MBS Day Ahead: Revisiting the Great NFP vs ADP Debate

Today brings the release of the ADP Employment Report.  At times, this data has been maligned as the a horrible predictor of the official number from the Bureau of Labor Statistics (BLS) released 2 days later, but at other times it's been a significant market mover.  How are we to know when it will matter and when it won't?  Unfortunately, there are not hard and fast rules in that regard, but generally speaking, it's been less and less relevant in 2014 after showing some good potential in 2013. 

Before digging any deeper into today's potential significance, here's an overview of this great debate from The MBS Live knowledge base.

Part of the reason ADP may be garnering less attention in 2014 is that 2013 saw high hopes that the recent change in the report's methodology would make it a more relevant early indicator of BLS prospects.  In other words, if ADP was moving higher or lower, markets increasingly saw NFP doing the same.  The improving correlations broke down over the past 6 months, for whatever reason. 

If you want to blame something, that blame is probably better placed on the government's numbers.  Government data collection is a huge bureaucratic mess of inefficiency.  ADP, on the other hand, is the world's largest payment processor and got that way by growing in the free market.  Their technology puts BLS to shame and they could easily produce a more accurate month-to-month reading on the economy.  In other words, ADP does a better job at tracking changes in payrolls.

Unfortunately, the ADP Employment report isn't designed to simply report changes in payrolls.  It's designed to predict how BLS will report payrolls.  In that regard, ADP's role is to try to think like a mad man and predict what he will do next.  As you can imagine, if you're not actually mad, it's hard to know exactly what a mad man might do. 

Here are some new charts to frame the evolving relationship between the data sets.

This first chart is simply the initial release (no revisions) for ADP and BLS.  Note: in all cases, we compare BLS's PRIVATE PAYROLLS as opposed to the Nonfarm Payrolls.  The outright numbers here are less useful to financial markets than the following chart which shows how often the two data sets "beat" or "miss" their respective forecasts together.  In other words, if ADP is stronger than expected, what are the chances that NFP will be as well?  (This chart also shows the magnitude of beats and misses so the higher the lines are above zero, the bigger the beat--vice versa for misses).

As mentioned above, the data is falling out of favor in 2014, and it's the sorts of big gaps seen in late 2013 and early 2014 that damage the correlation.  Again, to be fair to ADP, this is really BLS's fault for the big swings bet.  In general, ADP has always been the more even-keeled report.

The final chart is included for those who are curious, or for those who point out that much of the long-term correlation between the data sets is due to ADP being able to revise their numbers to better match BLS (it's named in honor of the man who does that more reliably than anyone over the years, MBS Live hall-of-famer and Victor Burek). 

In this chart, if the blue line is above zero, it means Friday BLS numbers came in above Wednesday APD numbers.  If the red line is positive, it means ADP was revised higher (or "moving toward BLS").  Basically, the more the two lines are on the same side of zero, the more ADP is revising to bring itself closer to BLS (keep in mind, that IS ADP's stated goal though, so it's not a scandalous discovery--merely interesting).

As you can see, there's a very strong correlation here (somewhere in Texas, Victor Burek is reading this and saying "well of course there is!").  With last month's BLS data absolutely shooting the moon (not charted), it's highly likely that ADP will be revising last month's reading higher today. 

Whether or not markets care, remains to be seen.  They certainly have bigger fish to fry with tomorrow's ECB Announcement.  Even later this morning, ISM Non-Manufacturing at 10amis arguably now a more important data-set than ADP.