S & P notes jump in RMBS, AVMs

[Stephen J. Vertin MAI]

NEW YORK--(BUSINESS WIRE)--Standard & Poor's--March 19, 2002--The recent wave of Automated Valuation Model (AVM) vendors entering the North American RMBS marketplace with their own variations on established model types will increase the need for detailed evaluation, according to Standard & Poor's.

"These variations have implications for the user in such areas as accuracy, geographic coverage, and hit rate (a measure, expressed as a percentage, of the total number of values a system was able to provide divided by the aggregate sample)," writes director Leslie Albergo in an article titled "RMBS Automated Valuation Models Necessitate Closer System Examination."

With the growing number of models available and their increased usage in the market, Standard & Poor's is continuing the evaluation of AVM systems it began in 1998. This ongoing evaluation is designed to address the credit risk associated with the potential use of this valuation method in the private-label residential mortgage market.

The increase in AVM vendors has been attributed to the Internet providing greater distribution capabilities and increased data availability.

"It is expected that 10% of all new originations in the residential mortgage market this year will have an automated valuation attached to them in some way," said RMBS director Susan E. Barnes. "AVM usage has evolved from being employed solely as a quality control tool to its utilization in the origination process. Utilization in the origination process ranges from its being employed as an unbiased 'reality check' on an appraisal to its use as the sole determinant of property value."

Fueling this streamlined origination process are the benefits derived by the seller, which include expedited approval, resulting in reduced origination costs, an accelerated closing period, the potential for inclusion in a risk-based pricing strategy, and competitive customer service.

The complete report is available on RatingsDirect, Standard & Poor's Web-based credit analysis system.

The report is also available at www.standardandpoors.com. To access it, click on "Resource Center"; then, under "Reference Tools," on "Structured Finance"; and finally, on the article title.

 

[Austin]

This quote is key to the AVM question: “the potential for inclusion in a risk-based pricing strategy,” Reply: What is a risk-based pricing strategy? It is simply an insurance policy in which the lender protects themselves by charging a higher rate or up front fee to the borrower risk rated to cover the potential loss. That being the case, would it not make sense that if a homeowner loses his property, then the lender absorbs the entire loss? They can’t have it both ways. They can’t charge an up front fee for insurance, then get a deficiency judgment if the property sells at foreclosure for less than the amount owned by the home owner.
This is a circular reference. If we do not have a system to insure a stable economic pecking order among properties, then the potential for loss is increased exponentially. This is basically the same scheme that FNMA is considering with the $50 fee in place of an appraisal and the same argument applies.
AVM is not an appraisal method because it does not evaluate the subject property, it evaluates the market segment in which the subject falls. In other words, it tells you everything but what you want to know. It is a good tool for evaluating market segments and drawing general conclusions, but one on one it is a killer, as some will soon see. My concern is that when they see, you know who is going to pay the price. That is why I want to remove the liability of the homeowners and tax payers before the big experiment. Or so it seems to me. Wouldn't you agree?

 

[Pamela Crowley (Florida)]

Accuracy and hit rate vary dramatically, at times depending upon the model type and the granularity of the factors encompassing the test. The hit rate among the seven systems varied from 50% to 90%. This large range indicates the inherent differences in the system types, with a higher hit rate attributed to Repeat Sales Index models and the lowest hit rate to Hedonic models. Accuracy ranged from a 0% variance to an overestimate of as much as 300% and an underestimate of approximately 85%.

This is the reason for the big push by AIRD (Appraisal Institute database) and software companies/lenders that are mining the data from our more complete appraisal reports.

It's absolutely sickening that some appraisers are not only involved but are promoting this. Even worse are the ones that have no idea what they are doing when they send a report "AI Ready".

I'll withhold ranting about the 'Appraiser assisted AVM' where someone that actually calls themself an appraiser signs that pre filled in crap for $40. Although, I doubt any self respecting 'appraiser' doing those knows what they are doing anyway - those won't really help the accuracy of any AVM so the joke is really on the companies getting those done by incompetents.

