Appraisal Statistics

[DWiley]

one brand new example just posted https://appraisersforum.com/forums/threads/is-this-appropriate-to-add-to-the-report.220099/

That thread is about behavior that agents have been exhibiting for over 30 years, and it has nothing to do with statistical modeling.

 

[George Hatch]

One thing I think most of us would agree upon is that no analytical mode can work if it isn't actually performed. So pitting a [didn't analyze] with a [did analyze] doesn't produce a fair comparison.

 

[Terrel L. Shields]

Are you going to say that you don't carry a "base figure" in your head for adjustments...
And just slide the scale up or down to suit your subject and comps?
I doubt most appraisers are re-inventing the "wheel" of adjustments....

Hmmm. While I may be prescient enough to guess the final outcome, after all, these do fall into categories according to size, age, etc. I do sensitivity on each and every report unless I am ranking the sales (normally 5 or 7 sales.) and in that case I make no adjustments. Amazing...it's $30/SF. ...MH BTW

 

[J Grant]

Hmmm. While I may be prescient enough to guess the final outcome, after all, these do fall into categories according to size, age, etc. I do sensitivity on each and every report unless I am ranking the sales (normally 5 or 7 sales.) and in that case I make no adjustments. Amazing...it's $30/SF. ...MH BTW
View attachment 35460

Very interesting...do you find after the first round of sensitivity/ line item adjustments are made for several factors...that we might find ourselves adjusting the adjustments? Nothing wrong with starting with a base amount for first round of line item sens tesing one has to start someplace.

A one line item adjustment such as sf, usually is made along with some other adjustments and does not exist in a vacuum, it relates to the rest of the property. In your example, we don't know the age/condition or upgrades of any of the comps...are they similar, inferior, superior to the subject? . Comp 3, the smallest house sold for a higher $ per sf...why.? Is it more upgraded or newer or on larger lot? If it Superior, then after being adjusted down superior, re apply the $ per sf adjustment and it would narrow the adjusted range of the comps..

Per the above $30 or similar sf adjustment applied, comp 3 the smallest adjusts to a higher price, and the question is why would someone pay more ( $91, 820 ) for comp 2 a 1456 sf house then they would pay ( 486 940 ) for larger comp 2 house of 2052 sf ?

Unless comp 3 is superior relative to subject and needs adjusting down or comp 2 is inferior and needs adjusting up...either one would tighten the range and then the adjusted sales "make sense" Once the adjusted range is narrower and makes sense, the value of subject emerges...hatching from its very own egg...

 

[George Hatch]

If $30 obviously produces the most convergence in these adj value indicators, the averages of these other nearby adjusted datasets don't vary by as much as many people might think (like less than 1% with this dataset if you're using anything between $20-$45 per sf).

 

[Terrel L. Shields]

Very interesting...do you find after the first round of sensitivity/ line item adjustments are made for several factors...that we might find ourselves adjusting

Yes, you have to adjust for differences. The first and most obvious as well as critical is site value. Then any non GLA item, then age-condition. My Sensitivity analysis is actually two part. I adjust out land, etc. Then based on the observed effective ages I make the age-condition adjustment, which transfers to the SF adjustment for a final adjustment to SF. I avoid quality differences like the plague. You basically are balancing the Age-con and SF adjustments and it may take several trials to see what fits best. Sometimes if using 4- 5+ comps I may pull the 3 most similar out for sensitivity rather than use all 4+ sales.

