Enter any search term in this field:

The Pliable Nature of Statistics: How Accurate Are Those Online Home Valuations?


“Facts are stubborn things, but statistics are pliable.”  Mark Twain used this phrase and others to emphasize the persuasive power of numbers and statistics.  While statistics can indeed be powerful and useful, they also can be misused and misleading.  Zillow, Redfin, Realtor.com, Homes.com, Trulia, Real Estate ABC and many others have developed statistical algorithms and now offer free home “valuations” that many, if not most, owners and buyers now are using to obtain preliminary estimates.

So, what is a real estate broker to do when confronted with the client/seller who uses these “facts” and declares that “Zillow values my property for way more than” the listing broker’s recommendation?  How does a buyer’s broker explain to his or her client that the value of a property may be worth more than the buyer is interested in offering, because the potential buyer went online and “Redfin says it’s worth 20% less than the list price?”  As more and more consumers look online for real property valuations before they actually speak with their brokers, it is imperative that brokers have at least a basic understanding of what these online home estimators are.  And are not.


Zillow, Redfin, Trulia, Homes.com, Realtor.com, Real Estate ABC, et al, provide automated valuation models (AVM’s) to attempt to ascertain a home’s value.  The AVM’s generally are proprietary mathematical models using public databases regarding both individual properties and community propensities.  While AVM’s have been in use for decades (e.g CoreLogic and Freddie Mac), their readily-available access to consumers is relatively recent.  Zillow, for example, has only been around since 2006.  Redfin has been around since December 2015.  Realtor.com was launched in the 1990’s, but its online use directly by buyers and sellers is a relatively recent phenomenon.

Rather than asking what these online home valuations are, it may be better to look at what they are not.  They are not appraisals; they are not CMA’s; they are not a substitute for the judgment of a qualified real estate broker.  Zillow takes great pains to explain that its “Zestimate” is a “proprietary formula” that is calculated from public and user-submitted data,” and is “a starting point in determining a home’s value” and “is not an appraisal.”  Zillow encourages buyers, sellers, and homeowners to “supplement Zillow’s information by doing other research,” such as getting a comparative market analysis (CMA), getting an appraisal, and “visiting the house (whenever possible).”  Redfin notes that the “Redfin Estimate is just a starting point – it is not an appraisal or a substitute for the expert pricing advice of your real estate agent.” Homes.com recognizes that a “Homes.com Home Value is not an appraisal; it’s an estimate of value based on certain information.  Use as a starting point, but we strongly recommend you contact a real estate professional for a more detailed evaluation.”  Real Estate ABC acknowledges that there “is no substitute for the deep local market expertise that the best realtors and brokers offer.”  The other online home estimators have similar warnings – perhaps for good reason, given the potential legal consequences of claiming the estimates are as good as or better than formal appraisals or even CMA’s.


Homeowners interested in selling their properties now follow these home estimates like the stock market.  But ask most real estate brokers, and you’ll find the general answer to be something along the lines of – “Not a chance in hell.  I run checks on properties and they’re not even close.”  (A recent check of a local property showed a valuation ranging from approximately $722,000 to $988,000 among Zillow, Redfin, Homes.com, and Realtor.com – a swing of approximately $260,000!)  Most sellers will claim that the online home estimates are way too low for their homes; most buyers will claim that the properties they’re looking at are valued too high.  Who’s correct?  It depends.

A standard computer industry axiom is “garbage in, garbage out” (GIGO) – the quality of the information coming out of any info processing system is only as good as the quality of information going into the system.  While the information “going into” the online home estimates generally is proprietary, it generally is accepted that the data primarily includes public information, including tangible, public features, such as lot size, square footage, property age, and number of bedrooms and bathrooms.  The data also may include community propensities, such as price trends.  What if these tangible items are inaccurate, or what if there are characteristics unique to the home?  While some of the online home estimators allow for an owner to input the correct or updated info, you cannot assume that this has been done.  Also, what if the property owner provides inaccurate data?  The controls regarding input are relatively open compared to the multiple listing services and, at least for now, may not be credible.  (Remember, GIGO!)

Also, these online estimates generally do not include the intangibles – buyer and seller motivations, property’s overall “appeal,” attitudes toward schools or community centers, foliage, condition, and what may be referred to as the “micro factors,” such as potential zoning issues, new construction, area traffic, school reputation, etc.  A 5% difference on a $400,000 home is $20,000, which is not an insignificant amount to most consumers and may be the difference between a closing and a failed sale.


