Two Takes On The ADP Number

Tyler Durden's picture

The following two takes on this morning's ADP number are from two opposites: Goldman, with its transition to a permabull, and Knight Capital, with its far more balanced outlook on the economy.

From Knight's Brian Yelvington:

ADP, the best (and really only) predictor of Friday’s monthly jobs data, printed at a very high 297K gain for December versus expectations of a 100K gain.  We have noted before that “best” here is a pretty low bar and the ADP report should be considered in its own right, and not just a forward look at the official numbers.  ADP overestimated November’s jobs data (by 43K), but underestimated the prior 6 months (average difference of 55K).
 
That being said, the 297K print is hard to argue with.  270K of the jobs were in the services sector, so this raises expectations for the ISM Non-Manufacturing number due out at 10AM.  We will closely watch this number for confirmation of the ADP data, but there is historically not a huge basis to argue with the number.  Even adjusting for holidays and noting the service bias, it is not out of line.  Service jobs accounted for about 97% of ADP December job gains and 84% of all ADP prints over the past five years.  A confirming ISM number at 10AM will significantly raise expectations and estimates for Friday. 

And from Goldman:

MAIN POINTS:
1. The monthly ADP employment report surprised sharply to the upside in December, posting a gain of 297,000 jobs versus consensus expectations of 100,000. This is the best ADP reading currently on record (the official data go back to January 2001, though the series has only been released publicly since 2006). Notably, the initial report for June 2006-since revised-was +368k, which substantially overstated the comparable BLS first print of +90k.

2. The bulk of the December 2010 surprise came from small- and medium-sized service sector firms (+120k and +123k respectively, up from +48k and +30k in November). Goods sector employment growth was still soft, with manufacturing up 23,000 (vs +15k in November) and construction up 6,000 (vs. +10k in November). ADP's estimate of financial sector employment was down 6,000, versus a gain of 2,000 in November.

3. The ADP data have a special quirk that could have affected today's report. ADP records payrolls based on the number of names on the payroll-regardless of how many hours they work during the week. Not every firm immediately "cleans" payrolls when an employee quits or is laid off; in some cases, it can take until the end of the year for the payroll list to be officially updated. This creates a lot of volatility in the December report in particular. In theory, one would expect greater purging in payrolls in bad years (like 2008 and 2009) and less in relatively better years (2010 was hardly spectacular, but at least payrolls were up on the year). Of course, the official report attempts to adjust for this behavior, but if 2010 saw relatively less purging than the sample period, it's possible some of today's improvement could be the result of this data quirk rather than genuine acceleration. Given the potential for an overstatement, we have put a -1 judgmental adjustment on our US-MAP reading, which still records a significant upside surprise.

4. Luckily, we will have a useful cross-check of the ADP report later on this morning. If service-sector employment really is accelerating sharply, we'd expect to see the employment index of the ISM non-manufacturing survey (which was 52.7 last month) post a meaningful increase. This report will be released at 10am and will help us gauge how much weight to put on the very strong ADP report.