We won’t have the BLS series for private NFP as of Friday. Can we use the ADP-Stanford Digital Laboratory series (which missed consensus by 84K on the downside)
I think the answer is yes…and no. As noted by other observers, the month-on-month correlation of ADP and BLS series is low. However, the quarter-on-quarter is fairly high. A regression of BLS series on ADP series (3 month log differences) yields:
Over this period, the q/q growth rates are highly correlated, and have roughly unit elasticity. Using these estimates, I generate the following nowcast.
Figure 1: Private nonfarm payroll employment, implied preliminary benchmark (black), nowcast (light blue), +/- 1 standard error interval. Source: BLS, author’s calculations.
While the point estimate suggests a decrease in employment, when the report comes out, the wide standard error interval suggests that a positive reading is as possible as negative.
