Today the market seemed to be spooked by the latest payroll employment report. According to the Bureau of Labor Statistics (BLS), nonfarm payroll employment rose by 115,000 in April. The official expectation was for 160,000, although the whispered number was 125,000 to 150,000. What is most interesting is that in actuality, nonfarm payroll employment rose by 896,000 in April.
Let me explain.
Payroll employment is an economic series known to be affected by seasonality, which refers to the tendency of a time series to exhibit periodic and predictable variations. For example, employment tends to increase in June because of recent graduates entering the job market and students getting summer jobs. Likewise, employment tends to decrease in January because of lower demand for temporary workers following the holiday season. When we analyze a time series affected by seasonality, we are mostly concerned with whether the latest data is more or less than what would be expected given the usual seasonal variations. This process is called seasonal adjustment, and it is used by the BLS in its payroll employment report. Using seasonal adjustments, the BLS estimated that 781,000 of the 896,000 payrolls actually gained in April was attributable to seasonality, and therefore reported a seasonally adjusted gain of 115,000 (896,000 minus 781,000). While it is important to analyze economic data “net” of seasonal changes, the results of seasonal adjustments are very sensitive to the assumptions underlying the models used to perform those adjustments. It is not uncommon for a small change in one of the inputs to make the difference between a gain and a loss in payrolls. That is why we prefer a different approach, which looks at the year-over-year change in the unadjusted payrolls. By comparing this year’s April payrolls with the ones from April of last year, we are able to remove the seasonal effect without having to make any arbitrary assumption that could bias the results. While the year-over-year change contains one year of data and does not isolate what happened last month, monitoring the trend in the year-over-year change over time allows us to gauge the recent direction of payrolls.
The chart below plots the year-over-year change in unadjusted payrolls since 1939 (red line). Historically, year-over-year payroll growth dipping below 0% has been a good indicator of economic recessions. After the steep recovery following the 2008-2009 recession, the number has been stuck between 0.8% and 1.6% for about a year, indicating that while payroll growth continues to be positive, it is far from the levels that have historically been associated with strong economic expansion. This seems to be consistent with the message we are getting from the other employment indicators we watch, which are pointing towards continued yet moderately-paced improvement.