Our regular commenter elbarcosr backed me up on how wacky the zestimates have been lately. They seem to be getting worse, which is hard to believe.
Being a Zillow homer now, I thought I better look into it.
Let’s serve up a nice big softball. Certainly the zestimates have to be accurate on recently-sold homes, don’t they? We saw how Redfin’s evaluator can cozy up close to a recent list or sales price, and you can’t blame them. After a few years, the database would look pretty consistent.
Does Zillow do the same? Wouldn’t it make sense to have your algorithm compute a recent sales price into the property’s zestimate? Because if you did, it would also help value the nearby homes that haven’t sold recently – because that’s how everyone would value them.
Evaluations in unique, non-tract areas is tougher. But if we are just looking at recently-sold properties, and their zestimates – the uniqueness shouldn’t matter as much!
I looked at 28 homes sold in La Jolla, Del Mar, and Rancho Santa Fe that closed between $2,000,000 and $3,000,000 in 4Q15, and compared their sales price (the definition of value) to their zestimates.
The average margin of error was 16%, and after removing the four that were wrong by more than $1,000,000, the average error was still 12%.
These are houses recently sold, and their sales price defines the actual value!
Even though the $2,000,000 to $3,000,000 range is the lower end for those areas and there are plenty of comps to help pin-point a zestimate, let’s consider an easier target.
Carmel Valley should be the hotbed of zestimate accuracy, especially when we look at the low-end where every data point is a pure tract house.
There were 57 CV sales in the fourth quarter between $1,000,000 and $2,000,000 that were considered.
The average margin of error was 3.6%, which is probably acceptable. But if it was any higher, there would be concerns – these are tract houses that just sold in 4Q15, and have a long history of steady comps around them!
My takeaway?
The only zestimates that might be close are in pure tract neighborhoods.
Disregard all others.
The day before we listed our townhouse, the Redfin estimate was $13,000 higher than our planned list price. OK, cool.
The minute we listed the property, Redfin’s estimate immediately plunged $24,000 — now Redfin users see an estimate $11,000 lower than our list price.
Jim, have you experienced buyers citing Redfin/Zillow estimates when making offers?
We as realtors continue to give your data to Zillow and Trulia for free. They in turn take our money to advertise with them, after we gave them free content. Realtors all over the country fight the “Zestimate” when listing a property. It also becomes a battle when taking buyers out and have them tell you the current price is too high based on the “Zestimate”. The “Zestimate” makes me Zick. Our MLS data is valuable, we give it away like candy to anyone that has a real estate based search site. It’s a double edged sword, if our companies don’t allow it on the internet our sellers would not list with us. The system is broke. We need our own national site to crush the Zillow’s of the world and stop giving our information away for free.
Nice idea but it is too late now.
We gave Zillow a 10-year headstart, then we sold realtor.com so agents are left with no national website in our control. To start from scratch would take years and big $$ and we’d still have to spend $100 million per year on advertising it to try to catch up. Upstream is going to be a national database – that, and about $5 will get you a cup of coffee.
http://www.notorious-rob.com/2016/02/a-quick-question-on-upstream-the-local-mls/
NAR blew it. Statewide and local associations blew it. Large franchise companies blew it. Now we are left with the rubble.
Agents can either join Zillow’s quest for world dominance, or wait for the inevitable steam roller to flatten you. Or go grab a salary at Redfin, the happy beneficiary of the agent fall-out.
My experience is that I don’t even believe it in a late 90s Carlsbad tract. First, my zestimate increased by _56%_ from Feb 12 to Nov 14 – not even two years! Then corrected and dropped 25% from Nov 14 to Aug 15. Now back up 20%.
Given that my exact model sold 2 months ago for $275k less than my zestimate means “Zillow is not even close.” And that house took almost 5 months to sell! And a year ago my exact model sold 5 doors down – again, $225k less than my zestimate but on a 2.5x bigger lot. Same house, totally different estimates. Zillow to me makes no sense – it’s pretty much random in our tract.
Zillow estimates are garbage. Any home estimate that only used numbers in a public record cannot take into account what needs to be considered to arrive at a reasonable estimation. It is a pity that people give them any credence at all. It is a poor shortcut for the lazy or uniformed. A guy like Jim, drawing on years of experience in these neighborhoods combined with 15 minutes on the MLS, can provide a more credible estimate than Zillow 100% of the time.
I’m surprised there hasn’t been a class action against zillow yet regarding prices. Not trying to do lawyers jobs for them but all you’d have to do is track zillow prices then follow up with sellers that had to deal with zestimates that were lower than the selling price.
I actually find it shocking how wrong the Zestimates can be. I’m a software developer and firmly believe that anything people can do, computers can be programmed to do. If Jim can be that accurate (and I know he can, I’m a customer), then I believe a computer can replicate it. Jim doesn’t operate on gut feelings, he operates on facts. Facts that can be fed into a program.
Not sure what the point of this comment is, except that Zillow apparently sucks at their primary mission. Oh, and “hire good help”. 😉
Let me add that 80% of the zestimates reviewed here were on the high side. The sellers that use their estimates for a starting point, and then add the seller-obligatory 5% to 10% to generate their list-price expectation are in for a rude awakening.