06-reference/research

jumbo portfolio overlays contingent liability

2026-08-25·research-brief·source: deep-research·by Ray Data Co (deep-research synthesis)·! medium

Portfolio Overlays and the Co-Signed PITI: The Agency Exclusion Is Real, the Lender-Level Answer Is Unpublished, and the One Jumbo Guideline I Could Actually Read Is Narrower Than Fannie

The question

Verbatim: "Do Florida jumbo portfolio construction lenders (Seacoast, TD, Goldwater, and the national OTC jumbo programs) honor Fannie's B3-6-05 'Debts Paid by Others' exclusion for a co-signed mortgage, or do portfolio overlays count the full contingent PITI regardless of payment history?"

Context: this is the follow-up [[2026-08-19-asset-depletion-mortgage-qualification]] left open. That brief found the co-sign path costs the founder roughly $597k of his own 2027 build capacity and that the B3-6-05 escape hatch is unusable on the current six-week closing sequence. This brief asks the prior question: does the escape hatch even exist for his lender class, because if portfolio overlays count full contingent PITI regardless, the twelve-month sequencing fix buys nothing and co-sign is simply off the table.

CALIBRATION NOTICE. Portfolio lenders do not publish underwriting overlays. I could not source Seacoast Bank's, TD Bank's, or Goldwater Bank's treatment of contingent liabilities, and I did not infer it. What follows is: the agency baseline (primary-source verified), the structural reasons portfolio underwriting deviates from it (verified), one verbatim non-agency jumbo guideline that happens to be public (verified, N=1, and not one of the named lenders), and a direct-inquiry script. Anyone reading this looking for "Seacoast's policy" will not find it here, because it is not knowable from public sources.

What we already know (from the vault)

What the web says

Convergences and contradictions

Synthesis for RDCO

The honest answer is that this question cannot be closed from a desk, and that is itself the decision-relevant finding. The agency baseline is solid and verifiable: B3-6-05 as of 08/05/2026 does permit full-PITIA exclusion on a co-signed mortgage where a co-obligated payer has twelve clean months of documented payments, and the parents-on-the-note structure satisfies its hardest condition. Everything downstream of that is unpublished. Seacoast, TD, and Goldwater do not post underwriting guidelines; portfolio credit policy is an internal document, and the honest posture is to say so rather than to manufacture a plausible-sounding answer about a specific bank. What is knowable is the shape of the risk, and the shape is unfavorable. When a lender keeps a loan on its own balance sheet, the discipline that forces guide conformity - rep-and-warrant exposure on a sale to a GSE - is absent. What remains is credit policy, examiner expectations, and ATR, none of which require the exclusion to be honored. The QM verification safe harbor gives a jumbo lender a reason to gesture at the Fannie guide, but a safe harbor is permission to be no stricter than a standard for verification purposes; it never caps how conservative a lender may be on qualification.

The single piece of hard evidence I could retrieve points the wrong way for the co-sign path, and it should be weighted carefully rather than dismissed or over-read. Axos Bank's Prime Jumbo guidelines (03/24/2025) enumerate three contingent-liability exclusions - divorce buyout, assumption, court order - and none of them describes "my parents are on the note and pay the note." The catch-all then forces the payment into DTI. That is one lender, it is not a construction-to-permanent product, it is a correspondent purchase program rather than a pure balance-sheet portfolio, and it is not on the founder's shortlist. It is N=1 and cannot be generalized to Seacoast or TD. But it is a real instance of exactly the failure mode the question was asking about, produced by a bank that explicitly defaults to Fannie when its own guidelines are silent and then chose not to be silent. That is more informative than a hypothetical. Prior distribution before this brief: unknown. After: the "portfolio overlay counts the full PITI" branch is live, documented, and should be treated as at least as likely as the honoring branch until a named lender says otherwise.

