Data Cancer found....

[Joyce Potts]

I just had a homeowner try to engage me a couple weeks ago to appraise his custom home/2ac property because he thought the appraiser who appraised it the week before had screwed up. I started doing the research and the comparables all pointed in the direction the original appraiser concluded to, so that's what I told the property owner (I basically talked myself out of that assignment).

I offered to consider anything he thought was relevant so he did the "search by price" thing in Zillow and came up with some sales in his target price range. Not any different than what a lot of brokers do. Obviously none of them panned out when I looked them up, which when I explained how/why that's when he agreed that there was no point in spending more money on another appraisal.

My point here is that the Zestimate itself is usually counterproductive but it's still pretty cool for a layperson to have direct access to at least some of the raw data points and to have the means to build their own searches. That wasn't possible back in the day.

In the last five years of my practice, I ran at least three (3) AVM's, usually Zillow, Realtor.com and Trulia, for every residential property I appraised and put it in my work file. If they were significantly different, I got out in front of it in the report itself. The bigger point being, that in ~85% or more cases, the AVM's were pretty damn close.

 

[George Hatch]

In the last five years of my practice, I ran at least three (3) AVM's, usually Zillow, Realtor.com and Trulia, for every residential property I appraised and put it in my work file. If they were significantly different, I got out in front of it in the report itself. The bigger point being, that in ~85% or more cases, the AVM's were pretty damn close.

I never thought of doing it that way but that looks like a good idea. Except for the part that if I'm doing an SFR at all it will usually be some orphan troglodyte thing that's only coming to me because the SFR appraisers don't want to hassle with it. Which in turn means it would also be pointless for me to run an AVM on those.

Still, that's a good tool to keep in our toolbox for future use.

However these other AVMs are performing, I'd be interested in seeing whether or not the GSE AVM that's being fed with the last 5 or 10 years of appraiser data is doing significantly better.

 

[Joyce Potts]

Nah, don't hassle her for having an interest. Agree or disagree, Joyce Potts has always argued her opinions in good faith. If it wasn't her making these arguments it would be someone else because there are plenty of appraisers who share these view

I never thought of doing it that way but that looks like a good idea. Except for the part that if I'm doing an SFR at all it will usually be some orphan troglodyte thing that's only coming to me because the SFR appraisers don't want to hassle with it. Which in turn means it would also be pointless for me to run an AVM on those.

Still, that's a good tool to keep in our toolbox for future use.

However these other AVMs are performing, I'd be interested in seeing whether or not the GSE AVM that's being fed with the last 5 or 10 years of appraiser data is doing significantly better.

AVM's are not for every residential appraisal. There are always going to be properties that are non-conforming in some way, but for the overwhelming majority of residential assignments (that are admittedly going away), the AVM's are faster, cheaper and are within a reasonable variance.

 

[George Hatch]

As with appraisals, they only actually need to be sufficient to purpose.

With the UAD thing, my thinking is that for the purposes of appraising that one subject property the quality/condition ratings and adjustments only need to be internally consistent. It doesn't matter in an individual appraisal if that appraiser always rates high or always rates low because the relationships between subject and comps remains consistent. That is, when the appraiser actually is being internally consistent and isn't trying to jam the square peg into the round hole for the purposes of expediency.

It's when the GSEs go to reusing those Q/C ratings in their AVM or their other analyses that they need the higher level of uniformity in these ratings from the appraisers. The additional uniformity doesn't add to the appraiser's ability to get to the reasonable value in that individual appraisal, but it does contribute to the utility of that data when the GSEs go on to use that data in their larger datasets.

 

[Dublin ohio]

According to the public records at the County this property also had prior sales
$350k in 09/2021
$240k in 06/2018
$179k in 12/2011

If I had access to the data and I wanted to see how this property has previously stacked up when compared to all the others nearby then each of these transactions can be used for such comparisons for those time frames. I wouldn't need to be engaging in any guesswork as to whether or not the location is of significant effect on the pricing.

These transactions all predate the 3yr limit in Zillow's reporting of the sales, but they'd be super easy to do for someone who has access to the local MLS.

FWIW here are 5 properties in the same sub that have comparable sales history per county records

$310K-12/24
$214K-3/18
198K-7/14

$359,900-10/24
$282K-9/21
$215K-12/19

310K-12/24
$282K-9/21
$215K-12/19

$397K-2/24
$350K-4/22
$202,500-5/16

$359,900-2/24
$213,940-6/18
$184,500-6/2016

 

[Alsie35]

In the last five years of my practice, I ran at least three (3) AVM's, usually Zillow, Realtor.com and Trulia, for every residential property I appraised and put it in my work file. If they were significantly different, I got out in front of it in the report itself. The bigger point being, that in ~85% or more cases, the AVM's were pretty damn close.

Smart business practice. They are also useful as an "idiot check" to alert you in case there may be something else going on that you haven't found yet. More information is ALWAYS useful.

 

[George Hatch]

FWIW here are 5 properties in the same sub that have comparable sales history per county records

$310K-12/24
$214K-3/18
198K-7/14

$359,900-10/24
$282K-9/21
$215K-12/19

310K-12/24
$282K-9/21
$215K-12/19

$397K-2/24
$350K-4/22
$202,500-5/16

$359,900-2/24
$213,940-6/18
$184,500-6/2016

Interesting. I completely missed the point that you would have access to the data. My bad.

 

[Alsie35]

As with appraisals, they only actually need to be sufficient to purpose.

With the UAD thing, my thinking is that for the purposes of appraising that one subject property the quality/condition ratings and adjustments only need to be internally consistent. It doesn't matter in an individual appraisal if that appraiser always rates high or always rates low because the relationships between subject and comps remains consistent. That is, when the appraiser actually is being internally consistent and isn't trying to jam the square peg into the round hole for the purposes of expediency.

It's when the GSEs go to reusing those Q/C ratings in their AVM or their other analyses that they need the higher level of uniformity in these ratings from the appraisers. The additional uniformity doesn't add to the appraiser's ability to get to the reasonable value in that individual appraisal, but it does contribute to the utility of that data when the GSEs go on to use that data in their larger datasets.

THIS. Is how I viewed all of my appraisals til about 2019 when Fannie informed me that the ratings are "Writ in Stone" across-the-board on all properties, on all appraisals. It was a removal of appraiser discretion/common sense and shoehorning of all pegs into round holes - for the statistical expediency purposes you discussed in your second paragraph.

 

[Dublin ohio]

Interesting. I completely missed the point that you would have access to the data. My bad.

Strictly from county records. They have a search by sale date and subdivision. Took me less than 10 minutes

 

[George Hatch]

Strictly from county records. They have a search by sale date and subdivision. Took me less than 10 minutes

I thought of doing that but I was too lazy. (Not to mention that it would have increased the amount of the invoice that I'm submitting to the GSEs for shilling for their anti-appraiser agenda.)

Besides, wouldn't you like to see the MLS listings? I would.

That 2nd one on the list tracks pretty well with the subject's 2021/2024 paired sales. $70k spread in 2021 and a $70k spread in 2024. As a percentage the $70k spread in 2024 is actually lower than the $70k in 2021.