This weekend, we noted that in terms of job creation, the 2002-07 post-recession recovery was the very worst in the post-WWII era. This was in response to the pundit commentary that job losses aren’t nearly as bad today as the 2001 mild consumer-led recession.
Since so many t-heads want to compare the two cycles, let us hasten to point out a critical difference between the two: Birth/Death Adjustment.
As we have discussed ad nauseum, prior to 2002, the B/D adjustment had a minor impact on total BLS reported job creation. Since 2003, the B/D adjustment has been part and parcel to BLS’ Current Employment Statistics (CES) program, the official measure of US employment. In brief, the Birth Death adjustment hypothesizes how
many jobs were created by companies too new to
participate in the CES survey.
Since this major change in modeling was effected, Birth Death jobs are much more significant, rising to the point where in 2007, the B/D accounted for over 80% of all BLS reported jobs. Thus, comparing the two periods is an apples & oranges affair.
Example: Consider the by-any-measure-ugly March NFP report. In the present environment of Construction and Finance problems, was it reasonable for BLS to hypothesize a Net Birth/Death jobs increase to 142k for March? This was an 11% improvement over March 2007 B/D adjustment of +128k. According to the BLS, new business creation is significantly more robust than last year.
And even more egregious, these hypothesized (i.e., fictitious) were unfathomably in areas that are furiously laying off people: +28k jobs in construction, and +6k in financial activities.
Bill King points out that in the same CES report of NFP, construction lost 51,000, manufacturing dropped 48,000, retail and trade lost 12,000, while professional services dropped 35,000. Yet the B/D Model somehow managed to create jobs in all these categories.
>
Previously:
NFP: Birth/Death Adjustments (December 2007) http://bigpicture.typepad.com/comments/2007/12/more-on-birthde.html
How Strong is this Jobs Recovery? (December 2005) http://bigpicture.typepad.com/comments/2005/12/how_strong_is_t_1.html
The Accelerating BLS Birth/Death Adjustment (August 2007) http://bigpicture.typepad.com/comments/2007/07/the-accelerat-1.html
>
Sources:
CES Net Birth/Death Model
http://www.bls.gov/web/cesbd.htm
Current Employment Statistics (CES)
U.S. Department of Labor
Bureau of Labor Statistics
http://www.bls.gov/sae/home.htm
Reality may not always be the first one to the party these days, but history shows she is without fail the last one standing. Drink up. She will wait to be “priced in”. No hurry.
Well, yah.
But there’s also no accounting for illegal workers, right?
So in construction, those 28K “adds” could also be the result of 35K illegals going OFF-payroll being replaced by 28K ‘legals.’
There’s a lot that’s not accounted for.
Barry,
How about a chart of B/D numbers going back to 2000? The BLS web site only has it going back to ’07 as far as I could find.
I think that chart would get a lot of links!
~~~
BR: This link takes you back to 1999
http://www.bls.gov/ces/cesbdhst.htm
Seems everything is really worse than reported these days. And if it is true what we hear that if they didn’t save Bear Stearns, we risked a financial implosion, then as Hussman says, we are going to get one anyway. It is so amazing to hear all these plans to save housing that can never, ever prevent what is to come. But I guess they have to make it look like they tried. It is rearranging the deck chairs on the Titanic and maybe helping Wall Street get on the few lifeboats available.
The reason the historical b/d data is important:
The bulls argue we should be losing 200k jobs a month as we did in previous recessions. But are we losing a comparable amount of jobs if you remove, from both periods, the B/D adjustment?
Barry
Good post.
Would there not be a lag in the B/D adjustment, in that if all these jobs supposedly were being created by new companies in 2007, should they not start filtering into the formal survey fairly soon?
Also, are more new businesses started at the end of an economic cycle, and is it fair to assume that the rate of new business formation accelerates when credit gets harder to get, as it did in the latter half of last year? The former seems a fair assumption, the latter does not.
T.
Sorry to say this..
Does it really matter what we think as long as the government thinks and states otherwise?
They are following the principle of:
Deny everything, dont admit anything, accuse somebody else..
dad29–why would 35k illegals quit or be fired only to be replaced by 28k higher-cost legals in an industry that’s in the crapper?
I agree with your basic premise about much being unaccounted for, but I don’t follow the particular ‘for instance’ example.
Thanks.
Barry, you misunderstand the birth/death adjustment. It is used in the INITIAL estimation of employment, which is based on a survey of employers. The estimates from the sample are augmented with the b/d adjustment, which as you point out is problematic.
But when employment is revised every year, the revisions are based on state unemployment insurance tax reports from ALL employers, including the actual births (and of course, the actual deaths are “included” at zero jobs). No b/d adjustment is made to the revised data. So revised 2006 data is apples-to-apples with previous years. Once 2007 is revised, same thing. Give a call to your local employment department economist (my job in another state) or to the CES technical number on the monthly press release, they’ll walk you through the process.
Your point still stands that the last “recovery” was far weaker than previous ones in terms of jobs, but it has nothing to do with the b/d adjustment. It has more to do with weak investment (run the same comparison with investment from GDP estimates and the curves look the same) and until recently the high-priced dollar.
~~~
BR: Scott, when various pundits are comparing the NFP announcements, they are comparing the monthly BLS announcements, like the one we got on Friday, versus the 200k monthly job losses in 2001/02.
