Actual Age/effective Age Vs. Condition

[Steve Owen]

I really enjoyed Greg Boyd's linked article. I remember studying some of this years ago and it all came back.

On the other hand, Austin's first graph doesn't seem to contain any of those errors. I had noticed years ago that size does matter and within sub-markets, I could draw the same kind of graph. However, the clincher here is the phrase "within sub-markets." In this area, within the sub-market of high-quality, custom homes in sub-divisions, larger homes tend to have amenities such as three-car garage, vinyl fence, inground pool, etc. So, the statement that trying to adjust for such things is a futile excercise might seem to be right. However, there are many exceptions.

If you looked at the entire market here, size alone does not predict sale price. When I run an entire closed sale search for the City of Joplin, there are 3,910 in the system. Sorted by GLA, after the twenty or so at the bottom with zeros are dropped, you get a progression that runs like this: 450sf=$11.5k, 500sf=$11k, 530sf=$15k, 550sf=$7k. Huh? If GLA was the primary determinant of value, the 550sf house should have sold for $16.8k.

There are three more 550sf houses after that one, selling at $12,100, $17,000, and $15,500. Maybe my market has a prejudice against 550 sf houses, because the next one, at 576sf sold for $24,500 (the statistical line indicates $18.77 for this one if the $7,000 sale is left out). More likely, the 550sf houses were in much worse condition, with the first one being the worst. The original question was about effective age. Effective age clearly has more to do with value in this segment of my market than size does.

Near the top end of my sorted search, there is a run that goes like this: 5,000sf=$595k, 5,005sf=$490k, 5,053sf=$700k, 5,100sf=$166.25k. Any appraiser trying to predict values based strictly on GLA in this market would obviously be in deep doo-doo. Of course, the 5,100 sf house... it has a rather odd looking design, and just happens to be 21-30 years old; all the others in that string had conventional design and were almost new. The point is, there are clearly things going on here other than size.

Very near the top of my sorted search, we get 7,047sf=$425k, 7,248sf=$600k, 7,398sf=$410k. What gives? Well, the 7,248sf house has "ALL NEW CARPET.MAGNIFICENT INTERIOR, INCREDIBLE KIT W/,GRANITE COUNTERS,FAB WOOD FLOORS,PLANTATION SHUTTERS" as well as 2 fireplaces, 3 garages, brick/stone exterior, etc., etc. It has a lot of those "futile" adjustments. :rofl:

To be fair to Austin, I could sort out sub-markets and then have a graph that looks a lot like his. Size does matter when it comes to houses. Indeed, GLA is a good starting point. But, I'm reminded that "In the long run, men hit only what they aim at" - Henry David Thoreau. And, I wonder, what would happen if you put together a graph to try and predict value based on number of bathrooms.

 

[Steve Owen]

Sorry... I was writing at the same time Austin was, and much of what I said is redundant based on his post and further explanation of his method.

How do you know how to code the data unless you have tested for size correlation first by ordering the data? You are doing it exactly backwards if that is the way you are doing it. How do you know the gravity of the quality and condition influence? How do you deal with correlated variables? The answer is doing it with the method you describe you are double dipping. You don’t know quality from size from condition. You are divining an answer.

I don't know how Lee Ann does it, but I sort for characteristics most like my subject before I even think about what I'm going to adjust for. So, if I was appraising a house in a moderate to high-value subdivision, the first thing I would do is just input basic characteristics, such as date of sale, GLA and (maybe) two story. If I only get three returns, odds are pretty good they are going to be the three comps used. If I get two dozen returns, then I might sort again, based on number of garages, fireplaces, etc. In this example I would not use age because, within that subdivision, they would all be about the same age. But, if I was in a low-moderate end of town, I might start with GLA and age.

Point is, I think Austin's method is pretty valid... now that it has been clarified as being within sub-markets. When you get that oddball return, such as my 5,100 sf example, you just have to know (based on experience, and of course, only if this is true) that it is not comparable, and throw it out.

 

[James Moore]

Maybe the appraiser should be bright enough to recognize what is important in the case at hand.

In some neighborhoods, size is the most important variable. Especially in newer areas where most of the homes have the same features and similar quality.

In our most desirable older neighborhoods with the fine quality older homes ($500k and up), remodeling may be the most important (assuming it was done right). Still, the larger remodeled homes will normally sell for more than a smaller remodeled home.

Age is also important, but not always the same amount per year. In my market, there was a huge change in houses in the early 1980's - went from dark woodwork to all white, small master baths to the "luxury" bath with whirlpool tub, separate shower, etc. I try not to compare houses from the different "eras" if possible.

