Blind Squirrel and Acorns

[Michigander]

You can always count on a geology guy to find gold in a pile of rubble. I tend to the sarcastic in these online threads, sorry to anybody reading it.

But yeah although the "D" reviewer I was referring to was an AMC's administrative person.

I guess what I am clumsily trying to say is for someone like the OP, working in a centralized office trying to make sense of appraisal reports in remote markets, stop concentrating on the unknowable (value opinions) and focus on the knowable (facts and reported analysis). If the facts are not accurate then the credibility of the opinion is diminished. As to analysis; if there is no connecting logic between the facts presented and the conclusion reported, the conclusion is not credible. Stop accepting appraisal reports that are not credible no matter what that final number might be.

Well put!!!

 

[J Grant]

The truth be known, a reasonable value is a heuristic judgment and all the facts in the world cannot actually "support" what is a judgment call. But we present them to appear to support our opinion. I am afraid that is the way the world is. "Rule of Thumb" is a verboten term...but reality is our value estimates are often just that. Intuition, sixth sense, educated guess...It applies a lot more than we want to either admit or recognize.

That may be true of opinions we make in "life", or expounding on subjects in conversation. But it is not supposed to be true in an appraisal.

Though our judgments itself is not "factual", it should be an unbiased and logical conclusion made by considering facts, and other influences on value that can be observed to show patterns, buyer and seller reaction to qualitative influences, as well as reaction to market forces.

That is where imo, a competent review is valuable. Behind discovering errors, ( I'm grateful if an error is found in a report by another) , a good review can and will point out when judgments and conclusions of appraiser are contrary to facts, or ignored relevant facts ( such as cherry picking comps, ignoring knowable, visible condition or location factors etc)

 

[Michigander]

If by "the system" you're referring to Fannie then I would like to point out their appraisal policies fall under the category of "assignment conditions" and as such are subordinate to the hardwired content of USPAP. Fannie is not an appraisal entity and they don't have anyone over there who's qualified to instruct appraisers on appraisal standards, much less overrule the ASB on what those standards are and how they work.

So no, when there's a conflict between the form monkeys at Fannie and the ASB I don't see any reason to allow their stupidity to interfere with your obligations as an appraiser or a reviewer. These are the same morons who said the appraisal process follows the forms and who stacked the deck against appraisers (in bad faith) on the URAR as well as their C#23.

Moreover, if you ever get around to working with non-residential datasets you'll be in for a rude awakening on what is and isn't possible in terms of "accuracy" in an appraisal.

The operative benchmark for assignment results in SR3-3 is "credible", not "accurate". And credible is not measured as an absolute, either; it's measured within the context of the intended use. That means there's a sliding scale.

Great post. Too many form appraisers venturing into non-lending work too, without adequate knowledge. Trained by form-filling underwriting driven appraisers.

 

[Terrel L. Shields]

But it is not supposed to be true in an appraisal.

It's not supposed to rain on my birthday but it does.

If appraisal was a mechanical process then a computer could do it. Why does the AVM fail so often? Because it cannot make those judgments that come with experience. We choose the value that we are guided to, then use the support we find as evidence our intuition was correct.

I just looked at some land sales. Prices from $1,800 - $4,500 per acre. I can adjust for the amount of woodlot, I can adjust for this and for that. But "location" is a judgment call. "View" is a judgment call. "Access" cannot be quantified in any but a general way. Some sales, I may cull simply because they don't look "right"....

All of my choices depend upon my judgment. None of the rest ultimately matters since nothing else is going to determine a value for me.

Another example. I have two parcels - one is 40 acres. one is 56 acres. The first has been on the market for a couple years and sold for $2,750/ac. It is down a long driveway. Less than ½ mile in the same community I just appraised another sale, I came in below the contract price. It closed anyway. It is down a long private driveway. It is as rugged as the other property and actually faces it on the side of large hollow. And it has maybe a few less acres of woods...not much. It was bought in auction by a realtor, who then put it up for sale. Sold for $3,900/acre. There is no market evidence to explain the difference. Both appear to be arm's length sales. Judgment. It is all about judgment.

 

[J Grant]

T, I agree it is judgment (among other things) that differentiates an appraisal opinion form a value estimate derived from AVM .

As you would (hopefully) agree, the judgments have be logically concluded from a series of both facts and observations (it is assumed experience leads to better analysis). When a series of disparate facts exist, such as you mention buyers paying very different prices per sf of land or other disparate facts/trends, the appraiser's explanation of why his/her judgment to rely on certain facts, such as certain closed sales and not others, is critical.

