"Quantifiable Market-Derived Methods" for adjustments required by FNMA/USPAP

[Mike Ault]

Of course, the unicorn "made up scenario to poopoo quantitative analysis". If homes with an extra bath sell for less in your market, then you'd adjust up for having the extra bath.

I always make bathroom adjustments. I just don't always have sufficient data available for reliable regression analysis support for those in small communities such as White Deer, Texas. Nobody else does either, with the possible exception of Danny. Give me the keys to CU, and I will get excited about automated analysis. That's why I like QuickSource, at least I have access to the origination appraisers data on some of those comparables so I "know" what the seller concessions associated with the sales are. An important field with values that are typically larger than the market demonstrates for an "extra bathroom", and one that is consistently misrepresented or omitted in the MLSs I use.

 

[alebrewer]

About 15 years ago, I was told that one paired sale is no more meaningful than none. While I disagree with that assertion in principle, I nonetheless understand its gist. To Terrel's point - one paired sale could be nothing more than a cherry picked pair to support an opinion (i.e. cart before the horse). However, if an appraiser allows the data to speak - instead of trying to make it say what they want it to - the paired sale is more meaningful. Of course, not as meaningful as multiple paired sales to support an adjustment.

 

[J Grant]

Exactly what I said - the goal is to reach a $0 range.

Nope. I do not agree that the goal is to reach a 0 range.

Again, the goal of applying adjustments is to make each comp more similar to the subject, to the extent possible; thus, we apply adjustments for negative or positive features. Not every feature differential gets adjusted for, and some are more important than others in any individual appraiser

Math is a tool, but apparisal is a Market value problem to solve, not a math problem to solve.

If the powers that be try to turn it into a math problem for speed or profiteering, then let them live with the possible skewed results. Of course, they are safe to experiment since it is the taxpayer on the hook for results.

 

[alebrewer]

Give me the keys to CU, and I will get excited about automated analysis.

Didn't realize we were discussing automated analysis. I thought we were talking about reducing the adjusted sales range?

 

[alebrewer]

Nope. I do not agree that the goal is to reach a 0 range.

So, then, you're saying the subject doesn't have an intrinsic value point? That it's somehow numinous? If the subject's value can't be known, then what is the intent of the exercise at all?

 

[alebrewer]

If the powers that be try to turn it into a math problem for speed or profiteering, then let them live with the possible skewed results. Of course, they are safe to experiment since it is the taxpayer on the hook for results.

Everything in life is a math problem - if you know how to solve it. Had to take a class on 'Mathematics for Economists' in my grad school journey - it would have been much easier for me to just say, hey - human behavior can't be quantified, so I'm not gonna hurt my head trying to do this math. Instead, I took the alternate path.

 

[Mike Ault]

Didn't realize we were discussing automated analysis. I thought we were talking about reducing the adjusted sales range?

I thought we were talking about "reliable" regression analysis. That means many analyzing many transactions & data points, which requires multivariate regression automation. No way to isolate and analyze a single feature such as "a bathroom" in a reliable manner without doing so. If you have found one, however, please share!

 

[alebrewer]

I think adjustments are more about conducting an experiment where you select appropriate adjustments that you think the market would make and see if the result makes sense

How do you determine whether the results made sense or not?

 

[alebrewer]

I thought we were talking about "reliable" regression analysis. That means many analyzing many transactions & data points, which requires multivariate regression automation. No way to isolate and analyze a single feature such as "a bathroom" in a reliable manner without doing so. If you have found one, however, please share!

Regression isn't the only technique. To J's point - experience is the factor that helps one determine which is the more appropriate technique. Unless you think all mathematical techniques are equally crap?

 

[J Grant]

So, then, you're saying the subject doesn't have an intrinsic value point? That it's somehow numinous? If the subject's value can't be known, then what is the intent of the exercise at all?

What does this even mean? I never said the subject's value can't be known!

But if there were such a thing as an intrinsic value point, and every property had a known benchmark value that everyone would pay the same price for it, always and forever, we would not need an appraisal, AVM, or any valuation.