Deep Dive - The Cost Approach

[Terrel L. Shields]

you only need to know contributory value

Really? So, this one I'm trying to finish up with a 80x40 shop, a 40x24 open shed and an identical enclosed shed with concrete floor, and 3 kennels, plus a 3,300 SF house with a basement garage in a county that has only 4 other properties on 100 or more acres PLUS 1 property that transferred from father to son, I can somehow extract contributory values from those 4 sales and maybe 12 other sales over the past 5 years in it AND the adjacent county. So, how far away do I go? And how valid is the contributory value extracted from distant counties? And I am bumping up against the state line of Kansas which is a non-disclosure state. Cannot confirm a thing there. A mere $100-200 difference in land value skews your analysis to boot.

 

[alebrewer]

Really? So, this one I'm trying to finish up with a 80x40 shop, a 40x24 open shed and an identical enclosed shed with concrete floor, and 3 kennels, plus a 3,300 SF house with a basement garage in a county that has only 4 other properties on 100 or more acres PLUS 1 property that transferred from father to son, I can somehow extract contributory values from those 4 sales and maybe 12 other sales over the past 5 years in it AND the adjacent county. So, how far away do I go? And how valid is the contributory value extracted from distant counties? And I am bumping up against the state line of Kansas which is a non-disclosure state. Cannot confirm a thing there. A mere $100-200 difference in land value skews your analysis to boot.

You keep providing examples of the convoluted BS western AR properties that you have to appraise, Terrel. In the words of my grandma - bless your heart. And I mean that sincerely. Back to my point, though, your examples don't address the lack of validity of the CA in residential appraisal. I still contend that you have no idea if your depreciated cost estimates reflect true market contributory value sans sales data, but I also understand you have to provide 'something' to show some kind of value. Even though - in my mind - it's kind of made up.

 

[CGinMN]

Historically, it has been acceptable that from time to time, when the data are too scarce to be able to quantify the market reaction to a feature, appraisers are able to use their professional judgment. I think because of the "appraisal bias" movement and because of the quants in charge at the GSEs, we're moving away from that. In many cases it is good, because lazy appraisers shouldn't be relying solely on their judgment for things that can be quantified in a number of ways. However, I think we take it too far when we dismiss approaches all together because we might not have perfect pairs to extract depreciation. It doesn't mean it's made up. The appraiser's judgment and reconciliation are worth something. If we buy into the argument that it is not, then what differentiates us from AVMs and data scientists who don't have local market knowledge?

 

[Joe Flacco]

Cost approach info can be useful in that it shows how cost lines up with adjustments in the sales comparison approach.

I only develop the cost approach for new or recent construction. For older detached single family homes I will develop the site value.

 

[Joe Flacco]

Even with only the site value developed for older homes, that provides a pretty good rough estimate of the contributory value of improvements. Then the reader can have an idea on the reasonableness of some of the adjustments in the sales comparison approach.

 

[alebrewer]

If we buy into the argument that it is not, then what differentiates us from AVMs and data scientists who don't have local market knowledge?

If, at the end of the day, the AVM estimates prove to be more valid than appraisals/appraisers, which should be utilized? And how do we quantify validity of an appraisal estimate? Subsequent sale(s) of the properties. Just read an article on AVM's about how much of a confirmation bias is introduced when list prices are incorporated into the AVM metrics - very interesting. Anywhoo - my question really is - if quantification of the elements of comparison leads to better results, then judgment shouldn't be part of the equation anyway - not saying it is, but certainly worth considering.

I think because of the "appraisal bias" movement and because of the quants in charge at the GSEs, we're moving away from that.

On the contrary - I've a sneaking suspicion that the CA may become the new golden child for the bias mongers. You can't really manipulate a SCA - the sales are the sales. You can, however (theoretically), manipulate a CA to account for neighborhood price differentials without it being so blatant.

 

[CGinMN]

If, at the end of the day, the AVM estimates prove to be more valid than appraisals/appraisers, which should be utilized?

In the instances where an appraiser would use depreciated cost for an item that suffers from PO and FO, which seems to be the example you keep going back to, AVMs typically aren't even attempting to estimate the contributory value of these features - barns, outbuildings, pools, etc. The argument that AMVs might even be more valid than appraisers is odd given the context of this conversation. You don't like the CA because it will never provide enough proof to you (educated reader) that it accurately quantifies market behavior. Yet, an AVM is not held to any such standard. So are you concerned with the final result, or how it was attained? If it is the final result, then why do you care how it was attained? If it is how it was attained, then I'd ask, is how the AVM is dealing with these variables (likely omission) more credible or accurate than deprecated cost + expert judgment?

And how do we quantify validity of an appraisal estimate? Subsequent sale(s) of the properties.

The SCA model is not any more intrinsically related to the subsequent sale price than is the CA model or IA model. We can quantify the predictive value of AVMs and all of these approaches using sale price as a measure. It doesn't mean sale price and sales comparison have a special relationship that the other approaches do not.

Anywhoo - my question really is - if quantification of the elements of comparison leads to better results, then judgment shouldn't be part of the equation anyway

This is wrong because it assumes you can only quantify something one way and get only one answer. If you quantify something multiple ways, which I do all the time, you have a range of outcomes from which to reconcile, and judgment plays a role.

On the contrary - I've a sneaking suspicion that the CA may become the new golden child for the bias mongers. You can't really manipulate a SCA - the sales are the sales. You can, however (theoretically), manipulate a CA to account for neighborhood price differentials without it being so blatant.

