Employment (NFP, private NFP) exceeds consensus slightly The recession is not here as far as I can tell. Nor does it seem to have been here in H1.

First consider the nonfarm payroll employment level measured in three ways:

Figure 1: Nonfarm payroll employment as reported (black), Bloomberg consensus as of 9/1 (red square), nonfarm payroll employment incorporating preliminary benchmark revision (teal), and civilian employment adjusted to nonfarm payroll concept (chartreuse), all in 000’s, s.a. Bloomberg consensus assumes no revision to previous months data. Source: BLS via FRED, BLS, Bloomberg, and author’s calculations (see post). 

While the NFP series was revised downward over the preceding two months, it was still continuing to rise throughout the months in which some observers have claimed there was a recession. One could fault the establishment survey, upon which the NFP series is based, because of a faulty birth/death model. In fact, it was allegations that the birth/death model understated firm growth (and hence employment level estimates during the GW Bush years) that drove in part the development of a household survey based measure adjusted to the NFP concept. This issue is discussed in detail here.

Another point – the benchmarking of the the level of employment involves use of the Quarterly Census of Employment and Wages, which in turn should account for the near-exact number of firms in the economy. The preliminary benchmark lifted the level of employment for March 2022, which I’ve incorporated into the NFP series graphed above (as Implied Benchmarked). I wedged in the increase into the previous twelve months, then assumed increases per the reported NFP from April onward (see this post).

Since the household survey measure of employment does not involve the birth/death model for firms, that series should be immune to mismeasurement of that type (it’ll have greater measurement error due to the smaller sample size, though). On a quarter-on-quarter basis (i.e., 3 month change on monthly data), no series has registered negative growth. Only on a month-on-month basis did one series (the household series) show negative growth (in April), equaling a -0.5% annualized growth rate. To put this into context, the standard deviation of m/m changes annualized is around 3.9 percentage points over the period 2021M06-22M08 (all percentages calculated in log terms).

What about completely independent measures of employment growth. The newly revised ADP-Stanford Digital Economy Lab measure registered slower growth than the BLS series (see discussion here). However, it was still rising.

Figure 2: Private nonfarm payroll employment as reported (black), Bloomberg consensus for BLS series (red square), nonfarm payroll employment from ADP-Stanford Digital Economy Lab (light blue), all in 000’s, s.a. Bloomberg consensus assumes no revisions to previous months data. Source: BLS via FRED,  ADP, Bloomberg, and author’s calculations (see post). 

The BLS series was rising q/q annualized 3.5% vs. ADP-Stanford’s 2.6%.

Finally, what about aggregate hours worked. Here’s a picture of that series, including one adjusted for the preliminary benchmark.

Figure 3: Aggregate Weekly Hours Index as reported (black), incorporating preliminary benchmark revision (teal), 2002=100, s.a. Source: BLS via FRED, and author’s calculations (see post). 

While aggregate weekly hours are not rising all the time, the are generally up over the past year. Interestingly, there is a big jump in January 2022, right in that period when some observers have argued a recession was beginning.

Finally, does the Sahm rule indicate a recession has started (this is usually interpreted as a coincident, not leading, indicator).

Figure 4: Three month trailing moving average of unemployment rate minus minimum unemployment rate in preceding year (3.5%) (black), and trigger value of 0.5% (red dashed line), both in %. Source: BLS via FRED, and author’s calculations. 

Note that we are nowhere near the trigger value of 0.5% in the above graph. Nor have we been.

More By This Author: