Zillow, based in Seattle, Washington, has made millions of dollars since its launch many years ago. The idea of allowing anyone to see the value of their home for free was a novel idea with extraordinary appeal to home owners. Zillow seeks to utilize public records combined with math formulas to calculate a home value.
So, how accurate is this “Zestimate®” value? At Country Club Lifestyle Realty we have analyzed 91 homes sold throughout Houston to show that Zillow will at times be consistent with the actual sales price of a home, but the overall results are not a ringing endorsement. In a few cases Zillow missed the sales value of a home by 25-30%. In one specific case the value was off by $121,000. That is a large chunk of change.
The study involved home sales ranging in value from $130,000 to $530,000 over a 30 day period beginning in December 2011. The majority of sales in the Houston and surrounding suburbs fall within this range. Higher end homes were specifically excluded from the study since luxury homes may tend to skew the data. The intent of the study is to determine an accuracy level for the vast majority of home owners. While the study includes the Houston and surrounding suburbs the author believes the values are consistent for the majority of the country. The results are similar to a study performed by The Wall Street Journal in 2007.
The actual data utilized in the study is presented below. Figure 1 provides the raw data. Figure 2 shows a linear comparison of the same data with 5% error bands. Figure 3 shows the normal distribution using a bell curve.
- Percent Error of Zillow’s Zestimate Against Actual Home Sales
- 5% Error Bands for Zillow Zestimate When Compared Against Actual Home Sales
- The Normal Distribution of Zillow’s Zestimate in Percent Error Against Actual Home Sales
Using a statistical analysis the results are the following:
- Zillow’s Average Percent Error: 6.14%
- Study Accuracy: +/- 0.86%
- Median: 6.70%
- Standard Deviation: 8.45%
As evident by first glance (plus the large standard deviation) the only clear trend is that Zillow generally underestimates the value of the home. For this study homes below $160,000 tended to not exhibit large swings in percent error when compared to their actual sales price. Beyond this value a specific home may be undervalued by 25% or overvalued by 15%. Overall 77% of the homes were undervalued. This is good news for buyers when they convince sellers that Zillow is an accurate representation of their home’s value. In other scenarios they may lose out on the house of their dreams by insulting the home owner with a low ball offer.
For a given home owner there is only a 1 in 5 chance their Zestimate is within 2% of the actual value and a 1 in 3 chance the value is within 4%. The same 1 in 3 chance also exists for a Zestimate above 10%. Overall these results are not impressive.
As reported by The Wall Street Journal, through an interview with Stan Humphries, Vice President of Data and Analytics, Zillow’s estimates utilize sales prices, past sales history, tax assessment data, and general home size. In addition they seek to obtain a value through similar comparable homes. Based on this information, why then is Zestimate wrong and why so much random variation when compared to actual sales?
The approach as stated by Mr. Humphries appears sound; however, after reviewing this statement the error becomes apparent. First, tax assessments can vary significantly from home to home. Some homes may be homestead while others may not have been re-assessed for several years. Second, past sales history (as most real estate agencies and banks will attest) provide no indication of current or future performance. The housing economy can change quickly as evidenced in 2008. Third, the general home size leaves out the amenities associated with homes and communities. For example, gated communities generally increase the home value. Upgraded kitchens and bathrooms on a 1970’s home can impact the overall value. The overall condition of the building and lawn clearly influences the sale value. Even the energy rating of a home can have a significant impact.
Clearly Zestimate is not a reliable number but sadly many home owners lean on Zillow to determine the value of their home. A quick run through blogs posted by home owners in Zillow’s Real Estate Advice section illustrates this fact. Many voice their frustration to Zillow when their Zestimate changes due to tweaks in the algorithm and the value declines.
So how should a home owner obtain a better representation of their home value? The most accurate method is to contact a local Realtor. However no longer are in home visits or long extended phone conversations required. Many Realtors (including CCLR) now offer a free market analysis through e-mail. Home owners can sign-up for monthly email campaigns that provide current home sales in their area. For anyone around Houston and the surrounding suburbs, CCLR utilizes a program called Market Snapshot that provides the latest home sales and listing details tailored for a specific home. Click here to enter your home’s address and information for your free Market Snapshot.
Very informative and a lot of effort put forth! I posted a blog about the accuracy of the Zestimate with a similar conclusion, and tell folks how accurate Zillow claims to be in their area.
http://www.trulia.com/blog/jimsweat/2011/11/is_it_a_zestimate_zappraisal_or_zaccurate
Our market has had major price moves in the last few years, which seems to make it even less accurate!
Thanks for sharing!
Respectfully,
Jim Sweat, ABR, CRS, CDPE, GRI, e-PRO, ILHM
REALTOR
Sandals Realty
Mobile: 941-306-7384
Hi CC Lifestyle, Brad from Zillow here, Zillow does indeed use public record information along with a multitude of other sources and resources combined with some pretty complex Algorithms to determine the starting point “zestimate.”
