Little noted is the fact that, while May’s payroll employment release surprised on the upside, the household series were providing conflicting indications.
Jim Hamilton pointed out in his post a couple days ago that the establishment and household series agreed on the May net job creation numbers. But what I want to focus on in this post is the short term trends in the levels and the changes, and what these divergent trends might suggest for what’s likely to occur in the future.
Figure 1 below shows the log levels of payroll employment from the establishment survey, household employment adjusted to conform to payroll employment (“adjusted” household), and civilian employment from the household survey.
Figure 1. Payroll employment (blue), “adjusted” Household employment (red) and Household employment series (green). NBER-defined recession dates in gray. Sources: FRED II, BLS, NBER, and author’s calculations.
What’s obvious, even from a distance, is a flattening of the gradient in the household series. In the past, I’ve put greater stress on the payroll employment series because the month-to-month — and even quarter to quarter — volatility of the household series is so much greater than that of the establishment (see this post). Indeed, Jim Hamilton suggests putting weights of 0.8 on payroll, and 0.1 on the household and ADP series, respectively, in this post, while Kliesen suggests (for a different purpose) slightly higher weight on the household survey. But both the household and payroll series seem to have some informational content. And here I think it’s of interest to observe that the household employment series is providing a different picture on where employment is going. Figure 2 shows the monthly change (not percentage change) in employment over 3 month periods.
Figure 2. Three month trailing moving average in change in payroll employment (blue), “adjusted” household employment (red) and Household employment series (green). Sources: FRED II, BLS and author’s calculations.
The household employment growth is at essentially at zero over the past three months. The establishment payroll series and the household series adjusted to conform to the coverage of the establishment series actually match in May. This suggests to me perhaps something interesting is happening outside of the groups usually focused upon. In particular, one might want to examine “… the unincorporated self employed, unpaid family workers, agriculture and related workers, private household workers, and workers absent without pay.” (see Box 1 of Summary comparison of household and payroll survey concepts, definitions, and
methodologies).[empahsis added -- mdc]
I don’t interpret the divergence in the two series to mean that we should now follow the household series. Rather, I think the right perspective is that the divergence might help us glean some understanding of what’s going on in the economy, specifically by looking at the unincorporated self-employed (see ). For instance, are construction subcontractors (who are sometimes taking the hit ) primarily unincorporated self-employed? If that’s not the answer, then is it Big Picture’s Ritholz’s story of mis-estimation of payrolls in the construction sector? (see also Jim’s take on payroll estimates of construction employment ).