Ignore the Employment Report Punditry!

Each month the investment community waits with bated breath for the Employment Situation Report. Bond and stock markets often move significantly. Pundits intone wisely about the “miss” from expectations. They delve deeply into the “internals” discussing small moves in various categories.

It is 95% wrong! While it provides a vehicle for filling air time and pixels, the analysis is often far from the mark. How do I know? The early reports are merely estimates. They are well-intentioned and executed soundly, but the methods have built-in errors.

No, the errors are not the widely-maligned birth/death adjustment, or seasonal adjustments, or political machinations. These are all fertile ground for pseudo-experts and conspiracy buffs, but completely misguided.

The key employment data represent estimates. This is equally true of the ADP report and the “official” BLS data, as well as other attempts by Street economists. At the time of these estimates, there is only partial data from survey responses. Even after two revisions there is still sampling error in the BLS report of +/- 100K or more.

Which Estimates are Best?

There is a way of “keeping score” on these estimates, but no one pays any attention. The actual count is not known for about eight months, when solid data from state employment offices is tallied. No one exaggerates the number of employees when it means extra tax payments.

Sadly, the market prefers instant and inaccurate data and ignores the later solid numbers.

We can compare both the ADP and the BLS to the eventual count from the Business Dynamics series, drawn from the Quarterly Census of Employment and Wages. Here is what it looks like for the last ten years.

We can observe three things:

  1. The ADP and BLS reports are highly correlated. When they are too high or too low, it is usually together.
  2. In the 2008 era the ADP report was off by a significant margin. They have changed methods since then.
  3. Most importantly, both estimates were way off – far too high – at the end of 2017. This is the last time for which we have actual data. This should be a big story, but it has gotten absolutely no attention.

Advice for Friday

For traders trying to “game the number,” good luck! Despite contrary claims, there is no good way to guess the sampling error. It is far better to track this as a long-run series, included with other economic data as part of an overall picture.

I have written scores of posts on the employment report, but this one is a favorite; I managed to include all of the key concepts and also to have some fun.

Another key post was my quiz, inviting you to test your employment report IQ. There are fifteen questions. No one who has tried it, including professors in labor economics, has gotten more than half right. If the Friday morning pundits had to answer 5 questions right before speaking, we would have a long period of silence!

Long-term investors can play the volatility by viewing the trading as an irrational offer from Mr. Market, buying or selling as indicated.


Thanks to Eric Lo for work on data analysis and charts.

You may also like