From the BLS:
And on the benchmark revision:
The total nonfarm employment level for March 2010 was revised downward by 378,000 … The previously published level for December 2010 was revised downward by 452,000.
The following graph shows the employment population ratio, the participation rate, and the unemployment rate.
Ryan Avent at Free Exchange at The Economist:
LOOK almost anywhere in the recent economic data and the signs point to an accelerating recovery. A solid fourth quarter GDP report contained a truly blockbuster increase in real final sales. Manufacturing activity is soaring. Consumer spending is up and the trade deficit is down. Markets are trading at their highest level in over two years. And so economists anxiously awaited the first employment figures for 2011, hoping that in January firms would finally react to better conditions by taking on lots of new help.
Instead, the Bureau of Labour Statistics has dropped a puzzler of an employment report in our laps—one which points in many directions but not, decidedly, toward strong job growth. In the month of January, total nonfarm employment grew by a very disappointing 39,000 jobs. This was not at all what forecasters were expecting. Earlier this week, an ADP report indicated that private sector employment rose by 187,000 in January; the BLS pegged the figure at just 50,000. There were some compensating shifts. December’s employment gain was revised upward from 103,000 to 121,000. November’s employment rise, which was originally reported at 39,000, has been revised to a total gain of 93,000.
But there is bad news, as well. The BLS included its annual revision of the previous year’s data in this report, and while job growth over the year looks stronger than before, the level of employment looks worse. In March of last year, 411,000 fewer Americans were working than originally reported. And thanks to a weaker employment performance in April through October, 483,000 fewer Americans were on the job in December than was originally believed to be the case. For now, the economy remains 7.7m jobs short of its previous employment peak.
The BLS press release makes this very clear in a box right at the top, which says that
“Changes to The Employment Situation news release tables are being introduced with this release. In addition, establishment survey data have been revised as a result of the annual benchmarking process and the updating of seasonal adjustment factors. Also, household survey data for January 2011 reflect updated population estimates.”
The effects here are large and unpredictable: the total number of people holding jobs in December 2010, for instance, was revised down by a whopping 452,000 — but despite that, the official December 2010 payrolls number now shows an even bigger month-on-month rise than it did before. More generally the size of the total civilian labor force was revised downwards by 504,000, almost half of which came from the Latino population. That has all manner of knock-on effects: the BLS warns that “data users are cautioned that these annual population adjustments affect the comparability of household data series over time.”
This is a messy report, then — even messier than you’d expect from a monthly data series which is mainly valued for its speed as opposed to its accuracy. At the margin, it’s bad for markets, which concentrate on the headline payrolls number, and it’s good for politicians, who tend to concentrate on the headline unemployment number. But for anybody who’s neither a trader nor a politician, it’s a noisy series which is best treated with a whopping great amount of salt — especially in January, and especially also when any big-picture message is so murky.
Does this mean that most of the “fall” came from discouraged workers dropping out of the workforce? That would explain the difference between this and the Gallup survey, which showed unemployment rising to 9.8% instead of falling. Or am I missing something?
At first glance I thought that was people dropping out of the labor force, but it seems instead to be the conjunction of two different things. One is upward revisions of the last couple of months’ worth of jobs data. The other is a downward revision to the baseline estimate of how many people there are. Basically, more people had jobs a month ago than we thought had been the case, and also there were fewer unemployed people than we thought had been the case.
The upshot is that the new data looks a lot better than the old data. But the new data doesn’t say the situation improved dramatically over the past month, it merely says that last month’s take on the situation was too pessimistic.
I want a trained professional to analyze this. It is not unusual for the series to do something odd around Christmastide. It is not unusual for the series to diverge. Not this much.