Two numbers leapt out from this morning's jobs report, and they seemed to point in different directions: On the one hand, payroll numbers went up to 163,000 in July — the highest number of jobs added to the U.S. economy since February. On the other hand, the unemployment rate also edged up, from 8.2 percent to 8.3 percent.
Democrats and Republicans have wasted no time in cherry-picking the numbers that work to their political advantage. But the mixed message is the outcome of an economy that's recovering in fits and starts, as well as inherent technical shortcomings in how we measure the labor market in the first place.
First, the numbers are drawn from two separate surveys of the labor market: The payroll numbers come from what's known as the "establishment survey" of about 140,000 businesses representing about 490,000 establishments, while the unemployment rate is drawn from a household survey that goes to 60,000 households.
Payroll numbers according to the establishment survey were modestly positive: They showed "a pace of growth which, if sustained, would lead to lower unemployment over the medium run," says Justin Wolfers, an economist at the University of Pennsylvania.
But in terms of measuring jobs, the establishment number isn't complete: The payroll survey is meant to capture private firms and businesses, but it isn't great at capturing self-employment, work for family businesses, and firms that have recently opened or closed, which the household survey can better measure, explains Gus Faucher, senior economist for PNC Financial Services. And according to latter, things aren't looking as great: Employment fell by 195,000 in the household survey, Wolfers points out.
"The two reports often move in tandem, but can move in opposite directions, especially in months such as July where there are big seasonal issues at play," explains the Wall Street Journal's Phil Izzo, who does a good job detailing how the unemployment rate is calculated from the household survey.
So where is the labor market really headed? Because the sample size for the payroll numbers is larger, it is "generally regarded by economists as being the more reliable indicator," says Wolfers, while the household survey that determines the unemployment rate tends to be more volatile and less reliable on a monthly basis.
Why, then, do we rely on the statistically less reliable survey to determine the unemployment rate? Because we have no other choice. "A survey of firm payrolls isn't going to have a list of the people that aren't working," says Jared Bernstein, former economic adviser to Vice President Biden.
Given these month-to-month movements — and the revisions to employment numbers that frequently crop up afterwards — economists prefer alternative means to measure the overall pace of jobs growth. Wolfers, for example, has two rules of thumb:
1. I like to focus on the average jobs growth in the payroll survey over the past three months. That's currently running at +105,000 jobs per month.
2. Or to focus just on July, I put 80% weight on the survey of firm payrolls, and 20% weight on the household survey, which yielded an underlying pace of +90,000 jobs per month.
In either case, "it looks like we're creating about enough jobs to soak up new entrants to the labor market, but not more," Wolfers concludes. "At this underlying pace, unemployment will neither rise nor fall (although it will bump around a bit due to statistical noise)."
In fact, a closer look at July's unemployment rate itself reveals a similar trend, as my colleague Brad Plumer noted this morning: Technically speaking, June's rate: 8.2165 percent, while July's rate was 8.2535 percent, which got rounded up to 8.3 percent — a difference that Bernstein describes as "statistically indistinguishable from no change at all."