More NCAB

[Austin]

Lee Ann: I think I understand what you saying, and if I do you have it exactly backward. My market is one of the most difficult markets in the country to appraise in for reasons you just explained. We have no cookie cutter subdivisions, our Realtors can’t measure, one blocks difference in location can put you in another world, you can literally drive down any street in this city and never see the same house plan used twice. It is amazing there this could happen, but that is the way it is. That is the very reason I was forced into regression methods and thinking. You can’t measure or detect these differences by any other method. To measure differences you need a huge number of sales and you need a diversity of data and certainly not perfect comp data as I described above. To measure these differences and spot all of these differences you have to have a system to sort all of this out, and using clone comps is not the system and the present sequence of adjustments in this situation is not even rational. Try it before you condemn it. After you finish, you can verify the accuracy of your appraisal by applying the regression formula and predicting the price of the comps to test the accuracy of method. If the regression formula predicts a price close to the comparable sale prices; what better verification can you ask for? With the present sequence of adjustments you could end up on another planet.
As to your comment about bad appraisers: Who was it that set the qualifications, tested, and licensed these crooks in the first place? The answer is the state appraisal boards. My response is if you want reform, why not go after the idiot boards that obviously dropped the ball and allowed these people to be licensed in the first place. You can’t expect the ones that caused the problem to admit they screwed up and then correct their mistakes.

 

[bradellis]

Austin,

OK, so I read all your stuff about regression analysis. It takes data to produce it- the more, the better. And I know enough about it to know that the more odd balls that go in, the more instances of deviations. Will it seek the correct central tendencies? Sometimes. Sometimes not. Mathematical crunching relies upon either a) a large data base, or b) less, but better data to be reliable.

Now, perhaps your personal skills are so great that you can apply statistics and make them more reliable than do the thousands of assessors out there whose value opinions are off the mark routinely- and not always because they do not have the relevant characteristics. If so, good for you.

But then, you chastize me for saying that a single example of board abuse- if that is actually the case, and we will find out, is adequate to jump to the conclusion that the board is a rogue board. This is directly contrary to all the other stuff you said. Why not apply regression analysis to this question and see what you come up with?

Can't let you have it both ways. Or doesw your belief in your techniques surface only when it is convenient?

Brad

 

[bradellis]

Steve,

I've been looking for an old version of USPAP but can only find one back in 1997. I did this to try to see if there was something differrent as any case against an appraiser must be prepared using the version of USPAP at the time of the assignment(s).

There are actually a couple of areas that could apply (from 1997 USPAP- I have not cross referenced it to today as it would be irrelevant). Here is my take on one of them, for what it is worth:

SR1-1(b) ...error of omission or commission...
Comment:.......an appraiser must be CERTAIN (emphasis added) that the gathering of factual information is conducted in a manner that is sufficiently diligent to ensure that the data that would have a material or significant effect on the resulting opinions or conclusions are CONSIDERED. (again, emphasis added).

Now we come to the crux of my argument- it may have nothing to do with your case, so consider it an academic approach. I contend that, without knowing what data is out there, an appraiser cannot either be "certain" or even know if he "considered" the effect upon the resulting opinion.

Given this, you, Tom and others may well be right about not having to consider ALL the data. so I'll grant you guys this. As to the specifics of your case, I will hold my ground.

So, the question is not whether or not the comps used in an assignment are the best, it is over what level of research is sufficiently diligent. And that is where we have a difference of opinion. For that type of group assignments you say it was, I say it was not. But please do not take this as a personal condemnation. I am well aware of what it takes to produce credible appraisals and am sure that this is your normal output. NOT commenting upon the specific appraisals in your case- only over methodology. And I am forced to raise it because you claim I am out there making this all up. I am not.

As to the NCAB case, you posted:

"Case in point, I read a transcript David Johnson sent me. The NCAB was arguing a man used the wrong type of depreciation in the cost approach. However, the guy used a cited cost manual, which has recognized authority in the industry. This type crap, is inappropriate behavior for Boards. Boards should not be able to fine, suspend, revoke, license because they preferred one recognized method over an other. This is abusive"

I fully agree that a board should not be capricious in this way. What I do not know from this case, and I'd suggest that you also do not know, is what type of asignment this was. if it was a commercial property and the appraiser used the depreciation chart in Marshall & Swift, that may be the appropriate technique. However, if this was for a single family assignment in which the client asked for a Fannie Mae compliant report, then using the chart would NOT be correct. Fannie Mae requires the use of the effective age/life method at a minimum and does not recognize the chart.

In the first instance, the board may be wrong. In the second the board may be right, since the requirements of Fannie are supplemental standards. Not going into whether or not the chart is right or wrong- only over what has been publically stated by Fannie.

It all depends upon the facts, and we do not know them.

