Open thread Feb.2, 2010

I was sitting at home, thinking my liberal thoughts. Then I turned on my T.V. and flipped around until I was lured in by this Fox network.  I tried to fight, but my mind went blank and then in my blankness they started making sense. So now the only time my mind is at peace is when I agree with Fox.

Good Grief! The Republicans just bankrupted the country. All the blah, blah, bullshit in the world shouln’t cancel that one out, should it ?

Oh yes it can.

How many of you actually compared the latest DOL non-adjusted and seasonally adjusted employment and unemployment data?  And how many news articles and television reports did you read that cited any of the non-adjusted figures?   
 
How many of you believe that the U.S. Postal Service hired 14,000 new workers in Jan 2010?  Or believe that retail employment rose by 42,000 in Jan 2010?  How many believe that the U.S. only lost 20,000 net jobs in Jan 2010?  It’s all BS, particularly when the DOL news release doesn’t explain that most of their cited figures are seasonally adjusted.   
 
If you didn’t look at it, here is a good place to start.   
 
Employment Situation Summary — JANUARY 2010  
http://www.bls.gov/news.release/empsit.nr0.htm  
   
Table B-1.Employees on nonfarm payrolls by industry sector and selected industry detail 
http://www.bls.gov/news.release/empsit.t17.htm

***It’s all BS*** MG

You’re right that the seasonally adjusted stuff is wrong, but you’re probably wrong in thinking that the non-seasonally adjusted numbers are better.  Reality seems to be that every year there is a bunch of hiring in October-December for the Christmas holidays.  And every January, most of those folks get laid off.

So we correct the data to try to remove the seasonal stuff and see the underlying trends.

… Except this year and last, Christmas retailling was weak to awful as was the hiring.  Which means that the seasonal adjustments based on Christmas past assumed more seasonal hiring than took place and made October-Novermber-December look worse than they actually were.  Now in January, we have assumed the de-hiring of the folks who weren’t actually ever hired and the adjustments almost certainly make things look more optimistic than they should.

This isn’t rocket science (and a lot of rocket science isn’t actually that complicated).  But it seems beyond the comprehension of the MSM and also of many people who I would think really ought to know better,

Anyway, the Feb numbers should be less hosed, and I’m guessing that March should give a fairly accurate picture of where we really are. … maybe

 I would love to read more about this topic.
http://www.forexqs.blogspot.com

MG,

I guess I don’t understand your objection to using seasonally adjusted data.  That’s the right way to do things.  The BLS uses the Census Bureau’s X-12 ARIMA model for seasonally adjusting data.  Here’s a description:
http://www.bls.gov/cps/eetech_seas.pdf
X-12 ARIMA is in the public domain, you can download it and just about every piece of time series software out there has X-12 ARIMA as an option.  Aside from all of the usual “differencing” and “lagging” stuff that every ARIMA model does to make the data stationary, the X-12 part also accounts for things that we don’t always think about.  For example, economic data is particularly sensitive to the fact that holidays fall on different days of the week and consequently have different employment effects; December doesn’t always have the same number of weekend days, etc.  That’s all part of the pre-adjustment stuff done in X-12.  And the post-adjustment stuff in X-12 has all of the diagnostics used to judge which ARIMA specification is best according to various weights that BLS assigns to all of the test diagnostics for autocorrelation, unit roots, outlier amplitudes, etc.

VtCodger,

…there is simply no way to know how many people were hired for seasonal jobs in this recession period and how many were laid off after the holidays. 

Well, you could ask them in the household survey questionaire.

Aren’t you conflating seasonal adjustments and cyclical adjustments?  We do know that stuff like the timing of Thanksgiving and the number of Saturday’s in December all contribute to seasonality effects in the data.  It just makes sense to correct for those known effects.  The effects of the recession are cyclical rather than seasonal. 

In the physical sciences you always have the option of not doing the experiment if you’re not comfortable with the parameter estimates.  That’s not the case in the social sciences.  The experiment is already done; but you do get chances to revisit and revise the data at a later date.  I guess the real question isn’t whether the point estimate of 9.7% is on target, it’s whether or not there is any unintentional bias in that estimate.  I don’t see how seasonality adjustments make the BLS numbers a biased estimate.

***Aren’t you conflating seasonal adjustments and cyclical adjustments?***

No, I’m merely pointing out that getting the cyclical adjustment right, doesn’t correct for using almost certainly incorrect numbers in the seasonal adjustment.  We don’t actually know what numbers to use in the seasonal adjustment this Winter.  Adjusting for windage isn’t going to improve your shooting much if the sights are grossly misaligned.

Probably what we ought to do is estimate error bounds and publish the error estimate along with the computed value.  But estimating error bounds properly is very hard because of the need to estimate the size of biases.  Besides which, very few people really care about reality.  What they want is fantasy that matches their preconceptions.

