I'm not a statistician, and nobody should presume too much from trying to "read" graphs -- but isn't it very odd that their accuracy was extremely high during rising markets, but very low during flat or falling markets?
I could understand that some indicators would lag, and so you would get more error when the market changed significantly. But it appears that they are only able to correctly assess rising sales? What is going on?
"NOTE: Graph start at 600 thousand units to better show the difference between the preliminary and final estimates."
CR,
I had noticed that your graphs properly started at zero. It's amazing how few publishers stick to that rule. I think the NYT used to at least give lip service to it, but no more.
I remember reading a book on this years ago, I can't remember the title but it was kind of a Strunk and White on the graphical representation of data.
Your attention to the matter is noted and appreciated.
I would guess it's due to cacellations. The builders have all mentioned very high cancellation rates. New home sales are counted when a contract's signed, but it get's fixed in later revisions, at least that's my understanding.
you are supporting islamic evildoers when you talk about a real estate bubble. Go out and buy a new house and suppport the war on terror. Or move to another country. But dont buy a house there. Just rent.
I think you're right -- but I agree with Name, there should be more accuracy (or inaccuracy) in both directions.
The book you are thinking of is: The Visual Display of Quantitative Information." Probably the best book I have ever read on graphics and conveying meaning. See more here:
the scale does make the rise in sales volume from July 02 to Jan 06 seem like a near-doubling, which it wasn't, but I accept the caveats and rationale for this presentation
quite separately, this graph contains a big warning about what could be expected -- a much further drop in sales volume, another 25-30% off the current pace seems an entirely reasonable prediction
watching the total stalemate, zero sales at all, in my neighborhood, this is not difficult to believe
Paul, I'm not sure why they err consistently when the market is weak. If you look at each month, the error can be somewhat significant in either direction, except when the housing market is weak. I presented a 12 month graph - that smooths out the month-to-month variations - and shows the consistent errors.
Bob_in_MA, I try to be careful with the graphs. For this graph it's the difference between the lines that matters - not the absolute values - but I added the caveat to make it clear.
Are you all saying the preliminary should be adjusted for anticipated cancellations? I don't see how that could be done, given the extreme variation in the effect.
They do note all over the report their RSE (relative standard error) and 90% Confidence Interval.
Change August from July 4.1%
90% Confidence Interval ±15.5 !!
Bob_in_MA, when I look at the data I try to take a several month view (in my mind) - and I use the reported numbers for all of my analysis (I don't adjust the numbers).
However, after reading the AP story and watching CNN - both got the housing story backwards and led with an increase in sales! - I thought I would post what is happening with revisions and why I expect August to be revised down.
As always, it's the trend that matters and not any single month.
CR, To question whether August new-home sale prices are flat or down yoy is moot. We're still at the very beginning of this r.e. correction.
Tim Iacono refers to a year old post of yours (link below) to suggest we ask, does BB still attribute our housing bubble to "strong fundamentals"? The answer's likely to determine how the FOMC reacts to our coming housing debacle. I think it would be indefensible for the FOMC to intervene to slow down a housing price reversion to long-term growth rates. Yet, this is exactly what ten year Treasury yields are telling us BB will do. It's time for BB to inform world markets - business cycles are NOT dead, they act for our long-term best interests. ttp://themessthatgreenspanmade.blogspot.com/2006/09/housing-report-that-mattered.html
CR,
That first graph is great (and I might poach it for my strategy article, properly footnoted of course). Any chance you have the data that goes back further than that. Interesting that there was a simalar trend toward lower revisions back in the last recession, even as mild as it was on the housing front. Just wondering what it might have looked like back in the Papa Bush Recession/S&L debacle, which I suspect was much more simalar to what we might be seeing this time around.
"However, after reading the AP story and watching CNN - both got the housing story backwards and led with an increase in sales!"
Perhaps this will lead to the proverbial "Dead Cat Bounce" in the RE market -- providing just a little more time for RE agents to find a job in a new field...
Dirk, as always poach away! I didn't check the data going back any further (it is available - no problem). It will be interesting to see if this is a recent problem or goes back a long time. I'll look into it.
