http://www.cnbc.com/id/45659547



Aren't sales actual countable things? How hard would it really be to just get real numbers and add them all up?

Silly.

Well, this is what happens when you have salespeople trying to do statistics and economics.

It seems that the numbers they report are not actually a count of houses sold. The numbers came from a model that predicted the actual number of sales based on the number of sales of MLS-listed homes. Since not all homes are listed in MLS it makes sense to adjust the number upward if you have a good model. For example, if 75% of houses are sold through MLS then you know that 25% are not. So if you see 3 MLS sales it makes sense in a way to report that there really were 4 sales. But it seems that a far higher percentage of houses are sold through MLS than 5 years ago, probably because it is harder to sell so people are more likely to use a real estate agent.

Add to that the fact that there was a lot of double-counting in many areas and you have a statistical nightmare.

The two most scary things about this are that they are NOT going to revise the median price data and that NAR does not have enough money to pay the trillions of dollars that their idiocy cost investors and homeowners who relied on their data.

http://www.realtor.org/wps/wcm/connect/ ... fda086cc0a



Well, if this is true, then they would not need to because they used actual transactions to get these numbers. Sounds like they don't use the same model for calculating number of sales as they do for calculating median price data. But aren't the Case‐Shiller indexes the de-facto standard anyway?

Well that would be fine if there were no double-counting. If a closed transaction were counted twice then that certainly would affect the median.

It's also a bit alarming that these transactions included in the median price numbers do not include private sales to avoid foreclosure, foreclosure auctions, and even some short sales. That likely means the median reported overestimates the actual sale prices. Normally using the median makes sense because it avoids skewing from very low or very high sales. But when there are a lot of "off book" sales occuring at very low prices it makes the median suspect.

I personally think NAR has lost what little credibility they had by letting this go on so long and then bungling the "fix" so badly.

Looks like the data is out and the error was on the order of a 15% overstatement. In the modeling world that is enormous. I'm not sure it changes the general conclusion that there was a big housing market bust. But it sure ought to embarrass the heck out of the NAR. That kind of error is really inexcusable. In all honesty I expected a correction on the order of 3-5%!

I'm not sure how this is not bigger news.

a. It's embarrassing
b. It shows a tremendous shift away from FSBO
c. It indicates the sheer volume of house flipping that's been going on
d. How many other models does this affect?
e. It has the possibility to affect the U.S. GDP
f. Others?

Looks like housing is consistently about 14% of GDP. If housing sales was 15% less, maybe GDP was 2.1% lower. Oops, that's HUGE. That's a big deal. Fortunately that 14% includes a lot, not just home sales. So maybe it's not quite so bad.

Still, this is big news. I also don't understand why the media is not all over it.