The Evolution Payroll Seasonality
In last week’s employment report September nonfarm payroll employment increased 254k (223k private), well above the trailing three-month average of 140k (103k private), which raised discussion on the seasonal adjustment. First, there have been questions whether the seasonal factor for September was unusually supportive compared to prior years, which we don’t think is the case. Second, there are some similarities in the monthly pattern between this summer and last summer, with smaller June August increases being bracketed by larger May and September gains. The timing is similar to possible residual seasonality in initial jobless claims, so the question is whether the same phenomenon could show in employment.
On the first question, two common ways to look at the seasonal factor are as the ratio of the NSA to SA data, or as the percent change in this ratio (the latter is similar to the difference between NSA and SA percent change, with a positive change meaning that the factor anticipates an increase in NSA employment). Under the first formulation (NSA/SA ratio) the value is the smallest ever, but because the ratios for each month need to average out to 1.0 over the year, changes in the seasonal pattern for any one month can influence all the other months. Since seasonal variation across the year has been falling over time ratios will need to move closer to 1.0. Consequently it’s more meaningful to look at the change in the seasonal factor between August and September.

For the private sector the factor change has moved above immediate pre-COVID levels and is close to values in 2016 and 2017, which makes sense given that NSA employment changes have returned to a similar pattern as those years (Figure 1). For total employment the factor change has moderated over the last few years, which all else equal will boost SA employment, in what is a gradual return to pre-COVID factors. Total NSA employment growth has been running a bit stronger than pre-COVID in September, though it likely makes sense that the seasonal factor is more muted, as government over the last year has boosted full-year employment growth by over 0.2% relative to the pre-COVID trend.
Turning to the second question, Table 1 shows the behavior of employment gains around the summer months in the last two years, with shading in cells where the monthly change is lower than both May and September. For total payrolls, June through August are lower than May and September in both years; for 2023 the June gain is almost as high as September, though the pattern was sharper in pre-benchmark values, before that month was boosted by an update to the birth/death forecast. Note that this pattern does not hold for earlier years.

Because different industries have very different seasonal patterns we might expect the pattern to show at the industry level if it represents a change not yet captured in the seasonal, although sometimes residual seasonality only emerges when individually adjusted series are added together. Only trade, transportation, and utilities have the pattern in both years, and even after removing it the pattern remains for 2024 (though not 2023). In 2024 the pattern is also present for professional services and healthcare, though if those are also excluded September remains stronger than earlier months. At this stage we consider this something to watch and analyze further, though would not yet conclude there is residual seasonality.