A quick Predictive Methods Conference summary

[hastalavista]

...I agree with you in that the last 5 years has been an extremely rolling boiled pot and that the market is about to simmer. I also think that in many markets there will actually be the cooling off cycle and that's where the decision makers are going to backpeddle a little about AVM useage. Especially when they start seeing foreclosures increasing and they review the docs and see so many with AVM's as the primary source for decision making. It might actually turn that curve ball back into a hard fast ball with only a straight down the pipe approach - get an appraisal.

I think so, too. There are a couple of other factors involved here:
A. We, as appraisers, are still our own worse enemy when we allow ourselves to get pressured and inflate values. Our "credibility" is dangling by a thin thread (IMO). What I hope does not happen is that its only inflated appraisals that "float to the top"; that may be a hit to our integrity that will be hard to recover from.
B. Collateral valuation accuracy is king, but Speed & Cost are a close second. We need to be able to take advantages of new technologies to increase our productivity and hold-down our costs when possible. Those appraisers who are providing such services as "AVM assisted appraisals" (You know anyone who is doing that?) are at the front of the curve.

I've said this before; I think Fannie's influence as the secondary guideline-setter is going to "wane" as private markets take on more and more of the securitization role; remember, Fannie & Freddie were created to maintain the liquidity of the market and provide lenders with funds to loan. They still play a very important part in that liquidity objective, but the private markets are becoming large and sophisticated enough to take more and more of the role. For the private markets, the rating agencies are the ones that are charged with evaluating the "risk" of a portfolio. If the rating agencies conclude that appraisals are the most accurate tool available for valuing properties, then they will reward those portfolios that use an appraisal to value the collateral (their reward is a higher rating; maybe AA+ instead of AA).

But, whether its Fannie, Moody's, Freddie, S&P, or the OCC, it will boil down to us as individuals to maintain value credibility and us collectively to do what we can to get those who use inflated appraisals as part of their service offering out of the profession.

OK, I'm off my soap-box!

 

[Brad Ellis]

Howard,

While that is a salient point about knowing about the AVMs being used, I am not sure that USPAP requires us to know the underlying algorithms (the black box stuff).

I am of the opinion that we do not but it would be interesting to hear from the other AQB instructors on their understanding of it.

One thing is for certan, in my mind anyway, and that is the fact that we, as appraisers, can certainly offer these limted scope products/services. The real question is more over how much we could charge and be competitive with BPOs.

Brad

 

[Ross (CO)]

So, what's the going rate for a typical BPO today ?.....$25 ?......$50 ?

Can one label themselves as being "competitive" by offering AAAVM's for less than $25 ?....less than $50 ?

Will the appraiser's signature still be required on such a product ?

Would the industry tolerate an AAAVM report which was submitted anonymously ?

Do those AVM's currently created, and used in substitution for a normal appraiser-sourced value opinion, have any acknowledgment of the person, the individual, who submitted that AVM for the loan decision ?

How does it work......when an AVM-sourced loan goes belly-up, perhaps 18 to 36 months after starting, and someone to "go after" can not be found ?

 

[hastalavista]

Do those AVM's currently created, and used in substitution for a normal appraiser-sourced value opinion, have any acknowledgment of the person, the individual, who submitted that AVM for the loan decision ?

How does it work......when an AVM-sourced loan goes belly-up, perhaps 18 to 36 months after starting, and someone to "go after" can not be found ?

Ross-

As I understand it, the only loans that AVMs are used as the collateral valuation tool is for HELOCs and 2nds. The reason behind this (as was explained to me) is that although the loan is secured by real property, these products are priced to investors as if they will experience 100% loss if the particular loan goes into default. Obviously, all properties that are encumbered by more than one trust deed don't go into default, and of those that do, many (if not most) when sold have some equity to pay off some or all of the remaining liens. Regardless, these mortgages are sold with a price that assumes a certain percentage will go into default with no recourse to the lien holder.

