Square footage adjustments

[Mike Garrett RAA]

That is the very reason appraisal is an "art" not a science! It never will be a science because "buying" is emotion, not logic!

 

[Fred]

Slacker,
The only thing that Mike did not tell you is which market that works in.

Austin,
Why not just take the data the way it comes instead of trying to force it into pre-conceived market hypotheses.
You write <span style='color:darkblue'>"…never in my appraisal career have I even seen three sales that ranging in size from 2,500 to 3,200 square foot after adjustments having the same unit price per square foot."</span>
I have answered any number of class, test and practice problems - and work assignments - that were nothing like anything I had seen before.

In the givens of the problem, I told you flat out it was a hypothetical "perfect market" in which the improvements for all three sales had exactly the same square foot contributory value, but the land values were different. I really do not know how to respond to the theme of your last post that which seems to be that if your estimated regression line does not fit, then the data must be wrong.

You also write,<span style='color:darkblue'> "That formula you cited gives me a clue though: Y=$92.5x + LV. Where did that equation come from? I have never seen anything like it before. The LV factor in your equation represents the X intercept."</span>

You have never seen anything like that before???"
"Y," the value of the property, is equal to the land value plus the "depreciated" or "contributory" value of the improvements.
That formula is within every improved property. Some folks might even refer to that formula as the cost approach. :!:
How about if I say Land Value plus Building Value (at $92.50/sq.ft.) is Total Value? Same formula.

If the market price of a property is: $92.50/sq.ft. plus land value;
and the cost new of the property is: $100/sq.ft. plus land value,
would not the $7.50/sq.ft. difference be the "accrued depreciation?"
Still never seen this "formula"??? :roll:

Thanks for the chat! I enjoyed it.

 

[Austin]

Interesting new article on this subject just published in the AI’s new online Appraisal Journal by Dale Leitzke. The title is: “Gross Living Area Adjustment Shortcut.” Leitizke has the same problem I have experienced with trying to communicate on this subject-it took me about ½ hour to down the article due to a FNMA form included. For those of you that can access the site you can get it there. It has to do with using the cost approach to support sales comparison adjustments. After I read it and absorb it I will get back at you. I want to use his data and see if he used the correct sequence of adjustments. This should be interesting. I haven’t even read it yet but I bet you we don't come up with the same solution. Says he has used it in 4,000 appraisals. He is still appraising so we know he ain’t from North Carolina.

 

[Austin]

Ok. I haven’t read the article, I took his marketing grid and in less than two minutes came up with the same answer he did. His subject sold for $60,000 and had 1,430 square feet. Sale 1 sold for $47,000 and had 950 square feet. Sale 2 sold for $54,000 and had 997 square feet. Sale 3 sold for $80,000 and had 1,830 square feet.
The subject was a cape COD, sales 1 & 2 were one story, and sale 3 was two-story. He made adjustments for things like design and appeal, basements, garage carport, etc.

All I did was to look at the tend line of GLA vs. sale price and knew the market was correct; the subject was worth $60,000. The trend line showed a size adjustment of $33 per square foot that make the slope of the adjusted sale trend line have a slope of zero indicating a perfectly adjusted sale. There is absolutely no justification for any other adjustments. He made adjustments of $2,500 for design and appeal, $2,000 for age, $4,500 for basement, up to $5,500 for garages, and his size adjustment was half what it should have been. Using the correct sequence of adjustments, there is no market justification for any adjustments other than size.
If you want to see it for your self, just go into Excel, put GLA in one column, sale price in the next column, and graph the results. Calculate the trend line and R^2 and you will see that this graph explains 93% of the factors affecting price are size alone.
I still haven’t read the article, so I will get back to you on that.

 

[Fred]

OK, I'll bite. I do not have the article. First thing puzzling me is that there is a "sales price" for subject. Subject sold? Should that be "contract price?"

I used your style of GLA analysis, I agree that with subject at $60,000 the R-squared is 93% (.9295). So, I guess I keypunched the data right.

However, when I make (what appears to me) to be the indicated size adjustments to appraise subject, I get a value of $66,266 (unrounded) for subject, and this bumps the R-squared up to 98%. If subject already sold at $60,000, that is low compared to the sales 1-3.