You are looking at reality — the adjusted actual data. That is not a place where the pundits play. They are not comparing the annualized, adjusted benchmarked data.
there he goes again, using fuzzy math! we all know that recession + inflation = GOLDILOCKS!!
ScottB – thanks. That’s very helpful. Translating,once the annual revisions are in the data’s no longer noisy and is clean ?
Within a year I think Barry has a good point though. But a key point that’s not discussed is that we’re very early in the business cycle and employment is a LAGGING variable. As the economy continues to slow my expectations are that so will employment. And that the talking head meme on 200K job losses will be seen in the future. For some charts and discussions, including net job creation above population growth and productivity growth try this: http://tinyurl.com/6yzw9u
Even just allowing for population/ labor force growth we’re not keeping up.
This is getting hilarious. Everything is or used to be a bubble: housing, employment, WMD, government, etc. Maybe you’re a bubble too! Stick yourself with a needle to be sure …
NOR . . .
I heard it was
“deny everything, admit nothing, make counter accusations”
I saw someone doing construction this weekend, so therefore there must be hiring going on in the sector.
-BLS
Actually, if the layoffs are so high, it is not unreasonable to assume that some laid off workers decide to start up a new, small company. My friend in financial analysis is doing just that with some co-workers. So, the net effect of the B/D adjustment is that it corrects for those that “jump right back in” to the work force. It can only be a fraction of those laid off, but I’m just saying….
NOR is right.
Credibility just doesn’t seem to matter to our government any more.
ScottB,
Thanks indeed. Do you know where one can find historical data on the accuracy of the B/D adjustment itself? Just a simple comparison of the initial sample’s accuracy to the after B/D estimate’s accuracy? ( for each full year that is ). If not, can you vouch for it?
We know that governmental agencies are releasing parolees early in order to adjust for a loss in property tax receipts. So, who’s keeping tract of that number being added to the labor pool and where is it reflected?
“Bill King points out that in the same CES report of NFP, construction lost 51,000, manufacturing dropped 48,000, retail and trade lost 12,000, while professional services dropped 35,000. Yet the B/D Model somehow managed to create jobs in all these categories.”
“Somehow” ?
Really ?
.
David,
“The bulls argue we should be losing 200k jobs a month as we did in previous recessions.”
Aren’t they precious ?
We didn’t have seven consecutive years of negative job growth previously. It’s like you fell 50 feet down a 100 foot deep well and got caught on a crevice. Then you fell another 10 feet and got caught on another crevice. They only want to count the second 10 foot drop, but you’re still 60 feet down in a hole, and losing your grip.
But at least you’re contained.
.
Thanks for pointing this out. Are you suggesting that the actual NFP number was -220k (B/D adjustment +140 minus sample -220 equals -80)? This is a big hitter in the light of the overall (80%) majority of the data coming from the adjustment model. The gap between actual job losses shown in the sampling data and new businesses creating new jobs must be a record and is highly suspicious. The B/D adjustments are computed as a time series and if we believe that we are at a crucial economic juncture therefore per definition unfit to detect turning points.
I just wanted to make sure that everyone knew that, as one poster asked, the historic data is clean. It’s the preliminary estimates that can go awry. This year’s adjustment for March 2007 was downward by 293,000 jobs, knocking about 0.2 percent off of the already-low-and-not-keeping-up-with-population-growth preliminary number.
I have asked BLS to send me data evaluating the past accuracy of their model, and will share it with you ASAP.
The total b/d adjustment for the first quarter of this year was -101,000. That compares to the +71,000 used in first quarter 2007. So they have made some adjustments, we just don’t know how accurate they are. And as someone who messes about with the actual micro-level firm data, I must say that there are lots of births (and expansions) even in the grimmest of times, same for deaths (and layoffs) even in the good times. The balance does shift, and there’s the rub.
Another BLS series which does measure historic births and deaths (the Business Employment Dynamics series) shows that the birth/death balance went negative in late 2000, predating the recession. The current series went very slightly negative in 2nd quarter 2007. So more negative revisions to nonfarm employment for the last half of 2007 would not be a surprise.
I just wanted to make sure that everyone knew that, as one poster asked, the historic data is clean. It’s the preliminary estimates that can go awry. This year’s adjustment for March 2007 was downward by 293,000 jobs, knocking about 0.2 percent off of the already-low-and-not-keeping-up-with-population-growth preliminary number.
I have asked BLS to send me data evaluating the past accuracy of their model, and will share it with you ASAP.
The total b/d adjustment for the first quarter of this year was -101,000. That compares to the +71,000 used in first quarter 2007. So they have made some adjustments, we just don’t know how accurate they are. And as someone who messes about with the actual micro-level firm data, I must say that there are lots of births (and expansions) even in the grimmest of times, same for deaths (and layoffs) even in the good times. The balance does shift, and there’s the rub.
Another BLS series which does measure historic births and deaths (the Business Employment Dynamics series) shows that the birth/death balance went negative in late 2000, predating the recession. The current series went very slightly negative in 2nd quarter 2007. So more negative revisions to nonfarm employment for the last half of 2007 would not be a surprise.
John,
why would 35k illegals quit or be fired only to be replaced by 28k higher-cost legals in an industry that’s in the crapper?
Those were hypothetical numbers. More realistic hypothetical plug-ins:
100K illegals leave, 28K legals fill in. That better satisfies the ‘crapper’ situation, as construction has not stopped dead, but there’s a lot LESS than there was.
I understand that there’s a slowdown and that there is a lot of number-massaging goin’ on out there.