Selling models and market structure
The Record Is Legal, Not Economic
6 min read
A recorded sale price states what was conveyed for consideration. That is the question a county recorder exists to answer, and the answer is a legal artefact.
It has never been an economic figure. What is new is how far apart the two are drifting, and that nothing in the data will tell you.
What a recorded price was built to carry
Concessions have always sat in the gap. That is why the seller contribution cap exists and why concessions are disclosed. The cap is drawn around money moving at or around the transaction, from seller to buyer, and within that frame it works.
A builder forward commitment moves earlier than the frame reaches.
A forward commitment is when a builder's mortgage company pays points in advance to secure a block of mortgage money at a below-market rate, then originates loans from that block for selected homes. The cost is committed before any individual buyer exists, which is why it is not a seller concession.
Read the timing rather than the mechanism. The money moved months before the transaction, from a party that is not the seller, to a lending facility rather than to a buyer. No seller contribution, so nothing to disclose and nothing for the cap to reach.
The result is a transaction that is fully disclosed, entirely legal, and economically different from what the record says.
A definitional edge rather than avoidance
It is worth being precise about what this is, because the loose version of the argument is wrong and will be corrected by the first lender who reads it.
Nobody is exploiting anything. The rule was drawn before the practice became common, around a shape of transaction that made sense then, and a structure appeared that sits outside it. Every party is doing something ordinary and disclosed.
What changed is volume. As of June 2025, around 64% of new homes sold by the largest builders used a permanent buydown, against roughly 13% for smaller builders. On 12 August 2026 the 30 year rate on loans from homebuilder mortgage affiliates sat at 5.23% on a weighted average against 6.6% among non-builder lenders, a spread of 137 basis points that had run as wide as 179 earlier in the month. Where a forward commitment combines with other incentives, some builders are spending 6.5% to 12% or more of the sales price on buyer assistance.
When a discount arrives as price, the record captures it. When it arrives as rate, the record captures nothing. The industry has shifted from the first to the second at scale, and a definitional edge that was immaterial at low volume is now carrying a substantial share of new construction economics.
The dataset drifted while its definition held still
This is the part that should concern anyone who works with comparable data rather than anyone who works with builders.
A comparable set assembled in 2021 and one assembled in 2026 both report recorded prices. Same field, same definition, same source. The second contains a far higher share of transactions where the economics sit somewhere the record does not go.
No field changed. No provider flagged anything. No data quality process caught it, because nothing about the data is of poor quality. The definition held, the practice moved underneath it, and the resulting drift is invisible to every check designed to look for errors.
That is a different failure from bad data and it is harder to see. A wrong figure gets corrected. A correct figure that has quietly stopped describing what it used to describe gets used with confidence.
For anyone building or buying valuation tooling, that is the more useful framing. The models are not degrading. The inputs are answering a slightly different question than they were five years ago, and the gap widens every quarter the spread persists.
Why do recorded sale prices differ from what buyers really paid?
Because a recorded price captures consideration conveyed rather than economics. Concessions, assumed liabilities, below-market seller financing and personal property included in a sale all sit outside it to varying degrees. A builder forward commitment sits outside it completely, because the cost was committed before a buyer existed and never took the form of a seller contribution.
What is a seller contribution cap and what does it not cover?
It limits what a seller may contribute toward a buyer's costs, expressed as a percentage of the sale price, and it governs money moving at or around the transaction. It does not reach a cost committed months earlier, by a party that is not the seller, to a lending facility rather than to a buyer. That is a definitional edge rather than avoidance, and the distinction matters because the first framing invites a rule change and the second invites a data field.
Is a builder rate buydown a concession?
Economically, yes. On the closing disclosure under a forward commitment, no. Both answers are correct, and the fact that both are correct is the problem rather than a technicality about either.
What else sits in the same gap
Forward commitments are the largest current entry and they are not the only one.
Assumed liabilities that transfer with a property. Below-market seller financing, where the price is nominal and the terms carry the discount. Personal property folded into a price. Family transactions never tested against another buyer. Each is a legal price that is not an economic price, and none is marked anywhere a comparable search will look.
What they share is that the information existed at the moment of the transaction and had nowhere to go. Not withheld, not hidden. There was no field for it, and the people who knew moved on.
Which suggests where the useful work sits. Not in cleaning the record, since the record is accurate, but in capturing what a participant knew at the time and holding it long enough to matter when the same property comes back around.
That is not a data problem. It is a memory problem, and it is unglamorous enough that nobody has treated it as infrastructure.
What we can and cannot see from here
One disclosure, since this is a piece about what a dataset does not contain.
CMAflow reads the same record everyone else reads and cannot see what is not in it either. No comparable search we run will surface a forward commitment, and no adjustment we compute knows about one. What a report can do is hold what an agent recorded at the time, and keep it available when the same house lists again in three years.
That is a considerably smaller claim than solving the problem described above. It is also the true one, and the distance between the two is where most of this category oversells.
This article is general information and analysis, not financial, lending, or appraisal advice. Verify any home value with a licensed professional before acting.
The Independent Agent
Substack | Spotify | CMAflow FAQ | YouTube | Free CMA | Home valuation | Insights | Blog