Brad Ellis, IFA,RAA

 

[Stephen J. Vertin MAI]

Brad:

You say:

"I am well aware of what it takes to produce credible appraisals and am sure that this is your normal output. NOT commenting upon the specific appraisals in your case- only over methodology. And I am forced to raise it because you claim I am out there making this all up. I am not. "

Help me, I do not understand what you are talking about. Please clarify.

Brad, this argument remains me of the following:

a) 1 + 2 + 4 + 7 + 1 + 2 + 5 + 3 + 6 -8 -9 = 14
b) 7 + 7 = 14

I believe (a) is a superior method in calculating 14 because it takes into consideration all of the numbers 1 through 9. Prove me wrong.

Steve Vertin

 

[Austin]

Brad: In reply to point # 1 about regression methods. Think of regression as a logic system to arrive at the truth. As to oddball or outliner properties: There are many steps in the regression process. The first step is to filter the data and get about 30 of the most comparable sales. Second: I graph the GLA vs. the actual sale price at that stage to spot the outliner or oddball properties. Once you see the graph and discern the pattern you know where you stand. If the cluster forms a highly linear line, you have an easy job and you know exactly the gravity of the value influencing factors. If there are sales out of the pattern, then you have a basis of investigating why they are out of line. There is always a reason. In the beginning I use to call Realtors or sellers to find out why, and there was always a reason for outliners. Examples I have heard: Well the seller was going into a nursing home and the government was going to get the money anyway so I told them to put a fast sale price on it or some such story. There was always a reason, so now, I just remove the outliners and filter the data down to a reasonable range. Then I do a regression one step at the time starting with age, GLA, total value of extras, etc., and at each stage check and verify what is going on by predicting the price of the comparable sales and checking the variance between the actual and predicted prices. The measure of success is when the total variance between actual and predicted prices drops sharply, you have just identified a significant value influencing variable. The key is using the correct sequence of adjustments. Common sense tells you that you have to solve for the physical differences first and anything else like view, time, etc., are residual in nature. For example, if you use 20 sales and one has a pool or beautiful view, and after you solve for the physical features the price is right in line with the other sales, you know the pool and view had no effect on price. If you use the accepted sequence of adjustments and assume the pool was worth $15,000 and the view $5,000, you screwed up. You have to work from within the data set and can’t import personal or even supported opinions of what else you “feel” affects price. Every factor is totally supported and in total harmony with the cost approach. Wouldn’t it make sense that if you put the property together one piece at the time in the cost approach that the same pieces would logically be the same value influencing factors that affect price in the sales comparison approach?
Now as to the tax assessor’s method: There is a good reason their methods and AVM’s produce wrong answers. For example, I have found my market contains five distinct property categories. Historical or classical style is level 1. Houses built between 1920 and 1945, are group 2. From 1945 to 1960, is group 3. From 1960, to 1990, is group 4. And after 1990, is group 5. Each group is distinct with a unique style, quality of construction, location, etc., and if you mix the groups selection, you will get the wrong answer. You can’t estimate the total population of properties accurately for this reason. You have to research and identify the value influencing factors. That is where the data base engineering comes into play. If I am appraising a group 3 property, I filter the database and use only group 3 properties. Another related factor in this regard is that I have found that group 2 properties have a flat GLA trend line, meaning size does not affect price. The price is driven by condition and location. Realtors really screw up when they price properties in this category because they always assume bigger means a higher price when actually it means a lower price. Again, you have to follow a logical progression to find out what factors affect price.
As to your analogy about regression and the OREB. We are talking about a board and not a group of boards and talking about people and not value influencing variables. If I committed a crime, would the police wait until I had committed 5 crimes and establish a pattern of criminal behavior before they took me in? The answer is no. There is no basis for your analogy. One crime makes a criminal but one outliner does not represent the market.

 

[Fred]

Austin, you write
"Think of regression as a logic system to arrive at the truth"

Some folks seem to think that truth is relative.

Steven Santora, IFAS

 

[bradellis]

Austin,

Your quote:

"Common sense tells you that you have to solve for the physical differences first and anything else like view, time, etc., are residual in nature."

REALLY?

What about location? And time is residual? I'll give you a little example of time:

Had my house appraised in January for a refi. One September sale of a two story across the street was at $236,500 (not comparable to mine- I have a one story, but just looking at the data). Duplicate model of that one sold a block away in December for $249,900. Seen them both and they are a match- right down to features and condition. (I love open houses). On the way back from the granite fabricator yesterday (refi to remodel the kitchen), I saw another duplicate open house and stopped in. It was already under contract after 24 hours, but the broker had already committed to the open house and could acept only back up offers. While she did not reveal the contract price and it obviously did not close yet, she did tell me that they got close to the asking price. Asking price was $284,900, and it was in truly poor shape- roof leak, settlement obvious from just walking the second floor, no updating at all. Built in 1979 like mine and all of them.

What is going on, is it time? Yep, but with a twist. You see there is a major freeway extension opening end October about 6 blocks away. Because of BOTH the time and location, values are escalating rapidly.