Sad that the people with the brains and money are the only ones that will know about this deception and will eventually be the only ones that would consider hiring a real appraiser for themselves - but, just for when they are purchasing. Gee, I'd love a home equity loan on a 300% above real value AVM. The 'general public' will never be informed.

And.... while I'm at it.... I'm willing to place a bet that all those borrowers are still being charged for a full appraisal even though they are not getting one. Just where is the 'savings' for the borrowers that they like to throw in to these articles???? Who is keeping the millions in appraisal fees that are being charged and not paid to any licensed appraiser??? The banks!

 

[Austin]

Quote from above: “This large range indicates the inherent differences in the system types, with a higher hit rate attributed to Repeat Sales Index models and the lowest hit rate to Hedonic models.”
Reply: What does that statement mean? It means that AVM works fine in cookie cutter subdivisions. So will averaging the last three sales from the cookie cutter subdivision. Second and more significantly, the reason the hedonic models are giving bad answers is that they don’t know which independent variables to use. They just pick the independent variables they think will work, load the model, and pull the trigger. Sorry, but hedonic modeling doesn’t work that way. You have to know which variables to use, in what order to use them, and progress your way through one independent variable at the time and know when to stop. There is also the problem of covariant or correlated independent variables that you have work around. Then there is the little problem that they haven’t figures out: how to test the model to verify its accuracy. I didn’t say it couldn’t be done, I said “they” have figured it out yet.

 

[jtrotta]

Austin;
How do you know what "Risk-based pricing strategy" is :?: If it is an insurance policy to cover a potential loss and they are charging a Fee for it, then I would assume it needs to be disclosed at the closing. If that is true and the Lender is using to protect against loss, then they will not be able to go after the homeowner for a Deficiency, as they would then put themselves in a situation to be sued. Insurance companys follow specific guidelines and are extremely cautious, as to their liability.

As for the deficiency issue itself, I have my own opinion of that and will not ramble here, as it is not pertinent.

8)

 

[Austin]

Jtrotta: Why do you always read something into my post that is not there? I did not say anything about an insurance company. You can make your own insurance policy and we all do it all of the time. If I get a high-risk appraisal, I charge a higher fee to compensate for the risk. Investors do the same thing with higher cap or yield rates to compensate for higher risk. Lenders do the same, that is why we have a prime rate for low risk borrowers. They don’t buy an insurance policy, they adjust the rate. They don’t have to disclose anything and they sure as the devil can still go after a deficiency for the losses thus double dipping.

 

[Anonymous]

I get calls all the time that tell me assessor values are
always low...like its a fact. I tell them when I worked
in the assessor's office the variance was usually
80% to 120% from sales prices after we did physical appraisals,
then it got worse each year after that.

85% to 300%?!!

Reminds me of the old proverb,

"A fool and his money are soon parted."

elliott

 

[Joe Birrell (NY)]

Spinners spining the spin - double speak by geeks.

 

[Bryan (OK)]

I include a copyright statement in the Addendum that basically states the appraisal report is the sole property of the author with approved use for normal preparation of a loan. The sale of information contained in this report is strictly prohibited and forbidden without the written consent of the author.

The exact wording is much longer and I don't know if it will hold water when a $400 / hour attorney attempts to rip it apart, but I am trying to keep the data from AVM's. How will I catch them . . . who knows??

Maybe someday I can hire a $500 / hour attorney on a contingency basis to reap a retirement windfall :!: :!:

Our product is being sold for other than its intended purpose without compensation, that is, to the appraiser.

Maybe we could find an appraiser / attorney with empathy to draft of rock solid statement we could all include in our reports :?:

 

[bradellis]

Pam,

Are you aware that the AIRD participants can deny FNC from using their data in AVMs? I am told by both their President and by another board member that this is possible as is the right to redact any information from the data sent to them. Just thought you would like to know- ar at least that is what they SAID.

Brad Ellis, IFA, RAA