Zillow’s "Zestimates" of home market value are often exaggerated and therefore may be no more accurate than homeowners’ own overblown estimates of value, charges a study in The Appraisal Journal’s Winter 2010 issue.
Zillow countered that the study used data that is "out of date and limited in scope."
The Appraisal Institute, a global association for real estate appraisers that publishes the quarterly journal, says any opinions and views in its articles are those of the authors and not necessarily those of the institute.
The study, "Zillow’s Estimates of Single-Family Housing Values," found that the property search and valuation site overvalued 40 percent of the 2,045 homes sold in Arlington, Texas, in the last half of 2006 by more than 10 percent.
"This was a very stable market in Arlington during the last six months of 2006. If Zillow thinks that prices increased by 10 percent over this period, they definitely need to review their source. (Prices in Arlington rose) only 1 percent per year for 2005 to 2006, and 2006 to 2007," said one of the study’s co-authors, Dr. Ronald Rutherford, professor of finance and real estate at the University of Texas at San Antonio.
The researchers said that Zillow’s valuation model did not seem to take into account occupancy and therefore overpriced vacant homes. Occupancy is not a factor in calculating a Zestimate, the company said.
Zestimates for the studied homes were a median 7.92 percent, or $9,717, and an average 11.66 percent, or $13,576, over the actual sale price. The average sale price for the sample was $138,064 and the median price was $124,000.
Zillow underpriced only 0.88 percent of the properties by 10 percent or more.
Citing a separate study, the researchers said homeowners overestimate the value of their homes by 5.1 percent compared to the sale price, while new owners overvalue their homes by about 8.4 percent.
Therefore, they said, "Zillow’s house value estimate and related information may be useful as a quick way to learn about the relative value of houses in a neighborhood, but it appears less precise than a homeowner’s own estimate of home value." …CONTINUED
The researchers chose to study Arlington in Tarrant County because Zillow gave the county a four-star "Best Zestimate" rating. At the time of the study, the authors indicate that Zillow said 99 percent of the homes in the county had Zestimates and that 60 percent of those were within 10 percent of the sale price, with a median error for all Zestimates in the county of 7.5 percent.
Overall, Zillow valued 59 percent of homes within 10 percent of the sales price, the study said. Nationally, as of its latest update in October, Zillow values 42 percent of homes within 10 percent of the sale price and its Zestimates have a 12.8 percent median of error. (Zillow’s data coverage and Zestimate Accuracy Table can be viewed here.)
The site’s current estimates for Tarrant County say 30 percent of the Zestimates in the county are within 5 percent of the sale price, 54 percent are within 10 percent, and 78 percent are within 20 percent, with a 9.1 percent median error.
In responding to the study, Zillow officials noted that every home details page on the company’s site report that the estimated valuations are not appraisals and should be used as a starting point to determine a home’s value. The company has clashed with critics in the past over the nature of the site’s automated value estimates (see story).
Zillow officials also faulted the researchers’ use of 3-year-old data.
"Regardless of the paper’s conclusion, it’s out of date and limited in scope," said Zillow spokesperson Jill Simmons. "What’s important to our users is our accuracy today, not our accuracy from three years ago. To that end, Zillow publishes updated accuracy for every county in the country several times a year, accessible on every page of the Web site, for consumers, professionals and researchers alike."
Zillow launched in February 2006. Simmons also pointed out that the authors compared data from two different time periods: sales-price data from a multiple listing service in January 2007, and Zillow’s estimates in the last week of January and the first half of February 2007.
"A reason to first collect the sales sample and then obtain the Zillow data is that this gives Zillow a chance to update after the sales had occurred so that we would be comparing their best estimate, which is why we waited a month to obtain the Zestimates," Rutherford said.
"It is preferable to examine this in a stable market. This year and last year have not been that stable, but we would be quite happy to re-estimate the models if Zillow will provide us with a database of their Zestimates with tax ID and property address during each year for 2006-09 including Dallas, Tarrant, Collin and Denton counties in North Texas," Rutherford added.
In response, Simmons said, "We welcome the opportunity for the researchers of this paper to speak directly with our data and analytics team about our Zestimate accuracy today."
What’s your opinion? Leave your comments below or send a letter to the editor.