The practical consequence is that the co-sign branch now carries two independent failure points, and the founder controls only one of them. Failure point one is timing, which the parent brief already identified: twelve months of seasoning against a six-week gap between a 2027-04-30 parents' closing and a 2027-06-15 construction-loan closing. That is fixable by calendar, at zero cost, today. Failure point two is lender policy, which is not fixable at all - it is a fact to be discovered, and it is discoverable in roughly one phone call per lender. Sequencing the closings twelve months apart is a real constraint on the whole plan (it pushes the parents' purchase to roughly mid-2026, already past, or the build application to mid-2028), so paying that cost before knowing whether the exclusion is even available in this lender class would be backwards. The correct order is: ask the lenders first, then decide whether the sequencing inversion is worth buying. That reverses the parent brief's implied sequence, and it is the substantive change this brief makes.

The cheapest read remains the one the parent brief already reached by a different route. Co-signing costs ~$597k of build capacity on the parent brief's arithmetic - not this brief's finding, and it belongs to [[2026-08-19-asset-depletion-mortgage-qualification]] - against a ceiling that [[2026-08-14-construction-to-permanent-loans-florida]] shows is already the binding constraint. If the exclusion is unavailable at the founder's actual lender, that $597k is permanent rather than temporary, and the co-sign branch should close. If it is available, the founder buys back the capacity only by paying a twelve-month calendar cost. Neither branch makes co-signing cheap. The structures that route around the question entirely - an intra-family purchase-money note, or H4P if both parents are 62+, both raised in the parent brief - do not create a contingent liability for any underwriter to have an opinion about, and this brief strengthens rather than weakens the case for looking at them first.

Direct-inquiry script

Five questions, asked of an underwriting desk or a loan officer who will route to underwriting. Every one is yes/no or a short factual answer, and the set settles the question in one call per lender. Ask the OTC/national programs (Goldwater, BuildBuyRefi-type aggregators) in writing, because their answer is a matrix lookup and email creates a record; ask Seacoast by phone and ask for the credit decision-maker, because their answer is judgment.

  1. "On your jumbo construction-to-perm program, do you follow Fannie Mae Selling Guide B3-6-05 'Debts Paid by Others' for contingent liabilities, or do you have your own overlay?" If they say they follow Fannie, go to Q2 anyway - Axos says that too and then overrides it.
  2. "Concretely: if I am a co-borrower on my parents' mortgage, they are also obligated on the note, they make every payment, and I bring you twelve months of their bank statements showing no delinquency - do you exclude that full PITI from my back-end ratio, yes or no?" This is the load-bearing question. Do not accept "we'd look at it."
  3. "If the answer is yes, what exactly do you need in the file - canceled checks, bank statements, or both - and does the twelve months have to be complete as of application, as of the credit decision, or as of closing?" The measuring date can be worth six weeks, which on the current calendar is the entire gap.
  4. "If the answer is no, is there any documentation package that would get to yes - a release of liability, a longer payment history, higher reserves, compensating factors - or is it a flat overlay?" Distinguishes a negotiable credit-policy position from a hard matrix rule.
  5. "Separately, does that property count against my financed-property limit or trigger any other program restriction even if the payment is excluded?" Fannie's B2-2-03 counts it regardless; find out whether the portfolio program does too, because it can carry reserve or LTV consequences even on the favorable answer to Q2.

Two calibration notes for whoever makes the calls. First, a loan officer's verbal "sure, that's excludable" is not an underwriting decision; ask for it in email or ask them to confirm with underwriting. Second, ask these before paying any application fee and before the sequencing decision, because Q2 alone determines whether the calendar inversion has any value.

Why this is in the vault

This closes (as far as public sources permit) Open Follow-up #1 from [[2026-08-19-asset-depletion-mortgage-qualification]] and it changes a concrete edit already queued against [[milestones]]: the parent brief recommended inverting the 2027-04-30 / 2027-06-15 closing sequence to a twelve-month gap as "the single highest-leverage change to the plan," and this brief shows that fix is worthless unless the founder's actual construction lender honors the exclusion - a fact obtainable in one call per lender and not obtainable any other way. It also converts the Q1 2027 gate's co-sign branch from "expensive" to "expensive and contingent on an unverified lender policy," which is a different thing to put in front of a decision.

Open follow-ups

Related

Sources