Why can't we band together on Feb 4th?

[J Grant]

I'm pretty sure that is what they want. Appraisers signing off on avms.

They want all the appraisers using these apps. And then they will add more features down the road like they will pick the comps for you. The apps will pick the comps for you, adjust the comps for you, put the graphs in for you, put the maps in for you, and then all that is left for you to do is sign it.

If "they" pick the comps for us, whoever "they" are then it is no longer our appraisal and we won;t be needed - an app choosing the comps already exists and it is called an AVM.

 

[Tim Hicks (Texas)]

They want to prove that their lower quality AVMs are just as good as an appraisal by making appraisals more like AVMs?

 

[TerryRohrer]

For now, they want to show support for our adjustments and in the market condition section, when we make a time/market condition adjustment want to see an exhibit of it in a chart or graph form. I am providing the chart or graph even in a stable market since not making a time adjustment is a judgment call as well and should be supported then,

I have a conversation within my mind based on years of prior appraisals, market knowledge etc.

All that and you don't have a clue what the requirement is that you are supposedly responding to.

 

[Tim Hicks (Texas)]

The GSE's want all homes to be same size, quality and condition so their AVMs can be a cheaper failsafe. Unfortunately, that is not how the real estate market works. But they so badly want it to be so. They want their appraisal waivers based on AVM values to be the gold standard. They are willing to accept houses with few updates or upgrades to be overvalued based on their data and don't like it when borrowers protest their AVM values because their home is worth more their AVM value.

The sad part is that they should not be looking at alternative valuations because nobody ever complains when they know their house is not worth the AVM value. The simple truth is unless you are in a new cookie cutter neighborhood, no two homes (much less three) are exactly alike and have the same value. But that is not the valuation world they want.

Never mind that the public data for the subject property is wrong many times. That is irrelevant.

So, what do they come up with? Hybrid appraisals to further drive a wedge between the appraiser and the best valuation model (appraiser inspected appraisal).

They will never admit that best course to value is a full appraisal by a licensed appraiser, because they continue to claim they have the data that makes them better. The most ridiculous part of that equation is that the key ingredient to home valuation is the subject property and its physical characteristics, not the data itself.

 

[alebrewer]

If "they" pick the comps for us, whoever "they" are then it is no longer our appraisal and we won;t be needed - an app choosing the comps already exists and it is called an AVM.

That's not how AVM's work - and that's (IMO) one of their biggest issues. Comp selection is HIGHLY correlated to value variance. So that, two appraisers can have performed two credible assignments - and yet their conclusions be wildly different - based solely on comp selection.

What AVM's DO - and what the GSE's want to move towards - is analyzing an entire market segment and not just 3-6 'comps'. When analyzing a market segment, outliers are tossed and the remaining pool of sales are ALL analyzed to extract adjustments and the corresponding confidence levels (if thinking in terms of multi-variate regression, what is extracted are the variable estimates, the Std errors, and the measurement stats (t, p, R, R2, etc.)

 

[J Grant]

That's not how AVM's work - and that's (IMO) one of their biggest issues. Comp selection is HIGHLY correlated to value variance. So that, two appraisers can have performed two credible assignments - and yet their conclusions be wildly different - based solely on comp selection.

What AVM's DO - and what the GSE's want to move towards - is analyzing an entire market segment and not just 3-6 'comps'. When analyzing a market segment, outliers are tossed and the remaining pool of sales are ALL analyzed to extract adjustments and the corresponding confidence levels (if thinking in terms of multi-variate regression, what is extracted are the variable estimates, the Std errors, and the measurement stats (t, p, R, R2, etc.)

Appraisers analyze the market; they put 3-6 comps on the grid as the most competitive and similar. I use 4-5 compes, usually 5 if possible. But that also means I reviewed perhaps 10-12 others as well.

 

[Tim Hicks (Texas)]

That's not how AVM's work - and that's (IMO) one of their biggest issues. Comp selection is HIGHLY correlated to value variance. So that, two appraisers can have performed two credible assignments - and yet their conclusions be wildly different - based solely on comp selection.

What AVM's DO - and what the GSE's want to move towards - is analyzing an entire market segment and not just 3-6 'comps'. When analyzing a market segment, outliers are tossed and the remaining pool of sales are ALL analyzed to extract adjustments and the corresponding confidence levels (if thinking in terms of multi-variate regression, what is extracted are the variable estimates, the Std errors, and the measurement stats (t, p, R, R2, etc.)

So, all the people with highly upgraded homes or homes larger than public records get penalized. Conversely, those that have nothing done to their home since purchase 20 years ago and the GLA is overstated on public records get rewarded. AVMs are about medians, averages and ranges that does not fit many homes in the market. It does not account for features like views, outbuildings, bad data for ADUs, etc.

Agents are always pushing to get max value for their sellers and sometimes they over push past reality. Sometimes they sell homes for too less relying on bad public data records. There are just too many variables that make AVMs only reliable in cookie cutter areas. Rural properties have so many features that are not accounted for in the data that AVM data is inadequate.

 

[Tim Hicks (Texas)]

Appraisers analyze the market; they put 3-6 comps on the grid as the most competitive and similar. I use 4-5 compes, usually 5 if possible. But that also means I reviewed perhaps 10-12 others as well.

In the city, I first run the entire subdivision, then 20% each way on the GLA. Price is never a selection for the search criteria unless I want to cut out all the double wides in rural appraisals for stick built homes.

 

[Tim Hicks (Texas)]

That's not how AVM's work - and that's (IMO) one of their biggest issues. Comp selection is HIGHLY correlated to value variance. So that, two appraisers can have performed two credible assignments - and yet their conclusions be wildly different - based solely on comp selection.

What AVM's DO - and what the GSE's want to move towards - is analyzing an entire market segment and not just 3-6 'comps'. When analyzing a market segment, outliers are tossed and the remaining pool of sales are ALL analyzed to extract adjustments and the corresponding confidence levels (if thinking in terms of multi-variate regression, what is extracted are the variable estimates, the Std errors, and the measurement stats (t, p, R, R2, etc.)

An actual appraisal and an AVM can be wildly different too? One can have all the factual data of the subject and one can have only the limited data of public records.

Even in current data, I will see a three year old tract home with the media room and game room addition upstairs, but the tax records only show the base home GLA. What really is a 2700 sf home is only 2300 sf on public records. AVM that correctly.

 

[alebrewer]

AVMs are about medians, averages and ranges that does not fit many homes in the market.

So is risk, is it not? In a pool of SF properties, and assuming a normal distribution, 50% are 'above value' and 50% are 'below value'.

The part that I think a lot of folks haven't thought through yet is the 'fudge factor'. RN, a LOT of appraisals come in at purchase price. Of course much of that is due to anchor bias - which goes away when there is no 'person' performing the valuation...