Flash Update
A L E R T
November 6, 2006
__________
October Jobs Data Appear Rigged
Unemployment Drop Statistically Indistinguishable from Increase
Jobs Gain Statistically Indistinguishable from Decline
__________
For purposes of full disclosure, my background is as a conservative Republican with a Libertarian bent. Shadow Government Statistics is not a political newsletter, but rather one dedicated to accurate economic and financial reporting. Irrespective of who is in office, I do my best to tell it as it is. -- John Williams
"WASHINGTON, Nov. 3 -- Republicans seized on a drop in the unemployment rate to assert on Friday that tax cuts were invigorating the economy, highlighting just four days before the election an issue that party strategists are counting on to off-set bad news about the war." So opened the lead article in Saturday's New York Times.
Bill Clinton's Labor Secretary Bob Reich discussed in his memoirs how polling surveys had shown that if you overstated economic growth, you could fool enough people possibly to swing a close election.
With continued Republican control of both the House and the Senate at risk, the Bush Administration had both the motive and the opportunity to manipulate the October labor report in its favor, to generate a reported five-year low in the unemployment rate. Beyond the reported gain in jobs and drop in unemployment being statistically indistinguishable from a drop in jobs and a gain in unemployment (based on BLS 95% confidence intervals), were the data rigged? While there is no smoking gun, a strong odor of cordite permeates the air. The level of pre-orchestrated hype and a wide variety of unusual reporting characteristics in the October labor data argue strongly in favor of manipulation.
Up front, there is nothing illegal about rigging economic reporting. The only prosecutable offense to my knowledge is if an individual (i.e. in the White House, Federal Reserve), who is outside the reporting agency (the Bureau of Labor Statistics in this instance), gets access to the data before 4:30 p.m. Eastern Time the day before release. There could be an interesting story here for someone in terms of when the Republican "party strategists" began counting on the data in question to offset the political pressures from the Iraq War.
The October employment report came against the background of rapidly declining economic activity in September, as seen in retail sales and industrial production. September help-wanted advertising, which leads October employment, crashed. New claims for unemployment insurance, net of annual comparisons to the impact of Hurricane Katrina, was little changed, and the purchasing managers October employment indices were little improved.
Our comments in the last newsletter as to the upcoming employment report: "The political needs of the Administration remain great in this last major economic release before the election, and risk remains high of political massaging. Happy news remains a fair bet, regardless of any miserable underlying reality and other indications that would suggest a possible jobs loss."
In general, both the payroll survey data (jobs growth) and household survey data (unemployment) are of poor reporting quality. As a result, the employment growth in the payroll survey often is sharply at odds with the household survey, as seen in the October report. Also, since the BLS began readjusting the seasonal-adjustment factors every month, the BLS has been able to report pretty much whatever it wanted to. Recent reporting has tended to show soft jobs growth, which pleased the credit markets, and improving unemployment, which has been a theoretical plus on the political front. Such was the pattern seen in October.
Payroll Survey. Here is how the numbers were reported. The seasonally-adjusted October gain in payroll employment was 92,000 +/- 108,000, which was statistically indistinguishable from a contraction. Yet, October was up by an extraordinary 231,000, net of revisions. September's previously reported 51,000 jobs gain nearly tripled in revision to 148,000.
The monthly bias factor (birth/death model), added for good measure into the jobs report for October, was upped to 73,000 from 57,000 in October 2005. That means that the 92,000 gain reported in October was just 19,000 net of the bias gimmick.
The revisions, however, were extraordinary, for the second month in a row. They had the effect of boosting total jobs created by the Bush Administration over time, without rattling the credit markets too much with the current month's gain. Much of the revised gain was in education, as the BLS suddenly picked a large number of teachers who somehow had not shown up in initial September reporting.
Also, as has been the recent pattern, the seasonal factors have been unusual. When properly adjusted, the year-to-year change in the seasonally-adjusted and not-seasonally-adjusted series should be virtually identical. Based on the revisions to the data. The unadjusted level of September payrolls was revised upward by 184,000, but the adjusted levels lost out again, only revising upward by 139,000.
Nonetheless, year-to-year changes in September and October adjusted payrolls were unusually inflated versus the unadjusted numbers. Adjusted versus unadjusted annual growth rates for September and October were 1.483% versus 1.408% and 1.469% versus 1.425%. The difference is that instead of respective monthly gains now of 148,000 and 92,000 for September and October, the gains would have been 126,000 and 60,000. For the two months combined, the current gain of 240,000 would have been 186,000 with consistent seasonals.
Household Survey. The household survey showed a seasonally-adjusted gain in employment of 437,000. That reflects the number of people with jobs, as opposed to the payroll survey, which reflects the number of jobs, where individuals can hold more than one job.
The popularly-followed U3 unemployment rate in October declined to 4.42% +/- 0.2%, from 4.58% in September, a change that was statistically indistinguishable from an increase in unemployment. On an unadjusted basis, U3 declined to 4.1% in October from 4.4% in September.
Unemployment rate U3, however, usually moves with the less-followed but broader U6 unemployment rate. Unadjusted U6 held steady at 7.6%, while the seasonally-adjusted U6 unemployment rate rose to 8.1% in October, from 8.0% in September. Since U6 is less subject to attention and, accordingly, to manipulation, it is the more credible number and suggests an upturn in October unemployment.
Politically, the Administration needed something dramatic in these numbers, like "a five-year low in unemployment." Aside from the number being ridiculous against the background of other economic data that show a deepening recession, Wall Street is going to start raising an equally absurd concept, that of the economy nearing full employment. Such a concept would pressure the Fed to raise rates. Accordingly, look for the unemployment rate to rebound sharply in the next report.
As to the election, Main Street U.S.A. usually has a pretty sense of how it is doing and usually does not believe the government's numbers, when they vary too far from reality. Betting on bogus jobs numbers to turn Republican prospects is a real long-shot, regardless of Bob Reich's polling studies.
Further detail will follow in November's "Shadow Government Statistics," targeted for release on Monday, November 27, 2006, following the Thanksgiving holiday. There will be at least one other intervening Flash Update.