Friday, June 5, 2009

Employment Shocker

by the Sandwichman

May nonfarm employment fell by 345,000, much less than even the Sandwichman's revised prediction (between 570,000 and 635,000). Moreover, the April and March employment totals were both revised upward, resulting in reduced job losses for those months of -504,000 (-539,000 preliminary) and -652,000 (-699,000 first revision) respectively. Taking these revisions into account nonfarm employment now stands at 132,200,000, or 214,000 below the preliminary April number.

The BLS birth/death adjustment added in 220,000 jobs in May, compared to 176,000 for May 2008.

Meanwhile, the number of UNemployed persons increased by 787,000 in May, compared to 563,000 in April. And the unemployment rate increased to 9.4% from 8.9%


Daro said...

My compliments for posting up the prediction miss on the "front" page...

juan said...

Not 100% current but close enough - from EconomicPicData, two charts comparing job openings and # unemployed:

Chart 2 [percent]

Chart 1 [absolute]

ProGrowthLiberal said...

And the employment-population ratio fell to 59.7%. It was at 63.4% as of December 2006. Yet Niall Ferguson is afraid of inflation? Excuse for my post just up.

Have dog food...will travel... said...

With math like that; who needs enemies.

The way they calculate these numbers could lead to a situation where nobody is employed and yet have a less than zero unemployment percentage.

At least a massage therapist is working these days, at least on the numbers.

Jack said...

Excuse my ignorance, but it leads me to this question. Each subsequent monthly report of unemployment is given in specific numbers, total of reduction in employment. If the pie keeps shrinking how can we not expect those reductions to decrease in raw data numbers? Why isn't an emphasis put on the percentage of change from one month to the next, which would seem to be more elucidating? That would allow for one to see a change of pace based upon the number of those still employed the previous month. It seems unreasonable to expect an increasinig loss in the number of employed, which would imply extremely dire circumstances.

Sandwichman said...


Reporting the employment data as a number of jobs gained or lost is a kind of popularizing shorthand. People can grasp "345,000" jobs lost more readily than they can .26 of a percent. Especially with the contrast with the unemployed percentage. Otherwise, there would be more confusion between the two.

It is true that as the pool of employed people shrinks, the same number of job losses represents a larger percent. It's also misleading that the month-to-month comparison juxtaposes two preliminary figures at different stages of data correction.

The dilemma recalls the question "do you want it done cheap, do you want it done now or do you want it done right?" The problem comes to the rush to interpretation, as well as the pressure to find the optimistic and predictive angle of interpretation. To put it bluntly, there is no predictive information in the most recent month's preliminary employment situation report, aside from the truism that next month is likely to resemble this month barring some catastrophic or miraculous event.

Jack said...

True enough that the raw figure of jobs lost (or increase of unemployment) has more graphic effect. Though, I wonder, what is the effect of seeing the same number (approximately) maonth after month and thinking, Oh it's no worse than before, but in fact it is given that an equal number lost from a smaller pool is a larger share. A percentage summarization would more accurately indicate any change of pace.

BillCinSD said...

What's the over-under on the revision of the May number?

Anonymous said...

Perhaps,at some point, someone can explain to me how 787K more folks are unemployed, but nonfarm payroll only falls by 345K.

From where did the people who were not unemployed last month, but who are now unemployed come?

Just as important, perhaps, will we be building more of them next month?

Anonymous said...

I would add this information to my above comment:

Sandwichman said... 787K more folks are unemployed, but nonfarm payroll only falls by 345K...

The numbers come from two different surveys. The unemployed figure comes from the household survey and the employment number comes from the establishment (employers) survey. The simplest explanation for the difference would be that 442,000 more people entered the work force in May than found jobs. That's just the simple explanation. A more complicated explanation might delve into the different methodologies of the surveys, etc.

Regarding the Chris Martenson site and skepticism about the birth/death model: I believe there are some flaws in the B/D model, namely having to do with seasonal adjustment factors and with using "normal" historical time series to model during a crisis.

But the CM alarm about "creating jobs out of thin air" misunderstands the B/D model. Those 220,000 hypothetical jobs mostly replace 220,000 real jobs that disappeared into thin air through non-reporting. The B/D model assumes that not all of those companies actually went out of business and thus not all of the jobs actually disappeared. The BLS is trying to correct for a known systematic bias from non-reporting. Some correction is clearly justified. I just think their model overdoes it. Barron's sounds a more judicious note on the question.