I remember my first real estate broker was also an appraiser. God rest his soul, he was a great guy. I remember him pulling up Zillow and looking at their automated value model (AVM) estimate (Zestimate). He commented that one day this technology may do all the appraiser’s work and/or put them out of business.
Fast forward to today and these AVMs are available on just about every real estate site you visit. More often I am finding sellers and buyers believing these are 100% indications of value, which, in turn have made some negotiations shift from positive to negative. Fortunately, lenders are involved in most of the deals and require appraisals which, still, answers the question of value better than an AVM.
Appraisals aren’t rocket science but they do take a lot of time. They find and match comparables to the subject property’s condition, location, size and style and then make adjustments for any variations. Appraisers also go through years of training to get their license and last I heard there was a shortage of them.
As an investor, buyer, seller or lender, is it okay to use an AVM or should we stick with appraisals?
AVMs have to know the above while also being intelligent enough to decipher between a home with an extra 500 feet in the garage or main living area. They also need to know the ever-changing adjustments that are used “locally”—which local appraisers know. Sometimes total bathrooms have a high correlation to value and other times it has very little correlation to value. This is where the AVM starts to have trouble, there are so many distinguishable features that need to be quantified.
While these AVM estimates are getting better, I don’t believe they are up to par for using them for business.
What exactly is an AVM?
The M, in the AVM, stands for model which is also called an algorithm. These models use an internal algorithm which runs thousands of statistical calculations in a regression model on factors in the comparable properties’ physical data. No emotional buyer or seller feelings are factored into this model, which do influence home purchases. That one algorithm may have a confidence score of 50% in one area and 95% in another. The lower the confidence score the lower the chance of that value being correct. Some are better than others but they all vary in confidence or error rate (or standard deviation for those math whizzes) depending on what locale they are used in. All of the AVMs have small print that declare it’s just an estimate but sadly, many investors believe these values to be true and have lost money. The larger that error rate is, the larger the value’s range will be and the less confident we truly are in the value.
I don’t make decisions based on large ranges. Sure, just list your property high and let the market sort it out, but did you buy it high? How long will it be listed high? Days on market matter. I’ve even heard some state it’s okay to add or subtract 10% of the AVM value in some areas and that will be closer to the true value. Don’t do this.
I do understand the usefulness in looking at AVMs from time to time, especially if I believe in their comparables and feel confident in their adjustment algorithms. Still, betting one of your life’s biggest investments on these models concerns me. Homes are not commodities. Even a certain model home, built by the same builder, can generate a different value if it’s in a different part of town (and not all models can see these changes) or there is some sort of deformity it can’t see electronically. AVMs do not factor the interior aspects of the house or again, the emotional value buyers may feel based on the way you maintained or present your property. One home that the AVM lists as a comparable to your investment property may have had holes in the walls and roof when it sold, consequently, throwing your “estimate” way off.
I also double as a private lender and one of my borrower’s properties has a AVM that the borrower loves from Redfin. Though, Trulia and Zillow give a different estimate that range 10%-20% less than Redfin’s estimate. Which estimate do I use as a lender? Neither. I use a local appraiser whom I have met and who knows my lending area well (and that’s even after I’ve done my own BPO as a Realtor).
AVMs are not wrong all the time. I think they can be a great way to start a dialog on price and they will only get better with time. Improvements on the models considering other factors besides physical data will make a huge difference.
Here are three reasons why your AVM may be wrong:
- There is simply not enough comparables for the algorithm to generate a quality estimate.
- You have a home that has no true comparable near it. An 8 bedroom 4 bath, built in 1920, may be a bit hard to use the comparable method in my market. In cases like that an appraiser is extremely useful because they can use other valuation methods (besides the sales comparison approach) that may be a better indicator of value.
- The MLS or public records have incorrect information. AVM values can really be thrown off if square footage or entire rooms are missing from comparable data. My brokerage listed over 100 listings in the first quarter of 2017 and no matter how hard we try, the MLS finds something minor that we need to correct, like not adjusting the date of a pending sale etc. As a member of a few MLSs I can say that some MLS systems are better at this than others, but one of the goals of the MLS is to have quality data which does help appraisers (and AVMs) be more confident in their values.
How do we remedy these issues?
Hire a licensed (local) appraiser or quality Realtor who will help you find one. Finding a home’s value is more of an art form than science. Quality appraisers and Realtors know their area and frequently know the homes that sold with problems. They have subscriptions to the local MLS and they read the comparable sales data of the sold properties. They can see if it was an arm’s length transaction or, for instance, if the selling Realtor reported that there was a giant sinkhole in the neighbor’s yard which decreased the comparables value (and should adjust yours higher).