Square footage adjustments

[Austin]

Slacker:
You have it backwards. Price is always the dependent variable because we are estimating price not GLA. You stumbled upon the approximate answer because you were equalizing size and not price. Make price your dependent variable and GLA your impendent variable and try again. Now, before you do that: Steven’s problem kind of threw me for a loop because Steven didn’t include a subject property therefore I did not have anything to equalize to. The formula you use to equalize is: (GLA of sale 1 – GLA of subject) x 1st guess of the adjustment per square foot. You create a graph of adjusted prices, GLA as independent variable vs. adjusted prices as dependent variable, which change every time you change your guess of $/sf for the adjustment. Keep doing this and watch what happened to the trend line. Generally if you lower the price per sf, the slope of the trend line decreases. If you go to far start raising the number and keep doing these iterations until the trend line is level or has zero slope. Then look at the variance of the data points about the trend line and the remaining factors are attributable to something else. Remember the size adjustment is not just a size adjustment. The price per square foot is a compound or pregnant number. It has elements of size, point of diminishing returns, quality of construction differences, market noise or variance, covariance of variables, etc. That is the point I was trying to get across to Steven in my last post. In my view, if the site values are different, not including excess land, which is a separate issue, you can’t take out the land value because you don’t know what it is and if you do take it out at its highest and best use price you are muddling up the remaining factors because land value mostly likely under his scenario has something to do with lack of comparability or highest and best use questions. If you try to equalize a data set and it takes some off the wall number like -$10 per square foot, you have a problem of lack of comp comparability.

Caterina:
Don’t feel intimidated. When I 1st started doing this stuff I found that the problem to be solved resulted from multicollinearity of variables. So, I went looking for information on multicollinearity of variables. I found a webb site forum like this one for statisticians, grad students, and college professors. They had a forum with 1,400 different opinions of what multicollinearity is and how to deal with it. I didn’t read them all but I read enough to know they didn’t know any more than I did. I found one post by an etymology professor that had it nailed in my opinion because we had both reached the same conclusion. Believe it or not, he professed at my alma mater.

 

[Fred]

Austin writes, <span style='color:darkblue'>"The size adjustment based on the data you presented would be the slope of the trend line, which is roughly $160 per square foot. One of your sales most likely sale 1 has a land value $11,000 higher than the other two sales proportionally, but it is hard to tell which one because there are only three data points. You omitted a subject property so the only answer I can give is the slope of the trend line."</span>

Actually sale 3 has the highest land value and once the analysis considers that land value, the slope of the trend line for improved space would be $92.50/sq.ft.

Sl	Sq Ft	Sale Pr.	Land	Land Val	Imp Val	Imp SF Val

1	2,500	281,250	half-ac	50,000	231,250	92.50

2	3,000	352,500	3/4-ac.	75,000	277,500	92.50

3	3,200	396,000	one-ac	100,000	296,000	92.50

It is not a just question of regression versus not regression.
I have posted before that all sales analysis, even what appears to be the basic get-a-pencil and dope-it-out approach shown on the above table is still regression. The $92.50 per square-foot contributory value of improvements created above by simple arithmetic does not become a more reliable indicator of value if I shift to regression-speak and call it the slope of the trend line Y = 92.5X + Land Val; that has an r-squared of 100%. However, the data indicate that for sites between one-half and one acre, for improved space of this (whatever) type, each square foot of space between 2,500 square feet and 3,200 square feet adds exactly $92.50 to overall value.


Given this solution, if the subject is 2,800 on a ¾-acre, the indicated value is the same whether one solves with the regression line or does the standard Fannie Mae vertical grid and adjusts for site size and building size. Each comp will adjust to exactly the same amount.

