7 things an automated or non-appraiser valuation won't tell you and a real life illustration of how inaccurate these values actually are and the consequences of relying on them:

Lenders and brokers using Automated Valuation Models (AVMs) and homeowners using "free online home values" to determine the value of a property need to know what those results aren't telling them.

  1. Whether the house is really there. A computer can't so much as drive by a house to see if it's actually located where it's supposed to be, has four walls and a roof, and really is a four bedroom split level and not a one bedroom shack.
  2. Whether unique features of a property might add to or detract from market value. So a computer returns an estimated value of $150,000. Did it account for the sewage treatment station next door? The railroad tracks nearby with trains that blow their whistles every night? The school district? The desirability of its tree-lined street versus the next street over? 
  3. Are they really estimating value or merely regurgitating data.  Many AVMs and free online services that pretend to estimated "value" rely on public assessment records and a few have direct access to local MLS services.  AVM's are really good at stating the obvious and making it look like they did something special.  One very popular AVM that I won't mention by name will give you a value that is one of three things:  The most recent assessment with an appreciation rate applied; The most recent list price with a SP:LP ratio applied; or the most recent sale price with an appreciation rate applied.  DUH!  Here is a real live example (one of hundreds that I could provide).  A property in Lexington, KY is assessed at $1,900,000 as of Jan 2014.  The AVM value is $1,300,000 as properties in this segment had been declining and at this time this particular AVM did not have access to local MLS data.  After gaining access to MLS data in 2015 "algorithms" were changed.  This property is listed for sale in Feb 2014 for $2,999,999. One month later the AVM value was $2,900,000 (up from $1,300,000 practically overnight).  It eventually sold for $2,400,000 in 2017.  The AVM after it sold was $2,428,000.  This is not the work of sophisticated algorithms.  This is a restatement of what transpired.  To hammer the point home, I listed my own home for sale by owner on the site.  The AVM went from $218,260 (the approximate tax assessment) to $258,924 just by listing it.  The value per the AVM  increased by $40,664 overnight simply because of the list price.  Now imagine lenders everywhere using these things as the basis of a loan.  That would be the beginning of an other market collapse that would make 2007 look like a hiccup.  This does not mean that AVMs are useless.  Appraiser developed AVM's can be quite good but that is because they require an appraiser's input verses a data entry person plugging in an address. 
  4. What makes the comparables comparable. Many of the AVM's that exist online were developed by computer nerds and not appraisers.  They are relatively good at predicting values of the typical residential property located in an homogenous tract development.  So is a 5th grader with basic reading and math skills.  Many of these programs select comparables by proximity.  The closest sales are designated the best comparables.  When there are no nearby sales then the programs select sales based on zip codes or other boundaries.  I recently appraised a property for $45,000 in a somewhat blighted urban area but in a zip code where the median home prices was well above $180,000.  According to the AVM that the lender provided they were expecting a value of around $100,000 which was apparently based on a prior sale at $87,000 and some sort of appreciation rate.  The property was listed for sale shortly thereafter and spent 158 days on the market and sold for $43,000 after going pending 3 different times.  An AVM can not see the property, does not know the market and can not differentiate between a "sale", a "market sale", and a "comparable sale".  A computer might compare your subject property to another property with similar square footage sold three months ago a quarter of a mile away. Even if that "comparable" property is in a different, less desirable school district, fronts a four-lane, 55 M.P.H. street, and is flood-prone. Or even if the property was sold under duress, such as in a divorce situation, or not at arm's length, such as to a family member. A computer simply does not know all the adjustments that might need to be made to a "comparable" property's sales price. 
  5. Whether a market is declining. Automated valuations use data from recent, nearby sales. If those sales were completed at the peak of a local housing market, the computer will think the trend is going up. Even if a professional appraiser knows that the overall neighborhood is beginning to experience a downturn. As a lender, don't get stuck with a property that's been overvalued by a computer.
  6. Whether there is a conflict of interest. Free online home values are often farmed out to real estate agents in your area, who use the service to get your listing when you decide to sell. The best way to do that is to impress you with their confidence that they can get a higher price for your property. If they tell you your property is "worth" the high end of what they believe they can sell it for, the theory goes, you're more likely to sign a listing agreement. With most things, it's best to "under promise and over deliver" — but the opposite is true when you use a free online home value service. 
  7. What qualifications, designations, experience and education the preparer of the value has. When you work with an appraiser, you can be confident we're highly qualified, ethical and prepared to complete your assignment professionally and with good judgment. Most of the time, you don't know the qualifications of whoever is behind those free online values, and they couldn't compare to an appraiser's if you did. And if you're relying on an automated valuation, you're cheating yourself out of an appraiser's education, experience and expertise.
  8.  

     

AVM USERS BEWARE:
The following is a sample of 20 consecutive sales in the Lexington Market that occurred on the same day and reported in the local MLS.  They were not cherry picked!  They were the first 20 reported closed sales for that day.  The first column is the actual sale price of the property and assumed to be equal to market value are least approximately equal to market value.  The second and third columns show the estimated values as indicated by two leading providers of AVM's on the internet (I wont mention their names). What is very interesting from these results is NOT that the majority failed to accurately predict value within a reasonable range of error, but that the resulting values from the two providers are so very different from each other despite both using "sophisticated" regression analysis programs to predict these values.  This is what many lenders are advocating as a better method of valuation.  Put yourself in the seller's shoes of Sale 1 and you used AVM2 to set your asking price.   You would have lost around $33,000.  If you were the seller of Sale 20 and you relied on say an average of AVM1 and AVM2 you would have lost $250,000.  The simple fact is AVM's don't work very well and will never work very well as the real estate market is an imperfect market.   

     

 

Sale # Sale Price AVM1 % Diff AVM2 % Diff
1 $152,900 none $119,573 -28%
2 $430,000 $387,539 -11% $348,565 -23%
3 $152,000 none $168,496 10%
4 $355,000 $501,317 29% $338,495 -5%
5 $114,000 $124,106 8% $121,317 6%
6 $111,000 $121,234 8% $125,800 12%
7 $337,500 $307,895 -10% $388,212 13%
8 $318,000 $284,608 -12% $251,158 -27%
9 $515,000 $477,360 -8% $460,618 -12%
10 $127,500 $116,175 -10% $123,126 -4%
11 $345,000 $256,206 -35% $316,389 -9%
12 $202,000 $203,606 1% $186,863 -8%
13 $250,900 $248,094 -1% $255,303 2%
14 $268,000 none $362,042 26%
15 $233,000 none $224,936 -4%
16 $144,000 none $137,214 -5%
17 $145,000 $124,396 -17% $60,080 -141%
18 $175,000 $133,026 -32% $155,923 -12%
19 $238,000 $240,946 1% $263,114 10%
20 $610,000 $311,861 -96% $404,317 -51%