Yes, the BLS is now producing such a steaming pile of waste matter that it might as well be made an arm of the DNC. Or in the alternative, some of the billions that US taxpayers have wasted on it over the years might be recouped from a sale to CNBC. After all, bubblevision’s monthly cheerleading session couldn’t do without it.
In any event, once again this month the labor department bureaucrats did not go out and actually count 242,000 new jobs or even extrapolate them from a valid, scientific sample survey of the Gallup variety.
Folks, they never left their cushy offices; they plucked these numbers from a computer model!
Even before the Census Bureau turned over to the BLS its ragged February data sheets from calling households which do not have phones and surveying businesses which may or may not exist, they more or less knew this month’s numbers.
How? Why they are just fitting a projection line to the trend of the last several years, and then gussying up the resulting hockey sticks with even flakier seasonal maladjustments, re-estimates, birth/death factors and a lot more statistical shenanigans.
So why waste $600 million per year on the BLS when a couple of staffers and a lap top computer at the Democratic National Committee could produce the same propaganda? So could an intern or two at bubblevision headquarters in Englewood Cliffs, New Jersey.
Either way Steve Liesman of CNBC and Barrack Obama of the White House would still be out crowing about how awesome things are on main street. On that score, our clueless lame duck President left nothing to the imagination with respect to Friday morning’s fiction,
“America’s pretty darn great right now,” Obama told reporters Friday as he celebrated a strong jobs report that he said proved Republicans’ “doomsday rhetoric” is little more than “fantasy.”
Let’s see. It would appear that a few pages fell out of Barry’s briefing folder before he rushed to the press room. That is, just about anything that measures what is actually happening in America is not “pretty darn good” at all:
We will dig deeper into the BLS’ trunk of junk below, but just try this one for a smell test. Is it really possible to believe that private “educational services” jobs——which encompasses everything from swank private days schools to boiler room driven for-profit tuition mills——-have been growing at the exact pace of 62,000 jobs per year since 2006?
In fact, the February 2016 number for this BLS establishment survey category was 3.51 million jobs. That was 64k higher than February 2015; and that, in turn, was 62k higher than February 2014, which was, yes, 63k higher than February 2013.
Then if we go back to February 2006 we get a level of 2.90 million jobs. So subtract that from the 2016 number, divide by 10 years, and, presto, the result is 62k new private education jobs per year for the last decade running.
C’mon. There are more cross-currents out there in the blooming, buzzing world of private education than Carter has liver pills. This huge jobs category covers literally hundreds of thousands of institutions from Podunk Iowa Day School through Harvard Law School. The odds that employment grew by almost exactly 62k for the last three years lies somewhere between zero and none .
In fact, this number is not measured; it’s modeled.
Likewise, try jobs in the BLS category called “outpatient care centers”. Here the magic 10-year trend is 29k per year and that compares to 31k jobs gained in February 2016 over prior year; and a gain of 33k and 29k in the two years before that, respectively.
Right. It puts you in mind of the lesser lights in your high school class who stole the answers to the final exam, and weren’t smart enough to get one or two questions wrong in order to cover their tracks.
Let’s say, for example, that outpatient centers have been steadily shifting labor from direct payroll to temp agencies and other outsourced contractors. Accordingly, the actual labor used in outpatient care would migrate to another BLS category, and the job count would not rise by exactly 29k per year and could actually fall if the outsourcing shift were extensive enough.
Shhh! Don’t tell the modelers. As I said, either the DNC or CNBC is welcome to have the entire BLS—– kit, caboodle and all.
And then there is the hideous stupidity of the seasonal adjustment factors—or something called ARIMA, which stands for “auto-regressed integrated moving average”.
Stated in plain English, they created an adjustment factor for February which purportedly embodies average seasonal variations going back five years and projected forward five years. Then they repeat the same ten-year smoothing process for each remaining month of the year.
As it turns out, however, the market-moving and CNBC-fooling headline “jobs” print for any given month is a complete creature of the ARIMA seasonal adjustment razzmatazz. For one thing, the seasonal adjustment factor is often 5X to 12X larger than the headline delta or job gain for the month. The adjustment swamps the adjustee.
