.comment-link {margin-left:.6em;}
Visit Freedom's Zone Donate To Project Valour

Friday, March 30, 2007

Now Really, Folks

Bloomberg headline "Construction Spending in U.S. Increased 0.3% in February"

These are Census figures for construction value put in place during a particular month. It is tracked under four different categories. This is where I got them - all official as can be. Tell me how construction spending increased in February? (These are the non-revised numbers for Jan, but I use non-revised against non-revised and revised against revised.)

Total All Categories
Jan 2006..: 1,194,547
Feb 2006..: 1,199,873

Nov 2006..: 1,181,274
Dec 2006..: 1,189,308
Jan 2007..: 1,180,212 (YoY -14,335)
Feb 2007..: 1,170,817 (YoY -29,056)

Federal
Jan 2006..: 19,441
Feb 2006..: 19,565

Nov 2006..: 20,885
Dec 2006..: 20,452
Jan 2007..: 22,440
Feb 2007..: 19,799

State & Local
Jan 2006..: 235,790
Feb 2006..: 239,974

Nov 2006..: 256,727
Dec 2006..: 263,766
Jan 2007..: 263,520
Feb 2007..: 266,639

Non-Residential
Jan 2006..: 277,893
Feb 2006..: 277,777

Nov 2006..: 311,586
Dec 2006..: 319,002
Jan 2007..: 318,865
Feb 2007..: 321,976

Residential
Jan 2006..: 661,423
Feb 2006..: 662,557

Nov 2006..: 592,076
Dec 2006..: 586,088
Jan 2007..: 575,387
Feb 2007..: 562,404
It's not the seasonal adjustment, either, because the non-seasonally adjusted figures show a YoY and MoM drop as well, and that release has the January revision in it.

Comments:
MoM, either they out and out lied, I do not see how they arrived at +3%, orf they divided Feb 2006, 1,199,873 by Feb 2007, 1,170,817 and got 1.025 which rounded off is 3%, but it is a negative 3%.
 
Attack of the absolute value virus?

Heh, I didn't think about that. I bet that's what they did.
 
ed in texas

Cynical analysis:
News that 34% of mortgage holders did not know what kind of mortgage they hold means that they obviously can't do the math, right? Release the numbers. Oh wait, somebody added it up and found the fault. Time for another round of 'education reform'.
 
It looks like you pulled the wrong number for Jan, 1,180,212.

The page you sent me to has Jan being
1,167,668. That's the revised number from what I can tell.

See what I meen?
 
Yeah, they revised Jan down sharply. But given the interpolations used, it's exceedingly likely that Feb will be revised down heavily too. The Census is revising all the constructon stats down significantly. That's why I compare revised to revised and preliminary to preliminary.

If you compare the non-seasonally adjusted numbers, Feb's preliminary is less than Jan's revised. Of course, it should go the other way at this time of year.

Jan (r) NSA: 80,830
Feb (p) NSA: 79,071

That's why seasonal adjustments can be so misleading. The combination of the SA and the downwards revision is making it look like sales increased when they didn't, and in fact actual sales decreased from Jan rather than increasing as is normal. It's a huge, huge discrepancy.

SA factor for Feb is 81.0 and for Jan 83.1.

The number I have listed is the unrevised SA number. Next month when Feb gets revised I roll the revised number in for Jan but keep the Feb preliminary to compare to March for MoM(for seasonally adjusted only).

They are comparing apples to oranges and it is numerically invalid. If you are on a downward slope, this can happen in any SA series which is why you should always use both sets of numbers. This data is not all that reliable in the first place for three months or so. It's really an estimate based on starts, which then gets updated on the basis of surveys.

(Given where the most building is, weather should have had little impact here.)

This is actually a very bad, bad report rather than a report that shows an upswing.

Construction methodology here:
http://www.census.gov/const/C30/methodology.pdf
 
Post a Comment

Links to this post:

Create a Link



<< Home

This page is powered by Blogger. Isn't yours?