 

[RCA]

Or, maybe diving too deep into statistics results in additional confusion. Don't get me wrong here, I am all for crunching numbers. Heck, I even like doing it. The thing is, and you have made this point for me in your posts, there is a serious reasonable doubt to the credibility of results when it comes to using statistics. My point is therefore, it is dangerous and reckless to promote their use too heavily, without also being up front about their reliability. This last part is truly the problem. You see it, I see it, plenty of appraisers see it but, do our clients see it? Do appraisers do a good job of communicating the credibility of their valuations to clients? I think the ugly answer to that is often not. The result is then, that clients believe these sorts of things are reliable, and then demand they be employed. The sales grid in lending assignments is the greatest example of this, and is even another degree further into the problem. The further problem is that Lending clients designed the forms and subsequent methods employed, not appraisers. Because these are the most widely used formats, most people are familiar with them and accept them as the norm. The result is pretty much everyone thinks adjustments are an accurate thing (dumb appraisers too), because too few of us point out that they are not. Why don't we speak up? LOL - Well, that goes right back to who designed the forms and most significantly, who pays for them to be used! USPAP says we must reconcile the methods and techniques used in the appraisal. It stops short of specifically saying we must communicate the reconciliation to our clients in our reports, however considering the rest of the USPAP, this ought to be a given (and considering appraisers like to split hairs when it comes to the meaning of words and text, it ought to be spelled out, IMO). When was the last time you reviewed a report that had a statement concerning the unreliability of adjustments on a sales grid? LOL - Pretty much never right? Instead, appraisers fluff it up with all sorts of nonsense, including citing the use of statistics. Appraisers like to complain about all sorts of problems from the client side. I think this is a case where appraisers ought to take a look at their own behavior and make some changes.

So, I'll end where I started. Statistics can be useful, but it is not a good idea to act is if they are the solution to the problem of adjustments, when all they really are is an available tool that works once and a while. I think you misrepresent the use of the word "many", especially when you admit there is no way to test the reliability. IMO, the word "few" is more appropriate. But again, I am open to being wrong.

And because I like to write too long I'll add this, we are called Independent Fee Appraisers, not Value Calculators.

Hmmm. I think I have some bad news for you. It's this.
1. Assuming you know all the features in a subject and its comps that contribute to value and
2. Assuming that you have accurate data on the features as either numeric amounts or string values (e.g. "C1", "C2", "C3", ..., for condition).
3. If you can find a set of adjustments that bring all adjusted comp sales prices to the same amount.
4. Then you have an "accurate" value. You have the one and only value.

Now, understand, once you have all adjusted sales prices equal, you can go back into your numbers, twiddle around and find that you can raise some adjustments and lower others and still get the adjusted totals to be equal. Adjustments can offset each other. But, you will wind up with the same adjusted value!

So, in essence you have an accurate price estimate if you can get all your adjusted totals equal, given assumptions #1 and #2 - and you have all the relevant comps for the subject.

Of course, nowadays assumptions 1 and 2 are rarely true. Yet, we keep getting closer. The AVMs keep getting closer. The appraiser still has the advantage that he has feet on the ground and the "human factor". Honestly, it will be some time before the AVMs get close to that. Yet, the actors who determine price, the seller, buyer, broker, agent, underwriter, appraiser and so on, have their limitations and tend to simplify things to match their own simplified opinions of value.

But, the point is, if 20 appraisers appraise the same house, using the same criteria (features) and data, then we can argue that the appraiser with the closest set of adjusted comparable sales prices has the most accurate value.

And, we don't need to concern ourselves with all features, but only those that are important across a significant number of potential buyers. For example, if there is only one potential buyer interested in black everything, then who cares?

So, conclusion, if I use my MARS or other regression tool and consistently pull in tighter comps than another appraiser, then I have an advantage. That's going to be a fact that the reviewers are going to have to contend with despite other things in the report they might disagree with. Also to note, if I can keep the rational for my adjustments out of the report and away from the purview of the AVM companies, so much the better. But of course, you might not get away with that. - Depends on intended users ( Read Ted Whitmer excellent article "Applying USPAP Fairly in Review" in the latest "Working RE".)