Slope of the Trendline
Y = 92.5X + LV
Y = 92.5(2,800) + 75,000
Y = $334,000

Classic (Fannie-type) Vertical Grid (no

xxxxxxxxx	281,250	352,500	396,000

Site Size	25,000	00,000	-25,000

Bldg Size	27,750	-18,500	-37,000

Adj Value	334,000	334,000	334,000

Note: The "x's" and extra "0's" are to help the columns to align

Austin, your write: "I think the answer to that question is when you extract the land out of an improved property then you don’t have a problem if the highest and best use is as improved or there is little obsolescence in the property as improved"
One of the primary reasons for accounting for land separately is precisely to measure HBU, comparbility, obsolescence, (synergy, chemistry, etc). Yes, the market is in the figures trying to get out and tell its secrets, but if we do not ask the right question, it does not matter whether we use plain old 'rithmetic or regression. The truth is in the numbers, not in the form of presentation.

Again, I stress my point that I do not see how to make reliable SF adjustments, without considering the marginal contributory value of the space, NET of the land value - because that contributory value is exactly what the size of BUILDING adjustment is supposed to reflect.

 

[slacker]

Slacker:
You have it backwards. ................The formula you use to equalize is: (GLA of sale 1 - GLA of subject) x 1st guess of the adjustment per square foot.

O.K. I switched the variables and came up with the 160.

But shouldn't formula for the equalization be?

=((GLA of subject - GLA of comp)x 1st guess) + sales price of comp

This way I can graph the adjusted sales price and adjusted price per square foot of each comp. By the way, WINTOTAL does this for you in the COMPS Powerview.

Using an assumed GLA of 3000 sft for the subject, if you subtract the subject GLA from the comps GLA as you suggested it will result in a negative adjustment for a smaller comp. Isn't that backwards?

Now, using the value of $160 per square foot with the formula above I get a nice flat line for my adjusted sales price but my adjusted price per square foot is all out of whack.

If I use the $43 psf my adjusted price per square foot is nice and flat but the adjusted sales price has the larger variance.

Don't I want to keep the variance in the adjusted price per square foot as small as possible and not the adjusted sales price?

Given this solution, if the subject is 2,800 on a ¾-acre, the indicated value is the same whether one solves with the regression line or does the standard Fannie Mae vertical grid and adjusts for site size and building size. Each comp will adjust to exactly the same amount.

I agree with Steve on this one:

In my example, using the $160 will result in a 28% gross adjustment for comp 1 and a 8% adjustment for comp 3. If I use the $43 my gross adjustment for comp 1 is 7.5% and 2.8% for comp 3. Using either is going to get me to the same estimate of value but by using the $43 I get there by using a smaller overall percentage of adjustment.

Isn't making fewer adjustments overall the name of the game?

In the end I guess I'm confused about how you can justify using a square foot adjustment of $160 that is far above the actual price per square foot of all the comps.

Either way, it is cool watching the line move.

 

[Austin]

Slacker & Steven:
I can tell that your minds are getting sluggish and your bodies need equalizing before your minds can take in the equalizing process. This works for me: Drink a pint of warm castor oil with a teaspoon full of lemon juice. Then go to the Mall and take a long walk and just relax. Tonight when your minds have cleared up and your bodies have equalized, we will address your concerns. Remember only one teaspoon full of lemon juice because the lemon juice is a covariant variable.

 

[slacker]

Slacker & Steven:
I can tell that your minds are getting sluggish and your bodies need equalizing before your minds can take in the equalizing process.

Actually I can take it in just fine. Just too many buttons to push in order to come up with the same number in the end.

It's pretty to look at though so you've got that going for you.

And for all it's worth, I really pefer a bran muffin and a cup of black coffee. Wait a minute, I had that about 20 minutes ago.

Gotta run...............

 

[Mike Garrett RAA]

Year and years ago, ERA Real Estate made a cardboard calculator for real estate agents to use in pricing a home for sale. It sort of looked like a slide rule. The center portion would slide out for style and size (square footage). Then there were little wheels that turned for the various features. It was amazingly accurate.

 

[Mike Simpson]

I've attended the classes & looked at the software. I've listened to the theories & pondered this question many times in the past myself. I count among my former employers & partners 12 appraisers with whom I've discussed this issue with time and again.

