Module 9 · Outlook & Benchmarking · Competing-hazards engine
Every HECM in this portfolio is racing toward one of two exits: a mandatory buyout at 98% of MCA, or a termination event (death, move-out, payoff, foreclosure). This page models that race for all 53,780 loans, benchmarks the seized book against the entire active HMBS universe, and projects the non-pooled inventory the public can't see. Calibrated to observed events; validation gaps published, not hidden.
The funding calendar
Each loan's balance accrues at its own note rate plus 0.50% MIP toward the 98%-of-MCA purchase trigger, while an age-calibrated termination hazard competes to resolve it first. Aggregated, that produces the program's forward funding requirement, under flat rates and ±100bp.
Expected values, $M per quarter. Bars: base-case buyouts. Lines: ±100bp accrual scenarios. Area: termination runoff (deaths, payoffs, moves, foreclosures).
Month-one model vs. May 2026 actuals: terminations 401 modeled vs. 391 observed; mandatory buyouts 510 modeled vs. 655 observed. The shortfall is deliberate. The model accrues balances but does not yet simulate future line-of-credit draws, and the book carries $2.95B of unfunded draw commitments that accelerate loans toward the 98% trigger. The 22% gap is an empirical measure of draw-behavior risk, and it brackets the truth: treat the base case as a floor and the +100bp path as a reasonable proxy for draw-adjusted timing. Each month's new disclosure re-tests this model in public.
Peer benchmarking against the full HMBS universe
Same files, all issuers: 243,413 active HMBS loans benchmarked against the 53,780-loan seized book, loan-level, same month. The result puts a common narrative to rest: this is not a distressed-servicing story.
Loan-level liquidation events, annualized from the May disclosure month.
Mortality is identical (2.84%/yr vs 2.86%/yr) and so is foreclosure (0.25% vs 0.24%). The borrowers behind the seized book are dying, moving, and defaulting at market rates, so servicing through the MSS arrangement is not distorting borrower outcomes. That is a defense of the current arrangement Ginnie Mae can take to any oversight audience.
Mandatory buyouts run 3.6× the market rate (13.7%/yr vs 3.8%/yr). The driver is age, not servicing: the book is 3 years older (7.1 vs 4.1 years average loan age) and 14 points deeper into MCA (64.7% vs 50.6%). The seized portfolio is further down the conveyor belt every active book is on. Refinance is the one real anomaly: zero refi events vs 1.18%/yr in the market. Seized-book borrowers are not being refinanced out, a marketing-period consideration for any sale.
The invisible inventory
Bought-out loans leave the pools, and public loan-level visibility, while they await FHA assignment. The flows are still knowable: buyout inflow (this page's engine), assignment outflow ($2.5B claimed in FY2025 per the audited AR), and the stock anchor (reverse UPB minus pooled balance ≈ $2.54B at FY2025 year-end). A stock-flow model completes the picture.
Warehouse balance, $M, under the three accrual scenarios, holding assignment capacity at the FY2025 observed pace (~$208M/month).
Implied dwell time today is ~12 months of inventory. Under the base case the non-pooled stock drains by ~Q7–Q8; even +100bp only extends it modestly. Implication for disposition strategy: the window in which a bulk non-pooled sale relieves a genuine operational burden is open now and closes within roughly three years, after which the warehouse argument for selling weakens and the hold-to-assign path gets cheaper. Timing carries most of the decision here. (If draw behavior accelerates buyouts ~20% as validation suggests, the window extends by roughly two quarters.)
Data-quality scorecard
Field-by-field completeness of the public tape, seized book vs. all active issuers. The honest headline: most gaps are universe-wide disclosure suppressions rather than seized-book defects. Geography is blank for everyone at loan level (recovered here via pool stratifications). The seized-book-specific gap is servicer attribution.
| Field (public loan-level tape) | Issuer 9281 | Active universe | Reading |
|---|
Every unknown on a bid tape becomes a haircut. The public tape's gaps (geography suppressed, servicer attribution thin, valuations origination-era) are what a pre-marketing data-remediation sprint fixes from servicing records: refreshed BPOs, title runs, occupancy verification, servicer-of-record confirmation. This portal demonstrates how much can be recovered analytically (state mix to ±0.14pp; value via HPI); the engagement assembles the rest inside the workflow environment the SOO mandates.
Run monthly, the scorecard becomes the remediation work plan's progress metric: completeness by field, by sale pool, trending toward bid-ready. It is a deliverable the engagement produces, not a complaint.
The engine re-validates against each new month's observed events, in public. That is the standard performance-based advisory analytics should meet.
Run the disposition economics →