 |
| automated valuation model, automated valuation method, robot appraisal, algorithm, math formula, avm, appraisal, mary cummins, los angeles, california, real estate appraisal, appraiser, zillow, redfn, trulia, realtor, realist, realavm, fha, corelogic |
Automated valuation models (AVMs) are statistically based computer programs that use real estate information such as comparable sales, property characteristics, and price trends to provide a current estimate of market value for a specific property.
Most people are familiar with Zillow and Zillow's Zestimate of home value. Zillow themselves have stated the Zestimates are not an appraisal. Below is their main disclaimer.
"The Zestimate® home valuation model is Zillow’s estimate of a home’s market value. A Zestimate incorporates public, MLS and user-submitted data into Zillow’s proprietary formula, also taking into account home facts, location and market trends. It is not an appraisal and can’t be used in place of an appraisal."
Real estate appraisers have always said "the 'a' in 'Zillow' is for 'accuracy.'" In light of Zillow's recent ibuyer program failure everyone now realizes they are not accurate. So what is the problem with AVMs?
The main issue with AVMs is the quality of the data. We all know Garbage In, Garbage Out or GIGO. Zillow and other AVMs such as Trulia, RedFn, RealAVM, Realtor, CoreLogic use a proprietary formula and public data to make their estimates. They also allow the public to edit their data and use false MLS data from real estate agents. They do not actually look at the home. They don't know the condition. Is it a full high quality remodel or ready to be demolished. They assume average condition. They don't know the amenities, upgrades, view (ocean, lake, freeway, back of an industrial building), additions, neighborhood boundaries (Bel Air proper or the flats of Los Angeles), exact location/neighborhood (on the ocean, a lake, in a gated community, built on an old landfill, next to the freeway and industrial properties)...and many other very important factors. They assume all properties in area have the same view and specific location. They don't even know if the home actually exists or burned down last year.
This of course makes their estimates vary widely and results in an inaccurate and biased valuation. Zillow's own research has shown that their accuracy is not as good in areas with little public data. Not all Tax Assessors, Building and Safety, MLS list data or more important recent accurate data online publicly. Some states don't report sales prices publicly. This goes back to GIGO.
I have found and Zillow has admitted publicly that their accuracy is lower in the extremes. Homes which fall on the lower and higher end of the price ranges for areas tend to have less accurate estimates. Zillow also admitted that they are less accurate in areas with older homes. There is no way for a computer program to tell if a home has been fully remodeled or needs to be demolished. This makes it less accurate in areas which are revitalizing. Some call revitalization "gentrification."
Inaccurate valuations are bad for everyone. If you're a homeowner and the bank offers to waive the appraisal in exchange for using an AVMs, this can cause problems. If you upgraded your property, you probably won't be getting added value for all your upgrades especially if most homes in the area aren't upgraded. If your home has one of the best views in the area or the largest lot, the AVM won't see that value either. You will get a lower valuation. The rate you pay for the loan is based on the loan to value LTV ratio. The higher the appraised value, the lower the LTV and the less risk for the bank. The less risk, the lower the cost of the loan, the lower the rate, the less likely you are to have to pay mortgage insurance. If you're looking to get some cash out, you would get less cash out of the deal with a lower valuation. Using the AVM just puts $350-$500 more money into the hands of the lender. It doesn't save YOU any money no matter what they say.
Now if you have the worst house in the neighborhood in bad condition, an AVM would be to your benefit. They will most likely over value it. If you are buying said home, you might incorrectly over pay for the home assuming it's worth more. This is why investors should never rely on AVMs. An appraisal by a licensed appraiser is like $350 insurance. You will get an unbiased independent full valuation of the property. $350 is a lot better than paying $100,000 more than what it's really worth.
I'll quickly go into how appraisers valuate homes so you can understand the basic AVM algorithm a little. This is a basic simplified search. We generally search for sales and listings as similar to the subject as possible. We use the computer to search for sales, listings within a 1/2 mile radius from the subject that have sold within the last 90 days which are +/- 15% difference in gross living area. We then choose the best comps based on location, size, bed/bath count, view, condition, amenities... Sometimes there are no recent sold comps so we have to go back a year or so and adjust for appreciation/depreciation over time.
This is where having someone who has inspected and actually viewed the subject property is so important. This is also why it's so important to have an appraiser with years of experience who knows the area very well. A math formula can't see the condition, view, real bed/bath count, upgrades, amenities, specific location in a neighborhood... They generally only see tax roll and MLS size, bed/bath count, number of garages, pool. Those are very rarely accurate. Tax roll is generally the original size. Real estate agents lie in MLS ads about everything. You won't know the real full bedroom and full bathroom count of the subject or comps without an appraiser.
