Selling models and market structure
The Count Moves First
6 min read
In September 2026 the Seattle metro had an estimated 84.1% more home sellers than buyers, and the San Francisco metro had 10.6% fewer sellers than buyers, according to Redfin. Both readings describe the market now. The comparable sales that price a home in either city describe the market as it stood when those contracts were signed, so a valuation built from closed sales leans high in Seattle and low in San Francisco until the sold record catches up with the count.
That is the whole argument, and the two cities make it unusually easy to see, because they share an industry and are moving in opposite directions at the same time. Redfin attributes the split to wealth from the AI boom in San Francisco and to layoffs at large employers in Seattle. The cause matters less here than the order in which the evidence arrives.
What Redfin's September count says about Seattle and San Francisco
Redfin publishes a monthly estimate of buyers and sellers across the 50 most populous US metros. A metro with more than 10% more sellers than buyers is a buyer's market, one with more than 10% fewer sellers than buyers is a seller's market, and anything inside that band is balanced. In September, 41 of the 50 were buyer's markets, 4 were balanced and 5 were seller's markets. Nationally there were an estimated 62.6% more sellers than buyers, and the estimated count of buyers, about 928,892, was the lowest in records dating to 2013.
Seattle sits deep on one side of that line and San Francisco sits just over the other. Three consecutive monthly releases show both moving away from balance:
| Month, as first published | San Francisco metro | Seattle metro | United States |
|---|---|---|---|
| July 2026 | 6.3% fewer sellers than buyers, balanced | 65.1% more sellers than buyers, buyer's market | 51.3% more sellers than buyers |
| August 2026 | 12% fewer sellers than buyers, seller's market | 72% more sellers than buyers, buyer's market | 57.9% more sellers than buyers |
| September 2026 | 10.6% fewer sellers than buyers, seller's market (2,378 sellers, 2,658 buyers) | 84.1% more sellers than buyers, buyer's market (12,014 sellers, 6,525 buyers) | 62.6% more sellers than buyers |
San Francisco is the weakest of the five seller's markets in September, behind Nassau County, Montgomery County in Pennsylvania, Newark and Milwaukee. Redfin noted in its August release that it was only the second month in four years that San Francisco had been classified as a seller's market, the other being June. Seattle recorded the second largest monthly increase in seller surplus among buyer's markets in August, behind Orlando.
How Redfin counts buyers and sellers, and what the count cannot see
The measure deserves credit before it gets examined. Redfin publishes the method, the thresholds and the metro table every month, which is more than most market indicators offer, and the measure looks forward in a way a median sale price cannot. Sellers are the number of active listings in the MLS. Buyers are estimated from Redfin's own data on the typical time between a buyer's first tour and closing, combined with MLS pending sales. The figures are seasonally adjusted.
They are also revised, and the revisions are not small. Seattle's July figure was first published on 13 August as 65.1% more sellers than buyers. The August release, a month later, described July as 55%. The national July figure moved from 51.3% to 52.1%, and the national August figure from 57.9% to 56.9% in the September release. None of that is a flaw in the work. An estimate built on pending sales will move as pending sales close or fall through. But it means the count is a model output with its own uncertainty, not a census, and any argument resting on a single month should be read with that in mind. This one rests on direction across three releases, which survives the revisions in both cities.
What the sold record shows in the same month
Now set the count beside the closed sales, which is what a comparative market analysis, an automated valuation and an appraisal all start from. The figures below are Redfin's September city level data, which is narrower than the metro geography behind the count, so the comparison is directional rather than like for like.
| September 2026, city level | San Francisco | Seattle |
|---|---|---|
| Median sale price | $1,579,248, up 14.9% year over year | $854,593, down 2.7% year over year |
| Sale to list price | 115.0% | 99.2% |
| Homes sold above list price | 69.7% | 22.6% |
| Homes with price drops | 10.1%, down 5.0 points | 32.0%, up 5.1 points |
| Homes sold | 1,411, up 7.5% year over year | 1,943, down 14.1% year over year |
| Median days on market | 17, down from 26 | 19, unchanged |
Read the Seattle column without the count beside it and it describes a market close to balance. Homes that sold went for 99.2% of their asking price, more than a fifth sold above it, and the typical sale took 19 days, the same as a year earlier. The median price is down, but by 2.7%. Nothing in the sold record looks like a metro with 84% more sellers than buyers.
The listing side tells the other story. Almost a third of homes for sale carried a price cut, and sales were down 14.1% on the year. Those two figures are where the surplus shows up, and neither is a sale price.
Why the sold record runs behind the count
There are two mechanisms, and they compound.
The first is timing. A sale enters the record at closing, typically 30 to 60 days after the contract was signed, and the price reflects the conditions on the day it was negotiated. The count changes the day a listing goes live or a buyer stops touring. A Seattle sale that closed in September was most likely agreed in July or August, when the seller surplus was somewhere between 55% and 72% depending on which vintage of the estimate you trust, not the 84.1% that applied by the end of September. The same arithmetic runs in reverse in San Francisco, where a September closing was most likely agreed in July or August, when the metro was balanced or only just a seller's market.
