UAD 3.6

[alebrewer]

Aside from that, true matched pairs with only a single different vairable are pretty rare in the real world.

That's not the only way to do matched pairs... another tried and true way is to adjust for other differences and what is left is the contributory value of the element of comparison in question.

 

[timd354]

That's not the only way to do matched pairs... another tried and true way is to adjust for other differences and what is left is the contributory value of the element of comparison in question.

I get what you are saying, but then what you are doing is not actually a matched pair analysis and the calculation of the the contributory value of the element of comparison in question will be a complete mess if the adjustments for the the other differences are off.

Additionally, deriving any adjsutment on a single paired sale is weak at best and is simply not statistically valid.

Your view of matched pair analysis (a/k/a Paired data analysis) does not align with the leading appriasal texts, including the The Appraisal of Real Estate:

 

[J Grant]

Readers of appraisal reports actually understand that biggest lie repeatedly told in appraisals is that an adjsutment was deirved from a matched pair.....while not always a lie, it often is a lie....how do I know? I know becuase when I ask an appriaser to actually provide the matched pair, they often cannot do it. Aside from that, true matched pairs with only a single different vairable are pretty rare in the real world.

Paired sales, not matched pairs. Since, as you say, a true match is rare, but paired sales are plentiful, they are used as the comps. Even if an appraiser used the word matched pairs, a reader can see clearly from the comps as adjusted on the grid, whether an adjustment made sense or not. How are they going to determine that from a software program? Especially since it is rare, (since a computer science degree is not a prerequisite) an appraiser does not understand the algorithms and lacks the time or means to vet the large data sets used.

Of course, adjustments are not the primary reason a valuation goes off, - there in an appraisal or an AVM, or other program, the comp choices determine results. The most well-supported adjustments applied to the wrong comps will result in a skewed value in an appraisal (or other product).

Statistics and larger data set derived adjustments should augment traditional, easy for a reader to see and thus understand appraisal methodology, rather than replace it. Traditional appraisal methodology is based on how the buyers and sellers behave with the relevant subject and a smaller set of highly vetted comps, filtered through the appraisal development. Vs large data sets, which can use not similar and sometimes inappropriate propeties to get enough data.

 

[J Grant]

Paired sales can isolate the item for adjustment by making all the other adjustments and leaving the one item , it works well, providing good comps are used. A bad comp will stick out when it is done well -and I will eliminate a comp when that happens.

 

[alebrewer]

I get what you are saying, but then what you are doing is not actually a matched pair analysis and the calculation of the the contributory value of the element of comparison in question will be a complete mess if the adjustments for the the other differences are off.

Only if the adjustments are 'off' as you say. It is a recognized technique and, if performed properly, elicits supportable results.

Additionally, deriving any adjsutment on a single paired sale is weak at best and is simply not statistically valid.

Never said it wasn't.

Your view of matched pair analysis (a/k/a Paired data analysis) does not align with the leading appriasal texts, including the The Appraisal of Real Estate:

This is a bit disingenuous, is it not? You conveniently left out the part about using 'primary' pairings first and 'secondary' pairings next. I think you know you were being deceptive, though.

 

[BRCJR]

Paired sales can isolate the item for adjustment by making all the other adjustments and leaving the one item , it works well, providing good comps are used. A bad comp will stick out when it is done well -and I will eliminate a comp when that happens.

I thought that process was sensitivity analysis......

 

[TerryRohrer]

I get what you are saying, but then what you are doing is not actually a matched pair analysis and the calculation of the the contributory value of the element of comparison in question will be a complete mess if the adjustments for the the other differences are off.

Additionally, deriving any adjsutment on a single paired sale is weak at best and is simply not statistically valid.

Your view of matched pair analysis (a/k/a Paired data analysis) does not align with the leading appriasal texts, including the The Appraisal of Real Estate:

View attachment 104804

Folks who do appraisals for a living, rather than opine about how they should be done, recognize very early on that "text book" examples are rare or nonexistent in the real world, and must find a means to finish the job. What Alebrewer suggests is nearly always how it must be done if the analysis needs to rely on "pairing."

Having said the above, it is also quick and easy to articulate that was how the issue was resolved, and there is no credible defense for appraisers who fail to make that explanation. I typically do the same to gauge the impact of differences in quality and condition on price, though I generally use significantly more than a pair of sales, rate them compared to the subject, and run a small regression. That process takes little more explanation than the previous sentence.

 

[TerryRohrer]

I thought that process was sensitivity analysis......

Grant doesn't believe in "statistics" or "math" in appraisal.

 

[timd354]

Only if the adjustments are 'off' as you say. It is a recognized technique and, if performed properly, elicits supportable results.


Never said it wasn't.


This is a bit disingenuous, is it not? You conveniently left out the part about using 'primary' pairings first and 'secondary' pairings next. I think you know you were being deceptive, though.

Not deceptive at all and the section 15th edition of the The Appriasal of Real Estate that i quoted from includes nothing whatsoever about primary or secondary parinings.....below is the entirety of that section (in 2 separate screen shots). You are apparently confusing paired sales analysis with grouped data analysis (see below).

 

[alebrewer]

Not deceptive at all and the section 15th edition of the The Appriasal of Real Estate that i quoted from includes nothing whatsoever about primary or secondary parinings.....below is the entirety of that section (in 2 separate screen shots). You are apparently confusing paired sales analysis with grouped data analysis (see below).
View attachment 104806

View attachment 104807

In the edition you're citing, they're called 'pure pairings' (i.e. primary pairings) and 'adjusted pairings (i.e. secondary pairings). The idea is the same, though, and does confirm that - contrary to your accusation - using adjusted pairs is a recognized technique.