Deep Dive - The Cost Approach

[alebrewer]

The sales on the grid are typically just a sample of a data set, so it is possible that they would widen, while the larger set of data from which the adjustments were developed would tighten. I am in agreement with what you're trying to say though. We shouldn't just plug adjustments in and then let the results speak; this is yet another reason why human judgment and expertise play an important role.

Great observation, but one that I would disagree with. The data set used (IMO) should always express a tighter adjusted range than raw range. If not, your adjustments are not 'working' (so to speak). Trying to imagine how this is possible... are you maybe saying that you use the large data set to extract your GLA adjustment, then apply it to the 'grid' data set (for example)?

 

[alebrewer]

Certain adjustments will tighten up the range by their nature. Sensitivity is accounting for a variable that makes all the comps fall more in-line. So that I understand why you're asking this, let me first ask: do you believe the point of adjusting the sales is to create a tight range or to make the sales more comparable to the subject?

For the most part - yes. I'm not saying the purpose is to make the sales more comparable to the subject, but rather to adjust for elements of comparison in such a manner as you distill a range of adjusted value within which the subject falls. If you start out with a raw sales range of, let's say, $100k, and you perform your analyses and end up with an adjusted sales range of $150k, what, exactly, have you told your intended user? That - before you analyzed the 'comparables' the subject's market range was $100k but after your analysis it's $150k?

 

[CGinMN]

That is an example of what I'm saying, yes. But instead, consider the margin of error. Let's say you have 3 sales that all sold for $300,000, but they are all different designs, different conditions, different size sites, etc.

 

[alebrewer]

That is an example of what I'm saying, yes. But instead, consider the margin of error. Let's say you have 3 sales that all sold for $300,000, but they are all different designs, different conditions, different size sites, etc.

In my (granted somewhat simple) mind, that is the market saying that whatever the elements of comparison are - the market doesn't recognize them as adding to, or detracting from, value. (assuming the sales are all arm's length sales exposed to the market with no undue stimulus of course). Are you saying that's not the case?

 

[CGinMN]

That would not be the case in this scenario. The market does recognize a difference for these differences, but when adjusted up for GLA and down for site size, the net effect is not exactly 0.

 

[Joe Flacco]

Sometimes I get the one that is $360k in that scenario.

 

[alebrewer]

That would not be the case in this scenario. The market does recognize a difference for these differences, but when adjusted up for GLA and down for site size, the net effect is not exactly 0.

If you see that IRL - let me know. In every market I'm familiar with, market participants pay a premium for superior features and vice versa. But to your scenario - I would agree - if the market paid $500k for 6 different houses that all had different elements of comparison - your argument would hold.

Is that why you believe the adjusted range isn't important?

While I see the benefit of hyperbole for the purpose of exposing a flaw in a hypothesis, I also think use of hyperbole can be dangerous if we try to normalize the hyperbole... it may lead folks to believe sensitivity analysis is meaningless (fore example).

 

[alebrewer]

That would not be the case in this scenario. The market does recognize a difference for these differences, but when adjusted up for GLA and down for site size, the net effect is not exactly 0.

BTW - you know this argument is logically flawed, right? Logically, it does not follow that (a) the market would pay a premium for a superior feature, and (b) two properties sold for the same price but had different features. Unless, of course, the 'net' difference between the features is $0.

 

[Joe Flacco]

How tight of an adjusted range are you guys typically working with?

I'm pretty happy if I can get to a 10% adjusted range.

 

[CGinMN]

I never said the adjusted range was not important. I already said I agree with you in most cases. You're asking about other cases, so I'm explaining. I can give you more examples too. I can probably drum some up from the real world. It happens not infrequently in rural and high-end markets.

BTW - you know this argument is logically flawed, right? Logically, it does not follow that (a) the market would pay a premium for a superior feature, and (b) two properties sold for the same price but had different features. Unless, of course, the 'net' difference between the features is $0.

You still don't understand, so I’ll give you a visualization to help you better.

How much would you tinker with sensitivity in order to tighten this up? Would you make the garage stall adjustment $12,400? Or would you reduce the GLA adjustment to $35/sf first? At a certain point, this is overfitting. Are you saying my adjustments are incorrect because they don't narrow the range? Sensitivity analysis is not meaningless. But may not be as useful in certain scenarios and can be susceptible to overfitting.