Data room / Run 1, stages 3 to 5: scenarios, model, verdict
Model assumptions
run-1/models/ASSUMPTIONS.md Open the file as written
Stage 4 Quant Model — ASSUMPTIONS
Every assumption the model (equity_flow_model.py) makes. Per RUNBOOK.md,
the model is held to these; anything the code needs that is not stated here
must be added here, not silently embedded. Results in RESULTS.md.
Deal parameters (from the spec, §§3–7, 11, 19, 29, 33, 36)
- Enrollment value V0 = $625,000 (spec §19 standard deal). Enrollment value equals the appraised value; the appreciation baseline is enrollment value, not purchase price (spec §3, §33).
- Monthly disbursement = $1,500; contracted term = 84 months (7 years); committed total = $126,000.
- Origination fee = 2.5% of committed total = $3,150, netted from the first disbursements; the capital-return basis remains the GROSS amount (spec §29 Flow 1). Homeowner's net cash received over a full term = $122,850.
- Appreciation share = 25% during the active term; +1 percentage point per completed ride-along year (years 8, 9, …), hard cap 50% (spec §7). Share at exit year T (T ≥ 7): min(25 + (T − 7), 50)%.
- Capital return = 1.0x of gross disbursed (no premium), Priority ahead of the appreciation share (spec §11, §29).
- Documented-improvement offsets (spec §15) are ignored (assumed zero).
Financing and market environment
- First mortgage: $350,000, 30-year fixed at 3.5%, current at enrollment; standard amortization. (Spec §19 implies ≈$350K payoff at year 7; we amortize explicitly — balance at month 84 ≈ $297,656.)
- Home-price appreciation (HPA) is deterministic and constant per scenario: grid −3% to +8%/yr in 1-point steps. No volatility, no path dependence beyond the pause rule below.
- Selling costs at exit = 7% of sale price (commission + transfer taxes + closing costs), paid from proceeds AHEAD of the Equity Flow claim — adopting the title reviewer's finding (title 2.2) that the spec's §29 Priority 5 ordering is unenforceable against third parties; real closings pay costs of sale first.
- Exit = voluntary arm's-length sale at fair value in exit year T; grid T ∈ {1,2,3,5,7,10,12,15,20,25,30}. No death/foreclosure/bankruptcy legal outcomes are modeled (those are Stage 3 scenario findings; this model prices the contract as written and enforced).
Market pause (the spec's contested mechanic)
- Two branches are modeled (the spec is internally inconsistent — investor
A10; SCENARIOS.md preamble):
- Branch P (pause operates): value is checked quarterly; disbursement occurs in a month only if the value at the most recent quarterly check was ≥ enrollment value. Under deterministic negative HPA the first check (month 3) already prints below V0, so deployment stops at month 3 (~$4,500 gross). At HPA = 0 the value never falls BELOW enrollment, so payments continue (tie goes to payment).
- Branch N (pause inoperative, spec §20 behavior): full disbursement through min(T, 84 months) regardless of value path. Headline tables use Branch P for HPA < 0 and note Branch N variants; positive-HPA cells are identical across branches (value never drops below enrollment on a deterministic up-path).
- Paused months do not extend the term and do not reduce the share (spec §29 Flow 3a).
Settlement (waterfall as practiced)
- Escrow available to the Equity Flow claim at exit = max(0, sale price − 7% costs − mortgage balance at exit).
- Investor claim = gross disbursed + share × max(0, sale price − V0). Settlement paid = min(claim, escrow available). Non-recourse: no deficiency (spec §14, §29).
- Homeowner net at closing = escrow available − settlement paid.
Customer-side metrics
- Homeowner cash flows: +$1,500/mo in disbursed months, with the first $3,150 withheld (fee netted against the earliest disbursements); −settlement at exit month. APR-equivalent = annualized monthly IRR of that flow (the effective cost of the cash received). Where the homeowner receives nothing (immediate pause) and pays nothing beyond received, the APR is reported as n/a.
- Taxes are EXCLUDED. Per tax H-1, exit taxation likely worsens every appreciating cell for the homeowner (gain computed on 100% of appreciation incl. the investor's slice); the model is therefore homeowner-favorable in appreciating cells.
- HELOC benchmark: identical monthly draws (same months as EF disburses, same net-of-fee amounts), balance accruing at 8.5% APR monthly, repaid in full at exit. No fees. "EF worse than HELOC" = EF settlement > HELOC payoff at exit (dollar test) — equivalently APR-equivalent > 8.5%.
- Cash-out refi benchmark: the homeowner refinances the $350K balance plus draws $122,850 cash at a 7.0% 30-year rate (destroying the 3.5% lock). Cost = extra interest paid vs. the old loan through exit year T on the refinanced balance, computed as (payment difference × months) − (principal-balance difference at T). This is a crude but directionally honest measure of the rate-lock destruction the product is designed to avoid.
- "Doing nothing" benchmark: zero cash, zero cost. EF is "worse than doing nothing" only in net-cost terms (cost > $0 in every cell except pure pause cells); we report EF net cost as the price of liquidity and flag cells where that price exceeds 15% APR-equivalent as economically indefensible vs. simply not transacting.
Investor-side metrics
- Investor cash flows: −$1,500/mo in disbursed months, +$3,150 fee at month 0 (treated as fund revenue; note investor A15 — the fee is double-promised to opex in the spec), +settlement at exit. Gross deal-level IRR (annualized monthly IRR) and multiple (settlement + fee) / gross disbursed. NO fund-level fees, carry, servicing costs, credit losses, pause-ops costs, or legal costs are subtracted — the spec defines none (investor O1); results are therefore investor-favorable upper bounds.
- Hurdle for "clears cost of capital" = 8% gross IRR (investor Stage 1 benchmark). "Negative" = IRR < 0, which under non-recourse occurs only when the escrow cap binds (settlement < deployed).
- Stress rows: fully deployed vintage sold at drawdowns of −10% to −45% from enrollment, reproducing the actuarial-crossexam recovery arithmetic (breakeven drawdown ≈ −27%).
General
- All flows nominal; no discounting other than the IRR calculations. Monthly compounding; IRR solved by bisection on monthly rate, annualized as (1+r_m)^12 − 1.
- Single-deal model. Portfolio effects (correlation, vintage mix, diversification) are out of scope; Stage 1 actuarial/investor reports cover them.
- No behavioral modeling: exit year is exogenous. The lock-in-selected book's actual exit distribution (mean ≈ year 24, actuarial 4.1) should be used to weight the columns when reading the tables.