As a real estate agent though--I feel often times appraisers try too hard to prove their adjustments (we can be over analytical at times). I understand why...we're often challenged, either in a review, by our clients or some disatisfied borrower.

However, I've never seen a buyer base their purchasing decision on a model derived by a mathematical genious--not yet. They tend to base their decisions on the total attributes a home (and neighborhood) has to offer compared to those in a similar price range.

I base my adjustments on my experience, and conversations with Developers, Builders & Real Estate Agents. Typically, that would be an adjustment of 1/4 - 1/3 (closer to 1/3) the construction cost new for the subject residence in the cost approach.

 

[Fred]

Austin writes, "Tonight when your minds have cleared up and your bodies have equalized, we will address your concerns. "

Sorry to see you stoop to the ad hominem. So much for the prrofessional discourse on size adjustments.

 

[Austin]

Wayne: Kill this thread fast before the E & O Insurance Companies see it, because if they see it our insurances rates will surpass the rates for MD’s. We will have to charge $4,000 per appraisal to pay the $3,500 insurance fee per appraisal.

Steven & Slacker: While I am writing this, I hope you guys are out getting your bodies equalized.
Steven: Referring to your example above with three comparable improved property sales with site values of $50, $75, & $100,000. If these three improved residential properties are indeed comparable then there is no way they could be improved in conformity to the highest and best use of the sites as though vacant unless they are in unique locations in which case they are not comparable or need a location adjustment. If these dwellings are comparable and the sites as though vacant are what you say they are, then you are admitting to huge amounts of obsolescence in at least two and possibly all three site improvements. Where I come from comparable properties don’t have a $50,000 range in site values unless it is location like on the lake or a golf course or something, in which case your three sales are in no way comparable. Then too, 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. If they are adjusted properly they all should have about the same price after adjustment, in which case if you divide the same the price by 2,500, 3,000, and 3,200, how can you come up with all three being $92.50 per square foot? 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 which indicates that you did the same thing slacker did, and that is to make land value the dependent variable and sale price the independent variable. You lost me in that last post.
If you were referring to your original problem you presented to me that could be a problem too. The way the problem was presented there is only one independent variable and that is size, therefore the only answer possible is the slope of the trend line. I don’t see how land value enters into this discussion in relation to that data set except to use the graph to spot the sale that is out of skilter with the higher site value. As I recall the R-sq factor was in the high 90’s and any land value difference was insignificant. If you had included a comparable sale with the land values we could have adjusted for LV that but that would not work either because if you adjusted non comparable sales for different site values all you are doing is further screwing up the already gross obsolescence in the improvements on the three lots. Again, wrong dependent variable. What you would be doing is adjusting improvements for obsolescence.

Mike Garrett: That slide ruler you described is a manual computer like they used on navy ships in WWII. I was on a battleship recently and these computers occupied entire suites on the ship and took scores of people to operate.

Mike Simpson: If what you say is true as to how buyers make buying decisions, then there is no pattern or reason to their decisions to purchase, in which case no procedure can make any sense of their irrational decision, or in other words, who needs appraisers? Appraising will never be recognized as a profession until the analysis of the appraiser can be duplicated with scientific methods. Years of vast experience and discussions with developers and or Realtors won’t help either because that is the blind leading the blind. If Realtors and developers knew what the hell they were doing we wouldn’t need appraisers.

 

[slacker]

I was on a battleship recently and these computers occupied entire suites on the ship and took scores of people to operate.

Hey, not unlike your regression model? :lol:

Mike Simpson: If what you say is true as to how buyers make buying decisions, then there is no pattern or reason to their decisions to purchase, in which case no procedure can make any sense of their irrational decision, or in other words, who needs appraisers?

There's a perfect procedure. It's called the "What did the house down the street sell for" model. I use it all the time. The underwiters I work with love it.

I still want one of those wheel thingies that Mike was talking about.

And yes Mike, I would have the Disto on the belt and the wheel around my neck on a big gold chain.