I am not aware of any loan program that uses an AVM as the source of collateral valuation for 1st TDs; an exception may be if a loan is held in a lender's portfolio, and it is refinancing with the same lender; however, even in this scenario, I believe there are certain guidelines that limit the percentage of loans that can be evaluated this way.

As to the cost of a BPO vs. an appraisal, that is a factor in deciding which tool to use. An AVM assisted appraisal is an alternative. Here are two quotes from the June, 2006 issue of DS NEWS (Default Servicing Trade Publication):

"It's an age-old problem: You get what you pay for. The income derived from doing BPOs is minimal, so for an agent to make money, they've got to do large volume and pay someone with little experience to do the BPO."
Dave McCarthy, CEO, Integration Asset Services, Inc.

"There is a qualitative aspect to the valuation process that goes beyond just the straight cost of obtaining an estimate... Frankly, this is something we've started taking a closer look at- comparing the values we get in foreclosure versus the values we get in REO. We're asking ourselves if we're giving up too much in our search for low cost."
David Tiberio, Vice-President and REO Manager at Bank of America

The truth is that in a rising market, inaccurate BPO valuations haven't been much of an issue. That is beginning to change.

I think its John H's tag line that says something like-
In a calm sea, it takes no great skill to keep the rudder straight

In a rising market, inaccurate BPOs haven't been much of a problem. Now, they are beginning to surface as an issue. The key is to see if (and, admittedly, it is a big "if") the appraisal industry can offer a viable alternative, producing a superior credible value, at a cost that is efficient for the client resulting in a profit that is sufficient for the appraiser.

Let's face it: No business client expects a service vendor to offer a product that isn't overall profitable. The measure of profitability cannot always be judged by the unit; sometimes it needs to be measured by volume, sometimes it needs to be measured on the whole (such as offering a loss-leader).

IMHO

 

[Howard Klahr]

"Black Box" Analysis

Hey Brad,

... I am not sure that USPAP requires us to know the underlying algorithms (the black box stuff).

Although this is not directly on point in regard to AVMs, Statement 2 regarding DCF analysis states the following:

"Computer printouts showing the results of DCF analysis may be generated by commercial software or by software prepared by the appraiser. Either way, the appraiser is responsible for the entire analysis including the controlling input, the calculations, and the resulting output. If using commercial software the appraiser should cite the name and version of the software and provide a brief description of the methods and assumptions inherent in the software."

In addition Advisory Opinion 18 - Use of AVMs states that:

"Understanding and Control of the AVM (AO-18)
When using an AVM in an appraisal, appraisal review, or appraisal consulting assignment, an appraiser should have a basic understanding of how the AVM analyzes data to determine whether the AVM measures and reflects market activity for the subject property. The appraiser does not need to know, or be able to explain, the AVM’s algorithm or intricacies of its statistical or mathematical formulae. However, the appraiser should be able to describe the AVM’s overall process and verify that the AVM is consistent in producing results that accurately reflect prevailing market behavior for the subject property.

AVMs differ in the number and type of data characteristics as well as in the volume of data analyzed. The appraiser should know which characteristics (e.g., size, location, quality) are analyzed and how the analysis is tested for accuracy and reasonableness. The appraiser should ascertain that the characteristics analyzed are those to which the market responds.

Some AVMs allow the appraiser to select the data analyzed on the basis of, for example, distance from subject, size, or age of the improvements. An appraiser’s ability to change the AVM’s selection parameters may affect the appraiser’s decision to use or rely on the AVM output.

The appraiser should be aware that the AVM may not perform consistently given the same input criteria. The appraiser should be confident of the AVM’s credibility when applied to a specific property. The appraiser decides whether to rely on the AVM output, regardless of the AVM’s overall test performance. In some cases, the appraiser may accept the AVM’s output, while in other cases that same AVM’s output would not be acceptable."

