BLS Stats vs. the Truth: Behind the Numbers

by: Tyler Durden

The confidence game in turbo boost mode. The market is oblivious of the underlying data behind today's "better than expected" BLS computer model output. Yet for the few remaining who do care about the increasingly irrelevant fundamentals, we provide some observations. Here are facts from John Williams' Shadow Stats:

July usually sees a regular pattern of planned automobile production line shutdowns to accommodate retooling for the new model year, but recent disruptions to the auto industry have changed pattern this year. Without the usual pattern of shutdowns, the government’s computers nonetheless responded by creating the usual offsetting boost in jobs, not only in the auto industry, but in supporting industries as well. The auto industry itself was alone among durable goods manufacturing industries in showing a reported, seasonally-adjusted monthly gain in July, up by 28,000 jobs. [Would anybody who recently got a job at Chrysler and GM please write us immediately]

...

While Wall Street likely will hype the July employment results as confirmation that economy has turned the corner, such hype and resulting overly optimistic expectations should be slammed in the months ahead, when the positive reporting distortions reverse out in a normal catch-up process.

...

The unadjusted annual declines in the June and July payrolls remain the deepest since a similar decline at the trough of the 1958 recession, but still shy of the 4.9% trough seen in the 1949 downturn. When the 1949 annual low growth is broken, most likely next month, the annual percentage contraction in payrolls will be the most severe since the production shutdown following World War II.

...

- Birth-Death/Bias Factor Adjustment. As discussed in SGS Newsletter No. 51, Birth-Death Model biases tend to overstate payroll employment during recessions. Never designed to handle the downside pressures from an economic contraction, the model adds a fairly consistent upside bias to the payroll levels each year, currently averaging about 76,000 jobs per month. The unadjusted July 2009 bias was 32,000, up from 25,000 the year before, but down from 185,000 in June.

And some observations from David Rosenberg:

U.S. nonfarm payroll surprise but less than meets the eye Today's employment report is being treated as a 'green shoot' of major proportions. While it was by far the best jobs performance of the year, much of the better-than-expected tally in nonfarm payrolls reflected the bounce in auto production as well as the distortion from the federal census workers. Combined, these two influences effectively "added" 100,000 to the headline number, so net-net, the consensus view of -325,000 was not as far off the mark as the market believed at first glance.

  • Payrolls came in at -247k in July — the consensus was at -325k

  • Upward revisions by 43k to the back data

  • Unemployment rate down to 9.4% from 9.5% (but as in Canada, due to a sliding labour force — down 422k in the U.S.)

  • We are going to see a big increase in industrial production for July — manufacturing workweek jumped 0.8% MoM

  • Income was surprisingly strong — average weekly earnings rose 0.5% MoM (after a 0.3% decline in June)

The auto sector added 28,200 to the industry payroll in July, which was the highest tally in 11 years. To show you just how big that really is, it is a 69% annualized surge. Normally, the industry, which is in secular decline, posts job losses of between 20,000 and 30,000 consistently, so this alone represented roughly a 50,000 swing. We estimate that there was about a 30,000 swing in the rest of the manufacturing sector due to the spillover from the current inventory adjustment in the motor vehicle industry. The 0.3% MoM increase in the workweek was also skewed by the 4.1% MoM jump in the auto sector.

[T]here have been large fluctuations in the federal government payroll too. After hiring a slew of Census workers in the spring, there were 57,000 layoffs in May-June and then we saw in today’s report that 12,000 federal workers were “hired” in July. Again, mathematically, this contributed about 20,000 to today’s headline number. In other words, and we have no intent on raining on anyone’s parade, there was about 100,000 non-recurring payrolls in that top-line figure. It may be dangerous to extrapolate today’s report into a view that we are about to fully turn the corner on the job market front.

To be sure, the drop in the unemployment rate was a surprise, but it was all due to the slide in the labour force — the employment-to-population ratio gives a more accurate picture of the slack in the labour market and the hidden secret in today’s report was that this metric slid to a 25-year low of 59.4% from 59.5% in June and 61.0% at the turn of the year. Of those unemployed, 33.8% of them have been unemployed now for over 27 weeks — a record amount (was at 29.0% in June and was at 17.5% at the start of this recession).

And to conclude with Rosie:

“Bear markets have three stages – (i) sharp down, ii) reflexive rebound, and iii) a drawn-out fundamental downtrend”.


We have little doubt as to which stage we are in today.