No bias here - keep moving

[CGinMN]

Whether they're right or wrong about their comparisons between conventional 1004 data vs PDR data, they're telling appraisers that they have been making these comparisons, and these are the results of those comparisons. They're also telling appraisers that their analyses included data from prior appraisals in their comparisons of the current conventional 1004s and PDRs, so their comparison isn't limited to just this one appraisal vs this one PDR. (As in, no presumption that either the appraisal or the PDR comprises the benchmark for the accuracy of the other)
View attachment 94815

Don’t you find it at all curious that the GSEs always seem come in at or above value when it is to their benefit?

 

[George Hatch]

Well, their day job IS packaging and selling MBS investments.

It seems to me that the only valid basis for equal comparisons would be the same tranches of borrowers (credit, risk factors, etc) and the same types of mortgages (also rated by their respective LTVs and other risk factors). So, excluding property types where an AVM would be used.

A-Credit borrower + AVM-eligible property might have a different loan performance than a C-Credit borrower + non-eligible property. That doesn't seem like it would be a direct comparison.

 

[chad hampton]

I get why they do it, because it pushes their agenda, but it makes zero sense to correlate appraisal quality and loan delinquency.

 

[Tim Hicks (Texas)]

The trouble with data is that it can be manipulated to say anything they want. Just look at their figures for GLA and values. How do we know which is the correct value trend?

I did the free trial for Appraisal Loft depreciated cost support for adjustments. You have to go into it and manipulate the quality ratings, condition ratings (without just pulling adjustments out of our assets as they say), but all you are doing is manipulating the data to say what you want it to say. The lot values were pulled out of the air it seems as it valued a similar size lot in a gated community as less than the subject.

We have one county where many of the lot sizes are completely wrong. They will have whole subdivisions with lot sizes smaller than the one story GLA. When you look at the Realist map, you can see the lot size is 2 to 10 times larger than tax records state. You have to do that for the subject and all comparable sales to see real lot sizes. Every time I review an appraisal where some out of town appraiser uses those tax records lot sizes, I know they are not geographically competent for that assignment. Yet, all that data gets fed into the Fannie Data sets.

 

[George Hatch]

I get why they do it, because it pushes their agenda, but it makes zero sense to correlate appraisal quality and loan delinquency.

Not trying to be argumentative about it, but isn't one of the primary frets about using less than the conventional 1004 that these alternatives will lead to an increase in loan delinquencies which will significantly hurt the interests of the taxpayers?

 

[George Hatch]

The trouble with data is that it can be manipulated to say anything they want. Just look at their figures for GLA and values. How do we know which is the correct value trend?

I did the free trial for Appraisal Loft depreciated cost support for adjustments. You have to go into it and manipulate the quality ratings, condition ratings (without just pulling adjustments out of our assets as they say), but all you are doing is manipulating the data to say what you want it to say. The lot values were pulled out of the air it seems as it valued a similar size lot in a gated community as less than the subject.

We have one county where many of the lot sizes are completely wrong. They will have whole subdivisions with lot sizes smaller than the one story GLA. When you look at the Realist map, you can see the lot size is 2 to 10 times larger than tax records state. You have to do that for the subject and all comparable sales to see real lot sizes. Every time I review an appraisal where some out of town appraiser uses those tax records lot sizes, I know they are not geographically competent for that assignment. Yet, all that data gets fed into the Fannie Data sets.

It's the appraisers (and now the PDCs) who have been feeding most of that data into Fannie's datasets. Based in large part on their own personal observations. Perhaps incorrectly, but my read of it is the GSE's AVM is not operating with all the same limitations in their data quality that the other AVM programs are forced to deal with.

Even if Fannie/Freddie aren't handling the loans on all the 2024 comps used in an appraisal they might still have appraiser-developed data on those comps from prior appraisals performed in conjunction with prior refi or purchase transactions.

The 2024 sale for S#3 was an all-cash transaction but Fannie also has in hand an appraisal on that property from a refi from 3 years ago so they know how big the GLA is, independent of the MLS or public record data that Zillow is forced to settle for. That is, they know what the real GLA is so long as that prior appraiser wasn't a donkey who got it wrong.

 

[CGinMN]

Studying delinquency rates of AVM-backed loans during a time of historically low delinquency isn’t going to yield anything useful. Let’s see what happens when there are actual delinquencies and market values turn downward. We all know AVMs will not perform well through a full RE cycle.

As others have pointed out, delinquency rates is only one metric. Studying the magnitude of a loss to the lender is another.

 

[chad hampton]

Not trying to be argumentative about it, but isn't one of the primary frets about using less than the conventional 1004 that these alternatives will lead to an increase in loan delinquencies which will significantly hurt the interests of the taxpayers?

I never really thought so, but maybe others do? I think primary purpose of an appraisal is that should the loan go into foreclosure, the bank will feel comfortable that they have a somewhat reliable idea of what the asset’s value is/was. In fact, if only the most risky of loans are now getting an appraisal, I would assume that the delinquencies would be significantly higher than the guy putting 30% down.

 

[TerryRohrer]

I never really thought so, but maybe others do? I think primary purpose of an appraisal is that should the loan go into foreclosure, the bank will feel comfortable that they have a somewhat reliable idea of what the asset’s value is/was. In fact, if only the most risky of loans are now getting an appraisal, I would assume that the delinquencies would be significantly higher than the guy putting 30% down.

That is a primary argument of the GSEs. As has been noted above, when you use AVMs on cookie cutter assignments for borrowers that pose no risk of default, and compare that traditional appraisals for risky loans made to sketchy borrowers on rural properties where no good data exists, you conclude that AVMs/waivers are more accurate that appraisals. And when you are inherently liars and thieves, you extend those analytical methods to things like appraiser bias and such.

 

[George Hatch]

... And when you are inherently liars and thieves, you extend those analytical methods to things like appraiser bias and such.

And there you have it. Those who think the GSEs are liars and thieves aren't going to take seriously anything they say. The presumption here is that they are lying about their analyses.

Not that it matters. True or not, the GSEs are showing the rationale they are using for their decision making. They want what they want and they don't want what they don't want.