Well, thus far that has not proven to be the case. Their golden child has been SCA. And "location" is just as relevant in the SCA as it is in the CA. Besides, whether an approach is more susceptible to fudging has never been a factor in determining whether or not the approach is applicable to an assignment or, when applied correctly, capable of producing credible assignment results. Nor should it be. Besides, I would disagree with the assertion that you can't really manipulate the SCA. Many MAIs make their living manipulating outcomes, and they would not be very successful if they just used the cost approach.

 

[alebrewer]

In the instances where an appraiser would use depreciated cost for an item that suffers from PO and FO, which seems to be the example you keep going back to, AVMs typically aren't even attempting to estimate the contributory value of these features - barns, outbuildings, pools, etc. The argument that AMVs might even be more valid than appraisers is odd given the context of this conversation. You don't like the CA because it will never provide enough proof to you (educated reader) that it accurately quantifies market behavior. Yet, an AVM is not held to any such standard. So are you concerned with the final result, or how it was attained? If it is the final result, then why do you care how it was attained? If it is how it was attained, then I'd ask, is how the AVM is dealing with these variables (likely omission) more credible or accurate than deprecated cost + expert judgment?

So my point was that 'appraiser judgement' has, for too long, been an acronym for 'I'm either too lazy or incompetent to do the right thing, so I'm gonna wing it' (present company excluded of course). And IF it turns out that AVM quantitative modeling is more sound (as measured by subsequent sales prices) than 'appraiser judgement', then it seems to me the whole 'appraiser judgement' argument just vanished like a **** in the wind.

The SCA model is not any more intrinsically related to the subsequent sale price than is the CA model or IA model. We can quantify the predictive value of AVMs and all of these approaches using sale price as a measure. It doesn't mean sale price and sales comparison have a special relationship that the other approaches do not.

Sorry for the confusion there. Wasn't saying that subsequent sales price is tied to the SCA - was saying that is the only way to validate the accuracy of a predicted value - subsequent sales price. Want to know how accurate a CA was? Compare it to a subsequent sale of a property. Same with validating an SCA.

This is wrong because it assumes you can only quantify something one way and get only one answer. If you quantify something multiple ways, which I do all the time, you have a range of outcomes from which to reconcile, and judgment plays a role.

There are, indeed, often several ways to get to the right answer. Sometimes you don't get the right answer, though, in which case the method used would not have been an appropriate method.

Their golden child has been SCA. And "location" is just as relevant in the SCA as it is in the CA.

Don't disagree with the theory here - just the application. Yes - theoretically location influences CA just as much as SCA. Difference is that, to manipulate the 'location' weight in an SCA, you'd have to blatantly cheat - you'd have to use sales from a different market. All's ya have to do to manipulate the CA to account for 'locational' differences is just throw in a couple more 'location factors' to even things out a bit. Much more subversive. As I said, though, at this point it's only a suspicion. Who knows how they'll end up solving for that.

 

[CGinMN]

So my point was that 'appraiser judgement' has, for too long, been an acronym for 'I'm either too lazy or incompetent to do the right thing, so I'm gonna wing it' (present company excluded of course).

Well in this case your beef is not with judgment. Judgement, when applied correctly, is useful.

And IF it turns out that AVM quantitative modeling is more sound (as measured by subsequent sales prices) than 'appraiser judgement', then it seems to me the whole 'appraiser judgement' argument just vanished like a **** in the wind.

Not necessarily so. How are we setting up this test? Are we averaging the results of every AVM across the country with every appraiser across the country? What happens when we limit it to the best performing appraisers vs the best AVMs? If the AVM is beating the appraiser, it is not because the appraiser used judgement in applying depreciated cost to develop some adjustments, it is because the carrot goes to the skippy and the stick goes to the appraiser who does the work. The principles and procedures of appraisal, when credibly applied, will always surpass that of AVM. If they aren’t, then they aren’t being applied correctly.

Therefore it’s not judgment, it’s not the CA, it’s incompetence and laziness, and I would say those are the inescapable results of the current residential appraisal marketplace. IMO

 

[alebrewer]

Well in this case your beef is not with judgment. Judgement, when applied correctly, is useful.

Not necessarily so. How are we setting up this test? Are we averaging the results of every AVM across the country with every appraiser across the country? What happens when we limit it to the best performing appraisers vs the best AVMs? If the AVM is beating the appraiser, it is not because the appraiser used judgement in applying depreciated cost to develop some adjustments, it is because the carrot goes to the skippy and the stick goes to the appraiser who does the work. The principles and procedures of appraisal, when credibly applied, will always surpass that of AVM. If they aren’t, then they aren’t being applied correctly.

Therefore it’s not judgment, it’s not the CA, it’s incompetence and laziness, and I would say those are the inescapable results of the current residential appraisal marketplace. IMO

Touché. Actually I don't even have a beef with the CA - I just think it's rather pointless in the residential world (well, in any world - but, again, I understand there are situations where its unavoidable). On the hypothesis that the act of appraising (when credibly applied) will always surpass that of an AVM - that is nothing more than conjecture. AI will overtake the appraiser - its just a matter of when (I know - that's conjecture also). As I see it, one of the biggest advantages that AV Models have over appraisers is the ability to 'wipe the slate clean' for each assignment. Like it or not, appraisers will always introduce some form of subjectivity into their analyses - that's part of being human after all. One appraiser believes the market acts predominantly on PPF - and his/her analyses will reflect that bias. Another appraiser believes year built isn't relevant - only quality and condition - and his/her analyses will reflect that bias. Maybe the bias affects the results and maybe not - but the AV model doesn't have that bias.