I’ve been meaning to analyze Zillow’s home price estimate accuracy for a while – especially since my brother keeps talking about those damn Zestimates. He’s not the only one, though. A lot of people put way too much stock in what Zillow has to say about their home values.
So I decided to look at some recent closings (like in the last week) in some of the key Chicago neighborhoods that we work in and to compare the price that homes sold for to what Zillow thought they were worth. The key idea is that the best indication of what a home is worth is what a buyer and a seller mutually agreed to sell it at. And I wanted to compare that to the current Zestimate in order to gauge Zillow’s accuracy before they have a chance to adjust their Zestimate based upon the latest sale. We ran the analysis for both condos and single family homes in neighborhoods like Lincoln Park, Near North Side, The Loop, etc… However, I excluded condo sales in buildings with over 40 units in them because those should be a lot easier to predict.
The results for the 36 closings we looked at with good data are in the table below but, to summarize, the errors ranged from 5552% (not a typo) to -52%. On average the error was 30.4% based upon the absolute values and when I throw out that one really big error.
You just can’t rely on a computer for something like this. A great example is my parent’s home in Dallas, which Zillow estimates to be worth $658K and we currently have on the market for $525K.
You’ll like this one–the property I own has two Zillow pages. One is the one created by the agent when we tried to sell a couple of years ago, and the other is the one I assume was created by Zillow using county assessment data. How far apart are the Zestimates? Only about $78,000.
Here are the links:
http://www.zillow.com/homedetails/4350-N-Broadway-St-APT-403-Chicago-IL-60613/87725354_zpid/
http://www.zillow.com/homedetails/4350-N-Broadway-St-APT-403-Chicago-IL-60613/2138695074_zpid/
So it actually has 2 different Zestimates. That’s weird.
Despite these being grossly wrong, people use them for or against whichever end of the deal they’re on. There’s a house across the street from me that’s listed as being 5400 sq ft when its really a 2700 sq ft ranch and they’re including a finished basement in the number to come up with a value of over $500k while neighboring 2 stories of 3200 sq ft are valued at around $425-475. And yet, the Trib and other news outlets are always quoting someone from Zillow on market values and trends. They’re really a poor source of information and nothing more than a self-serving dotcom trying to create buzz to sell advertising and partner redirects.
The only thing I use them for is aggregate price trend data. I’m assuming that they are wrong in a consistent manner so that the trend data is basically OK.
Your sample size is too small to use average error as an accurate assessment. The median error would more like 3.5%, however you have a large variance to account for in the numbers. If you throw out the large outliers which are most likely the result of foreclosures, then the average becomes more like 5% which isn’t bad.
However, I get your point. People have to exercise some logical sense of where the real estate market is at in their area when using Zillow. Zillow does not have a mechanism for adjusting based on number of foreclosures in the area because pricing accurately essentially go out the window in that situation.