Keywords: "Bureau of Labor Statistics" BLS "Birth / Death adjustment" "ghost jobs" "Chuck Butler" "John Williams" unemployment jobless jobs

..... the Birth/Death Adjustment. http://www.bls.gov/web/cesbd.htm

Explained simply, this adjustment tries to take into account the number of jobs quietly created by companies being born, or jobs lost because companies go out of business. The government assumes that newborn or dying companies are missed by its surveys.

August 23, 2006:

Message from Chuck Butler to Bill Murphy:

I was reading Dr. Richebacher's latest letter yesterday, which is something you don't want to do when you have sharp objects laying around! I kid around here because the good Dr. is always on top of whatever it is he is talking about... Anyway, I think he did a great job this month explaining the Bureau of Labor Statistics' (BLS) Birth / Death adjustment... Recall, that I've talked about this stupid practice and have even called them "ghost jobs" in the past...

Anyway, the thing I wanted to highlight as proof that the U.S. economy is going south quickly was this observation by Dr. Richebacher... "Over the three months April-June, the BLS reported the creation of 329,000 non-farm jobs. But this has come about with a stunning contribution of 657,000 from the net birth/ death model, accounting thus for about 200% of the total job creation. Without this adjustment, reported employment would have slumped. Please note that these net birth/death jobs would, if annualized, add up to more than 2.6 million new jobs."

WOW! I hadn't checked on the "ghost jobs" lately because, quite frankly, as Dieter says... I had grown tired of the babble, we dance now! Seriously though, I had beaten that topic to a pulp, and still nobody cared... Well, now that the Dr. Richebacher is talking about it, maybe someone will take notice! Of course that's not cut in stone... The markets seem to want to pick and choose data that they want to focus on... But... I think they will rue the day they failed to do the math on the "ghost jobs"....

John Williams, a noted alternate voice on U.S. Government statistics, has a somewhat calmer but nevertheless (to me) disturbing take on this. Original at

http://www.shadowstats.com/cgi-bin/gra/article/id=556 about half way down the page.

Here is the relevant part, with a few boldings and underlinings by me (FNC):

John Williams Responds

For continuity's sake:

(1) Last Wednesday (5/11[/2005]), I received an e-mail from a client. This e-mail contained a link to a Morgan Stanley research piece. The e-mail asked, "Will John comment on this article?"

(2) The Morgan Stanley article (also dated 5/11 and contained in the firm's Global Economic Forum publication) was written by Ted Wieseman. It was entitled, "'Birth/Death' Confusion Redux," and in part, its preamble stated:

"Spring is in the air, which in addition to warmer weather now seems to mean conspiracy theories running amok about the Bureau of Labor Statistics' 'birth/death' adjustment to the monthly payroll figures. ..."

JOHN'S RESPONSE:

There is no conspiracy theory in this matter. What has evolved and is going on is well documented by the Bureau of Labor Statistics.

I also have had first-hand experience in discussing the theories behind, and the changes made to, the bias-factor process as the system evolved, going back to the introduction of the precursor to the current concept in the years following the double-dip recession of the early 1980s. I also refer you to the material written in the Shadow Government Statistics background piece on employment. [See "Concluding Comments" section.]

Here are the basics. Jobs growth was underestimated coming out of the double-dip recession, and that led to a revamping of reporting biases so as to account for jobs created by start-up ventures that were missed in regular BLS surveying.

There is some justification for adding an upside bias to the surveying during periods of strong economic growth, but even BLS economists acknowledged that the use of bias factors is questionable during times of slow growth or recession.

The new bias factors in the 1980s were just plug numbers and were recast every quarter or so. But generally, they were little changed on a month-to-month basis and were added into the unadjusted data.

Aside from serious sampling flaws in the payroll survey and its annual benchmarking, the recent move from “Bias Factors” to a “Birth/Death” model has made the monthly reporting unnecessarily volatile. If the payroll data were properly seasonally adjusted, monthly biases in the reported data should not be predictable, but they are.

The current net annual birth/death factor is running about +900,000, in line with historic annual net bias factors. What the overly sophisticated and nearly worthless (in terms of statistical significance) models accomplished was to provide big monthly swings to the biases, and they enabled the BLS to justify the previous gut-instant plug factors with impressive looking statistics.

As a quick aside, in terms of broad results, three of the last four benchmarks revisions have been to the downside, and that does not account for multiple changes. For example, the last regular seasonally-adjusted payroll level of April 2002 was revised downward by 523,000 in May 2002, and then again by 265,000 in May 2003, for a total downward revision of 688,000 due to benchmarking.

To see the current birth/death problems, consider the actual monthly bias changes as used, where they swung from +66,000 in December 2004, to -280,000 in January 2005, to +100,000 in February. While such changes are added to the unadjusted data, the overall seasonal factors are reasonably small on a base of 130,000,000 or so. As a result, a swing of +380,000 jobs in the unadjusted January to February numbers from “without bias” to “with bias” ends up with a bias shift of similar size in the seasonally adjusted numbers. Accordingly, as predicted by Shadow Government Statistics, the February jobs change was an upside surprise to the markets, as January had been a downside surprise. This pattern has been consistent since the new models have been in use, and again, should not be detectable or predictable if the BLS system were in balance.

As to conspiracy, I think the markets are just disquieted by the realization the last highly touted payroll gain was little more than the “guesstimated” -- the model results are really no better than that -- plug factor added to survey data. What the markets have not recognized is that most recent reporting, in terms of statistical significance (confidence intervals published by the BLS) has not been distinguishable from no growth.

The markets read much too much significance into the monthly swings!