Since the June figger will be revised again (almost certainly downward) and the July figure twice more, the red line in CR's graph will likely become even steeper.
BR,
We are starting to see some weakness in estiate revisions over the last two weeks, will be running fresh numbers tomorrow. The ratio of earnings estimates being revised up to being revised down has fallen to 0.79, it was over 1.50 about 6 weeks ago (for 2006). The fall for 2007 has not been as steep, to 0.84 from about 1.35 six weeks ago. However, we are now in a period when the total number of revisions is at a seasonal low point. The total number of revisions will pick up in a few weeks as 3rd Q earnings start to be released. I want to see if the recent trend towards lower revisions holds as revisions activity picks up before making a big call on the subject. Feel free to click on my homepage, it provides a lot of this info on a regular basis, as well as the weekly report that tracks the revisions (comes out usually Monday PM's). Currently the median year over year growth expected for S&P 500 firms in the 3Q is 9.1%. Energy and Materials stocks should post the highest yr/yr growth, Consumer Staples and Tech the lowest.
Wow! Amazing graphs. I am so impressed with your outstanding work!
David
Bubble Meter Blog
I'm not a statistician, and nobody should presume too much from trying to "read" graphs -- but isn't it very odd that their accuracy was extremely high during rising markets, but very low during flat or falling markets?
I could understand that some indicators would lag, and so you would get more error when the market changed significantly. But it appears that they are only able to correctly assess rising sales? What is going on?
Very good summation.
"NOTE: Graph start at 600 thousand units to better show the difference between the preliminary and final estimates."
CR,
I had noticed that your graphs properly started at zero. It's amazing how few publishers stick to that rule. I think the NYT used to at least give lip service to it, but no more.
I remember reading a book on this years ago, I can't remember the title but it was kind of a Strunk and White on the graphical representation of data.
Your attention to the matter is noted and appreciated.
Paul,
I would guess it's due to cacellations. The builders have all mentioned very high cancellation rates. New home sales are counted when a contract's signed, but it get's fixed in later revisions, at least that's my understanding.
you are supporting islamic evildoers when you talk about a real estate bubble. Go out and buy a new house and suppport the war on terror. Or move to another country. But dont buy a house there. Just rent.
Bob in MA,
In that case there would be a three year span of repeated "unexpected" cancellations of about 5% per year from Jan 99 to Jan 02.
Its monthly data, so thats 36 straight high misses.
I'm not calling it a conspiracy, but at a minimum its a bad strategy. Create an "allowance for cancellations" account.
Bob-
I think you're right -- but I agree with Name, there should be more accuracy (or inaccuracy) in both directions.
The book you are thinking of is: The Visual Display of Quantitative Information." Probably the best book I have ever read on graphics and conveying meaning. See more here:
Edward Tufte: Books - The Visual Display of Quantitative Information
Highly recommended to anyone who hasn't read this genius before.
the scale does make the rise in sales volume from July 02 to Jan 06 seem like a near-doubling, which it wasn't, but I accept the caveats and rationale for this presentation
quite separately, this graph contains a big warning about what could be expected -- a much further drop in sales volume, another 25-30% off the current pace seems an entirely reasonable prediction
watching the total stalemate, zero sales at all, in my neighborhood, this is not difficult to believe
Why even bother with "intial" numbers if you going to be incredibly wrong all the time. Might as well wait..........
David, thanks! I enjoy your blog too.
Paul, I'm not sure why they err consistently when the market is weak. If you look at each month, the error can be somewhat significant in either direction, except when the housing market is weak. I presented a 12 month graph - that smooths out the month-to-month variations - and shows the consistent errors.
Bob_in_MA, I try to be careful with the graphs. For this graph it's the difference between the lines that matters - not the absolute values - but I added the caveat to make it clear.
Best to all.
Are you all saying the preliminary should be adjusted for anticipated cancellations? I don't see how that could be done, given the extreme variation in the effect.
They do note all over the report their RSE (relative standard error) and 90% Confidence Interval.
Change August from July 4.1%
90% Confidence Interval ±15.5 !!
Definitely giving themselves a lot of room there.
Pual, that's it!
