Monday, October 8, 2012

Is the BLS Fudging Employment Numbers for the Prez?; The Real Test is Coming 4 Days Before the Election

Were the employment numbers released by the Bureau of Labor Statistics cooked? It's possible that seasonal adjustment factors had something to do with the gains. Why and how adjustment factors are adjusted, only the BLS insiders know for sure, but here's an interesting find by Ed Yardeni:
 While I doubt that anyone at BLS tampered with the household data for political motives, I’m certain no one even thought to bother with the payroll employment numbers. September’s increase was a measly 114,000. I give much more weight to the revisions to the previous two months, which tend to be upwards when the economy is expanding. They totaled 86,000 during July and August, raising their monthly average gain to 161,500. The oddity here was that upward revisions occurred at the local-government level--mainly the hiring of school teachers (up 77,000)--which nearly matched the revision to overall payrolls.
Here's what I wrote in 2008 about the seasonal adjustment problem with teachers around this time of year:
As far as unemployment numbers, on September 7 [2008] we commented on the much weaker than expected payroll numbers released at the time:

An odd outlier was employment in local government education which fell by 32,000 in August, as seasonal hiring was less than usual. This unusual education jobs number accounted for approximately 30% of the difference between consensus forecasts and the actual payroll number. This would be the first time in history that a recession was led by summer school teachers!

Almost a month later, with the masses looking for the next distorted piece of government data, WSJ comments on the outlier we discussed on the day of the release:

Many economists suspect the drop in August payrolls was exaggerated by a fluky fall in local government payrolls, and new data from the Bureau of Labor Statistics supports that.

On Tuesday the BLS released state payroll data for August. If you sum up the changes across the 50 states and the District of Columbia, the total rose 159,000, compared to the decline of 4,000 in the national data.

That doesn’t mean the national tally is wrong; sum-of-the-states data are hampered by differing response rates by states and the fact the BLS seasonally adjusts each state separately, notes Ray Stone of Stone& McCarthy Research Associates.

That said, he says the difference is unusually large. In part that may simply be catch-up, since the sum of the states total has lagged the national total for some time. But he says the principal source of divergence in August appears to be in government payrolls. The national tally of government jobs fell 28,000, while the sum-of-the-states tally rose 88,000.

Mr. Stone takes this as evidence that the national data have been distorted by quirky seasonal adjustment, which is made difficult by “the timing issues.surrounding academic years, and the inconsistency in how teachers are paid over the summer. Some get paid on a 12-month basis, others on a 9- or 10- month basis. Any shift from year to year in the relative incidence of “months-paid” will play havoc with the seasonally adjusted teacher payrolls.”
Thus, this may explain the quirkiness in the payroll employment numbers. But what about the huge jump in the household data (different data than the payroll numbers)? Yardeni has this to say:
 According to September’s controversial household survey data, employment rose by a whopping 873,000, with the labor force up 418,000 and unemployment down 456,000. Monthly gains of 500,000 or more in household employment are rare, but there have been eleven of them in this volatile series since 2000. They are often followed by sharp reversals in the next monthly report. 
Bottom line: While there is clear evidence that incorrect seasonal adjustments for teachers may be skewing the payroll numbers, there are likely unknown (deep in the data) adjustments that are skewing data in the household survey also.

Writes Yardeni about other oddities in the numbers that suggest a correction to the data should be coming:
The employment gain [in the household data] was attributable to an increase of 838,000 in full-time employment, while part-time employment fell 26,000. What’s odd is that among those working part-time (which edged down slightly), there was a 582,000 increase in those working part-time for economic reasons. In other words, lots of people found full-time jobs, and lots of people who wanted to work full time could only find part-time jobs. Got that? Even odder is that the payroll survey showed that employment in the temporary help industry edged down by 2,000 in September. 
As highlighted above, according to Yardeni, these kinds of outliers are usually corrected in the following month's data. But, here's where things get really interesting. The next employment data from the BLS is scheduled to be released on November 2, four days before the presidential election.

That release may tell us a lot more as to whether the BLS is playing footsie with the president or not. Will they take a knife to this months data and readjust it downward, as often happens after such a huge quirky gain?

5 comments:

  1. Nice find on the data inconsistencies!

    ReplyDelete
  2. Think you mean the next payrolls report is out November 2nd, not October.

    ReplyDelete
  3. The answer to the accuracy of the numbers will be disclosed in the November report which, due to some sort of glitch, will be released on November 9th.

    ReplyDelete
  4. How dare you insinuate that the government is cooking employment figures! This is our beloved government! the government of the You Nighted States of America! It would NEVER lie to the Ameri-I-Can people for mere political purposes! Never! Quit your tawdry blog postings!* There. With that off my chest i can get back to watching my football game.


    *did I mention that in college I majored in sarcasm, with minors in cynicism and bowling?

    ReplyDelete
  5. Normally, the new unemployment claims numbers are revised upwards a week after the official release of those numbers. That's how it's been going for years now. Where are the new, revised upward numbers now, including California's reports? Are we just not going to get that updated report now, since Obama is needing to be reelected and it might hurt his chances?

    ReplyDelete