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

[alebrewer]

Market derived adjustments which are "within reason", that judgment being based on decades of observing the typical motivations of buyers and sellers within the local market. I looked at his "zero dollar range appraisals" and the adjustments were simply crazy and considerably out of line with his peers. Sensitivity analysis should have constraints applied and those bounds should not be exceeded to arrive at "the goal".

And regression? How does that work?

 

[Mike Ault]

And regression? How does that work?

Not that well. Especially in Texas. We don't have accurate data for too many variables. Starting with "site values". Undisclosed seller concessions and "smoke and mirrors" manipulation of list prices to hide them in the MLS foul the data set as well, leaving any automated analysis suspect. Nondisclosure states have too many vulnerabilities for unethical real estate practitioners to exploit.

 

[J Grant]

Explain to me what you think sensitivity analysis encompasses?

Sensitivity analysis isolates a feature for an extracted adjustment.

I was being a little sarcastic, referencing that I never saw a goal of 0 (zero) as the adjusted range of the comps.

 

[J Grant]

"Motivation" being the primary one. 10 different buyers and sellers would most certainly arrive at 10 different sales prices for the exact same property. The property attributes would remain the same, so it's hard to quantify adjustments with that degree of accuracy and my adjusted ranges have never arrived at "the goal". More and better data, tighter ranges – less and weirder data – wider ranges, but never a zero $ value range.

Exactly. I sold RE for over 5 years before training to be a developer, and this is what happens in negotiations.

 

[alebrewer]

Not that well. Especially in Texas. We don't have accurate data for too many variables. Starting with "site values". Undisclosed seller concessions and "smoke and mirrors" manipulation of list prices to hide them in the MLS foul the data set as well, leaving any automated analysis suspect. Nondisclosure states have too many vulnerabilities for unethical real estate practitioners to exploit.

Sorry - poorly worded question. How does regression work theoretically? Isn't the goal of regression to minimize the squares of the observations? (Hint: it's the same as having a $0 range as the goal). To your point, though, again - the more homogeneous the market, the better the regression results.

 

[Elliott]

The thing about RA that gets me is that we seem to accept the premise that a bunch of buyers with some or limited market experience and sellers with the same level of market knowledge will somehow buy and sell and determine through a complex mathematical process come up with a reliable indication of what a particular adjustment should be. Even when I'm a buyer or seller, I'm not that sophisticated to figure out that "seller has closed the fireplace" to add a brick accent....I mean there's a fireplace in back of the accent brick. Should I as a buyer give it some value?

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, and if not, make another set of adjustments, trying to anticipate where the market has been or is headed. YMMV.

 

[J Grant]

The thing about RA that gets me is that we seem to accept the premise that a bunch of buyers with some or limited market experience and sellers with the same level of market knowledge will somehow buy and sell and determine through a complex mathematical process come up with a reliable indication of what a particular adjustment should be. Even when I'm a buyer or seller, I'm not that sophisticated to figure out that "seller has closed the fireplace" to add a brick accent....I mean there's a fireplace in back of the accent brick. Should I as a buyer give it some value?

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, and if not, make another set of adjustments, trying to anticipate where the market has been or is headed. YMMV.

That is what I believe the skilled appraisers do. An adjustment is a reasonable model of the most probable market reaction that most typically motivated buyers would pay for x positive feature, and expect as a discount for Y adverse feature.
The reality is that individual, real-world buyers value features or defects a bit differently, even among similar properties, and sellers have their own set of motivations for accepting certain prices or rejecting others.

 

[Mike Ault]

I know the

Sorry - poorly worded question. How does regression work theoretically? Isn't the goal of regression to minimize the squares of the observations? (Hint: it's the same as having a $0 range as the goal). To your point, though, again - the more homogeneous the market, the better the regression results.

I know the trouble I have with it, and when I see someone in my market arriving at a zero dollar value range, I know most of those adjustment factors were PFA'd. You have to start with a good data set. Datamaster wouldn't even offer their service in my market because there was no consistency in the MLS for too many important variables. If you have to manually research and input the lot size in the proper format for automated analysis for every transaction and you still have no idea if there was seller concessions or if the list price was manipulated to conceal them without extended research then you're just spinning your wheels.

 

[alebrewer]

No doubt mathematical analysis of human behavior is intrinsically incomplete due to exactly what folks are saying - buyers and sellers do not act 100% objectively - there is always subjective motivation in a transaction. Let's say you discount mathematical modeling for said behavior, though. How do you come up with an MV? I mean - as soon as you apply that first adjustment - regardless of the technique, you're using mathematical analysis to model human behavior, right? And once you've crossed that line, folks are going to want you to use modeling techniques that are supportable. And regression is more supportable than experience (IMO).

 

[alebrewer]

I know the

I know the trouble I have with it, and when I see someone in my market arriving at a zero dollar value range, I know most of those adjustment factors were PFA'd.

you side stepped the question, my friend...