Module 7 · Methodology, sources & integrity
This page is the audit trail for the site: what was used, how it was processed, how it reconciles to Ginnie Mae's audited statements, and the rules HSG operated under.
Everything on this site derives from sources available to any member of the public. No information from any HSG engagement, government data room, or non-public channel was used. Ginnie Mae's disclosure terms of use are honored in full: the data is used solely for analysis of Ginnie Mae MBS, no attempt is made to identify any individual, and no loan-level records are published here, aggregates only. Disclosure data is reproduced per Ginnie Mae's terms; HUD sale results are public documents.
How the reconstruction works
The Ginnie Mae Issuer File (issrinfo) names the books: Issuer 4277 "Reverse Mortgage Funding, LLC" (Inactive) and Issuer 9281 "GINNIE MAE - REVERSE MORTGAGE FUNDING 4277" (Active), plus the full 283-record defaulted-issuer lineage in the 9xxx range.
HMBS loan-level files hllmon1 + hllmon2 (May 2026; 10,275,181 records) parsed against the official v2.5 fixed-width layout; L-records filtered to Issuer 9281: 1,873,138 participations → de-duplicated on Disclosure Sequence Number → 53,780 unique HECM loans.
The HMBS Pool/Security file's 3,882 Issuer-9281 pools sum to $11,253,247,759 securitized UPB, identical to the loan-level build. Pool Supplemental stratifications (564,894 rows joined on pool ID) recover state, payment-plan, property-type, purpose, age and rural splits.
Twenty-seven monthly Pool/Security files (Mar 2024 – May 2026) produce the runoff trajectory: $17.44B → $11.25B, −$238M/month, −18.3%/yr annualized.
The series crosses Ginnie Mae's audited FY2025 figure for the defaulted-issuer share of HMBS (≈$12.9B at 9/30/2025) in the same month our file shows $12.910B. The reconstruction and the financial statements describe the same book.
Each loan's origination appraisal is rolled forward with the FHFA state house-price index (purchase-only; all-transactions fallback) from the loan's own valuation-date quarter to 2026Q1. Loan-level state is blank in disclosure, so each loan's state distribution is estimated as the participation-weighted mixture of its pools' published state distributions (avg 35 pools/loan). The estimator was adversarially validated: a naive Bayesian-product version over-concentrated into California and was rejected; the mixture reproduces the disclosed state totals to within ±0.14pp. Metro view uses record-07 stratifications + FHFA metro/division indexes the same way. Outputs are estimates rather than appraisals, and aggregates only are published.
Termination hazard: age-banded monthly rates calibrated to observed loan-level liquidation events, pooled across the full 243,413-loan HMBS universe (justified by the peer test: the seized book's mortality and foreclosure rates match active issuers to within a few basis points). Buyout clock: per-loan accrual at note rate + 0.50% MIP to the 98%-of-MCA trigger, ±100bp scenarios. Validation, published: month-one terminations within 2.5% of observed; buyouts 22% low because future LOC draws are not yet simulated, so the gap is reported as the empirical measure of draw risk. Warehouse model: stock-flow on audited anchors (FY2025 reverse UPB minus pooled balance; $2.5B/yr assignment claims). Every layer re-validates monthly as new disclosure posts.
Reconciliation
| Measure | This site (disclosure build) | Ginnie Mae pool file | Audited FY2025 AR |
|---|---|---|---|
| Securitized UPB, Issuer 9281 (May 2026) | $11,253,247,759 | $11,253,247,759 | – |
| Defaulted-issuer HMBS share (Sep 2025) | $12.910B | – | ≈$12.9B (22.7% × $56.8B) |
| Reverse mortgage loans, UPB (9/30/25) | pooled view $12.91B | – | $15.447B incl. ~$2.1B unsecuritized + buyouts |
| HMBS obligations (9/30/25 → runoff) | $11.25B @ May ’26 | – | $13.313B @ Sep ’25, consistent at observed $238M/mo |
| Buyout flow | 44,455 mandatory-purchase participation events in May ’26 | – | $2.7B FY2025 buyouts · $2.5B assignment claims |
The loan-balance view ($12.93B at May 2026) exceeds securitized UPB ($11.25B) by construction, since HECM balances accrue interest and MIP above the securitized participations. The AR's $15.45B reverse-UPB additionally includes bought-out, not-yet-assigned whole loans that have left the pools. All three views are consistent.
