AVMs… The Not So Accurate Magic Bullet
Folks, in all the diatribes against appraisers by people who believe “we” have built-in systemic bias, the topic of AVMs has come up. Some of the diatribers apparently think that using an AVM will magically produce an accurate property valuation, while eliminating bias, and appraisers.
Well, not so fast.
A while ago I saw a ‘consumer facing’ article telling people how they could get the value on their home. That article mentioned three on-line services which claim to provide accurate property values. In alphabetical order, those services were REDFIN, ReMax, and Zillow.
Just for grins, I decided to access these sites to see how my home property would be valued. I was so thrilled to find out there is a nearly 10% spread in those values, equating to almost $57,000 between the low and high prognostications!
So much for accuracy.
I recall that the services actually showed which properties were used to derive my property value. The AVMs actually had built-in bias present because most of the properties selected for comparison were in no-way similar to my property in terms of location, view, site size or home design. Some even use listings! In effect, the value presentations were a load of bovine substance.
Then on 7/14/22, an article titled AVMs: Why One Size Doesn’t Fit All appeared in the MReport, which you should peruse for greater understanding about what AVMs are, how they are designed, and put to use.
But here are some excerpts from the article:
“…the various modelers are using different data and modeling techniques to test if their assumptions can produce more accurate valuations, a higher hit rate (in which the model can deliver a score on a higher percentage of properties), or increased confidence scores…”
“…more sophisticated [Lender-Grade] AVMs frequently layer in multiple sub-models, including regression analysis, appraisal emulation, data mining, current and historical market performance in the area, and the latest machine-learning modeling techniques.”
That last line is important because the people advocating AVM use instead of appraisers apparently believe that the AVM can be used to divine value of a subject’s property “here” by using properties from “over there” in order to increase the subject’s value. Talk about bias!!
Well, if an AVM is properly designed for lending purposes, it should be sophisticated enough to use properties near the subject, from all the data mining, appraisal emulation and other inputs it relies on. That’s no different than what competent and experienced appraisers do.
The AVMs used for my property (and others) on-line research are considered ‘consumer-grade’ and have far less sophistication and accuracy. They are basically marketing tools designed to draw in customers.
But the take away in all this is AVMs are not going away. And in fact, they are getting better at approximating property values close to what appraisers present.
Dave Towne on e-AppraisersDirectory.com
- New UAD Overhaul: What Appraisers Can Expect in 2025 & Beyond - September 19, 2024
- Cindy Chance Terminated - September 16, 2024
- Key Part of USPAP Not Available from TAF - July 19, 2024
Thanks for the article Mike. So, if they are soooo acurate why did Z shut down it’s home-flipping division with $540 million dollars in loses and axing 25% of it’s workforce??? And I don’t buy their BS reasonings.
Went to a property for a refi report about 8 months ago and asa I walked in the door the owner pushed her phone in my face showing that one of these sites stated her home value was $950,000 (bought it 14 months earlier at $725,000). I politely asked her to scroll up and when she did the site had “similar sales” with the top value at $675,000. At the same time I said to her when did that website have one of their employees out to her property? You should have seen the look on her face while trying to process everything. Then when she got it there was a big sheepish smile on her face then said to me good point! I estimated the property value at $835,000 with 5 comparables that bracketed that value.
So, yeah lenders good luck with that! But I think they could care less cause their banking (:P) on us taxpayers to bail them out again and again and again. Crazy!
Last appraisal I got from one client redfin was pushed in my face. They estimated 800k….the highest comp redfin used was $625k which I had in my report and was a net down in adjustments. Easy appraisal with all comps in the community. Can’t say for sure, but i’d guess because I didn’t use the borrowers cherry picked sales in superior areas.
I did the same experiment with my property and got the same results, I cringe every time I hear “I checked my property on Zillo”
AVMs, if you notice, also use non-arm’s length transactions. If they are going to use AVMs, the same appraiser (USPAP) search rules should apply. These sites use different perimeters, thus all have different conclusions.
“I was so thrilled to find out there is a nearly 10% spread in those values, equating to almost $57,000 between the low and high prognostications!”
Actually a 10% is not bad. I’d bet if you hired 3 appraisers they would have a similar 10% +/- spread. A 10% spread on a $550K to $650K home isn’t half bad.
I don’t like AVM’s anymore than most appraisers, but they can be a useful tool in some cases.
AVM’s are off from 5-20%. On a $100,000k home 80-120K valuation a $40,000 spread. On a $500,000 home 5-20% is $125,000 guesstimate. As appraisers everyone knows that if the borrower/owner has great credit; they take impeccable care of their homes (stop laughing). But lenders are willing to go with this UNTIL the market goes south THEN who are they going to blame?
Somehow us appraisers still…lol
Ive seen them off much more than that. Just appraised one that they had around 30-50% less.
AVMs accuracy is a joke. I just appraised a condo for $295k. There were 5 recent sales of the same model in the same project and none sold above $305k.
However, AVMs average estimated values were 14 to 28% higher than the appraised value and 10 to 24% higher than the highest sale price.
First American est. value is $341,791
Remine est. value is $336,335
Redfin est. value is $377,012
Zillow Zestimate is $341,500
The $40,677 difference doesn’t matter to the lending institution as long as a loan is made and no one ever defaults, loses a job, gets divorced, spouse dies, gets repossessed etc…….it’s on one example. Same thinking back in 2008 when millions upon millions walked away from their homes, and condos.
