The Big Picture

« Microsoft eyeing Yahoo? | Main | NFP: 88k (and I don't believe even that) »

NFP: The Return of the Over?

Friday, May 04, 2007 | 07:14 AM

Some time ago, we retired the Over/Under bet on the monthly NFP. The initial number gets so significantly revised, that the first release of it is nearly pointless.

This week, I was tempted to revive the over/under bet. Given the ongoing loss of manufacturing and construction jobs, the weak housing and auto data, and the very disappointing ADP payroll report, one should expect a very moderate NFP number.

However, as the astute Bill King observes, the past three Aprils have seen a huge accounting sleight of hand known as the Birth/Death adjustment. This is the BLS guesstimate of how many jobs were created but not measured. It is supposed to capture small firms outside of the establishment surveys direct measure. This guesstimate was ramped up in 2001, and has been a point of major confusion ever since.

The B/D adjustment has become very significant. It has had a major impact on BLS Non Farm Payroll total numbers. In 2006, for example, of the 2.26 million new jobs that BLS reports as being created, 964,000 -- nearly half (42.6%) -- were due Birth/Death adjustments.

For some reason, April is a big month for this statistical twist. We seem to get very strong B/D adds in April, which are then revised downwards.

Here are the recent April B/D adjustments:

April 2005, Preliminary B/D add was 257k. Subsequent revision 206k. NFP = 340k.

April 2004, Prelim B/D add was 270k; Revised to 225k. NFP = 287k.

April 2003, Prelim B/D add was 228k; Revised to 128k. NFP = -48k.

April 2002, Prelim B/B at 66k; Revisid down to 45k. NFP = -83k.

As King observes, "BLS has consistently overstated the Preliminary B/D job creation. After tax, state and other data is available, the number of B/D jobs added have been revised lower."

~~~

The point of this is that a significant upside surprise is quite possible -- and if history holds, that strong number will eventually be revised downwards.

Of course, its a wildcard as to what the actual BLS guesstimates for B/D will be. Understanding how the numbers are constructed will give you a better view on what the reality is beneath the released data. What you do with that is up o you . . .