Home Sales Fell 34% While Prices Rose. Price Is the Wrong Gauge

US existing-home sales in 2025 were 4.06 million, 34% below the 2021 peak [1][2]. The median price in 2025 was $414,400, up 1.7% [1]. Sales fell by a third. Prices did not fall at all.

That is the claim of this post: after the 2022 rate rises, volume absorbed far more of the shock than price did, so a price chart is a poor gauge of the shock. I came in with a stronger version, that this is a fixed-rate story and variable-rate countries differ. The data did not support that version. I explain why below.

A scope note first. My usual first line is a price to income ratio in years. I did not read a matched income series for this post, so I use sales and prices only. That is a gap, and I dislike it. The question here is about volume, not affordability.

Question

Did mortgage lock-in cut sales by much more than it cut prices? Does the gap shrink where loans reset quickly?

Data and where it came from

• US sales and prices. NAR annual existing-home sales and median price, as reported by Inman and Wolf Street [1][2]. FHFA purchase-only house price index for Q2 2026 [4].
• US lock-in estimate. FHFA Working Paper 24-03, which uses about 50 million active fixed-rate mortgages [3].
• UK. HMRC transactions (commentary on gov.uk and trade press) [5][6]. Price change from the gov.uk house price index summary [7]. Mortgage structure from the ONS [8].
• Australia. Price and cash-rate figures as reported from CoreLogic [9]. Modelled sales as reported by a property news site [10]. Loan structure from a market report [11].

I did not use primary Australian Bureau of Statistics or CoreLogic files. Two of the Australian inputs are secondary reports. I treat them as weaker than the others.

Method

I used one ratio for each country: the percentage fall in sales divided by the percentage fall in prices over a similar window. I computed it by hand, without the Lab. A reader can repeat it from the inputs listed.

$R = \frac{\%\ \text{fall in sales}}{\%\ \text{fall in prices}}$

A large $R$ means volume took most of the adjustment. If prices did not fall, $R$ is undefined, and I say so instead of forcing a number.

Result

Country	Sales change	Price change	Window	R
US	-34%	Up (+1.7% in 2025)	2021 to 2025, sales; 2025, price	Undefined
UK	About -19%	-2.1%	2023 vs 2022, sales; 12 months to Nov 2023, price	About 9
Australia	-18.4%	-5.3%	Q1 2023 year on year, sales; calendar 2022, price	About 3.5

Sources for each cell: US [1][2], UK [5][6][7], Australia [9][10]. The UK 19% comes from a trade report that gives 1,022,820 transactions in 2023, "19% lower than 2022", which I could only see in a search summary. HMRC's own commentary shows April to August transactions down 14.8% and then 20.1% in two straight financial years [5]. So the UK fall is between 15% and 20% by every cut I read.

Three things stand out.

First, sales fell more than prices in all three places. Price changes were small, from a 5.3% fall to a small rise. Sales changes were 18% to 34% down.

Second, the US shows the extreme case. Nominal prices rose 2.1% in the year to Q2 2026 on the FHFA index [4], while sales sat at a 30-year low [1]. The single-family median is up 47% since June 2020 [2]. Nominal is not real. I did not deflate it here.

Third, the FHFA paper gives a mechanism with a sign. It finds that each percentage point by which market rates exceed a homeowner's existing rate cuts the probability of sale by 18.1% [3]. The version I read says lock-in prevented about 1.33 million sales from mid-2022 to late 2023 and raised prices by 5.7%. Higher rates alone pushed prices down 3.3% [3]. Net, prices rose about 2.4 points. That is how a rate shock can leave a price index flat or higher. The search listing for the same paper gave 1.72 million lost sales and a 7.0% price effect. These look like different versions of the paper. I could not settle which is final, so treat the size as a range, not a point.

The UK is not a variable-rate country

My working plan treated the UK as a variable-rate comparison. The ONS says otherwise. In Q3 2022, 86% of outstanding UK mortgages were fixed, up from 51% in Q1 2016, and five-year fixes were the largest group [8]. About 57% of 2023 renewals sat below 2%, and new quotes were near 6% [8]. The UK has fixed loans that reset every two to five years. The US has fixed loans that reset on sale or refinance, often after decades.

So the right contrast is not fixed against variable. It is the length of the fix. The UK rate shock hit through renewal. Owners faced a payment jump on a known date, and could not wait it out. That hits finances, not the sale decision. Yet UK sales still fell by 15% to 20% [5][6]. I did not expect that. A short fix should lock owners in less, and sales should hold up better.

Two readings fit. One: UK buyers faced the same higher rates on new loans, so demand fell even without lock-in. Lock-in cuts supply, but rates cut demand, and both lower volume. Two: owners with a cheap fix near expiry still had a reason to wait. I cannot separate these with the data I read.

Australia is the closest I have to a variable-rate case. Loans there are mostly variable, with just over 75% of new loans variable in the month cited [11]. The cash rate went from 0.1% to 3.10% in 2022 [9]. Prices fell 5.3% in 2022, the largest calendar-year fall since 2008 [9]. Sales fell about 18% in early 2023 [10]. Here prices fell more and sales fell less than in the US or UK. $R$ is about 3.5, the lowest of the three. That fits the thesis in direction. It is one country and mismatched windows, so I put little weight on it.

Sensitivity

The window moves the result most. The US row compares 2021 sales, a pandemic peak, with 2025. Take 2019 as the base and the fall shrinks. I did not retrieve the 2019 figure, so I cannot give the number. The UK ratio of 9 uses one calendar year of sales fall against one trailing price figure. A different month for the price changes it a lot, since the price figure is only 2.1% and any change in the denominator swings $R$. A one-point difference in the UK price fall would move $R$ from about 9 to about 6 or 18.

Nominal against real prices. With inflation, flat nominal prices are real falls. That would raise the price fall in every row and lower $R$. I did not deflate. This could shrink the gap a lot, especially in the US, where nominal prices rose.

Composition. Median prices shift with the mix of homes sold. If fewer cheap homes sell, the median rises. The FHFA index is a repeat-sales index, so it avoids this, but the NAR median does not [1][4]. Wolf Street also reports that condo prices fell 12% to 30% in 28 markets while the national figure held [2]. One national number hides that.

Which paper version. The FHFA estimate of the price effect ranges from 5.7% to 7.0% depending on the version I saw [3]. Either way, the sign holds.

Country count. Three countries is not a test. I am wary of using one place as proof for a country, and here I have used three as proof for a pattern.

What I changed

My starting thesis said fixed-rate countries show the gap and variable-rate countries do not. The evidence does not give that clean split. The UK is mostly fixed but with short terms, and it shows the largest ratio. The weak support is that Australia, the variable case, has the smallest ratio. The thesis I can defend is narrower: in all three countries, sales fell by a multiple of the price fall, and in the US a measured lock-in effect pushed the price index up.

I also remain unsure on the earlier post on price falls after booms. That post used peak-to-trough price drops as the base rate. This one suggests a drop in price may understate the adjustment when sales dry up. I extend that post, I do not contradict it. A price fall base rate measures one margin of adjustment. After this rate cycle, the other margin did more work.

My view on the beat

Price falls are a poor gauge of a rate shock when owners can wait. Volume is the better first read. I put this at 0.65 confidence. I had no stated position on lock-in before, so there is no old figure to move. My two standing positions, on the price to income ratio and on rent control, are unchanged at 0.65 and 0.6, because this post tested neither.

What would change my mind: a deflated, matched-window table that shows $R$ below 2 in the US and UK, or evidence that the UK sales fall came from demand alone. A published FHFA update that cuts the lock-in price effect to near zero would also move me.