July 15th Update: Redfin trickled out an extra two weeks worth of data today, so this article has been updated accordingly since it was written yesterday.
Today is the Thursday following the second Saturday of the month! For those that don’t watch housing market data for fun, that means it’s Redfin’s data release day, which used to occur once per month but is temporarily being released more frequently to track the latest developments in the housing market. After taking a look at the data, I realized that it did not reflect as much of what I feel like I’m seeing in my market (Greater Boston Area) which is rapidly climbing levels of inventory and an increasing number of price cuts compared to the spring market.
Housing market data represents lagging transactions. The median time to close on a home in my market is about 38 days according to Redfin. What today’s housing market data release by Redfin through July 10th represents then, is homes which went under contract around the last week of May and the first week of June.
The above is something I’ve talked about in the past when writing about housing. What I haven’t explored yet is how selective representation of that lagging housing data can even further hide the reality of what is happening in the housing market right now.
Today I am going to take a look at Middlesex County, Massachusetts. Middlesex County is an excellent representation of the Greater Boston Area — it excludes the city of Boston itself, but includes some close-in urban areas like Somerville and Cambridge, suburban towns between I-95 and I-495, and some further flung towns bordering on being rural up by the New Hampshire state line. It is the most populous county in both Massachusetts and New England and thus represents the experience of the modal New Englander.
If the housing market is going to plateau, correct, or crash, one indicator would be a massive increase in the number of price cuts by home sellers, indicative of large scale demand destruction on the part of buyers. So let’s take a look at this metric in Redfin’s Data Center for our area of interest, which includes today’s July 10th data point (click to enlarge):

On the surface, this data release does not look good for those predicting a housing bubble! Under 6% of active listings have seen price drops, a small increase from the pandemic FOMO frenzy of 2020 and 2021. But we can also infer from this graph that price cuts have just barely inched above the levels seen at the same time of year in 2019, the earliest year for which Redfin provides data.
In July 2019, 30 year mortgage rates were around 3.90% according to Mortgage News Daily. So when contrasted with the environment facing today’s buyers with 30 year mortgage rates around 5.70%, price cuts being at the same level as 2019 rates would indicate that sellers still hold much more power over buyers in today’s market than they did pre-pandemic, and that the vast majority of buyers are still giving sellers exactly what they want.
Well, that’s what it would indicate, to people who don’t dig into how this data is calculated. You see, the lowest frequency that Redfin offers to calculate their data on is a rolling 4 week average. For the record, I’m not necessarily saying that it’s an intentionally malicious exercise in lying with statistics. But it is strange that these iBuyers are data-driven companies, yet their presentation and analysis of said data is incredibly juvenile.
Rolling averages can help to visually present trends in data which are otherwise disguised by short-term fluctuations. The downside is that one loses fidelity in the data. Thus, sudden, volatile moves within the data which should otherwise be flashing alarm bells will end up looking not so threatening when averaged in with the normalized data in the remainder of the period. For example, the most recent four data points that Redfin presents in this price drop dataset for Middlesex County, MA are:
Date | Percent Active Listings with Price Drops (4 Week Average) |
6/19/2022 | 5% |
6/26/2022 | 5.9% |
7/3/2022 | 5.7% |
7/10/2022 | 5.8% |
Note that when calculating one of these four week rolling averages, the new weekly data point pushes the old one out of the set. So we can’t, for example, take the 0.9% increase in the rolling average the week of June 26th and multiply it by the duration to conclude that week had a 3.6% increase in the week-over-week percentage of homes with price drops. What we can conclude is that the four week period ending June 26th summed to a total which was 3.6% higher than the week of June 19th.
The rolling average can serve to dampen what may be a significant data point in comparison to historical data if weekly fidelity remained available. These weekly measurements themselves used to derive the rolling average are not publicized.
You can further verify the dampening effect of a rolling average by switching the chart to the 12 week duration, where Redfin’s data point for July 10th is 4.5%, in comparison to 5.8% for the 4 week average.
Is price cut data measured on a weekly basis really so volatile that it needs to be smoothed with a rolling average for an effective presentation? Doubtful.
Instantaneous Weekly Percent of Active Listings with Price Drops Measurement
Ready to learn how much uglier today’s housing market actually is than Redfin’s data would suggest? Let’s take a look at the total number of active listings in Middlesex County, MA:

For context, active listings is the number of homes available for sale. It excludes listings which are under contract (contingent or pending) as well as Redfin’s “coming soon” listings. We see that there are 2,298 homes for sale in all of Middlesex County.
Next we can apply a useful filter to dial in on only the listings which have price drops. To do this, head to the All Filters button, then scroll down to the Cost/Finance section and enable the Price Reduced filter, choosing one of the available time periods. We’ll do one week for our purposes:

After doing this for Middlesex County, I’m left with the following filtered results:

192 active listings in Middlesex County have had a price reduction over the past 7 days. Out of 2,298 active listings in the county, that means the percent of active listings with price drops in the past week is 8.4%!
The disparity between the instantaneous data in the market versus what Redfin’s lagging data smoothed out with a rolling average purports to report is pretty large. 1.44 times as many active listings today have price drops as compared to what the four week rolling average reports.
Just for fun, what if I download Redfin’s data and create an updated graph with today’s instantaneous measurement included, ignoring the rolling averages?