Source register
| Source | What it provided | Cadence |
|---|---|---|
| Ginnie Mae HMBS loan-level disclosure (hllmon1/hllmon2, layout v2.5) | The loan-by-loan book: balances, MCA, rates, vintages, payment reasons, LOC | Monthly |
| HMBS Pool/Security + Pool Supplemental (v5.3) | Pool table, factors, validation total; state/MSA/demographic strats | Monthly |
| Ginnie Mae Issuer File (issrinfo) | Issuer identity, status, defaulted-issuer lineage (9xxx range) | Monthly |
| Ginnie Mae Annual Reports FY2023–FY2025 (audited) | Balance-sheet series, buyout/assignment flows, defaulted-issuer counts, HMBS shares | Annual |
| Ginnie Mae Global Markets Analysis Report | GNMA II yields/spreads, SDQ, MSR holder shares w/ CPR-CDR | Monthly |
| HUD Office of Asset Sales results (HVLS 2025-2/2025-3/2026-1, HNVLS 2025-1, HLS 2026-1; Report to the Commissioner Mar 2026) | Execution benchmarks, buyer universe, occupancy discount | Per sale / semiannual |
| FY2025 MMI Fund Actuarial Review, HECM (Dec 2025) | Hold-path modeling conventions: FCRA SER discounting, termination/severity structure, ±NPV scenarios | Annual |
| FHFA House Price Index (hpi_master.csv: state + metro/division, quarterly) | Localized valuation roll-forward (state level) + metro appreciation context (through 2026Q1) | Quarterly |
| HUD sale results, complete program history (17 results summaries, HVLS 2017-1 → HVLS 2026-1, + Report to the Commissioner Mar 2026) | Full execution record incl. pool-bid era, post-sale resolution outcomes, REO disposition, buyer roster, nonprofit-policy history | Per sale / semiannual |
| CourtListener / RECAP (public dockets) | TCB v. Ginnie Mae appellate-status monitoring | At build |
| FRED live series (OBMMIFHA30YF, DGS10, DGS1, MORTGAGE30US) | Same-day rate context, stamped at build time | Daily |
| FRED / Optimal Blue (OBMMIFHA30YF) | Daily 30-yr FHA primary rate | Daily |
| MBA National Delinquency Survey (press releases) | FHA delinquency context | Quarterly |
| SEC filings: PennyMac, Rithm, Onity | MSR fair-value assumptions; reverse-servicing economics | Quarterly |
| Court record & trade press (TCB v. Ginnie Mae; New View Advisors) | Tail-title litigation posture; HMBS issuance trend | As published |
| FY2026 HUD Congressional Justification (GNMA) | Operational scale (>$500M monthly transactions), staffing posture | Annual |
Property valuations in disclosure are origination-era; the equity module rolls them forward with FHFA HPI to produce an estimate carrying geographic uncertainty (a mixture rather than a known location) and no property-condition information; BPO/appraisal refresh remains step one of any sale. Loan-level geography is suppressed universe-wide in HMBS disclosure (this is not a seized-book defect); geography is recovered at pool level. The termination hazard is calibrated to one observed month (re-calibrated as each month posts); the engine omits future LOC draws, so its published 22% month-one buyout shortfall measures that omission. The warehouse model rests on audited aggregates rather than tapes. The hold-vs-sell engine and program optimizer are scenario tools with disclosed simplifications, well short of valuations.
Disclosure posts on the 6th business day monthly; this site's pipeline re-runs the full build (extraction → validation → funnel → series) from the new files. Sale results and GMAR fold in as published. The same discipline, inside the engagement boundary, becomes the SOO's reporting deliverable.
Sources Sought APP-T-2027-125 · NAICS 541618 · Responses due June 23, 2026, 4:00 PM ET
SBA 8(a) federal financial services · founder-led Ginnie Mae & HUD asset-markets practice (OCM, OIPM and OAS engagement history) · Washington, DC & Tampa, FL. Contact: Jelani House, Managing Director.
jelani.house@housestrategiesgroup.com