UPDATE July 2019: Zillow changed how they calculate median errors. They now calculate the median error of their Zestimates for houses “not listed for sale” separately from houses “listed for sale.” Nationally, the typical error is $18,000 for houses that are not listed for sale.
UPDATE: Zillow greatly improved the accuracy of their Zestimates in June 2016 so I updated this post in July 2016. Post was originally published in July 2015.
Note: I haven’t updated the video to reflect Zillow’s 2016 improvment in accuracy or the 2019 changes.
The Problem
You have a home for sale and Zillow’s Zestimate for the house is way too low.
To make matters worse, Zillow puts their (low) estimate of your home’s value right below your list price so everyone that looks at your house on Zillow thinks your home is way overpriced.
You’re starting to wonder if the low Zestimate explains the lowball offer you got or why your home hasn’t sold yet.
Low vs. High Zestimates
Low Zestimates can certainly complicate things for home sellers but, counterintuitively, high Zestimates can be even more damaging to home sellers, that is, if the sellers actually believe those high Zestimates.
Low Zestimates. Home sellers with low Zestimates will tend to immediately dismiss them as inaccurate.
High Zestimates. Some sellers with high Zestimates will tend to immediately anchor on the high estimates and think they’re accurate, holy and just.
There’s this natural emotional response – home sellers kind of fall in love with a high estimated value. They tend to get starry-eyed and start imagining how they’ll spend all that money. So they tend to overprice their homes. Overpricing is very expensive to home sellers but that’s a post for another day.
Most sellers tend to accept high Zestimates as gospel.
If you’re reading this, however, you probably have a low Zestimate. When people get low Zestimates, they start to search Google.
Top 20 search terms with “zestimate” include: “zestimate accurate,” “is zestimate accurate,” and “zestimate accuracy.” I bet the vast majority of the people making those searches have low Zestimates.
Visualize the Inaccuracy
As a former economist, I’m going to visually show you how accurate/inaccurate Zestimates are. If more people understood the true level of accuracy/inaccuracy of Zillow Zestimates, the damage to home sellers and buyers would be lessened. That’s my goal here, anyway.
I know I can at least help you better understand the accuracy of Zillow Zestimates.
That will help you when dealing with home buyers or sellers that throw inaccurate Zestimates in your face during contract negotiations.
Be sure and read my suggested response below.
I’m afraid, however, that I’m not going to be able to stop home sellers from falling in love with high Zestimates and I’m not going to be able to stop home buyers from feeling fear when offering fair market value for homes that have low Zestimates.
Zestimate Accuracy Table
Zillow is actually very good at disclosing the accuracy of their Zestimates.
I think it’s partly a CYA move to protect Zillow against complaints about misleading people when their Zestimates are inaccurate. I’ll paraphrase, “We told you in detail right on the website how inaccurate Zestimates are, so stop complaining about how inaccurate Zestimates are!”
Nevertheless, I love their honesty and transparency.
Zillow calculates the accuracy of Zestimates by comparing the actual sale prices of homes to the Zillow Zestimates for those same homes right before they sold.
Zillow’s typical error. You can see from the table above that the accuracy varies from city to city but overall the median error rate for all Zillow Zestimates in the U.S. is 6.1%, according to Zillow.
I applied their median error of 6.1% to Zillow’s own estimate of the median sale price in the U.S. in May 2016 of $229,737 and got a typical error of $14,000.
UPDATED July 2019: I applied their median error of 7.7% for off-market/not-listed-for-sale houses to Zillow’s own estimate of the median sale price in the U.S. in July 2018 of $234,900 and got a typical error of $18,000.
UPDATED July 2019: The typical Zillow Zestimate error for typical houses that are NOT actively “For Sale” is $18,000, but you don’t know if it’s $18,000 too high or $18,000 too low. And it gets worse because HALF the time Zillow Zestimates are off by MORE than $18,000, sometimes a LOT more than $18,000.
On one hand, it’s amazing that Zillow can get so close to the actual sale prices just by looking at public and other data on the houses.
But on the other hand, for home buyers and sellers, the estimates are far too inaccurate to be usable for pricing homes for sale.
Visualize Zillow Errors
I took Zillow’s error rates for a sample city, Denver, from the table above and created the graph below so you can visualize the accuracy of Zillow Zestimates.