 

[Mike Simpson]

Just took another look at this post...wow lotsa debate!!!--Good

Mr. Garrett has it right...appraising is an art--not a science. That's why all these rediculous models will never work. They attempt to measure purchasing emotion--won't happen in my lifetime.

Take a client into the field & observe their actions when it comes to purchasing a home. Take many clients into the field and you may actually come to draw a conclusion regarding purchasers in your marketing area!

-Mike

 

[Austin]

Steven:
When the trend line has that little variance there is nothing to adjust, the trend line tells it all. The only reason you equalize the trend line is to see if the equalized trend line has a zero slope which means all remaining independent variables are just noise or some minor missing variable. You averaged out size (which is a compound covariant variable) with least sum of the squares. We are breaking the independent variables into component parts. When you get a variance about the equalized trend line with that little variance the noise drowns it out then you equalize the trend line. You adjust to the trend line for obvious physical differences, but you equalize the trend line after you adjust out the variance about the trend line. Just look at the GLA point of the subject on the X-axis and where it crosses the trend line is the price. In this example, no adjusting was required and all we had to do was equalize out the size influence to point out the remaining variance or variables not accounted for. For example, I can look at the equalized graph and tell that sale # 1 is still $3,000 superior to the subject but that could be noise. Sale 2 is still $8,000 superior to the subject and sale 3 is $7,000 superior to the subject. Appraising is about finding the MV and not finding adjustments that do not matter. If you want to waste time trying to find out what the missing value factors are, go for it. I got my MV, it is perfectly supported, and that is when I stop.

 

[Fred]

Austin,
If your post answers the question of why you have the subject listed as sold at $60,000 I am not seeing it. So, I am still operating under the assumption that this was an exercise to appraise the subject property.

You wrote,
"the subject was worth $60,000. The trend line showed a size adjustment of $33"
You went on that the r-squared of 93%,
"explains 93% of the factors affecting price are size alone)."

Actually, the best-fit trend line for the three-comp market you posted indicates a value of $66,266 for subject at an r-squared of 98%. How it works is that a 98% fit is a better than 93% a fit.

You already posted,
"If my equation has less variance than your equation (and every accepted method is an equation), then my answer is better than yours."
The equation with subject at $66,266 has the least variance and is the best answer.

If you would just carry out your own square-foot adjustment of $33 (even though $34.27 fits the data better), you would see that the adjustmed values surround $66,000 - not $60,000.

----------	Sale 1	Sale 2	Sale 3

Sale Price	47,000	54,000	80,000

$33 sf adj	15,840	14,289	-13,200

Adj Values	62,840	68,289	66,800

 

[Caterina Platt]

Austin,

Looking at the large relative difference of the size of the comparables versus the subject, hearing your mention of basements, garage differences, etc. I would in no way buy the theory that it is size alone that affected the value. I've had too many buyers in my vehicle, and been one myself a few times. As with many things in life, size isn't everything.

I have a great respect for your knowledge and methods, and feel they could be very useful in measuring the affects of a dwelling's features in the market from a scientific approach. I am also of the mindset that this approach would be far more defensible when challenged. However I'm afraid you are, in this instance, simplifying and generalizing these homes' differences much more than a buyer would. The buyer's and seller's reactions are what we are supposed to be measuring here.

I'd need more information about these dwellings before I'd take the leap that size meant everything. Condition, parking structures, number of baths, storage (basement), school district, functional utility, effective age, visual appeal, it all makes a difference. The hypothetical buyers are talking the pros and cons out of the items I mentioned as they wiegh which home would work the minute they step into the property.

First off, these comps don't look all that comparable due to the relative difference of size alone. Are we talking a ruralish, small town market much like folks such as Jo Ann Meyer Stratton has to deal with? Is that the best you have? (eeew! I'm beginning to sound like an underwriter here) Granted, this is probably a textbook case and we're supposed to assume many things are relatively equal. What happens in the real world when they aren't??

Perhaps the regression analysis would show different GLA adjustments were you to graph 30 sales, as you normally do in your own market? I think that may be the flaw in this particular instance. You are measuring an incredibly small set of data here with a method in which you have stated yourself works best with more data.

Would be interesting to see your method applied to a market such as Jo Ann's with it's vast distance between sales and minimal data.