 

[glenn walker]

Inspected a property last week - Asked the agent how she arrived at her price and she handed me a list of 6 comparables and said well it's kind of simple I total the sales prices and divide them because I found appraisers don't understand buyer and seller motivation and my method automatically factors this into the equation.The interesting part was I worked on the appraisal and after making all the typical adjustments my final point of value was almost the same as her listing and sale's price except my adjusted sales were about $10,000 higher. On a $500,000 house a 2% variance is pretty tight. I know we are not supposed to average but I have to admit in this case her value was spot on: )LOL

 

[RCA]

Inspected a property last week - Asked the agent how she arrived at her price and she handed me a list of 6 comparables and said well it's kind of simple I total the sales prices and divide them because I found appraisers don't understand buyer and seller motivation and my method automatically factors this into the equation.The interesting part was I worked on the appraisal and after making all the typical adjustments my final point of value was almost the same as her listing and sale's price except my adjusted sales were about $10,000 higher. On a $500,000 house a 2% variance is pretty tight. I know we are not supposed to average but I have to admit in this case her value was spot on: )LOL

Experienced agents have a lot going for them and are an important influence on price. I have often considered running stats on agents - because I am convinced they do have a large influence on price. You know you can' t use the stats in the report, but agents are often important. Good agents will work with a client to sell their house at an acceptable price, and if it doesn't sell, then tell them how to upgrade at a reasonable cost to stand a better chance of getting the desired sale price. Then maybe a year later, after upgrades are completed, that same agent sells the house for the higher price. All that on-the-ground experience adds up. And woe to the appraiser who doesn't seek input from reliable and experienced agents.

However, as part of the valuation picture, agents are probably best considered as "confounding" variables - they alter the relationship between market reaction to home features and the sale price. They can suppress certain information (if they feel they can get away with it), emphasize good points to make them seem more important than they are, give advice on staging and advertising that affect perception, and so on. So, then, the question is whether a good agent lends value to a home. You might agree that they can add to the sale price; otherwise ... if we could only make an adjustment for the sales agent ... but we can't ... she/he might belong to a group that is considered minority ... if it were even allowed for other reasons. Not that a collateral analyst really cares ... they want to get a good idea of what the house is likely to sell for in the future, and I suppose they might figure they could always find a "good" agent.

Buyer and seller motivation affects homes on individual basis and should not affect market value. It appears you are focusing on one sale here, as well. So, given that you are being vague, I can't really connect your dots (what does 2% have to do with the sales agent?), not much to say without making a lot of assumptions about what you might mean.

 

[Artemis Fowl]

Experienced agents have a lot going for them and are an important influence on price. I have often considered running stats on agents - because I am convinced they do have a large influence on price. You know you can' t use the stats in the report, but agents are often important. Good agents will work with a client to sell their house at an acceptable price, and if it doesn't sell, then tell them how to upgrade at a reasonable cost to stand a better chance of getting the desired sale price. Then maybe a year later, after upgrades are completed, that same agent sells the house for the higher price. All that on-the-ground experience adds up. And woe to the appraiser who doesn't seek input from reliable and experienced agents.

However, as part of the valuation picture, agents are probably best considered as "confounding" variables - they alter the relationship between market reaction to home features and the sale price. They can suppress certain information (if they feel they can get away with it), emphasize good points to make them seem more important than they are, give advice on staging and advertising that affect perception, and so on. So, then, the question is whether a good agent lends value to a home. You might agree that they can add to the sale price; otherwise ... if we could only make an adjustment for the sales agent ... but we can't ... she/he might belong to a group that is considered minority ... if it were even allowed for other reasons. Not that a collateral analyst really cares ... they want to get a good idea of what the house is likely to sell for in the future, and I suppose they might figure they could always find a "good" agent.

Buyer and seller motivation affects homes on individual basis and should not affect market value. It appears you are focusing on one sale here, as well. So, given that you are being vague, I can't really connect your dots (what does 2% have to do with the sales agent?), not much to say without making a lot of assumptions about what you might mean.

I've actually done a simple/quick-n-dirty-type agent analysis in the past. Guy trended almost 15% higher SP on his listings vs. similar.