By the way, I didn’t pick a city with unusually bad Zestimates, their Denver estimates are actually more accurate than average.
UPDATE: In June 2016, Zillow greatly improved the accuracy of their Zestimates.
I created the graphic below to show the improvements. I think the error spread in 2016 is a lot tighter and more focused on the bullseye of the actual sales price.
Click graphic to enlarge
Shotgun! Using Zillow’s own numbers, you can see their Zestimates are scattershot. Although the 2016 estimates are much better, they’re still way too scattered to use when pricing your home. They’re a great way to find out the general ballpark value of your home but that’s about it.
Example. Let’s say a Denver home has a fair market value of $300,000. According to Zillow’s Zestimate Accuracy Table, 10% of their Zestimate prices were off my more than 20% from the actual sale prices. Half of that 10% are Zestimates that are too high by 20% or more, and half are Zestimates that are too low by 20% or more.
That means you have a 5% chance Zillow will give you a Zestimate of $360,000 OR MORE, and a 5% chance Zillow will give you a Zestimate of $240,00 OR LESS. Yikes!
Takeaway
Here’s a talking point you can generate to give to buyers who bring up low Zestimates.
Zillow makes it very clear that Zestimates are NOT appraisals. Zillow says that for a home in the [$X] price range in [Your City], their typical error is $X [[Median Error for Your City in Zestimate Accuracy Table] x [Your List Price]], but you don’t know if it’s $X too high or $X too low. And it gets worse because HALF the time Zillow Zestimates are off by MORE than $X, sometimes a LOT more than $X.
Here’s how it would look for a $300,000 home in Denver.
Zillow makes it very clear that Zestimates are NOT appraisals. In fact, Zillow says that for a home in the $300,000 price range in Denver, their typical error is $16,800, but you don’t know if it’s $16,800 too high or $16,800 too low. And it gets worse because HALF the time Zillow Zestimates are off by MORE than $16,800, sometimes a LOT more than $16,800.
If you don’t live in a metropolitan area, the Zestimate Accuracy Table also has Zillow’s error rates for states and counties, see the blue “States/Counties” link near the top-left of the table.
I sure hope this information has been helpful to you.
I know having a low Zestimate is a tough problem to have.
ADDED: To make this easier for you, I took Zillow’s own data and created a Zillow Zestimate Accuracy Chart which includes the Typical Zillow Error in Dollars for 25 U.S. cities.
# # #
“Zillow” and “Zestimate” are registered trademarks of Zillow, Inc.
16 Responses to Zillow’s Typical Error Is $18,000 (2019)
The general consensus of professionals in the industry is that Zillow has a better accuracy rate in active suburban and city areas where there are lots of recent comparable sales. Excellent article to test the assumptions and examine the hard data. I think your article helps emphasize that Zillow is a tool to find your answer today, not necessarily THE answer today. John, how do you see Zillow improving its accuracy over time? One option for them might be to utilize their huge network of agents (who advertise with them for example) to help value homes. One challenge of that would be preventing human bias, but perhaps some sort of triangulated crowd-sourcing program that requires three impartial assessments would work, and it could certainly prevent any agent from trying to value their own client’s property. Technology is awesome, I’m rooting for Zillow. I don’t advertise with them or anything, but I think its a smart service and I like seeing American innovation.
“John, how do you see Zillow improving its accuracy over time?” That’s a fun question!
I don’t see them using agent opinions or otherwise crowdsourcing opinions of value. Too human. Too hard to maintain quality. Too variable.
A possibly crazy but sufficiently wonky area they might explore would be to do computer analysis of listed homes’ photos or all home’s Google street view. Maybe the could fine-tune their Zestimates by incorporating photo data. Okay, that’s pretty wild but I couldn’t think of anything else! 🙂
I think you’re on to something — photo recognition is already in mainstream use. And one thing we keep relearning about computers is they keep surprising us. I think eventually the computer screen will be a hologram. This whole up-down scrolling we do now is going to be a thing of the past. With touch screens today, you tap the screen to click, but you keep your finger pressed to scroll – our minds do this automatically. With holograms there is going to be a whole new hand language, so watch out all you Italians out there who like to speak with your hands – the next generation of computers may want to borrow some of your slick moves!