Year..Jan….Feb….Mar….Apr….May….Jun….Jul….Aug….Sep….Oct….Nov….Dec
2000 881.6 877.7 878.3 881.6 875.0 873.6 872.3 881.2 871.2 871.8 880.9 911.7

Year..Jan….Feb….Mar…Apr….May…Jun….Jul…..Aug….Sep….Oct….Nov….Dec
2000 881.6 877.7 878.3 881.6 875.0 873.6 872.3 881.2 871.2 871.8 880.9 911.7 

Year..Jan….Feb….Mar…Apr….May….Jun….Jul…..Aug…Sep….Oct….Nov….Dec
2000 881.6 877.7 878.3 881.6 875.0 873.6 872.3 881.2 871.2 871.8 880.9 911.7

Tom,
 
I asked a simple question regarding U.S. Postal Service employment: “How many of you believe that the U.S. Postal Service hired 14,000 new workers in Jan 2010?” 
 
What would be the basis for believing that the U.S. Postal Service increased employment in January 2010 on the order of 14,000 or 14,300 employment, the latter figure being the BLS seasonally adjusted data?  If one believes the BLS non adjusted figures, then the U.S. Postal Service increased employment from 667,500 in December to 684,600 in January, an increase of 17,100.  Again, this is pure nonsense based on the seasonal hiring patterns employed by the U.S. Postal Service.
 
The latest BLS data is, in fact, inconsistent with other available data.  There is not one year since 1939 in which the December employment of the U.S. Postal Service was lower than the subsequent January employment.  The historical tables provide the evidence.   
 
Access to historical data for the “B” tables
of the Employment Situation News Release
http://www.bls.gov/ces/cesbtabs.htm
 
Data Retrieval: Employment, Hours, and Earnings (CES)
http://www.bls.gov/webapps/legacy/cesbtab1.htm
 
Series Id: CEU9091912001 Not Seasonally Adjusted
Super Sector: Government Industry: U.S. Postal Service
Data Type: ALL EMPLOYEES, THOUSANDS
 
Year..Jan….Feb….Mar…Apr….May….Jun….Jul…..Aug…Sep….Oct….Nov….Dec
2000 881.6 877.7 878.3 881.6 875.0 873.6 872.3 881.2 871.2 871.8 880.9 911.7 
2001 879.6 878.0 874.6 875.4 872.6 871.5 868.7 866.6 869.1 863.4 865.7 891.2 
2002 860.7 854.8 851.1 847.2 860.8 852.6 833.1 830.0 828.6 824.3 823.2 842.5 
2003 820.4 815.9 813.1 813.0 810.7 808.6 805.7 803.3 800.6 799.6 796.6 816.2 
2004 790.6 787.7 785.1 782.6 780.6 778.6 779.4 777.7 777.2 777.4 779.5 788.8 
2005 773.7 771.3 770.4 770.0 772.9 773.6 773.4 773.3 773.7 773.9 778.0 786.7 
2006 769.4 768.0 767.7 766.5 766.5 768.1 774.7 766.0 767.1 768.1 770.7 783.3 
2007 763.1 761.1 760.6 762.6 766.2 765.8 766.6 768.9 770.9 770.6 777.4 795.5 
2008 753.8 754.6 755.6 757.2 752.4 747.0 754.1 740.8 738.7 744.7 733.2 737.0
2009 736.3 721.8 717.5 723.0 705.0 700.7 711.8 690.4 687.6 695.5 680.9 667.5(P) 
2010 684.6(P)
P : preliminary
 

MG,

What seems more striking is the unusually low number for December, not the January number.  First, the fairly steady drop in Post Office employment is consistent with the Post Office’s earlier announcement to shed jobs…lots of jobs, mostly through retirements.

http://www.chron.com/disp/story.mpl/front/6327214.html

And don’t forget that due to certain peculiarities in the government’s compensation rules, a lot of folks choose to retire in December…especially if there is a retirement bonus.  That kind of thing could result in an unusually low December figure.

But clearly things have changed because the Post Office is now on a hiring binge.  So if the government is actively hiring people for Post Office jobs, then I would expect employment numbers to go up, wouldn’t you?  If you’re interested in a new carerr opportunity, see here:

http://postalexamoffice.com/?gclid=CMHizrWt4J8CFQUeDQodeHo9GQ

So what could explain all this?  Well, we are in a recession and I can’t imagine that a President would not want to do the one thing over which he has a fair amount of control, and that’s pumping up government employment.  But the postal business is itself highly susceptible to the business cycle, and given that the recession probably bottomed out about 6 months ago, it makes sense to begin hiring.  There are other reasons.  For example, DHL is no longer a player, so that puts added burden on the Post Office.  And even UPS is going through some restructuring.  A month ago they announced job cuts of 1800:

http://www.reuters.com/article/idUSTRE6072P220100108

And the other day I heard something on the radio about UPS possibly laying off another 6000.  I think it’s fair to assume that the Post Office probably had some advance knowledge.

I agree that the retirements would have had an impact, whether reflected in the December or January data.  Or both.  If the number of retirements exceeded the level anticipated, the number of replacement hires might show up as a bubble internally though net employment should be down.  

That doesn’t fully explain why the normal December employment gain at some level didn’t occur; those are temp gains.  Moreover, the January number indicates that the U.S. Postal Service probably hasn’t hit its planned reduction targets announced earlier in 2009.  I would have expected the Jan employment to be lower than Nov employment. 

The question I have is whether the employment hires include postal route contractors.  I wouldn’t think so.