Principle of Substitution, ....

[RCA]

And yet, those properties get appraised and the users make their decisions based on those appraisals all the time. Isn't that weird?

The thread is about the legitimacy, and correctness of the Principle of Substitution - and the associated technique of matched pairs. It is about the correctness of the assertions made by AI manuals and courses supporting the Principle of Substitution and matched pair analysis. It must be an important issue - as the AI reportedly spends hundreds of thousands of dollars on lobbyists to promote these concepts and techniques as reliable and legitimate.

Buying decisions occur before appraisals are generated (in most cases). Buyers rarely depend on appraisers to make decisions. - The users you are referring to are mostly banks and lenders - just looking for a rubber stamp, E&O insurance, a fall guy - and an appraisal that will not get them into trouble. -> Per the posts in this forum going back many years.

So your assertion that "users make their decisions based on those appraisals all the time" is really this: Lenders decide to accept appraisals if they meet certain standard requirements, e.g. having the right checkboxes checked or not, the value does not cause a problem for them, the wording does not create red flags for issues such as bias and so on. Fly-by-the-seat adjustments for unmeasured characteristics are generally accepted if they are not too large or small or contradict other things in the report.

Now, clearly, I am not too concerned about what gets accepted by the Lenders or the GSEs - the most common and critical users. Of course, they need the rubber stamps to function. They have moronic processes that the majority just simply work their way through without much if any critical thinking. Again, this forum pretty much supports that point of view.

But suppose you wind up in court with a lawyer intent on ruthlessly undermining your appraisal. How far can he get? I would say, almost all appraisers are skating on thin ice. Even the best.

You are probably not worried. You would just be very bland and state in court that you are just doing things the same as your peers. And you would probably be right. At least in the lower courts, a good show is more important than real substance.

-- So yes you are right that I can't expect a very "robust" discussion on these topics on this forum. But what little discussion is provided is a nonetheless good exercise.

As far as Dell - I have talked with him and been to one of his online classes. They are somewhere between where I am and the members on this forum are - at least last time I checked. Dell agrees it appears that the AI is retarded, although he doesn't say that exactly

... But then, at least as of last summer, - he really isn't into non-parametric statistics or MARS - and for me, that is the starting point for discussion.

 

[Surf Cat]

Dude.... either you're a genius or you belong in a straight jacket.

 

[Dublin ohio]

Dude.... either you're a genius or you belong in a straight jacket.

Intelligent and educated. But far from a genius. But I will give him credit for trying to reinvent the wheel. Even though the new wheel will have limited acceptance. Since it will cost 3 times as much and in reality. Nobody will want it because they can get a cheaper product that has wide acceptance and serves the same purpose

 

[George Hatch]

As far as Dell - I have talked with him and been to one of his online classes. They are somewhere between where I am and the members on this forum are - at least last time I checked. Dell agrees it appears that the AI is retarded, although he doesn't say that exactly

... But then, at least as of last summer, - he really isn't into non-parametric statistics or MARS - and for me, that is the starting point for discussion.

[shrugs] Someone has to be first. And if you believe you have no peers then I think that's great.

As for me, I have never considered myself to be some sort of thought leader or original thinker in this business, so I've got no dog in that fight. I'm just another lone gunman, toiling away in my mom's basement, same as most appraisers. In my view competent appraisal practice comes down to the fundamentals. Of being first and foremost honest in my handling of the facts and objective as possible in developing my analyses, opinions and conclusions. To finding different ways as may become necessary to apply the underlying principles and concepts to the appraisal problem at hand. I aspire always to write for my users, not for myself.

So far that has served me well. To my knowledge I have never come close to getting sued. But, maybe tomorrow will be different.

 

[Joe Flacco]

Anybody that believes that a larger set of less similar properties is more reliable than a very small number of very similar properties is incompetent.

 

[TerryRohrer]

Anybody that believes that a larger set of less similar properties is more reliable than a very small number of very similar properties is incompetent.

How do you measure that reliability? And what do you do in the cases where there is not one very similar property?

 

[Joe Flacco]

How do you measure that reliability? And what do you do in the cases where there is not one very similar property?

I would probably rank. It's better than this and inferior to that.

 

[George Hatch]

I think data qualification is the thing. That's what enables the use of the smaller datasets, both in terms of ranking and applying those "not-used-by-the-market-participants" line item adjustments . Data qualification is what we spend most of our SR1 time/effort on. IMO

 

[Mejappz]

Ignorance is bliss.

 

[George Hatch]

A few of you might remember the ZAIO wars way back when. As an appraisal application their plan had some good points. The appraiser would personally qualify each of the datapoints prior to its usage, then they'd build however many models they needed for a market area including the development of their adjustment factors for each market segment. That appraiser-qualified data was then fed into their AVM. As I recall they had protocols for model testing and calibration to refine those models.

There were other aspects of their program that were wholly unfeasible and unmarketable but those are arguably separate issues that had little/nothing to do with the analyses, opinions and conclusions of the appraisers who were signing off on those results.