Hybrid appraisal report

[alebrewer]

This just in from Keith Wolf (Collateral Risk Innovation Manager at Freddie):
"AVMs do not adjust for property features. AVMs classify properties by features and then weight those features into an estimate of value. A regression model does the same thing with correlation weights. Both methods have errors in the approach but the errors are factored into the model. As long as the errors are consistent in a data set the estimate of value is close enough for mortgage lending work to approve a loan. When the errors are inconsistent like in a thin data set in a less populated and/or nonconforming property and/or area those estimates get less reliable. Roughly 30% of properties in the US fall into the thin market bucket."

 

[George Hatch]

Real estate is not "data" - so it will never be consistent because teh way RE is bought and sold or financed varies because of teh humans involved - unless all humans become replaced by robots, in which case none of this matters - a world of no feelz and all cold data - that is what some envision so if there is a God and that is what they want, to have humanity wipe themselves out in the mae of efficiency, then that will be our fate. A few of us are preserved to be exhibited in a zoo for the robots to look at and laugh about how smarter they are than the primitive humans with their messy feelings.

Think it through. A property's attributes can be rated (Q2.3, C3.7) on a uniform scale, and types of attributes that can be rated are not just limited to quality and condition of the whole. Ceiling heights, floor coverings, moldings, landscaping features, the distance and sweep of a view amenity. If someone was asking us to extend the UAD ratings to these other attributes we could do it. The main reason most of use don't do it isn't because it can't be done, but because nobody has asked us to do it to that extent.

(and yes, I am not so isolated from the GSE trade to realize the UAD doesn't operate in fractions. But it could)

With respect to the emotional needs element in the SFR markets, that's obviously true to some extent (albeit only in the SFR/Condo markets) but at the same time it's also only one of the variables that contribute to the imperfection. When we analyze the others that isn't a silver bullet for curing the imperfect market but it does serve to reduce the pixie dust factor. Wherein less > more.

The other comment I'll make is that 20 years ago the buyers and sellers were compelled to rely solely on the judgement of the brokers. They didn't have access to the information, meaning the level of transparency was a LOT lower than is the case in 2024. We can fairly attribute the difference to the rise of the technology and apps. We ignore that trend at our peril. IMO

 

[George Hatch]

This just in from Keith Wolf (Collateral Risk Innovation Manager at Freddie):
"AVMs do not adjust for property features. AVMs classify properties by features and then weight those features into an estimate of value. A regression model does the same thing with correlation weights. Both methods have errors in the approach but the errors are factored into the model. As long as the errors are consistent in a data set the estimate of value is close enough for mortgage lending work to approve a loan. When the errors are inconsistent like in a thin data set in a less populated and/or nonconforming property and/or area those estimates get less reliable. Roughly 30% of properties in the US fall into the thin market bucket."

The main selling point for appraisers is that we almost always work with very small datasets, consisting of data points that have been individually analyzed and "rated" prior to comparing them to our subject. And we spend a lot of time and effort to get to those small datasets.

 

[alebrewer]

The main selling point for appraisers is that we almost always work with very small datasets, consisting of data points that have been individually analyzed and "rated" prior to comparing them to our subject. And we spend a lot of time and effort to get to those small datasets.

That, combined with the potential for physical/functional/external issues, should allow appraisers to remain relevant at least for the foreseeable future. They've got a decent handle on the physical component via the C/Q ratings. As LiDar becomes increasingly stronger, they'll be able to solve for the functional component. And then, as Google Earth Recognition technology gets better, they'll be able to solve (model) for the external component.

 

[djd09]

the shills will try to gaslight for the unethical stakeholders...and ask you to drink the kool aid

 

[Tiffany Gibbons]

Paschal is their poster child for the fraud. Loan officers coming up with values for waivers and criminals doing appraisal inspections for dope money. Standard operating procedure for the snake oil salesmen. Keep the party going. I need to go to one of those 15th street parties. I like steak and bourbon. But I prefer women in my own age bracket. That probably disqualifies me.

No, being a real appraiser is what would disqualify you.

 

[J Grant]

Think it through. A property's attributes can be rated (Q2.3, C3.7) on a uniform scale, and types of attributes that can be rated are not just limited to quality and condition of the whole. Ceiling heights, floor coverings, moldings, landscaping features, the distance and sweep of a view amenity. If someone was asking us to extend the UAD ratings to these other attributes we could do it. The main reason most of use don't do it isn't because it can't be done, but because nobody has asked us to do it to that extent.

(and yes, I am not so isolated from the GSE trade to realize the UAD doesn't operate in fractions. But it could)

With respect to the emotional angle, that's obviously true to some extent (albeit only in the SFR/Condo markets) but at the same time it's also only one of the variables that can be analyzed. When we analyze the others that isn't a silver bullet but it does serve to reduce the pixie dust factor. Wherein less > more.