 

[Austin]

I realize you guys are at a disadvantage not knowing the nature of this problem, which had a lot to do with my scope and method of dealing with this problem. This article was aimed at people that live in rural areas and do not have a wealth of comparable sale data with which to justify adjustments on the marketing grid, so the author gives a method of using the cost approach to support adjustments. That explains why the subject is a 1.5 story cape COD, two comps are 1 story, and one is a two story. I would never consider these sales as comparables in my every day work in my market. I got the sale price of $60,000 from his marketing grid and I assumed it was the contract price and assumed you would assume the same thing, as we typically don’t know the answer before we work the problem unless the subject is under contract. Therefore, to some extend we are not appraising the subject as much as we are verifying the contract price. If it were an equity, valuation, or refinance I would have reported it differently. Purpose, intended use, intended user = scope=methods.
If I did not have the contract price and this was a normal set of comparable data, I would have handled it like Steven is suggesting. My equalized trend line showed a value indication for this data set of $64,500 roughly, meaning this data collectively shows $64,500 as a least sum of the squares reflection of properties similar to the subject. The prices ranged from $60,000 for the subject to $68,300 for sale # 2 after equalizing the trend line. The trend line based on data that lacked comparability says $64,500, the market said $60,000, the gravity of the data said the market was correct in this instance.
Now why didn’t I make any further adjustments given all the other property differences and report a price of $64,500? Two answers: 1. General market randomness or noise in the best of markets is plus or minus 5% under the best of conditions. Once your adjusted sales fall into the noise range of + or – 5% of the trend line, there is no way you can measure anything else because all other independent variables are below the noise level, meaning there is no evidence of their existence. An example is a radio station that is below the noise level. It is may be there but you can’t hear it due to the noise. When I looked at the trend line equation of contract and sale prices vs. GLA, I could see that all adjustments excluding size were in the noise zone of + or – 4% with terrible comps, so there was nothing I could justify adjusting for. 2. As I just stated, given the level of lack of comp comparability, any further differences were just a wash. For example, on the marketing grid the author made total adjustments excluding size on sale 3 of about - $12,000, but if you look at the trend line of raw unadjusted sales vs. GLA, there was only about $1,000 difference between sale 3 and the trend line to begin with. The author imported data into the data set when it clearly was not justified as the trend line clearly shows. Remember the standard of a perfect appraisal is the equation that explains all of the relationships between dependent variable price, and independent variables or value factors. A perfectly adjusted data set is a trend line with zero slope with all data points on the trend line. If the equation can predict the price of the comp sales, doesn’t that give you confidence that the price it predicted for the subject are well supported?

Now to Caterina’s question: “Size doesn’t explain everything?” Actually in this problem it pretty well does depending on how you define size. I repeat size is a vicarious variable meaning that it represents much more than size. The graph of size (GLA)vs. price tells us that bigger cost more than smaller; it reflects the quality of construction; it reflects economies of scale derived from the size as in the cost manuals; in the sales comparison approach it reflects the synergy or obsolescence of independent variables; it reflects general market randomness; and it reflects a host of other minor independent variables that fall below the noise level. All of these factors are covariant variables and can not be broken out and treated separately, at least not in this example. So, how do we deal with them? We deal with them in the aggregate or as a unit conveniently covariant and co-existent with GLA or size. When we equalized the trend line we treated them in the aggregate or as a unit factor and averaged them out along with size with least sum of the squares. Then, the remaining variance or distance of the points from the equalized trend line fell within the noise range, meaning there can be no justification in attempting to measure them.
An example of covariant variables is a championship football team. Who can say which player contributes most to the team and estimate how much he is worth? He is worth nothing without the rest of the team. Some days he is worth a lot and some days he is not worth much with the team. If we took the athletic ability of each player and added it up, would it = a championship team? The answer is no. It is the synergy or lack thereof of the players working together that makes a champion or losing team. It is a team effort either way and you can’t base it on one player just like you can’t say among these comparable sales a carport =$3,000, a deck = $2,000, etc. All together they contribute X number of dollars to that property’s value so you have to measure their contribution as one unit. It is a team effort. Best player in football Mike Vick + Atlanta Falcons does not = Super Bowl. Not enough synergy. Would Mike Vice be worth the same to each team in the NFL? Apparently conventional appraisal theory thinks so.