Bold and Underling added for emphasis

IMHO, AVMs are very much like a DCF analysis, there are many appraiser's that have a complete and thorough understanding of the underlying concepts while others just input data and rely upon the output with no understanding of how the financial model arrived at the answer. I think that unless you can explain the process which would require some knowledge of the mathematics then the appraiser becomes just a data input clerk for the computer model rather then imparting professional knowledge and experience.

 

[hastalavista]

Howard-

I won't answer for Brad, but I'll give you my read-

I think the significant factor is "What degree does the appraiser rely on the AVM in concluding the Value Opinion?".

I'm not a commercial appraiser, so excuse me if I make an idiotic statement (at least, in this regard), but if an appraiser uses DCF software, the appraiser is relying on the results and the value opinion is significantly dependent on the accuracy of the results.

For an AVM, if the appraiser was relying on the AVM output and signing her/his name without independent verification, then the appraiser would need to have intimate knowledge of the process (knowing how the black box works); even directly inputing the parameters as you outlined into the AVM's search criteria wouldn't satisfy the need to know "how the thing worked" (IMO).

If the appraiser reviews the AVM valuation, and then completes an independent market analysis that concurs with the AVM (either consistent, supports, etc.), then the AVM becomes one of the tools the Appraiser is using in forming his/her value opinion. The Appraiser is not certifying the AVM. The Appraiser is providing his/her value opinion, of which part of the support is provided by the AVM.

My second AVM scenario is consistent with what is commonly referred to as "AVM assisted appraisals".

So, if I were able to devise an AVM that replicated my analysis process, I could probably sign off on its valuation conclusion without completing additional research (which, in this case, should be duplicating the exact same process).

However, if I take the AVM's value and complete my own independent research, I can then form an opinion that may or may not be consistent with the AVM. I am relying on my research, of which the AVM is part of that whole.

 

[Carnivore]

Howard,

Great point! The last time I checked, USPAP applies to all of us, even little ole residential guys.

 

[The Warrior Monk]

The DCF software we use is simply a calculator; we put in the numbers and the program spits out the number. It's similar to taking out a calculator, inputting some numbers, and hitting the equal sign. The data and methodology are transparent.

With AVM, the data is processed behind the scenes. It's not simply adding or subtracting numbers; it processing it in a proprietary fashion, which can be any number of ways, and it is completely hidden.

IMHO, the appraiser cannot just verify if the number is correct. If the appraiser is attempting a review, it cannot be done without knowing the manner in which the data was processed. If this information is not available, how is the appraiser to know if the methodology is correct? It's similiar to trying to review a restricted use report; not enough information is presented to perform a review.

The appraiser could also not review the results of the AVM and do a separate appraisal, but that defeats the purpose of doing an AVM in the first place.

 

[hastalavista]

IMHO, the appraiser cannot just verify if the number is correct. If the appraiser is attempting a review, it cannot be done without knowing the manner in which the data was processed. If this information is not available, how is the appraiser to know if the methodology is correct? It's similar to trying to review a restricted use report; not enough information is presented to perform a review.

David-

I'm looking at this differently. I don't see the process as the appraiser performing a "review" (SR-3) on the AVM (and, since an AVM is not an appraisal, I don't think an SR-3 review is possible). I see it as the appraiser using the AVM as a tool in his/her own valuation process.
The AVM is just another "data point"; the appraiser is the one that provides the opinion of value. It may or may not be consistent with the AVM. If it is consistent, there may be little need to add anything, other than a signed certification and SOW. If it is different, then there may be additional comments/data required, given the intended use.

If a client uses an AVM only, the client is relying on the AVM for the valuation of the subject.
If a client uses an appraiser who is incorporating an AVM as part of the analysis and presentation, then the client is relying on the Appraiser's Opinion of Value and not the AVM. The AVM is merely part of the data.

But, I'll shut up for a while and see how others see it!

 

[hastalavista]

(One quick addition- I don't know how M&S or Bluebook works, but I use those tools as part of my data analysis).