Thus, the ARIMA seasonal for February was +1.555 million jobs—-a figure 6.4X the headline print of +242k. During January the headline seasonally adjusted (SA) print was +172k or just 8% of the massive ARIMA adjustment of +2.163 million jobs.
Obviously, just alter the seasonal adjustment factor by a hairline fraction of the underlying NSA jobs total for any given month, and you get an altogether different answer. That is, the CNBC jobs panel would go gumming off in a wholly different direction about the headline print’s meaning for the state of the economy and for the putative state of mind in the Eccles Building based on nothing to do with the real world, but merely the statistical machinations of some GS-16 at the BLS. Or the DNC, as the case may be.
And this is no far-fetched theoretical point. The smoothed 10-year ARIMA adjustment number for February 2016 was +1.555 million, representing 1.095% of the 142.005 million NSA jobs modeled for the month. For February last year the SA adjustment factor was 1.545 million jobs or 1.109% and the year before that it was 1.512 million or 1.110% of the NSA total. Now how do you like that for scientific precession?
For one thing, weather variations in each February are far greater than implied by these tightly clustered ARIMA factors; and for another, seasonals could easily change over the course of the business cycle, with bigger seasonal effects early in the recovery bounceback and smaller ones from a more normalized level of operations late in the cycle.
Likewise, seasonals could also shift owing to secular changes in economic function, such as the shift of retail sales from brick and mortar stores to on-line e-commerce venues. In the latter, a lot of the economic function gets captured in BLS warehousing and package transportation categories, as opposed to retailing.
Needless to say, job counts, average weekly employee hours and load smoothing business operations are totally different between Sears and UPS, for example. Yet the ARIMA factor for way back in February 2006 was 1.523 million jobs——a number that is virtually identical to the 1.555 million used in Friday’s report. There is, however, not a snowballs chance that the US economy’s seasonal structure has remained absolutely unchanged for the past decade.
In any event, if this February was a seasonally warmer month than normal and the evident secular shift toward reduced seasonality has proceeded apace, the proper ARIMA factor for this February might have been, say, 1.001% versus 1.095%. In that event, the headline print would have been +109k, not +242k. And even though the difference is all statistical noise, Friday morning’s “all is awesome” chatter would have had a far different tone.
That is, the Jobs Friday revilers are cackling about statistical shadows. The emphasis everywhere and always should be on the fact that the headline “print” or delta for the month is meaningless noise because at the end of the day it is a double derivative. To wit, the delta in the ARIMA factor for February compared to January was -608k jobs, reflecting a drop in the seasonal adjustment from 2.163 million jobs in January to 1.555 million jobs in February.
On the other hand, the model generated NSA number for February was 142.005 million jobs, representing a +850k gain from the 141.115 million jobs modeled for January. So just subtract the arbitrary big minus (-608k) in the month-over month change in the ARIMA factor from the month over month gain (+850k) on the NSA modeled number and you get the magic +242,000!
As I said, you get statistical noise. And you need look no further than the 55k gain in retail jobs reported for February or nearly one-quarter of the entire “upside surprise”. Except……except the modeled NSA number for February was a loss of 142k jobs!
But don’t sweat the small stuff. The ARIMA seasonal adjustment factor for February was +295k jobs, thereby helping mightily to get the February headline into the awesome category.
No, you don’t have to be a philistine to make this point, and you can even agree that perhaps February is a more “seasonal” month for retail than is January, where the ARIMA adjustment was only +98k jobs or just one-third of the boost to February.
But here’s the thing. What is the likelihood that true seasonality in February 2016 was identical to the 295k ARIMA adjustment for February 2014 and nearly identical to the 301k adjustment way back in February 2006?
There has been a veritable retailing revolution since then. Just ask Eddie Lampert or Jeff Bezos, as the case may be.
In short, the monthly headline jobs number is the most useless data point imaginable. It reflects the capricious result of adding a statistical artifice called ARIMA to a modeled NSA trend that most definitely does not capture the real world jobs market, and most especially not the inflection points when the business cycle is shifting.
