September Employment: Weak or Strong?

The September employment numbers reported Friday by the Bureau of Labor Statistics (BLS) tell two different stories - one weak, one strong. Which are we to believe?

Before delving into the data, let's look at the differences in the Bureau's two major employment surveys. One is the payroll survey of nonfarm wage and salary jobs collected from a sample of about 140,000 businesses and government agencies, and which requires a monthly change of at least 96,400 jobs to be statistically significant. The other is a broader survey of about 60,000 households focusing on the civilian population age 16 and over, which requires a monthly change in employment of 436,000 or more to be statistically significant.

Since not all monthly changes in civilian employment from the household survey exceed its comparatively wider error range, the media unfortunately largely ignore these data even in months when the change is statistically significant. Never mind that the BLS reminds data users that the results of both surveys should be looked at to assess the employment picture. Even then, when the two surveys tell different employment stories, as in the September report, sometimes only one of the stories is highlighted in the BLS summary press release.

The leadoff sentence in Friday's release says that "Nonfarm payroll employment edged up by 103,000 in September..." with the next sentence adding that the increase "partially reflected the return to payrolls of about 45,000 telecommunications workers who had been on strike in August." Thus, the weak picture was highlighted.

Further down the press release in paragraph five, about the household survey data, we read that "Both the labor force and employment increased in September." But, oddly, how much September employment increased is not mentioned. Let's mention it.

Since the payroll and household surveys use different definitions of employment, in order to fairly compare one series to the other, it's first necessary to make adjustments to the data to make them comparable. The BLS obligingly does that, though the adjusted series lurks deep within the Bureau's Web site and doesn't make it to the sunshine of the monthly press release.

In its monthly online release entitled "Employment from the BLS household and payroll surveys: summary of recent trends," the Bureau adjusts total civilian employment from the household survey to the definition of payroll employment. (The series is also seasonally adjusted.) It does that by subtracting agricultural employment, nonagricultural self employment, unpaid family workers, and workers absent without pay from total household survey employment, and adding in nonfarm wage and salary multiple jobholders.

For the month of September, household survey employment so adjusted rose by a whopping 880,000 - a statistically significant increase certainly worthy of mention and wholly inconsistent with the minute rise in the highlighted payroll series. Of course, one of the problems with mentioning inconsistencies is that you're likely to be asked for an explanation.

It could be that the small rise in September payroll jobs is understated and will be revised upward in the next month or two, as was the job count in the past two months. Also, a chunk of the payroll employment data is estimated by BLS to compensate for the survey's lag in picking up jobs from new small businesses. These estimates are subject to later correction, sometimes substantial. And data errors certainly can't be ruled out.

It could be that the September increase in the adjusted household employment data is overstated, a possibility supported by the lack of improvement in the unemployment rate. On the other hand, the unemployment rate is to some extent elevated by the jobless receiving extended unemployment insurance who are required to be looking for work. Some of them otherwise might have dropped out of the labor force.

However, the labor force participation rate continued to drift upward last month, often a sign of encouragement to the jobless waiting on the sidelines, which supports the interpretation of a stronger job market. Additional support comes from BLS household survey data adjusted for seasonality but not to a payroll definition. They show that nonfarm wage and salary employment (including self-employed workers whose businesses are incorporated) rose a staggering 1.05 million last month.

The estimated increase of 880,000 definitionally adjusted jobs in September would normally be expected to lower the unemployment rate by about four tenths of a point, to 8.7 percent, taking into account growth in the working-age population as well as some of the employment going to the jobless outside the labor force. October should be a clarifying month.

Though the September jobs picture is confused, there is some reason for optimism, certainly more than a superficial assessment of the data seems to imply. At a minimum, the sharp rise in adjusted household employment to some extent deflates the expectation of imminent recession. It should also put on hold any further thoughts of yet another federal government temporary job fix.