FHFA to allow alternative appraisal methods on purchases up to 97% LTV

[George Hatch]

A lot of low LTV loans have been getting done with appraisals, not waivers. Presumably for not meeting the other qualifying criteria for that program. There's no saying the "other criteria" for the lowest LTVs is the same as for the highest LTVs.

 

[Essexfenwick]

I already get a bunch of easy orders

Then I notice it’s the highest sale in the neighborhood

They tried the easy way but it came in low… order an appraisal

 

[shrubberyvaluation]

A lot of low LTV loans have been getting done with appraisals, not waivers. Presumably for not meeting the other qualifying criteria for that program. There's no saying the "other criteria" for the lowest LTVs is the same as for the highest LTVs.

I imagine for a 97% supposed LTV they would want a higher credit score, and hopefully have some sort of initial review of certain factors that could widly throw off their AVM estimates. A garden condo unit in a large community it probably won't be too far off unless its something odd like 2 units combined together. I just now these things can creep up and have relaxed requirements over time.

 

[George Hatch]

No matter what, this change amounts to another brick in the wall.

 

[CGinMN]

We might not have any alternative data providers as an alternative to Corelogic's MLS business, but I'm not voluntarily patronizing any company that is subsidizing or supporting the PAREA program. I also think it improper for TAF to be handing out any cash-money grants to education providers unless its the result of some consent decree with the feds because appraisal texts 70 years ago went along the then-prevalent lending practices involving redlining. TAF doesn't hand out grants for USPAP QE or CE vendors.

If the feds want to subsidize the Appraisal Institute and other QE vendors in their PAREA programs then let them do so directly.

Vote with your dollar. I used to use Total, but I didn’t know any better. Lots of appraisers who decry AVMs gladly send corelogic $1000/year because of the hassle to switch software providers. Come to think of it, now is a good time to make that change.

 

[Digger88]

Vote with your dollar. I used to use Total, but I didn’t know any better. Lots of appraisers who decry AVMs gladly send corelogic $1000/year because of the hassle to switch software providers. Come to think of it, now is a good time to make that change.

I tried switching just a month ago but with silver membership it's $730 and the others werent much cheaper

 

[George Hatch]

I don't criticize anyone for doing what they have to do to get by. I care that people do right by their users and not commit crimes in their work.

I just think supporting the use of PAREA in lieu of experience is contrary to our long term viability. And is counterproductive to the function of acclimating people to standing up to the IRL challenges to our role. We all face these challenges on a regular basis, even daily. There's no way to convey to a raw recruit the extent to which they will be challenged or how critical it is for them to say no if/when necessary, except by IRL experience where another appraiser has to assume the full responsibility for how their trainee responded.

IMO

 

[Essexfenwick]

^^^ just helps the SRA. The Institute is advocating watering down the licensing for that purpose.

 

[alebrewer]

I wonder if perhaps Corelogic may also have access to some of the appraiser data because they're Corelogic and they own an appraisalware package which includes appraisal report storage on their computers.

Based on this logic, First American would have similar privilege....

 

[alebrewer]

Four things:

On the generic basis the Big-Data valuation modeling still has many or all the challenges that existed with that usage in 2005, including the usage of data that nobody has rated for quality and condition and appeal and such. Some advances but perhaps not enough to make a difference. Zillows Zestimator is often unreasonable.

The other modeling the lenders and GSEs used - which you are referring to - wasn't about individual appraisals at all. They were separating and selling off tranches of those loan bundles based on the model's risk ratings. That's not an appraisal failure as such except to the extent the actual 1004s were out of line, it's arguably an underwriting failure in not identifying the problems with what some of the appraisers are doing. Whatever they're doing in 2024 with selling off tranches of their MBS trade still has the similar potential for problems and it would still amount to an underwriting failure.

​

What's a little different now​

​

The technology to suss out more of the physical attributes off a pic is a far more recent development which is wholly unrelated to the problems the commercial AVMs have with using public record data; and

When it comes to the GSE AVMs they aren't limited to the same data sources that the commercial AVMs use - they have access and are using appraiser-developed data including the pics that software is analyzing. And that appraiser-fed data isn't just limited to however many appraisals they have on hand for the subject property but also includes the data the appraiser collected/qualified for most or all of the comps that they presented as comps (inclusive the appraiser-vetted data in each of THEIR respective appraisals). I daresay the appraiser-vetted data is what makes the GSEs AVM a whole different beast of its own, to which many of the other AVM criticisms might not apply. Leastwise to the same degree.

TBH, I have to cringe a little every time someone complains about the alternatives that the CU system spits out because it implies - among other possibilities - that perhaps other appraisers conveyed different ideas about the comps they were using in their own appraisals.

All very true. The challenge for FHFA will be: how do you maintain the integrity of the data when you reduce the 'live' data stream by 75%? IOW - they have good info on quality and condition of properties for the simple fact that they're getting that data from thousands of appraisals a day. How solid is that dataset in 2 years when the incoming appraisal data is reduced by 75%? What about 5 years?