Isn’t it a goal of most real estate brokers and their clients to obtain the best deal possible?  If so, isn’t it a goal to reduce the pricing errors as much as possible?  So how much error is acceptable?  All the online home estimators will readily acknowledge that there is room for error – perhaps significant error – in their computations.  Some of the big guys now are engaged in a battle to claim ownership of the “most accurate home-value feature.”  A 2017 “independent study, commissioned by Redfin” (a bit of an oxymoron, but I’m not here to judge), found that Redfin had the most accurate online home estimate feature.  Not surprisingly, Zillow claimed that the methodology was flawed.  (See GIGO.)  The study found that 64% of homes in the national sample sold within 3% (plus or minus) of the price predicted by Redfin using the final sales price against valuations at or about the closing date.  This compared with 29% for Zillow and 16% for Homes.com.  The median error rate was 2.06% for Redfin, 5.95% for Zillow, and 10.26% for Homes.com, according to the study.  The median error rate means that half of the home values in the area are closer than the error percentage.

Also, keep in mind that all real estate is local and subject to the nuances of each market.  In Deschutes County, Oregon, Zillow reports that 61.9% of the homes in its database sold within 5% of the Zestimate; 82.1% sold within 10%; and 93% sold within 20%.  So, for example, if the Zestimate on a particular date was $400,000, and the property sold that date for either $440,000 or $360,000, it would be within the 10% statistical category.  Some might believe that’s a rather large swing.  (This data is accurate through June 30, 2018.)

Another important consideration is the “date point” for the home estimates.  The foregoing statistics are based on a comparison between the final sales prices and the AVM’s from the same or proximate dates.  So, for example, if the closing date was August 1, 2018, then the AVM used for the statistical analysis would be on or about August 1, 2018.  Redfin is analyzed in the same manner, but also includes the list price in its analysis.  (Of course, that certainly may have a bearing on the estimate.)  To put it another way, the estimators appear to disregard the AVM’s that most buyers and sellers are reviewing throughout the sales process.

So, what about addressing the real, false expectations of buyers and sellers?  What about those sellers who meet with a broker and use as their ammunition the valuations provided by Zillow, Redfin, and the others before the property is even listed (a common occurrence)?  An unscientific study of Deschutes County properties suggests that a qualified real estate broker may be worth his or her weight in gold.  We looked at a sample of 100 homes in Deschutes County, and compared their final sales prices with the Zestimates and Redfin Estimates at the time of the original listings.  (2 of the homes did not have Redfin estimates.)  Only 29% of the Zestimates were within 5% of the sales price (high or low), 58% were within 10%, and 83% were within 20%.  59.18% of the Redfin Estimates were within 5% of the sales price; 87.75% were within 10%, and 98.91% were within 20%.  Using our $400,000 home as an example, a 10% difference (high or low) is an $80,000 swing; a 20% difference (high or low) is a potential $160,000 swing.  Most sellers would not be comfortable relying on an online home estimate if they were aware of the potential, extreme swings between what they are finding online and what the ultimate sales price may be.  (The foregoing is for discussion purposes and may not be used as a substitute for the reader’s own research.)


Online home estimates are not going away and, with further technological advances (such as using artificial intelligence), these estimates likely are going to become more and more accurate.  Most buyers and sellers will do extensive online research regarding their own properties and those they’re interested in, well in advance of meeting with a real estate broker.  In some instances, the online home estimates will be extremely accurate.  (Consider many established housing subdivisions.)

However, it should be abundantly clear that while the online home estimators may be a valuable tool, they are merely a starting point.  Even the online home estimators would acknowledge this.  There potentially is an extreme swing between the original online estimate and the final sales price, and real estate brokers should be able to provide informed explanations to their clients about the potential swings.  These AVM’s, like statistics in general, may be helpful, but they should not be considered as substitutes for the reasonable, well-reasoned judgments of qualified real estate brokers.


Disclaimer: this column does not constitute the giving of legal advice, and your reading this column does not create an attorney/client relationship. You are encouraged to consult a lawyer or accountant should you have questions about how this information may be applicable to your particular situation.
Previous Article Next Article