I saw this happen in Chicago some 10 years ago when a major tollway opened. Knowing that, I made my purchase in November, 2000 for $216.5, knowing about the freeway.

Yes, I will ackowledge tht regression could isolate this, if it were due to either one or the other. If you could isolate the location factor, and you would need to have homes right ON the freeway unaffected by time. Or conversely, isolate the time factor to get the residual location factor.

But, in this case, there is not adequate data. Only 6 homes have sold in the neighborhood since I bought.

Regression is a fine tool, but it is not the be all and end all. It takes skill as well, both in knowing how and WHEN to use it and knowing when it will not yield the correct results. In this case, you do not have your 30+ data sets.

Austin, the simple fact of the matter is that regression as a stand alone tool is simply another tool. In your market, you have been able to classify homes to refine the analysis. But it does not happen all the time. If it did, no one would need us at all. The computers massaging the numbers would conclude appropriate values and we'd all be out of work. Neither is regression anything new.

Why do most appraisers not use it? I do not know. It is a legitimate tool. Could it be that by the time you have done all the analysis, it might not yield any better results or might take just as long? I guess that would be the case sometimes. In my market, I can buy an already populated data set, run the regressions, even choose the data to go in (ignore some and include others). It only works sometimes. Perhaps your market is much more well defined.

As to your analogy on the OBRE, I agree that if one commits a criminal act then they are a criminal. But that was not your contention. Your premise is that they are a rogue board because of one case in which there was a difference of opinion. I say opinion, because we have 2 past presidents of the AI chapter disagreeing over whether or not the board's action was appropriate. The court will decide that.

But in the absence of any other data, this is simply a leap of faith on your part. Looking at adequate amounts of data will yield different results.

Brad Ellis, IFA, RAA

 

[bradellis]

Steve,

You are way too transparent. Let me respond in kind.

I'll make you a bet- White Sox vs. the Cubs. You can have the White Sox- a better record. I'll say that if I ADD all the runs the Cubs score in each game that the total will be greater than if you multiply all the games scores for the White Sox.

Now, I know you will not be taking that bet because if the Sox get shut out even once their total will end up being zero and you'd lose. I am sure you saw that right away.

So to your numbers: If I say the longer string is better because it considers all the numbers, you can say that the 2 number string yielded the same result; therefore, it is not the number of pieces of data but their accuracy or relevance. Granted.

But my question remains, how do you KNOW the 2 numbers are the more relevant OR more reliable? If you did not consider the other number BEFORE you tossed them, how could you know they were not critical?

Brad Ellis, IFA,RAA

 

[Austin]

Brad: I hate to belabor a point, but you keep making and re-enforcing my points for me in your attempted rebuttal. You start off by saying I am out to lunch because I advocate solving for physical differences first and then attack me for using regression, which is what the sales comparison approach is. Then you give a long example by first explaining that the physical features of your house and recent sales are identical (solving for physical differences first to demonstrate the residuals) so the price increase must be due to time or a change in supply and demand. Isn’t that doing exactly what you just accused me of doing? If you hadn’t established the fact that there were no physical differences in the properties being discussed first, your example would have had no meaning.
The existing sales comparison method is a crude unsupported form of regression analysis. Each accepted approach to value is a regression (equation). The entire appraisal process is a form of regression. Regression is an equation that explains the correlation between value influencing factors and price. Why is doing it correctly by computer such a problem for you?

 

[Stephen J. Vertin MAI]

Brad:

I do not know what you mean by "I am to transparent"?. Please clarify. Furthermore, you never clarified your above comments I did not understand.

Additionally, I do not understand your baseball analogy. I do know you are looking at my presented equation one dimensionally, because you have presented it back to me one dimensionally. The problem I presented has more to it than just the symbolism of the number of numerators in the equation. It also has the results from simple addition and subtraction (which is what the sales comparison approach is all about after market supported adjustments have been determined) and the never ending circular argument (my method is better than yours, when in fact, in theory, the results are the same). As Austin has politely pointed out, you are approaching the problem backwards.

You have been reduced to providing an advisory opinion out of some 1997 USPAP edition that does not even come close to your provided opinions. In fact your first stretch of USPAP had more credence than your second. You, for some reason, which has never been explained, believe unadjusted data is the key to the sales comparison approach. You also seem to believe that you can twist USPAP to say this. Furthermore, your underlying argument hints, that if we, as appraisers, looked at the same sales we would all choose the same comparables. Or that some how sales selection is USPAP dictated. With these premises you have come to the conclusion that the only way to come to a reasonable value conclusion is by looking at all the data in the market. Brad, three people have told you it is never possible to see all the sales. We have all given you reasons why. Furthermore, there are probably 100 more reasons appraisers in non-disclosure States can tell you. USPAP never said this and never will. Real estate is an imperfect market and as all economist will tell you one of the defining issue of an imperfect market is imperfect information. You have been told the mathematical facts of sampling and you keep on insisting or requiring us to validate your reasoning.

Steve Vertin