Forty-six competing sales. Forty-five of them reported every finished square foot as above grade.
That was the result of a comparable search I ran recently in an Adams County submarket, on a bi-level. Exactly one sale out of forty-six recorded any below-grade finished area at all. The county assessor’s records did the same thing, with no split between above grade and below grade for bi-levels and split-levels, just one finished number.
That is not a rounding difference. It is a structural disagreement between the standard appraisers are required to measure by and the only comparable data that exists, and it is why bi-level square footage keeps landing differently on an appraisal than it does on a listing sheet.
What the measuring standard actually says
Since April 1, 2022, appraisals on conventional loans sold to Fannie Mae have had to report square footage under ANSI Z765-2021. The Appraisal Institute noted at the time that only a small number of MLS systems had adopted the standard. The rule that matters here is short and leaves no room to maneuver:
Any space that is partially or completely below grade is required by Fannie Mae to be reported as basement area.
Partially. That single word is the whole issue.
A bi-level is built so that the lower level sits partway into the ground by design. It has full ceiling height, finished walls, carpet, windows, and in most Denver-area bi-levels a family room and often a bedroom and a bath. People live in it exactly the way they live in the rest of the house. None of that matters to the standard. The floor is partially below grade, so it is basement area, and it comes out of gross living area. If you want the longer version of how that classification works, I covered it in understanding basement square footage in Colorado appraisals.
On the bi-level I measured, the split was almost dead even: 853 square feet above grade and 813 square feet of finished area below. Call it 51/49. Under ANSI, roughly half the finished house is not gross living area.
The bi-level square footage problem in the data
Here is where it stops being an academic classification question.
The MLS does not measure this way. In REcolorado, bi-levels and split-levels are listed with all the finished living area reported as above grade finished area. That is not sloppiness by any one agent. It is what forty-five out of forty-six competing sales did, which makes it the convention rather than the exception. Adams County records follow the same practice.
Fannie Mae is aware of the gap. Its own guidance says GLA in local MLS systems and assessor records may not have been derived using ANSI, that the appraiser often cannot know which method produced a listing’s number, and that the appraiser has to decide through research whether data from those sources needs adjusting. The obligation to report the subject under ANSI does not go away either.
So the appraiser ends up holding two numbers that were produced by two different rules, and only one of them is negotiable.
Why the usual adjustment breaks
The sales comparison grid wants two separate lines: one for gross living area and one for finished basement area, each at its own rate per square foot. That works fine on a ranch or a two-story, where the MLS and the appraiser broadly agree on which space is which.
It falls apart on a bi-level, for a reason worth being precise about. Extracting a rate from the market requires the market data to distinguish the two categories, and for bi-level square footage that distinction was never recorded. When forty-five of forty-six sales report a single blended finished number, there is nothing in the data to separate an above-grade rate from a below-grade rate. The distinction the grid is asking about does not exist in the source.
Applying a standard GLA rate to an MLS figure and then adding a separate basement adjustment on top would produce a number, and the number would look perfectly ordinary on the form. It would also be unsupportable, because it would assign to above-grade area a rate derived from sales whose reported area was never separated in the first place. That is the kind of adjustment that reads fine until somebody asks how it was derived.
What holds up instead
The approach that survives scrutiny is to stop fighting the data and adjust on the basis the data actually supports.
On that assignment, the gross living area and finished basement lines were both set to zero, and all size-related adjustment moved to a supplemental line for total finished square footage at a single blended rate. One number, one rate, applied to the total finished area differential, matching how the market reports and prices these homes.
The rate was derived rather than assumed. A regression model gave one indication and two clean paired sales gave two more, with more weight on the paired sales, since paired evidence from the actual submarket beats a model’s central tendency when the two disagree.
It was then cross-checked from a different direction, using component rates for above-grade area, below-grade area, and below-grade finish, drawn from property types where the market does report the split. Weighted against the subject’s 51/49, finished space carried the same rate on both sides of the grade line. Sensitivity testing confirmed the rate was not a swing factor, and pushing every adjustment to its extreme at once kept the spread inside normal tolerance.
That last part matters more than it sounds. An adjustment you can derive one way, confirm from another direction, and then stress test is an adjustment that holds up in front of a reviewer, an underwriter, or an attorney.
I am deliberately not publishing the rate itself. A figure extracted for one property type in one submarket is not a figure for your house, and published numbers have a way of being quoted as though they were.
If you list or own a bi-level
A few things follow from this, and they are worth knowing before an appraisal lands on your deal.
Bi-level square footage on an appraisal will probably come in lower than the MLS number, and that is not an error. On a bi-level with an even split, gross living area can land at roughly half the finished area the listing showed. The house did not shrink. It got sorted into two categories instead of one.
Do not price a bi-level off a price per square foot pulled from mixed property types. Comparing a bi-level’s blended MLS figure against a ranch’s above-grade figure is comparing two numbers built by different rules. If you are going to use price per square foot at all, keep the comparison inside the same property type.
The finished lower level still carries value. Nothing here says that space is worth less to a buyer. It says it gets reported on a different line. Reclassification and devaluation are separate things, and conflating them is where the conversation with a seller usually goes wrong. Getting the numbers straight before the listing goes live is the cleanest fix, which is the case for pre-listing measurements.
If it matters to your file, ask how the size adjustment was derived. On property types where the data will not split, a single supported blended rate is more defensible than two tidy-looking rates that the market never actually produced.
The measuring standard is not going to change to match the MLS, and the MLS shows no sign of moving to ANSI. Denver has a lot of bi-levels and split-levels, which makes bi-level square footage one of those quirks worth understanding rather than arguing with. More of the common questions are answered in the Denver appraisal FAQ and the guide to residential appraisal in Denver.
Charles Volk, SRA, is a residential appraiser with 24 years of experience across Denver Metro and Boulder County, and serves on the Board of Directors of the Colorado Chapter of the Appraisal Institute. VolkHaus Appraisals works directly with lenders, attorneys, CPAs, and private clients, with no AMC in between.