The 40k increase in May ADP payrolls is difficult to equate to any specific Nonfarm Payrolls figure, since what was previously a fairly consistent downside bias in the ADP figures relative to payrolls has switched to a highly volatile upside bias. The ADP figures overshot payrolls by a hefty 129k/month through Q1, and posted an April overshoot of 39k. Yet, there would need to be a huge 120k bias in May to have these figures be consistent with the drop of 60k expected in Friday’s payrolls data, given a typical 20k government job gain. Thus today’s ADP figures signal notable upside payroll risk on Friday.
The industry breakdown of a 37k decline in May goods employment — with a 26k drop for factories — and a 77k service job gain leaves a broad-based overshoot of these figures relative to most payroll estimates.
Note that the ADP figures are poor predictors of monthly as-reported payroll changes, with an average absolute error since the start of 2007 that has widened to 56k, versus average absolute errors for payrolls overall of 39k for the survey median and 43k for a forecast that is the 6-month moving average.
Even still, the ADP figures are good predictors of eventual payroll revisions, and the persistent overshoot of ADP relative to payrolls since late-2007 implies that the next round of annual payroll revisions may be upward. Until 2007 the ADP figures were also useful for predicting revisions each month, but the current divergence in the ADP and payroll figures now looks likely to persist until the next benchmarking, with diminishing likelihood of interim revisions that close the gap.