The unemployment rate and jobs total went down? What’s up?

A job seeker talks to a recruiter from UC Health at the Colorado Hospital Association job fair in Denver, Colorado, on October 4, 2017. Photo by REUTERS/Rick Wilking

The unemployment rate dropped from 4.4 percent to 4.2 percent in September, while the economy lost 33,000 jobs. Impossible, right?

Well, no. As we've explained here for years, the two numbers, released Friday by the Bureau of Labor Statistics in its September jobs report, come from two completely different monthly surveys: one of households (a sample of 60,000), and the other of "establishments" — places that employ folks (a sample of 147,000 businesses and government agencies). The unemployment number comes from the household survey; the jobs created figure, by contrast, comes from the establishment or "payroll" survey.

So why did they differ so dramatically last month? Two probable reasons. One may be that September's storms and hurricanes put many jobs on hold, but didn't eliminate them permanently. Employment in "food services and drinking places" alone dropped by 100,000, after adding an average of 24,000 a month since last fall. But those jobs-on-hold presumably didn't show up in the household survey, because people on temporary hiatus due to disruptions like a storm are still reported as employed.

The second reason for the discrepancy is one we've pointed out again and again: you can't trust any one month's numbers. In fact, this message is drummed home every month when the Bureau of Labor Statistics itself not only reports last month's data, but revises the data from the previous two months. The revisions Friday? As the BLS put it: "The change in total nonfarm payroll employment for July was revised down from +189,000 to +138,000, and the change for August was revised up from +156,000 to +169,000."

In other words, the supposedly gratifying payroll total reported in August was actually paltry, and in the following month it was better than reported.

Another number pops out in the latest household survey, a number somewhere between weird and implausible: 900,000 more Americans were reported as "employed" in September than the month before. Really? When the number of jobs went down by 33,000? So when interpreting data from any given month — and especially one battered by extreme weather, it's best to exercise extreme caution and not overreact to supposed trends or anomalies.

One of our go-to economists on the monthly numbers, the University of Michigan's Justin Wolfers, tweeted: "That six-or-seven year run — the longest ever run of consecutive positive payroll growth — looks to be over."

This last comment refers to frequent accusations President Donald Trump made as a candidate that the BLS data willfully camouflaged a sagging economy under former President Barack Obama. And though he's now changed his tune and taken credit for the jobs reports under his administration, Mr. Trump may have succeeded in cowing the agency. This month's written summary of data included a curious sentence about the dramatic reversal in the payroll number: "Total nonfarm payroll employment changed little (-33,000)," the agency noted (italics mine). For an economy that has added well over 100,000 jobs a month for years now, "changed little" could be viewed as timorous understatement.

But again, let's not overreact. Wolfers had tweeted that he anticipated "a not-very-interpretable jobs report," and that's just what we got. Which means we should also be careful not to wax enthusiastic over the apparently healthy wage increase included in the September numbers.

Finally, as regular readers know, when the jobs numbers come out the first Friday of every month, we have for years added our own more inclusive reckoning of un- and underemployment, U-7, which includes people who say they want a full-time job but can't find it. The U-7 was 11.4 percent in September, down from 11.8 percent the previous month.