As you’re probably aware, Texas is a non disclosure State making it difficult to gather data and information by not only Zillow, but by the consumer as well. That’s just one good reason they’re coming to our site. Since there is no information publicly available for them to start their data and information search, Zillow is an awesome place to begin their journey, gather what they can that we’ve been able to provide, then they can contact a local real estate expert or professional appraiser to complete and obtain the most accurate as possible “fair market value” price of the PIQ. (Property in question)
Zillow is a great place to get the Zestimate, and a local agent who is willing, able, and wants to work with our visitors and potential buyers is the place to get their Zactimates. While our information may not be 100% dead-on accurate, we are at least trying to provide as much information as we can (or what we have access to) to help buyers and sellers make the best choices possible.
Let me ask you this, why is Texas a non-disclosure state? If we could provide consumers with the data that is being withheld, (non-disclosed) wouldn’t you think and agree that our zestimates in your area could be even more accurate?
We are very transparent about our accuracy grading as seen in these charts, “http://www.zillow.com/howto/DataCoverageZestimateAccuracy.htm” where can your readers and subscribers find accuracy ratings on non disclosure sites? You’re welcome to learn more about zestimates and Zillow at Zillow Academy dot com if you’re interested.
We’d love to have you as our guest.
Hello Brad! Thanks for adding to the article and I will respond soon.
CC Lifestyle, thanks for sharing an analysis using professional common sense and, it appears, subjecting the numbers to good metrics. We have similar challenges w/ Zestimates in northeast Indiana. My personal analysis – “snapshots” of my own listings compared to Zestimates – over the last few years, has revealed a 7% error at best and 28% at worse. While it does not stand up to the rigor of thorough statistical analysis w/ standard deviations and per cent accuracy, Indiana does collect sales data at closing and I’m fairly confident that all of it is made available to the data aggregators. If that is the case, per Brad Andersohn from Zillow, shouldn’t we expect Zestimates to have a MUCH better accuracy than a 7% – 28% error? In addition, that has the potential for being a double swing, meaning 7% above or below a REALTORS market analysis would result in a potential 14% variance.
Brad, let me try and answer your questions and provide my take on Zillow.
Let me openly state that I appreciate Zillow as it does have a place in the housing market. It provides information and a forum for Agents, Brokers, etc. to communicate with the public similar to many other sites. It also provides a large amount of data for consumers to help them better understand how their home compares to others.
My article is not attacking Zillow specifically but instead providing publicly additional information (similar to The Wall Street Journal study of 2007) regarding accuracy of Zestimate. Realtors witness daily the challenges in discussing value with home owners that utilize Zillow. This is a continuous battle as consumers are placing a larger amount trust on information they find on the internet particularly with large companies such as Zillow. This behavior clearly indicates that Zillow is not putting forth a decent effort in identifying the inaccuracy of Zestimate. While Zillow does have a page with limited information devoted solely to this topic, it by no means is readily apparent to the average user. I had to search actively to find the information myself. Also, Zillow seems to lack a detailed breakdown of data as it only touches on a few large markets with rural areas themselves totally ignored.
The study we performed is meant to help anyone seeking to understand the accuracy of Zestimate. It hopefully illustrates the value of agents (especially in non-disclosure states) by showing how corporations seeking to calculate a home value is limited in their ability to do so accurately.
As for Texas and the non-disclosure policy; in December 2008 when the House Select Committee considered property tax relief and appraisal reform the concern of non-disclosure was raised to the committee. Appraising property becomes far more difficult when sales price is not publicly available. However, privacy and confidentiality concerns were very evident by citizens. Many business owners do not wish for the public to understand the amount paid for their property. Cost of operation is directly tied to overhead and this information should remain private. Another specific issue was that appraisers would utilize the highest sale price of a comparable to determine the assessed value of a home.
I understand that Zillow desires to have the latest sales in order to calculate a Zestimate value. However, based on The Wall Street Journal report, Zillow’s own published results, and the public opinions of Realtors, the non-disclosure issue alone is not what is causing the large swings in value of Zestimate. The real issue lies in Zillow’s inability to understand markets down to the community level and at times street level. Additional modifications to the algorithm might fix the issue, but honestly most Realtor’s believe that Zillow should take a more proactive stance in supporting and promoting their expertise as the true authority of home value.
[...] by on January 13th, 2012 Country club Lifestyle Realty(CCLR) from Sugarland has published a report about their recent analysis/survey on Zillow’s Zestimate for Houston area homes. The results [...]
“The real issue lies in Zillow’s inability to understand markets down to the community level and at times street level. Additional modifications to the algorithm might fix the issue, but honestly most Realtor’s believe that Zillow should take a more proactive stance in supporting and promoting their expertise as the true authority of home value.”
I completely agree.
Let me start by saying that I think Zillow is an awesome website for other purposes and I do plan to participate in their exclusive agent offers in the future. However, I would like to see the chore of explaining “Zestimates” to potential buyers and sellers eased.
