Explaining Yesterday's Seasonally Adjusted Nonfarm Payroll "Beat"
Since there still is confusion regarding yesterday's whopping "surge" in non-farm payrolls, which represented a 243K jump in the Establishment survey (of which 490K was temp jobs, same as in the Household Survey where temp jobs soared by a record 699K), yet only to arrive at an employment number last seen ten years ago, when the US population was about 30 million lower (think about that: 30 million increase in population and no change in the total employed), here is the final explanation of what happened yesterday.
As everyone knows by now, January is when the BLS imposes its annual seasonal adjustment revision (more on that in a second) for the previous January-December period. What this manifests itself most directly in, is the divergence between the Nonadjusted January number of the establishment survey of any given year and the Unadjusted number. And while the January adjustment is always substantial, it is the fact that the so-called beat was entirely based on assumptions that makes yesterday's NFP number so meaningless, and hardly the basis to assume that the US economy has taken off.
The chart below shows the adjusted and unadjusted employment survey data for total Nonfarm Employees. The annual January overadjustment is more than evident. Just as evident are the subsequent under-adjustments as seasonal data is lowered to account for volatility in the NSA data. What is very notable is that in January, absent BLS smoothing calculation, which are nowhere in the labor force, but solely in the mind of a few BLS employees, the real economy lost 2,689,000 jobs, while net of the adjustment, it actually gained 243,000 jobs: a delta of 2,932,000 jobs based solely on statistical assumptions in an excel spreadsheet!
So how does this data fit in specifically in the context of the just passed NFP whopper of a number? Simple. The chart below shows the January seasonal adjustment for the past 4 years, since 2009. The number of jobs added for "seasonal" purposes to the NFP number were as follows: 2009 - 2,006,000; 2010 - 1,970,000; 2011 - 2,129,000, and the all important 2012: 2,146,000. Once again, this is the number added to the NFP unrevised baseline to get a "final" number which is then blasted to the media. The chart below shows the historical January adjustment, to the NFP data, as well as the 2012 reported adjustment, and also what the statistical adjustment would be for the NFP number to have the NFP number come in line with expectations of a 140,000 beat.
Here is the kicker: the market mood yesterday would have been far more somber if instead of a seasonal fudge-factored statistical addition of 2,146,000 jobs, the BLS had decided on a number that is merely the simple average of the statistical adjustment of the past 3 years, which comes down to 2,035,000. In fact, had the BLS used this seasonal adjustment, the final NFP headline number (SA) would have been +132,000, or a miss of expectations of 8,000 (the Seasonal Adjustment number to get to consensus January expectations would have to be +2,043,000 to the NFP number). In other words, the difference between a + and - 2% move in the stock market is based on less than a 5% variation to the entire January seasonal adjustment, as had the BLS add just the simple average, the BLS report would have been a disappointing miss, and the market would have likely dropped (although with 5 momos in charge of the entire market, the thesis would have likely promptly shifted to "more QE coming" so who really knows). And now you know how the BLS' seasonal adjustment, which as Charles Biderman pointed out yesterday is guarded as secretly as Coke's recipe, defined the tone and the mood of the market for at least one month.
Finally, as to some newsletter and namesdropping blogs allegation that the Labor Force did not, in fact, increase by 1.2 million in January, we have one simple question: just how does one "refute" a statement with an assumption? Because last we checked, the BLS did not provide a smoothing breakdown of how it applied its seasonal adjustment for the "population control effects" which saw the population increase by 1.7 million in January and those not in the labor force rose by 1.2 million. Quite the contrary , what the BLS did provide is Table C: "December 2011-January 2012 changes in selected labor force measures, with adjustments for population control effects" which does show how on an apples to apples basis the adjustment factors did in fact impact the two key components in determining the unemployment rate: the amount of Americans employed, and those not in the labor force.
And while one can try to say it is inconceivable to say that the US population jumped by 1.7 million in one month, we reply that this is coming from the BLS whose admission of the "population control effect" adjustment merely confirms that it has been misrepresenting the actual labor force participation rate for at least a year. In other words, while one may pander to semantics, and believe that a data series is not a data series because of one's mastery of sophistry and assumptions, this is totally irrelevant: the end result is that in January, those "not in the labor force" did in fact rise by 1.2 million (whether compared to December or to 2011 - please, go ahead and check as many times as needed), and the labor force participation rate dropped to a new 30 year low of 63.7%, a number which incidentally only has to drop by 5% more percent for the BLS to report zero, or even a negative, unemployment rate.