The other comment I'll make is that 20 years ago the buyers and sellers had to rely solely on the judgement of the brokers. They didn't have access to the information, meaning the level of transparency was a LOT lower than is the case in 2024. We can fairly attribute the difference to the rise of the technology and apps. We ignore that trend at our peril. IMO

I nunerstand Q and C ratings etc. Your posts ignore that apps and tech are vulnerable to criminals or just plain greed, - tech and data can be falsified, manipulated, or these days with AIs and deep fake videos and photos created for fraud - imagine a whole suite of fake properties with equity lines of credit created- for a start.

I agree the particulars of a property can be seen in photos, videos, drawings, and measurements. What is lost with no personal appraisal inspection is the site, influences, the quality ( houses can look either or worse in a photo ) how the house relates to the area, the neighborhood itself driving it, and no information to be gleaned from on site interaction with an owner, RE agent or builder.

An appraiser can do a very good job with PDC as the intel source in some cases, but not all, IMO. I did two proxies of this during covid when some owners did not want to let the appraiser in - the owner provided their own photos, got us a floorplan etc - the data itself was fine, but not having seen or walked the property it took me far longer to analyze the exhibits and there was no way to know if the value truly reflected the positives or negatives of the property as I might have first hand observed them.

Buyers can search Zillow and online properties and guess what - they make as dumb or personal-related decisions as they ever did, such as overpaying, etc.

RE is getting more and more expensive, and there are new issues with it with Airbnb, etc - a surfeit of data is as dangerous as not enough data. No offense to You know who on the board, but the Byzantie model of regression seems to get quite whacky results at time and it is not producing a market value as we understand it - more of a perfect net sale price.

 

[djd09]

at least we should call it what it is...unlicensed home inspections

 

[George Hatch]

IDK- it depends on how borrowers, RE agents and the lenders themselves see these products - the fly in the ointment for GSE work is it is taxpayer backed, therefore the lenders do not get to just decide we like AVM's better - it depends in future perhaps how the WAIVERS perform over a long period of time - they still are relatively new/less total volume.

As long as the AMC system is around, they will make sure they profit on any product used for valuation, be it an AVM, a WAIVER, an appraisal, a PDC, or a combination thereof. Imo there will always be a need for apparisers but it will shrink depending on how much control people want to give to tech ( which can be hacked, manipulated or just plain wrong ) and how much society in general wants to see tech barons and anonymous corporations control every aspect of their lives with no qualified independent people around to keep things on track.

We see how well removing editorials and opinions from journalism worked out—now social media, once thought to be a haven for exchanging ideas, has become saturated with foreign nations impersonating Americans, creating fake content and posts to influence our political outcome and economy.

The downside of tech is pretty dark and that is why we see increased hacking and ransomware and fraud and manipulation of it.
Too many people with no vested outcome removed and entire databases of properties can be hacked and sold to criminals, and a slew of fake mortgages or equity lines created - using AVM;s

Why not take tech and everything third party a step further and create fake houses and condos to finance with the proceeds going to criminals' slew of mortgages based on using a VA<, with a ty Casi Cuba floorplan, and AI-created photos and signed off on by a false identity PDC "person."

The way I see it, people form their own opinions about credibility. I use this example a lot, but everyone knows what the credibility factor is for the supermarket tabloids. They still sell but most people don't take any of it seriously because they know better. The disinformation game will eventually teach the public to refrain from jumping to conclusions about the credibility of all content coming in from new-to-them sources.

The reason the journalists are pissed off right now is because they relied too much on the appeal to authority (we're journalists so that makes us an authoritative voice).

The lie by omission (not reporting it) has become just as detrimental to the public's perception of the journalism ethics as has the lie by commission. And that factor is totally their own fault, wrought by their own hand. You can complain about the bots but IRL the one-guy+one-mic pundits have been decimating the audience share of the mainstream press and media. And the reason for that is that the podcast and commentary crowd (on both sides) don't allow the mainstream's lie by omission to stand unexamined or unchallenged.

CNN isn't attempting to trend toward objectivity in their editorial bias for no reason. They're doing it in order to improve their credibility factor.

And then there's the elephant in the room that nobody wants to talk about. The political polls showing up are noticing the distinct break in political orientation between male and female voters. Particularly in the younger age ranges. You can blame it all on the women and their adjacents allies being mad about the baby murder issue being returned to the states if you want to but if you watch people being interviewed on the street or in the panel discussions most of them touch on topics other than baby murder. You'll also see that a lot of men see things differently on these other topics than do a lot of the women. That trend seems to be rapidly picking up steam, especially among the 18-30 crowd. Men are trending one way and the women are trending the other way.

 

[George Hatch]

at least we should call it what it is...unlicensed home inspections

Agree completely. The act is the act. And that seems like a vulnerability that could be approached directly. In those states which regulate their home inspectors.