I am no expert at “Zestimate” calculations, but here are just a few of the issues I have to explain when buyers and sellers place a lot of confidence in “Zestimates”.
First of all, I don’t think being a disclosure state makes a big difference concerning the accuracy of “Zestimates”. SC happens to be a disclosure state.
In my market area of Myrtle Beach, SC, values can vary greatly not only at the street level, but in the type of property being used for comparison. In residential areas which also have oceanfront Condotel buildings, the algorithm mixes the values of single-family homes with vacation condos!
Zestimates are inaccurate for most of the residential properties located east of the ICW in Myrtle Beach for many reasons. Its calculations can’t take into account the local factors which significantly affect variations in value. In our coastal area, Zestimate values are frequently overestimated and underestimated by a large margin due to local factors.
Using tax valuations in any way to determine real value is a problem for all the obvious reasons. Calculations for property taxes vary greatly not only from state to state, but even within the counties in the states.
Assessed values for real property taxes are all over the board due to our point-of-sale tax valuation laws. Each residential property valuation is reassessed at each Assessable Transfer of Interest in the property (basically, title transfers). South Carolina State Law requires that each County within the State conduct a countywide reassessment every five years. In the year in which a reassessment is implemented, each parcel of real property must be adjusted for improvements and losses, and cannot increase more than fifteen percent every five years unless an assessable transfer of interest occurs. On a reassessment year, such as 2010, property values are reassessed using the market value of the real property as of December 31, 2008. And that’s just for starters….
In our distressed market, houses right next to each other can have significantly different tax assessed values depending on the last time there was an Assessable Transfer of Interest! The countywide reassessment every 5 years is imposed to provide equity among taxpayers. Still, that doesn’t change the fact that neighboring properties can have inequitable tax assessed values in between reassessments especially when market values are changing as quickly as ours have been over the past 5 years.
My point is that with so many variables involved in calculating tax values in any state, why use them to calculate current real market value for consumers? I understand how property taxes are calculated in my state, yet I still have to call our County tax assessor’s office for an estimate on property taxes when a buyer is considering a property! Using the amount the current owner just paid can be off by a few thousand dollars or more depending on the final sale price, the last time the property was reassessed and whether the property will be owner-occupied.
In NC, Counties are required to reassess at least once every 8 years. Its left up to each County to adopt a policy to complete a revaluation more frequently if they choose. If wide time gaps exist between reassessments, you could be looking at tax assessed values which are not even close to current market value. I recently sold a home in NC with a tax assessed market value of $370,000 (based on market value December 31, 2007). The sale price, supported by recent comps, was $160,375, December 2011. The Zestimate was and still is overestimated in value.
“Zestimate” calculations are simply not reliable. It’s difficult to explain real comps to buyers and sellers when they place so much confidence on the accuracy of a “Zestimate”.
Even a house next door may not be a “usable” comp in an appraisal if a buyer is using a federally-underwritten mortgage to purchase (which most financed buyers use). Guidelines for federally-underwritten / secondary market type mortgages state that a comparable should be within 25% of the subject property’s GLA (Gross Living Area) in size, and should not require more than a 10% single adjustment or exceed 15% net or 25% total adjustments for characteristics that differ from the property being appraised. In a nutshell, just because two houses have 4 bedrooms, 3 baths and a garage in the same neighborhood doesn’t necessarily mean they are comparable for appraisals.
The author is right, “Zillow should take a more proactive stance in supporting and promoting expertise as the true authority of home value”, not a computer-generated estimate.
I think zillow is way off on its valuations here in the Culpeper, VA area; which is a market 1 hour outside of Washington, DC. I have done comparisons on my listings and zillow is always low. I have elected to remove my listings from zillow databases. It is a headache, but saves more headaches when trying to sell a property.
[...] The results are similar to a study performed by The Wall Street Journal…. for complete article: http://homes-in-sugar-land.com/blog/2012/01/how-accurate-is-zillow-new-study-of-actual-home-sales-sa… This entry was posted in Financial News and tagged market, mls, real estate, sellers, zillow. [...]
[...] Study on Zillow [...]
@Steve Da’Mes – You do have the choice to market and advertise your sellers listings in whatever locations, websites, magazines, billboards etc that you choose, but remember that the seller knows where potential buyers are looking and searching too. Disclosing to your clients that you’re removing listings (their homes) from a network and marketplace that has over 22 million visitors each month on average might not be the best choice in their eyes for marketing their property.
Exposing your clients listings to potential buyers on Zillow has nothing to do with the Zestimate. The importance of strategic distribution is something every agent and broker should consider and take seriously. This article explains my point in much greater detail. Your clients that list with you may just want their homes available on Zillow, read this:
http://activerain.com/blogsview/2697254/the-importance-of-strategic-distribution
Zillow is highly inaccurate in our area of Fort Bend County. It has my home (and that of most of my neighbors) at least 20% below market value. A few short years ago, they had it 20-30% above market value. This isn’t CA, this is TX, we don’t have price swings back and forth like that. Things stay pretty steady.