Bob_in_MA, when I look at the data I try to take a several month view (in my mind) - and I use the reported numbers for all of my analysis (I don't adjust the numbers).
However, after reading the AP story and watching CNN - both got the housing story backwards and led with an increase in sales! - I thought I would post what is happening with revisions and why I expect August to be revised down.
As always, it's the trend that matters and not any single month.
Best Wishes.
CR, To question whether August new-home sale prices are flat or down yoy is moot. We're still at the very beginning of this r.e. correction.
Tim Iacono refers to a year old post of yours (link below) to suggest we ask, does BB still attribute our housing bubble to "strong fundamentals"? The answer's likely to determine how the FOMC reacts to our coming housing debacle. I think it would be indefensible for the FOMC to intervene to slow down a housing price reversion to long-term growth rates. Yet, this is exactly what ten year Treasury yields are telling us BB will do. It's time for BB to inform world markets - business cycles are NOT dead, they act for our long-term best interests. ttp://themessthatgreenspanmade.blogspot.com/2006/09/housing-report-that-mattered.html
CR,
That first graph is great (and I might poach it for my strategy article, properly footnoted of course). Any chance you have the data that goes back further than that. Interesting that there was a simalar trend toward lower revisions back in the last recession, even as mild as it was on the housing front. Just wondering what it might have looked like back in the Papa Bush Recession/S&L debacle, which I suspect was much more simalar to what we might be seeing this time around.
Bob in Ma,
I think the book you are refering to is "How to Lie with Statistics"
"However, after reading the AP story and watching CNN - both got the housing story backwards and led with an increase in sales!"
Perhaps this will lead to the proverbial "Dead Cat Bounce" in the RE market -- providing just a little more time for RE agents to find a job in a new field...
wunderbar!
Fabulous work, CR, as always.
Excellent job as usual CR.
Dirk, as always poach away! I didn't check the data going back any further (it is available - no problem). It will be interesting to see if this is a recent problem or goes back a long time. I'll look into it.
Best Wishes.
Dirk,
Didn't you once say that upward earnings revisions were a bullish indicator for the market? How is that looking now?
Dirk, I just checked and the initial releases are only archived back to 1995. So the data isn't available online to go back any further.
Sorry. Feel free to use this graph.
Best Wishes.
It's called M-A-N-I-P-U-L-A-T-I-O-N.
The re industry is great at spinming the news.
The public backlash is going to be awesome in the future.
Looks we might be having a stock bubble as well ... oops sorry forgot the stock rally is based on fundamentals.
The headline on the sales "increase" also was wrong on Marketwatch. It's pretty amazing how badly these outlets do their jobs.
For anyone who really wants to poke fun at that report,dig down to actual sales on a regional basis.
Somehow the Census Bureau has new home sales rising significantly in the NE (more than 20%!), even topping Aug. 05.
Even if I didn't know a pile of hurting builders I don't think I could swallow that number. I can't imagine where that's coming from.
Since the June figger will be revised again (almost certainly downward) and the July figure twice more, the red line in CR's graph will likely become even steeper.
BR,
We are starting to see some weakness in estiate revisions over the last two weeks, will be running fresh numbers tomorrow. The ratio of earnings estimates being revised up to being revised down has fallen to 0.79, it was over 1.50 about 6 weeks ago (for 2006). The fall for 2007 has not been as steep, to 0.84 from about 1.35 six weeks ago. However, we are now in a period when the total number of revisions is at a seasonal low point. The total number of revisions will pick up in a few weeks as 3rd Q earnings start to be released. I want to see if the recent trend towards lower revisions holds as revisions activity picks up before making a big call on the subject. Feel free to click on my homepage, it provides a lot of this info on a regular basis, as well as the weekly report that tracks the revisions (comes out usually Monday PM's). Currently the median year over year growth expected for S&P 500 firms in the 3Q is 9.1%. Energy and Materials stocks should post the highest yr/yr growth, Consumer Staples and Tech the lowest.
Thanks for the info, Dirk!
insurance medical payments insurance medical payments insurance medical payments // bossier city casinos bossier city casinos bossier city casinos