06-reference/research

construction loan structure rate lock carry

2026-10-22·research-brief·source: deep-research·by Ray Data Co (deep-research synthesis)

The Agency Framework Permits a Rate Modification at Conversion, So Single-Close Rate-Rigidity Is a Program Term and Not a Structural Rule; But No Published Jumbo One-Time-Close Float-Down-at-Conversion Exists, and the Carry Question Is Worth Roughly Half the Rate Question

The question

Verbatim: "What financing terms apply to the construction-loan structure choice (construction-to-perm vs separate construction loan + take-out refi) for the Sunset Park rebuild — rate-lock float-down options, interest-only carry during the 10-12 month build?"

This is Critical Open Question #4 on [[01-projects/home-rebuild-2027/README|home-rebuild-2027 README]], and it is the named open follow-up #1 of [[2026-08-14-construction-to-permanent-loans-florida]], which mapped the structures but left float-down availability explicitly unresolved. Scoped to the structure choice and the two named mechanics only: not general construction-loan mechanics (covered by the 08-14 brief), not the flood-zone layer (covered by [[2026-08-04-flood-zone-ae-construction-loan-requirements]]), not qualifying capacity (covered by [[2026-08-19-jumbo-construction-perm-qualifying-envelope]]).

What we already know (from the vault)

What the web says

Convergences and contradictions

Synthesis for RDCO

Size the two mechanics against each other first, on one consistent basis, because they are not the same order of magnitude. Derived from the vault's inputs at a $1.2M note and a 7.0% assumed note rate (U; anchored to MIDFLORIDA's posted 7.000% as of 2026-08-03, which is a $300,000 example and will not be the jumbo price), with the $415,622 payoff financed at day one and the remaining $784,378 disbursed linearly: average outstanding is ~$807,800, so interest-only carry runs ~$47,100 over a 10-month build and ~$56,500 over 12 months, a $47-57k band carrying $10k of duration risk. The monthly payment ramps from **$2,425** in month one to ~$7,000 at full draw; at a $1.5M note the 12-month figure is $67,000. Now the rate mechanic, stated as present value rather than an undiscounted sum so it compares honestly: on a $1.2M 30-year, 7.00% amortizes to $7,984/mo and 6.25% to $7,389/mo, a $595/mo difference whose present value at the 7% note rate over a full 30-year hold is **$89,400** (the undiscounted total is $214,000, which is the number not to use). So the rate question is worth roughly 1.6-1.9x the entire construction carry, on an assumed 75bp move and a no-refinance hold. One lever beats both: financing the $415,622 payoff rather than clearing it in cash adds ~$29,100 of gross interest over a 12-month build ($56,500 vs ~$27,500), before netting the yield given up on the cash. That belongs in the sources-and-uses conversation, not this one, but it is 2-3x the entire two-close duplication cost computed below.

Two-close's cost side is now partly computable, and Florida is the part nobody had priced. Because B5-3.1-03 makes the take-out a refinance with a new note, Florida's recording levies fire a second time. On a $1.2M obligation: §201.08(1)(a) note stamp is $4,200 but capped at $2,450; §201.08(1)(b) then charges the uncapped mortgage levy on the excess, $4,200 less the $2,450 already paid, so $1,750; which means the celebrated $2,450 note cap is cosmetic and total doc stamp lands at $4,200 either way. Add §199.133 intangible at 2 mills, $2,400. That is ~$6,600 of Florida recording tax per recorded obligation, derived from the statutory rates established in [[2026-08-25-florida-intra-family-purchase-money-note]]; gross, because the §201.09 renewal exemption is not checked and could reduce it (see follow-ups). Stack the $3,000-5,000 of duplicate closing costs the 08-14 brief cites as a reported figure (U) and two-close carries roughly $9,600-11,600 of duplication on this deal, plus a second lender's title policy that I could not price. Against that, two-close buys one thing: you shop the perm rate at completion, an option worth the ~$89,400 present value above if a 75bp improvement actually materialises. On those two numbers the option is worth buying unless single-close can capture most of the same move.

And the honest answer on whether single-close can is: the agency text says the mechanism exists, and no published jumbo program I could reach says it is available. B5-3.1-02 contemplates exactly the thing two-close is bought for, a rate reset at conversion on a Form 3179-equivalent modification, which is why the 08-14 brief's rigidity claim should not be carried forward as settled. But I could not find a single jumbo one-time-close product publishing a float-down or rate modification exercisable at conversion. The U.S. Bank 360-day lock looked like it until the page turned out to be builder-and-developer-facing, protecting a builder's buyers rather than a custom one-time-close borrower, with the lock listed separately from the custom-construction bullet. So the deliverable here is a yes/no question, not a recommendation: ask each lender whether a float-down or a rate modification operates at conversion, and get it in the term sheet. If yes, single-close dominates on every axis in this brief. If no, two-close's option retains real value and the ~$9,600-11,600 of duplication is the honest price of it. Do not let the question go unasked and then default to single-close on closing-cost arithmetic; the rate mechanic is the larger number.

Two terms to set at term-sheet stage regardless of structure. First, ask for 18 months of construction period rather than 12. The agency ceiling is 12 months per single period and 18 total with "exceptions will not be granted"; among the products the vault has inventoried, Goldwater's listing shows a 12-month term while BuildBuyRefi's jumbo one-time-close runs interest-only to 18 (U — two data points from an aggregator page, not a market norm). A 10-12 month build against a 12-month term has no slack, and in an AE-zone elevated build the schedule risk is documented: [[2026-08-04-flood-zone-ae-construction-loan-requirements]] puts two separate elevation-certificate deliverables and inspection-gated draws in the critical path. A longer term may carry a longer-lock or rate cost (U — unpriced here), but it is cheaper asked for up front than requested mid-build. Second, do not assume single-close buys one underwrite. The 18-month document-age allowance that would deliver that outcome is conditioned on a DU Approve/Eligible, which a jumbo cannot have, so on the portfolio product the re-verification posture at conversion is an open lender-specific term (U) and worth asking in the same breath. What does transfer is the shape of the risk: B5-3.1-02's value-decline trigger forces a new appraisal and requalification if the 1004D comes in low, and a 2028 completion appraisal under the subject-to-completion number is a scenario worth pricing rather than dismissing given the AE-zone insurance repricing tracked in [[2026-07-22-florida-ae-zone-insurance-repricing-tampa]] (the insurance-to-appraised-value link is (U), asserted as a mechanism and not sourced here). On a two-close the full re-underwrite is unconditional, so single-close is still the better posture on this axis; it is just better by a margin this brief cannot size.

Why this is in the vault

This closes Critical Open Question #4 on [[01-projects/home-rebuild-2027/README|home-rebuild-2027 README]] and open follow-up #1 of [[2026-08-14-construction-to-permanent-loans-florida]], and it retires that brief's inherited claim that a single-close permanent rate is immovable once set. It is the input to the Q1 2027 go/no-go gate's financing leg: the structure cannot be chosen on closing-cost arithmetic alone, because the rate mechanic is the larger number and its availability is a term to be negotiated rather than a property of the product.

Open follow-ups

Considered and not proposed: whether the construction-period interest-only payment counts in DTI, already answered on the agency baseline by [[2026-08-25-temp-housing-rent-dti-construction-perm]], whose own open follow-up already owns the portfolio-overlay residue. Also not proposed: "call three Tampa lenders," which is a task rather than a research question.

Related

Sources

Vault

Web (primary)

Web (secondary / weak — builder-facing or marketing collateral, undated or stale; used directionally only, and none of it a jumbo custom one-time-close term sheet)

Not retrieved