AVM's and statistical runs tend to blend or average disparate prices , whereas appraisal judgment can differentiate among disparate prices and explain why some are market representative to assignment while others may not be.

 

[George Hatch]

You're confused. You need to slow down for a moment to think about what you're saying.

AVM's and statistical runs tend to blend or average disparate prices , whereas appraisal judgment can differentiate among disparate prices and explain why some are market representative to assignment while others may not be.

You're commingling different modes of analysis with one of the steps in analysis that all of them share, namely the qualification and quantification of the data.

Every valuation model "differentiates" the data and correlates the differences in certain factors to the differences in prices. When an AVM looks at 5 sales and breaks down the differences in GLA relative to the prices that is a mode of data qualification and data quantification. It's the same as when an appraiser does it.

What makes appraisals "special" is not the fact that we qualify and quantify our data - every model does that - but the extent to which we qualify and quantify that data. We personally inspect, we read through the listings and look at all those interior pics and make phone calls on our favorite listings to further investigate them. That is what enables us to refine our various ratings for terms of sale, location influences, design/appeal, quality, condition, additional features, etc.


Just the fact that the quants are pushing for appraisers to incorporate more statistical analysis to their data should demonstrate to you that this isn't an either-or choice.

Zaio's program had it's weaknesses but one thing I really appreciated about it is that their appraisers were qualifying all of the data their AVM was using; and their appraisers were doing their own model specification and model calibration. That means that the one individual who was using that data for their reports was rating and qualifying all their data at one time and using one objective rating scale in isolation of an appraisal assignment involving a specific subject property.

The comparison of those data against the subject came later, during the assignment itself. In that way their ratings weren't subject to contamination by the appraiser's anchor bias toward the subject.

Don't kid yourself that these offsite valuation models can't do a reasonable job for some assignments. The main thing holding them back is the limitations of the database they're using.

The killer app for enabling the AVMs will be when some title company or third party database company uses a team of "raters" to go out and do a driveby inspection of every property that gets listed in the MLS; reads the listing, and then rates that property against a specific rating scale. Then if/when the property does enter into contract or does sell later on, a clerk can make phone calls to "verify" the terms and enter that info into the database. And this can all be done using semi-skilled unlicensed technician types, not licensed appraisers.

 

[George Hatch]

BTW, there are already a couple of such database companies who send employees out to suss out the raw data and verify sales with brokers and principals for their database. One of them is named COMPS, Inc. They don't get into actually rating the properties but they could.

 

[J Grant]

Every valuation model "differentiates" the data and correlates the differences in certain factors to the differences in prices. When an AVM looks at 5 sales and breaks down the differences in GLA relative to the prices that is a mode of data qualification and data quantification. It's the same as when an appraiser does it

How does an AVM "look" at sales? It's an algorithm model (far as I know, I never ran one) .How in the world is this "the same" as when an appraiser does it?

As far as to the extent that an AVM's or statistical programs would average prices probably only the software designers know.

 

[J Grant]

BTW, there are already a couple of such database companies who send employees out to suss out the raw data and verify sales with brokers and principals for their database. One of them is named COMPS, Inc. They don't get into actually rating the properties but they could.

That's fine, might be a resource for appraisers )but of course one has to rely on third party results and how literally these employees consider RE agent info is a whole other matter....I have interviewed agents and considered as garbage some of their input because the facts of a property or the market say otherwise...I have had them mis present properties I actually inspected!

Regardless of these companies' efforts, someone still has to make sense of it all and draw logical conclusions and judgments (or they can just try selling their service to "enhance " AVM's.

 

[J Grant]

The killer app for enabling the AVMs will be when some title company or third party database company uses a team of "raters" to go out and do a driveby inspection of every property that gets listed in the MLS; reads the listing, and then rates that property against a specific rating scale. Then if/when the property does enter into contract or does sell later on, a clerk can make phone calls to "verify" the terms and enter that info into the database. And this can all be done using semi-skilled unlicensed technician types, not licensed appraisers.

So what? It still won't be an appraisal. If at some point the lenders convince regulators this is a better system, have at it.... even without regulators, I don't' see how any prudent investor in loans would prefer this, or what it would do to interest rates. I don't even see a big cost savings (especially since the consumer pays for the appraisal).

It costs to hire teams of unlicensed technician types and property inspectors and property raters and then correlate the info and produce estimates , and it assumes a lot of reliability of the info coming from these unlicensed technical types. (whoever they are, they won't have much vested in this...if they are low wage people what do they care if info is correct or not, and what training will they have to know the difference?. (Maintaining and updating databases is expensive and time consuming. I don't see much cost savings, or superior results, but there are always entrepreneurs trying to make money so efforts in this direction will continue.