The appraiser also views the sold and listed comparables. Are they tear downs selling for land value or totally remodeled with new additions that don't yet show up on the tax roll? Does the subject have an ocean view but sold comparables face a loud ugly freeway? Is it a full bedroom with it's own door or a walk through bedroom or just a den? Is it a half bath, full bath or just a toilet in the basement? These are very important factors which can make the value differ by up to 100%. The AVM will never know all of those things about a property. This is why they are not accurate.
Recently I saw a property with AVMs that varied from $750K to $1.9M. Zillow, Trulia, RedFn came in way too high at about the same price. Realtor, CoreLogic RealAVM came in at market based on my own valuation after looking at the property. I realized the problem with the values when I saw the comps that Zillow, Trulia, RedFn showed for the subject. They were using comps from 1.5 miles away in a neighborhood that sells for twice as much as subject's neighborhood. That neighborhood sells for so much more because the homes are very high quality, they have the best school system in the state, lots of local shopping and a low crime rate. This was a Beverly Hills versus Watts comparison. Clearly Zillow, Trulia, RedFn algorithms are wrong. They cannot define a comparable neighborhood.
The issue had to do with the size of the subject and recent sales. It was larger than most homes that had sold recently. The homes directly around subject were all the same size built at the same time but they hadn't sold recently. Instead of going back in time to find an older sale of a similar size and time adjusting for appreciation, Zillow kept widening the search until it found comps of a similar size that had sold recently. That was a huge mistake. An appraiser would never base their value on homes from a totally different neighborhood which generally sells for twice as much. Remember, the three main indicators of value are location, location, location. This is why AVMs are so inaccurate and should not be used for valuations.
AVMs are just an algorithm based on a math formula. Some of the math formulas like Zillow are deeply flawed. They don't know neighborhood boundaries or the true characteristics of the properties. A math formula is only as good as the data used in the calculations. Because the AVM is not a live experienced licensed appraiser who has actually inspected the property and neighborhood the resulting value will never be accurate.
Please, do not rely on AVM valuations for real property values. If you are refinancing your home and your home is upgraded or has superior features than most homes in your area, you will be better off requesting a full appraisal. The lender will be paying for the appraisal so it costs you nothing. The lender just wants to save a few dollars for themselves by going with the free AVM. If your home is almost a tear down, use the AVM ;-)
References
List of different AVMs
Freddie Mac: Home Value Explorer® (HVE®)
Zillow: Zestimate
Realtor.com: Collateral Analytics, CoreLogic Total Home Value for Marketing, Quantarium
Redfn: Redfin Estimate
Trulia: Trulia Estimates
RealAVM™ is a CoreLogic® product
Fannie Mae: AVM 1 (data assessment/
integrity checks, comparable selection, comparable adjustment,
and reconciliation), AVM 2
CoreLogic actually has a few patents related to their AVMs as does Zillow. Here is the patent for their main AVM model. The description is very interesting. It deals with the issues of geographical neighborhoods and number of sales during set time periods, i.e. spatial and temporal distinctions. This is why they're more accurate than Zillow and had the correct AVM for the Marin, California property noted above. They have a better spatiotemporal understanding of the nature of real estate. They even stated that a "property value given by an appraiser can vary, sometimes erratically, depending on the comparable properties chosen in performing the appraisal." They stated sometimes there are no sales during the short range of time used by appraisers. This is why it's better to go back in time and time adjust if there are no sales rather than widen the geographical search area. The location factor carries a lot more value weight.
https://patents.justia.com/patent/20130144798
Here's one Zillow patent which goes into their algorithm, math formula. They're clearly mainly relying on median and average calculations.
https://patents.justia.com/patent/8676680
Great article on the issues and problems with AVMs. They're not standardized especially their reported confidence scores, Forecast Standard Deviation FSD and error rates.
https://vegaeconomics.com/webfiles/Principles%20for%20Calculating%20AVM%20Performance%20Metrics.pdf
Mary Cummins of Cummins Real Estate is a certified residential licensed appraiser in Los Angeles, California. Mary Cummins is licensed by the California Bureau of Real Estate appraisers and has over 35 years of experience.
- Mary Cummins LinkedIn
- Mary Cummins Meet up
- Cummins Real Estate on Facebook
- Mary Cummins Real Estate blog
- Cummins Real Estate on Google maps
- Mary Cummins of Animal Advocates
- Mary Cummins biography resume short
- Mary Cummins Real Estate Services
- Animal Advocates fan page at Facebook.com
- Mary Cummins
- Mary Cummins Animal Advocates on Flickr photos
- Mary Cummins Animal Advocates on Twitter.com
- Mary Cummins on MySpace.com
- Mary Cummins on YouTube.com videos
- Mary Cummins of Animal Advocates on Classmates
- Mary Cummins on VK