If today’s instantaneous measurement of the percent of active listings with price cuts is at all representative of the rolling average for July — and I have no reason to believe it’s an outlier based on the frequency with which I’ve been receiving price cut emails recently — Redfin’s data releases in a couple of weeks will look something like the above graph.
The instantaneous weekly data allows us to get a more accurate pulse on the market than the rolling average. Using it, we can conclude that the percentage of listings which cut their price in the past week is already higher than at any point in 2019.
Total Percent of Active Listings with Price Drops
When I first drafted this article I was confused by Redfin’s data presentation. I took the title of their graph to mean that they were measuring the percentage of active homes for sale which had a price cut at any point in their history, and just recording that data once per week. But no, I was wrong and they were just measuring the percentage of active listings which had a price drop in the past week, then smoothing that out with the rolling average. This error on my part did lead to an interesting metric to look at though, which we can dive into below.
Unfortunately, Redfin doesn’t allow us to filter “all listings with price drops” despite reporting this metric in their blog posts, however we can create a replica of this data set by filtering for homes which reduced their asking price over 30 days ago, then adding it to the number of homes which reduced their asking price within the past 30 days. First we filter by price drops within the past 30 days:

We see that 469 homes in Middlesex County have cut their prices in the past 30 days. When I change the filter to look at price drops which occurred greater than 30 days ago, the result is 172 homes.
Now we add the two figures together, and we end up with the total number of listings which remain active for sale that have had a price cut at any point in their listing history. That is 641 homes, a whopping 27.9% of the active listings in Middlesex County, MA which have seen a price cut.
This is an interesting metric because it shows that over a quarter of the homes on the market have reduced their listing price at some point, but still remain without buyers. Even if the weekly price cut percentages plateau, if this backlog of the total percentage of active listings with price drops continues to grow, it will represent a continuing shift toward potential favorable conditions for buyers.
We’re finally seeing the effects of quantitative tightening coming through in the data
As recently as March of 2022, home buyers could still get 30 year mortgages around the high 3.x% to low and middling 4.x% range. But March 2022 is also when the US Federal Reserve approved the first increase in their benchmark rate, officially kicking off quantitative tightening. Since then, several factors have masked what would be an obvious — to anyone with a crumb of understanding of economic theory, at least — impending correction in home prices.
Such factors included a final gasp of FOMO buyers rushing to lock in record high home prices at whatever interest rate they could get, fearing that home prices would plateau at best and the rising rates would permanently price them out. Stack on top of that the lagging nature of home sales data points while transactions are pending. Then add further obscurity with the use of rolling averages by real estate data providers like Redfin, even in data sets where it is not necessary.
The summary effect of all of these factors resulted in many months delay until the official data would reflect the new reality for home prices. At least for this month, those who can be tricked by statistics (and those motivated to fool them) have Redfin’s data to point to and proclaim that there’s no evidence of a housing price correction.
“Don’t fight the Fed” is solid advice. Unfortunately, many people have forgotten that it works both ways, during easing and tightening.
I’m privately tracking current levels.
https://docs.google.com/spreadsheets/d/1HOm5gc-3WZqDacxUGiwWIRg9K3Mq-cCvmOCn3msMYeo/edit?usp=sharing
Sam,
Thanks for sharing and great little tool you’ve put together! Apologies but your comment was automatically marked as spam (likely due to the link) so it took me a couple of days to find it. Glad to see that your percent of active listings with a price reduction in MA is 22.5%, which would ballpark with my calculation and suggest that Middlesex County is on the high end of average for the state with my figure of 27.9%. Intuitively this makes sense, as the median price in that county is on the high end for MA.
Interesting that the percentage of active listings with price drops in MA has doubled since the start of May. Looks like if the trend continues, we’ll see a doubling in this metric on a national level as well shortly.
I hope you continue to keep this updated, I’ll be happily following along!
isnt them just doing percent over a week? ur doing over a month?
Hi, thanks for the comment.
I downloaded Redfin’s data when writing this article which anybody can do for free from their Data Center. They do take data points every week, on Sundays. My original belief is that what they’re measuring is the total percentage of homes listed for sale in active status that have had a price drop, and they just take the data point on a weekly basis.
They trickled out an additional 2 weeks of data today so I will update the charts in this article, and also for some reason the period ending June 26th for Middlesex County was retroactively changed which now says 5.9% of homes had price drops, when I grabbed the data it was 6.1%. Just very strange as you’d imagine the input variables when taken are what they are.
I’ll also take a closer look if potentially what they’re measuring is just the number of price cuts in one week. If so that would change my graph to be less extreme but not the conclusion.