Are Zestimates as Inaccurate as We're Being Told They Are?

[Michael Goldsmith]

One of the first things I learned when I started studying REI, was that the estimates on realty sites like Zillow, etc. were wildly inaccurate. But are they, really? And even so, does it really matter?

Let's go over some of the basic principles.  

How are zestimates formed?  They take an average of previously sold properties in the area with roughly the same specs.  It doesn't take condition etc. into account.  Yes, I know.  It also doesn't exclude outliers.  The example of ouliers that I've heard a lot, is that if someone dumps a property super low for some extraneous circumstance, it will affect the estimated price on nearby properties.  But what would that actually do to the estimated price of a nearby property?  It would drag the amalgamated zestimate DOWN, so that a property would seem like LESS OF A DEAL than it actually was, right?  Therefore, wouldn't that mean that at first glance there should be hidden gems of deals that we're collectively writing off?

But does this happen in the other direction?  What would make a deal appear better than it actually was?  There would have to be an outlier in the opposite direction- someone would need to sell a property for WAY more than it's value, in order to drag an estimate up enough to create a false positive for a deal.  I find that scenario far less likely to occur.  Likewise, consider the following three scenarios:

1. The asking price of a property is less than the estimated price.  This is a "good deal."
2. The asking price  of a property is equal to the estimated price.  This isn't a "good deal."
3. The asking price of a property is higher than the estimated price.  This isn't a "good deal."

Only one of these three scenarios is ideal for our intents and purposes.  

Let's look at a theoretical property with an asking price of $110K.  The actual value of this house is $100k as well.  We don't know that yet for certain, but it makes the math easier.  We have 10 comps of equal actual value that have been sold in the last 6 months that is creating our estimate.  9 of those comps went for $100K, right on the money with the value, and we have one outlier that sold for $110K, a 10% increase in value.  Someone got very lucky on that deal.  That translates to an average of $100,100- a 1% change in the estimate.  

You'd have to have a lot of outliers (either positive or negative), or they'd have to be incredibly significant price increases to affect an average estimated value for a property.  Again, I find this far more likely to occur negatively where someone sells off a property really cheap, than one selling it really expensively.  

So, what would make this $110k property look like a good deal?  In actuality, it's a bad deal, right?  We'd have to have the numbers so skewed, I think it's virtually impossible.  You'd have to skew the numbers from the negative point, past the break even point, to a significantly positive estimated valuation that it would make it seem like a good deal when it wasn't.  

Perhaps I'm missing something here, but I find it hard to believe that an estimate could create such a skew that would create a false positive.  I could understand it skewing to make a deal look less attractive, but to make a bad deal look like a good deal?  I find that a little difficult to wrap my head around.  

I would like to point out that my conjecture is by no means in lieu of drilling down into the data and actually comparing comps by hand.  I still feel that doing one's due diligence is important.  However, I feel that the negative view that these estimates have on realty sites is a bit unwarranted.

Thoughts?  What am I missing?