In Riverside county CA, the doc stamps are $220, not $110 per thousand. I have seen numerous AVMs which use “comps” with sales prices double the actual sales price. If they can’t even get that right, how can they accurately value the myriad of other influencing factors?
My parents got an AVM appraisal from their lender for a refi. The comp selection was awful. One was across a major road into a different market and was 40% smaller in GLA. It could of used recent sales in the community.
Dave you spend how much time on this article. I appreciate the article. Combat the lyers. Zillow, redfin lying yes ok. but many more lying so much worse. make aware the fact articles saying appraisers devalue neighborhoods based on race is again lead news. Despite fact article proven false. without facts. 2 months prior. yet again its paid to be front page news. i guess news no longer has to be accurate to be news as a story proven false months prior is again paid to be lead internet news. Dave focus on the lying.
Oh man, you’ll have to get away from mainstream corporate news and their partner corporate partnership social medial platforms to get that perspective. It’s out there, dare to look.
I’ve said this before and will say it again, but pick a percentage, 50, 60, 70, 80, or 90%. Even if an AVM 100% matches an appraiser 50, 60, 70, 80 and or 90%, which homes are the match, and witch home is off 10, 20, 30, 40, and or 50% off?
Seek the truth.
Coincidentally THE Wall Street Journal published an article today, about controlling values thru government process, they cited a six-acre parcel which could be built out as a 1500-unit site project and interpreted this as an unfair mistake for NEW YORK taxing purposes.
All Real Estate values are local OUR “Newcom in California” is trying to dictate city, county and state zoning and uses.
Land sites vary for all kinds of reasons, Maybe a flood plain, an avigation, or a view easement, Maybe surrounded by chicken, or dairy farms, some stink, some don’t.
Stink or not distant politicians are pled to locals.
Our local Appraiser-mortgage group should Address the REAL problems which can influenced by US, and our Politicians, or elect new ones.
Did they perhaps mean to control price, not understanding the difference between price and value?
If only we had elections, it’s the selection program now. Have you seen the news out of Brazil or stopped by Frank Speech lately? Thank you.
Politicians have to concern themselves with Value, they frequently use prices to illustrate.
Appraisers and economists use prices to illustrate trends and prices. Most politicians have good education, but training is done in house.
Every appraisers has a valuation concept based on historical and market activity data. AMC’S SUC. My opinionated valuation of AMC’S.
Confidence scores vary too greatly, home to home, in many locales. Observed one the other day that was 200k off, something about being anchored by a long term owner and no recent finance or sales activity.
Avm’s only work on the upward swing in a stable rate market, and only work partially at best. Consider how recently we’re dealing with purchasing power variations not seen in a decade if not longer. What monthly payment used to buy someone a 700k, now barely swings them a 400k, etc, etc. The twist is the data is deceptive, as new builders continue with projects apparently unphased. Builders play the long game and can push stock, pre orders, and seemingly discounted compared to resale in a rising market for like an entire year before they feel the pinch of more expensive lending terms. Many an agent are speculating the current projects, despite being full speed ahead and producing many new units, will be many a new construction builders last projects. Could be a great time to zip up discounted stock on the closing, if you can time it right, they’re running out of pre orders and rate lock shoppers.
As many a buyer whom was on lock and shop or some rate lock extension program continued to flow through the data sets as if they were borrowing at lower rates than current, because they are, often resulting in MLS data which is reflective of 4 months ago rate locks. (many rate locks are 60 days to shop, another 60 to close.) Then you’ve got personal motivations, stressed agents needing deals as they go under left and right, and a myriad of good to poor advisement on the real estate circuit channels.
Competent representation is more important than ever, the data disparity proves that as well. As dom’s draw out and price slashes continue. This is a great event for our real property markets, some slide back to more affordable pricing, thereby reducing the margin of difference in home step up categorical pricing, increasing buyer affordability to get into better homes nationwide. We may only be observing a blip in listing excess though, won’t be long now before maximally leveraged people are frozen immobile, clutching to the low rate. The 4 month window of rate lock delayed data came to an end last month. So now we really get to see the effects of a sustained high rate. You can’t Everone shopping know is locked substantially higher than quite a few recent sales indicate. So, you know, the difference between price and value. Do not adjust based on rate conditions, even if you could know, because there are more variables than ever. Responsible home owners flipping equity, they’re going to get good deals and better values. People leveraged to the hilt will have to hang on or bail out. We’ll know it’s really unstable when the note of short approval follows enough new listings on a steady basis.
Tell me the avm programmers are ‘coding this in’ with ‘current market conditions’. Not a chance, not a chance. AVM confidence scores are manipulatable too, fyi. The amc guys are not going to be of assistance for appraisers pondering these conundrums, they were hardly qualified to even use computers, much less give complex advisement on shifting markets. Running the valuation services industry like the stock market won’t help either, despite success rates, the guys at zero hedge are still just basically speculative advisement in all positions. There was a story someone bought on speculation 4 million the rate would dip again next june, that’s a risky proposition. My mortgage guy from the cu just revised his estimate of rate reduction from next spring to ‘4 to 7 years out, possibly more’.
Think it’s a coincidence the intended appraiser replacement program is happening right now? You’re the quarterback, you make the call. I’m simply too fascinated with all the returning voices on twitter to spend much time on it, lol. Beast mode engaged. It’s a madhouse. Still not going to make an account though, lol.
Forget neighborhood boundaries. Take comps from where ever you want. Disregard all land values. But at least no Bias!!!