Comments on: January Case Shiller Housing Data Now Available

By: nullpointer

nullpointer — Sat, 11 Apr 2009 00:24:01 +0000

Not as effective != biased

It is impossible to separate a cube into two cubes, or a fourth power into two fourth powers, or in general, any power higher than the second into two like powers. I have discovered a truly marvellous proof of this, which this margin is too narrow to contain.

By: TigerAl

TigerAl — Fri, 10 Apr 2009 21:16:20 +0000

Please read what i wrote again because you have not understood my point on bias. Also, if I had to ask questions about index construction, it certainly would not be as basic (or as useless, IMO) as what you have stated. I have worked through the C-S and disagree that it is as effective as people think it is but it would take dissertation level research to prove that with simulations and come up with a better alternative. Maybe when i retire i will work on it but for now, this is the last i will say on the subject.

By: Nullpointer

Nullpointer — Fri, 10 Apr 2009 19:26:46 +0000

If you are a statistician and you’re making a technical point about bias, then shouldn’t you be asking whether the expected value of the delta in the three month moving average for sampled repeated homesales is equal to mean delta in three month moving average of all repeated home sale prices?

In that case it might be good to read the PDF containing the methodology (http://www2.standardandpoors.com/spf/pdf/index/SP_CS_Home_Price_Indices_Methodology_Web.pdf). In this PDF it is noted that they are not performing sampling but instead full measure of all SFH in the area minus some adjustments for suspected special cases. In that case the notion of bias does not apply to a paramater of a population, because a parameter cannot be biased.

By: TigerAl

TigerAl — Wed, 01 Apr 2009 02:52:52 +0000

Dee,
I am not an “industry professional” if that’s what you are suggesting and I don’t care about NAR or MLS. I am a statistician though and I usually want to know what metrics are being used to make a decision or a case for a point of view.
BTW, per statistical definitions, consistency is about process variation and bias is just an “offset” of the process mean from a target or value…one does not imply the other.
The index conclusions only apply to property values of homes that satisfy the same criteria as those of the data used to generate the index. IMO, any extrapolations beyond that are a crap-shoot.

By: Dee Snutz

Dee Snutz — Wed, 01 Apr 2009 02:07:15 +0000

me-thinks TigerAl is a little defensive about the NAR or MLS…i’m with geldpress on this one. we get way too much BS info from the industry “professionals” trying to cover up the actual numbers. Tiger (AL) i agree, there is some limitation to the index, however, the limitation is consistent and thus unbiased.

By: TigerAl

TigerAl — Tue, 31 Mar 2009 23:55:15 +0000

So what’s your definition of bias and what metric are you using to compare them?

By: geldpress

geldpress — Tue, 31 Mar 2009 23:35:50 +0000

Yes you are correct on the Case-Shiller limitations. But I still stand by my use of the term “unbiased”, especially when comparing the data to anything released by the NAR or MLS.

By: TigerAl

TigerAl — Tue, 31 Mar 2009 20:58:46 +0000

Careful with the term “unbiased”, Geldpress. The Case-Shiller only uses a certain set of data in calculating the index and the conclusions only apply to that group (unless you can prove otherwise). It excludes new home and sales data on homes that have not sold twice in what is called the “arm-length” period.
From the methodology documention: “The monthly S&P/Case-Shiller Home Price Indices use the “repeat sales method” of
index calculation – an approach that is widely recognized as the premier methodology for indexing housing prices – which uses data on properties that have sold at least twice, in order to capture the true appreciated value of each specific sales unit.”