the Fed's legacy Mortgage Debt Outstanding table stopped updating (last release March 2020), Z.1 absorbed mortgages held in FDIC receivership after the September 2023 reclassification noted December 19, 2023, and from the 2025 release (September 2026) FHFA retires PUDB TXT output and Single-Family National File A, merging both Enterprises into one flag-split file.

Straight answers

What are the best alternative data sources for investing in mortgage finance?

Mean quality is 8.80 versus 7.81 across the 1,744-dataset catalog.

Where can hedge funds get satellite imagery data on housing markets?

Not in this slice - all 10 records are government, GSE or catalog data, with no satellite, card-panel or web-traffic feed. The closest granularity proxies are FHFA HPI at census-tract level and AHS microdata (55,669 household records in the 2023 file). Imagery-driven housing signals sit in other industries' qualifying sets.

Which mortgage source comes closest to a point-in-time fundamentals database?

FRED Mortgage and Housing Data API. Ginnie Mae's daily new-issue filings are likewise stamped when published. FHFA NMDB and Fed Z.1 aggregates are revised; Freddie Mac's monthly performance tape lands as disclosed.

How do quants use commercial & residential mortgage finance data?

In five patterns: prepayment and credit models on Freddie Mac's roughly 56-million-loan tape; rate-factor timing on weekly MORTGAGE30US and MBA indices; credit-cycle gauges from Z.1 holder-type debt back to 1945; metro stress screens on NMDB performance tables (2002 Q1-2026 Q1); and CUSIP-level checks on Ginnie Mae Disclosure Plus before trading a pool.

Rows before rollout

Sample rows from any shelf entry — the field dictionary and coverage notes ride along. If the shelf misses what you need, say so; sourcing requests are half our job.

Talk to us