In their coverage of government and other economic reports, the business press routinely tells readers that the figures they are relaying are "seasonally adjusted." That is, raw results are smoothed out to supposedly "remove normal, recurring variations" in data.
There's one notable exception: The government's monthly employment report.
As has been their habit for as long as I have been following these things, various Associated Press reports in the past several days (examples here, here and here) and at yesterday's New York Times failed to tag yesterday's reported gain in payroll jobs of 142,000 as "seasonally adjusted." There appears to be no reason for this other than to discourage curiosity among those who might otherwise be inclined to dig deeper. Doing so sometimes reveals that the seasonally adjusted figures understate the job market's strength. Other times, they appear to overstate it. As seen below, especially in the private sector, the poor seasonally adjusted results of the past two months are nowhere near as bad as the underlying raw data:
As seen above, in comparison to the same month in previous years, April, May, June and July weren't too bad (not good enough in terms of getting the millions of people who are genuinely unemployed who are not included in the official figure of 5.1 percent employed, but at least mildly heading in the right direction). July's overall job loss of 964,000 (in a month when many teachers, professors, and others in public education are between school years) was the lowest in at least the past decade, and could defensibly have converted to a seasonally adjusted result of roughly 300,000.
But the wheels started falling off in August, and it got even worse in September — in both cases worse than the seasonally adjusted figures indicate.
Total nonfarm job additions in August and September of 197,000 and 558,000 were both the lowest respective months since 2010. While the seasonal conversion to 142,000 in September seems reasonable (at worst overstated by 25,000 or so), the August conversion to 136,000 significantly overstates the underlying poor result by at least 100,000. How can raw job additions have declined by over 200,000 from last year (from 391K to 197K) while the seasonally adjusted result only dropped by 77,000? Even if the seasonal adjustment was technically done correctly, it's not correctly informing us.
The situation is worse in the private sector. August's 24,000 jobs added is the lowest August result since 2009 — when the economy was still losing jobs shortly after the recession's official end in June of that year — by 130,000 jobs. Its current seasonal conversion to 100,000 jobs is about 75,000 jobs higher than the underlying data can support. (I should also note that when they were released last month, virtually everyone, including "experts" cited by AP, thought that August's original results would be revised upward this month because that has been the case in most previous years. Oops: It turns out that August's original results were revised significantly downward as was July.)
September's private-sector result showing 525,000 not seasonally adjusted jobs lost is the worst performance of them all. It's over 200,000 below September 2014, and the worst September since 2009. The seasonally adjusted result, compared to the conversions seen from 2010 to 2013, could easily have been to zero or a barely positive figure.
Thus, even though the seasonally adjusted figures the press is reporting are being described as "meager" (AP) and "grim" (New York Times), they're really far worse. At least readers here know that, even if the establishment press won't tell anyone else.
Cross-posted at BizzyBlog.com.