[…] To see a detailed discussion of Zillow Zestimate errors, including two graphs that clearly show you how inaccurate Zestimates are, see my previous post. […]
[…] than Zestimates. Weiss says their median error is less than 4%. For comparison, the median error for Zillow’s Zestimate of home values is greater than […]
Another reason was that Rascoff s home was located on an arterial road while nearby recently sold homes sat on quieter streets. Zillow continues to research how to program Zestimates to account for such factors, but we haven t fully cracked the nut on that one yet, Humphries said.
With the June 2016 change in the algorithms our zestimate dropped almost 200,000.00! We had only purchased our home in June 2015 where the zestimate was right about on target and continued to go up form there. I contacted Zillow and sent them the appraisal which was 1 year old.. They told me nobody’s home should be affected by that much. I agree completely, and yet it was!
We live in a waterfront community in a rather rural area, with many lower priced homes. I know this will negatively impact our home should we decide to sell . Who is going to pay that much more? I never would! We are just sick about this.
That’s the worst Zestimate story I’ve ever heard! $200K! Yikes!
It sounds like you don’t have any immediate plans to sell so Zillow may very well solve the problem before you decide to sell.
If you were to put it on the market with such a bad Zestimate on Zillow, you might want to include a list of recent comparable sales in all marketing material. Or get an appraisal right before you put the house on the market and mention up front and repeatedly that the price is based on a recent appraisal. That should help allay some of the fears from potential buyers.
“I don’t see them using agent opinions or otherwise crowdsourcing opinions of value. Too human. Too hard to maintain quality. Too variable.”
haha…spoken like a true salesman.
FYI, the more variable the more valuable the training data becomes and the more accurate the learning and its prediction…but ok
I agree with you.. It is hit and miss with Zillow estimates on houses and they pay no consequences, when they make huge erroneous mistakes.. They were off by $50,000 when I sold my house in 2 days & then they tried to correct their mistake by having the house go up by $50,000 in one day.. I have the screen images to prove it… The owners of this web-site should be liable for posting erroneous information because they can really hurt someone’s chances to sell their house for what it actually is worth, when potential buyers for your house in a particular are influenced by the information that they post… If you don’t believe me… Someone should do something to close down this website..
Howdy I am so excited I found your weblog, I really found you by
error, while I was browsing on Aol for something else, Anyways
I am here now and would just like to say many thanks for a tremendous post and a all
round interesting blog (I also love the theme/design), I don’t have time to read it all at the minute but I
have saved it and also added in your RSS feeds,
so when I have time I will be back to read a lot more, Please do keep
up the great work.
We live on what is known as an “Airpark”, there are 116 homes built around a small airport. The “zestimates” on homes on the airpark are typically much lower than what they actually sell for. The vast majority of the “comparables” that Zillow uses in the “zestimates” for houses on the airpark are not on the airpark and much less valuable. There appears t be no way to correct this.
I’ve been tracking our zestimate for several months and it’s been fairly steady, changing by about -100 to + 100 per day. Suddenly last week, our zestimate shot up by about $12,500 in one day! What would cause that change overnight? It seems like it could be a change in the algorithm or a data update of some type? I’m dismissing it for now.
Sounds like they updated the algorithm or got some new data. Or maybe it’s an error. My guess is that’s it’s an improved Zestimate.
I’m searching due to an over-inflated zestimate and thought I’d share. 1016 Washington st in Gloucester MA. For past 5 years value has bounced between 350 and 500. Today sitting at 1.02 million. Refi appraisal just this month came in at 387 due to lack of current comps. Property worth perhaps 550. But clearly zesrimate is incorrect by perhaps a factor of 2.
Its all very quantitative – houses of similar size (BR, BA) and location are grouped together. Then SF becomes the driver and lot size. SF is a very reliable factor, quantitatively, when looking at similar #BR and #BA. But, Zillow might not see town property lines, or know that this side of the road goes to this school vs that school. (interesting, it does know traffic patterns, thanks to apps like WAZE, so it must discount busy streets somehow) Zillow is also autocorrecting. It learns over time. It needs dozens of transactions to improve, but what we witness is small changes. Example – It reconciles prior list vs sale prices vs old Zestimates and refines its calculations, but requires a large population of new data to make noticeable change and these markets don’t move that quickly. But, RE valuation is not exact, and a Zestimate is just an estimate (vs appraisal). For a more exact number, use a local realtor or appraiser.
Comments are closed.