Nonfarm payrolls rose more than expected in the initial estimate for November (+266,000), with upward revisions to the gains for September and October (a net 41,000 higher). In contrast, the ADP estimate of private-sector payrolls rose more modestly (+67,000). What to believe? The unemployment rate has been little changed over the last three months, at around 3.5%. How much slack remains in the job market? The Fed is widely expected to remain on hold this week, but how will the labor market data influence Fed policy in 2020?
As a reminder, the employment report consists of two separate surveys. The establishment survey covers about 142,000 businesses and government agencies, representing about 689,000 individual worksites. Survey responses generate estimates of nonfarm payrolls, hours, and earnings. The household survey, which samples just 60,000 representative households each month, yields estimates of labor force participation and the unemployment rate. There is statistical uncertainty in each of these surveys and seasonal adjustment is often difficult, adding noise to the figures. The monthly change in nonfarm payrolls is reported accurate to ±110,000 and the unemployment rate is reported accurate to ±0.2%. The data are benchmarked once a year – the household survey data in early January, the establishment survey data (incorporating actual payroll tax receipts) in early February. The Bureau of Labor Statistic has indicated that the upcoming benchmark revision will lower the March 2019 level of nonfarm payrolls by about 500,000 (or 0.3%).
In looking at a graph of monthly payroll changes, one can clearly see that they are volatile. One should take any particular month with a grain of salt and focus on the underlying trend. Statistical noise and seasonal adjustment issues will tend to balance out over time. Taking the three- month average reduces the impact of noise, but does not eliminate it. The three-month average for private-sector payroll gains was +200,000 in November, vs. +149,000 for the three months ending in August. Private-sector payrolls averaged a +165,000 gain for the first 11 months of 2019 (vs. +215,000 in 2018 and +172,000 in 2017).
One caveat is that the payroll trend tends to miss at turning points. The BLS uses a birth/death model to account for business creation and destruction. That model does a very good job in normal circumstances, but will mislead at the start and end of a recession. The birth/death model added 274,000 to the unadjusted payroll in October, about the same as a year earlier, which seemed suspicious given anecdotal evidence of softness. The birth/death model was less of a factor in November, subtracting 13,000 (vs. -5,000 a year ago).
Seasonal factors were at work in November, reflecting the start of the holiday shopping season. Unadjusted payrolls rose by 622,000 (vs. +522,000 in November 2018), with most of that in retail and couriers (up 566,000, vs. +587,300 a year ago). Leisure and hospitality added 45,000 jobs on a seasonally adjusted basis, but fell by 200,000 before adjustment.
In a speech given in early October, Fed Chair Powell noted that nonfarm payrolls had significantly underestimated the rate of job losses in the early part of the 2007-2009 recession. Had Fed officials known, they would have taken more aggressive action. Since that time, the Fed has worked with the payroll processing firm ADP to construct a weekly payroll estimate to complement the official BLS figures. Those weekly data are not available publically, but ADP publishes a monthly estimate of private-sector payroll growth, using methodology that is similar to that of the BLS. ADP estimated a 67,000 gain in private-sector payrolls in November, with a three-month average of +104,000 – half of the BLS trend (note that we need a little less than 100,000 job per month to absorb new entrants into the workforce).