 

[Austin]

Many moons ago, I devised a very complex mathematically pure data set to test the capability of using regression. I did things like have a range of sizes based on a curved equation, different quality subdivisions, and other features that contributed different amounts depending on the combination of extras like for example, fire place, fire place + carport, fp + carport+basement, etc. The problem was very complex and it would take hours just to visualize it in your head. I did a regression analysis and graphed the results. The regression did a perfect job of sorting it all out and showing a clear picture of the market segments.
Now that I can post pdf graphics I may do that again to show you where I am coming from.
Someone mentioned above predicting prices by number of baths. That can be done and the same can be done with number of windows, shingles on the roof, etc. The point about that is these variables are all correlated to size. Bigger houses have more shingles on the roof, more baths, and more windows, so essentially what you are doing is mis allocating the value influence to the wrong value factor. That is why we must be consistent is using good appraisal theory when selecting variables and the same reason the additive model the GSE requires is a hoax method. (It consistently over values property because it double counts value factors like counting baths in the GLA and baths as baths.) If someone makes up their mind that the cost approach is not relevant and ignores the ratios from that method, they will never get this straight in their heads. It is like going to the doctor and deciding before you get there that nothing is wrong with you and if some quack doctor says something is wrong with you, well, what do quack doctors know anyway. There is nothing wrong with me, the doctor's methods are wrong if he says otherwise.

 

[Farm Gal]

Sigh:

Yeah in some of my sub-markets CONDITION is king and size is way secondary. Homes with less GLA can sell for 2 to 3 TIMES what the larger oens go for and both of similar age, a few blocks away in the 'same' location and the ONLY difference is conditon...

Places where I am adjusting say $6.00 per foot - IF AT ALL - for size and counting the new/newer HVAC, roof, siding windows and even the damn carpet for far more than the 'value added' for more or less GLA of house.

Austin GLA and number crunching does matter... in most places. To a high degree. Just not ALL nor all the time.

Sometime it just isn't THAT easy. Sometimes quantifying what DOES matter is tougher and takes more time than flipping a few sales into SPSS or any other numbercrunching program. and saying AH HA! Which is why I differe with you about verifying sales to insure you are comparing apples and apples not oragnes and tomatoes.

 

[Steve Owen]

I agree with most of what you're saying, Austin. I only mentioned the baths example to show that whatever you decide to allocate on in a math model will produce results... the decision to allocate on size makes sense most of the time if you've properly selected the comps to start with.

the additive model the GSE requires is a hoax method. (It consistently over values property because it double counts value factors like counting baths in the GLA and baths as baths.)

I understand where you've coming from there, too. Of course, I almost always have used a cost approach, which will give you a value indication without double dipping. However, I use the method of adjusting for GLA and also adjusting for baths without overvaluing. The reason it doesn't overvalue is probably that I put less emphasis on size than you would. We get to the same bottom line, but I think my way is more descriptive because it shows what the market is paying for amenities. For example, if we have two 3,000 sf houses but one has a three car garage and four baths and the other just has a two car garage and two baths I will have an opinion of value for the first house that is higher than for the second house... and so will the market (at least around here).

So... that kind of matches my comment earlier about double dipping if you adjust for both age and condition. Maybe you are and maybe you aren't... it depends on how you do the adjustment. When I adjust for GLA, it isn't the entire adjustment, like on your first and second graphs... by adjusting less than the total for GLA, you leave yourself room to tweak for the other amenities that often come along with GLA. And, guess what? When I find those almost perfect comps with all the same amenities, the same lower (read partial, if you wish) adjustment for GLA I would have used with less perfect comps still works. What that tells me is that this market does pay extra for things like an extra garage or an extra bath.

Still, I understand that often, you could do a linear analysis of price versus size and be very accurate if you had selected good comparable properties at the front end. I have done it that way before (it's been a long time ago) but, the analysis I was doing was for the value of a small house to be moved from a commercial lot. I would probably not try to sell that valuation method to a GSE, just because I doubt if many over there would get it.

Also, concerning your comment that the GSE method overvalues... I suspect that is true in practice, at least for many appraisers... but depends on the appraiser. If the appraiser knows their stuff, doing an appraisal on a form (even a 2055) should not overvalue the property. However, sadly, I suspect that in many instances it does, for exactly the reasons you pointed out.

 

[Fred]

Austin,
It’s hard to figure out if you can’t read or just can’t stop distorting

If size and price do not correlate, them obviously

Again, no one, I repeat no one, said size did not correlate, but that they might not correlate well. You always seem to have rev yourself up, by finding a straw man to argue against.

Steven seems to me to be saying: “I already knew the answer by divination before I started the analysis and I just presented that graph to make Austin look bad.”

See what I mean? Not only do you make up a position and attribute it to me, you stooped to using quote marks.

Someone mentioned above predicting prices by number of baths

Where?

The fact is that you make false assertions and math errors. I point them out – not because I give a hoot how you look – but because someone might buy into a mountebank version of statistics.

How do you know how to code the data unless you have tested for size correlation first by ordering the data?

By doing a Pearson test, instead of banging the data to make it fit a preconceived ordering theory on square footage.

The point being, without that graph you would have no guidance of where to start looking.

That’s not true. I only need the spread sheet the graphs came from.

Now that I can post pdf graphics I may do that again

I’d rather have the raw spreadsheet.