The second is selection, and it is the larger of the two in a buyer's market. Comparable sales contain only the homes that sold. In Seattle in September that was the subset of listings priced close enough to clear, at 99.2% of asking. The listings that sat, were cut, or were withdrawn never become comparables. The sold record in a market with a large seller surplus is therefore a record of the listings that were already priced correctly, which is why its sale to list ratio looks healthy while a third of the inventory is being cut. In a seller's market the bias runs the other way: list prices become the weak signal, because 69.7% of San Francisco sales cleared above them, and a valuation anchored on asking prices or on spring comparables starts below the market it is pricing.
This is a different problem from the one in the adjustment arrives late, where the sale price is sound but the buyer behind it faced a different cost of money. Here the price is sound and the buyer pool behind it was a different size. It is also distinct from the complaint that automated tools run on stale data, covered in why the AI market read is months old, because refreshing the data daily does not help. A comparable has to have closed, and closing is the lagging event.
Same model, opposite errors
Put a closed sales valuation into both cities in the same week and it leans in opposite directions, for the same reason. The method is identical. What differs is which side of balance the market crossed while the comparables were in escrow.
| City | What the count says | What closed comparables show | Direction a closed sales valuation leans |
|---|---|---|---|
| Seattle | Seller surplus widening, from 65.1% to 84.1% across three releases | Sales near list, at 99.2%, from the listings that cleared | High, because the record holds the sales that cleared and omits the third of listings being cut |
| San Francisco | Balanced in July, a seller's market by August and September | Sales well over list, at 115.0%, agreed while the market was cooler | Low, because the comparables were negotiated before the buyer pool outgrew the listings |
The third column is derived reasoning from the two published series, not a measured error. Its size depends on the property, the submarket and how old the comparables are, and in a single neighbourhood it can be swamped by everything else that separates one house from another. The direction is the point. It is a clean case of Context Blindness™: the gap between what the record can measure and what the people working the market already know, because the agent who has watched a Seattle open house go quiet, or a San Francisco listing draw a stack of offers, has the count before Redfin publishes it.
The tech headlines do not change any of this. Microsoft's offer of buyouts to around 60,000 employees was reported on 26 October and covers selected teams in the United States, Canada, India and the United Kingdom, so none of it is in the September figures, and its effect on Seattle housing, if any, is months away from appearing in a sale price.
What a valuation does with this
The practical answer is to stop treating closed sales as the only evidence and to read the listing side as a second witness. Price cuts, the share of homes sold above list, sale to list ratio and withdrawn listings all move before the median sale price does. In Seattle that means weighting the most recent comparables by contract date, noting how many active listings near the subject have already been cut, and widening the range for a stated reason rather than publishing the comfort of a 99.2% sale to list ratio. In San Francisco it means treating the asking price as a floor rather than an anchor, and saying on the page that the comparables were agreed in a cooler market.
CMAflow carries the same limit. Its report is built from closed comparable sales, so it inherits the lag described here, and the count of buyers and sellers is not one of its inputs. What it does provide is a place for the reasoning: the agent commentary prints what the agent knows about the current market as its own section of the report, so a stated reason for a wider range travels with the number rather than staying in the agent's head.
Is Seattle a buyer's market right now?
Yes, by Redfin's measure. In September 2026 the Seattle metro had an estimated 84.1% more sellers than buyers, 12,014 active listings against 6,525 buyers, up from 72% in August. Redfin classifies any metro with more than 10% more sellers than buyers as a buyer's market.
Is San Francisco a seller's market in 2026?
Narrowly. Redfin's September 2026 count put the San Francisco metro at 10.6% fewer sellers than buyers, just past the 10% line that defines a seller's market. It was one of only 5 seller's markets among the 50 most populous US metros, and the weakest of the five.
Why do comparable sales lag a change in the housing market?
Because a sale enters the record at closing, typically 30 to 60 days after the contract was signed, and the price reflects conditions on the day it was negotiated. A count of sellers and buyers changes the day a listing goes live or a buyer stops touring. Comparable sales also include only the homes that sold, so listings that sat, were cut or were withdrawn never appear in them.
How does Redfin calculate buyers versus sellers?
Sellers are the number of active listings in the MLS. Buyers are estimated from Redfin data on the typical time between a buyer's first tour and closing, combined with MLS pending sales. The figures are seasonally adjusted and revised: Seattle's July 2026 figure was first published as 65.1% more sellers than buyers and restated as 55% a month later.
The count will be revised again, and the sold record will eventually agree with whichever version holds. Until it does, the most useful thing a pricing document can carry is the date its evidence was agreed and what the market looked like that week.
For a brokerage that is a record worth keeping on its own terms. A closed file holds the price. It does not hold what the room felt like when the price was set, how many offers were due, how many listings nearby had just been cut, and that is precisely the evidence that runs ahead of every public series by a month or more. An office that writes it down on the deal, sale by sale, owns the leading indicator rather than waiting for one to be published.
Sources. Redfin, buyers vs sellers, September 2026 data, published 14 October 2026. Redfin, buyers vs sellers, August 2026 data, published 10 September 2026. Redfin, buyers vs sellers, July 2026 data, published 13 August 2026. Redfin, San Francisco housing market booms while Seattle slumps, published 2 September 2026. Redfin, San Francisco housing market and Redfin, Seattle housing market, September 2026 data. TechCrunch, Microsoft buyouts, 26 October 2026.
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 | Insights | Blog