As we have indicated previously, the only meaningful number on the short-run employment trend is the daily withholding taxes collected by the IRS and faithfully published for all citizens to see. No business in America withholds taxes for phantom seasonally adjusted employees, and if they hire 20 hour per week part-timers versus 40 hour per week regulars, the withholding data reflects half jobs, not whole jobs.
Likewise, only real businesses send taxes to the IRS, not birth and death statistical constructs. So did the real employers of America sending real wage withholding taxes to Washington convey an economy that was “pretty darn great”.
No they didn’t. In light of wage rate growth of about 2% per year, the implied quantity of labor hours generated in the US economy during February was about negative 2%. The real world data, therefore, reflects not awesome——-but an economy slipping into recession, as Lee Adler’s latest 4-week moving average clearly suggests.
Needless to say, the BLS trend-modeled/ARIMA adjusted/noise ridden monthly deltas have no capacity to anticipate the inflection points when the business cycle rolls over, as it is at present. After years of trended gains during the last cycle, the BLS headline jobs numbers suddenly went into cliff diving mode well after the recession was underway.
But based on the BLS’ monthly noise emission, the Fed minutes for May 2008 and the White House chief economic spokesman both claimed there was no recession in sight.
Then this happened!
There is a further point. The only thing that really matters on the jobs front is the trend of jobs growth over reasonable periods of time, the mix of employment opportunities and the metric for what is actually sold in today’s labor market. To wit, hours and gigs, not standard 40 hour work weeks punched into a time clock.
As to the mix issue, you need look no further than the February numbers themselves. On the one hand, there were 86k new jobs in education and health, 48k in hospitality and leisure and the aforementioned 55k in retail. These categories alone add up to 190k or nearly 80% of the putative gain—even though for the most part they represent part-time work and bottom of the scale pay levels.
By contrast, there was a loss of 19k jobs in energy and mining, a drop of 16k in manufacturing and only hairline gains in the other principal full-time, full pay categories. For example, there are a total of 8.2 million jobs in FIRE (finance and real estate), but the gain was a miniscule 6k.
Likewise, there are 6.6 million jobs in construction, where the gain was just 19k, and all of that and then some was due to the seasonal adjustment factor. And in the huge business and professional services category (20 million jobs), the gain was a scant 23k.
In fact, notwithstanding the noise-ridden nature of the monthly numbers, the February report makes it plainly evident that there is nothing awesome about the US labor market at all. We have focused consistently on what we call “breadwinner jobs”, which on average pay about $50,000 on an annualized basis. Yet once again the February level is still one million jobs short of where it was way back in December 2007 on the eve of the financial crisis.
The point here is that the business cycle expansion is not going to last forever, and at 81 months it is already extremely long in the tooth compared to the 61-month post war average.
In fact, at the 58k rate of breadwinner job gain recorded in February it would take another 31 months just to get back to the level achieved in early 2001! The odds that we will not have another recession during that span are not compelling, to say the least.
So Washington and Wall Street can celebrate the Jobs Friday report all they wish, but it doesn’t change the fact that massive money printing and unspeakable growth in public debt since the turn of the century have produced what amounts to a bread and circuses jobs market. Compared to the shrinkage in breadwinner jobs, here is the picture for barhops, bellboys, wait staff and hot dog vendors.
Indeed, since last February, the US has added 360K waiters; in the same time period, a paltry 12k manufacturing workers have been added as shown in the chart below.
The same pattern holds for the longer term trend since the start of the Great Recession in December 2007. The US now has 1.6 million more waiter/bartender jobs. As to manufacturing jobs—- not so much. In fact, there are now actually 1.4 million fewer such jobs.
Even that understates the reality, however. The manufacturing jobs lost represent about $80 billion annually in reduced wages. The waiter/bartender jobs gained, by contrast, represent only $30 billion in annual wages.
Now, it is true that Barry went to Harvard Law School where there is not much emphasis on math skills. It’s hard to say what Steve Liesman’s excuse might be——except flogging the Wall Street narrative that it’s always time to buy stocks would seem to be the essence of it.