Ev Ehrlich's Everyday Economics

The Sorcerer’s Apprentice

The Bureau of Labor Statistics announced on Friday that the economy added 192,000 jobs, which is only modestly above the level needed to keep job growth over the growth rate of the working population, but still the largest monthly number since the “false dawn” of February through April of last year.  Back then, the economy suddenly produced over 300,000 jobs a month, but then promptly ran four job-losing months in a row, partly because the initial gain included the hiring of Census workers who left those jobs after the Spring count, and partly because the crisis in the Eurozone sent a new wave of fear into financial markets.  So while most economists – or at least the ones I agree with -- expected the employment number to be higher, at least the direction is right – up.

With all that said, and mindful of my role earlier in life as Defender of the Statistical System (when, as Undersecretary of Commerce, the statistical system was my bailiwick), I don’t believe the number.  The economy has to be generating more jobs than these numbers suggest.  And there are a bunch of reasons to think so.

The first is simply that, going back to September, the initial estimate of new jobs created has been revised upwards in each of the two following months.  (An “initial” estimate of the number of jobs created in any month, like last Friday’s estimate for February, is revised in each of the next two monthly reports, whereupon it sits for the better part of a year, when it is given a final revision as part of an annual review.)  September, 2010, saw job losses of 95,000 when first reported in early October – early December BLS took this loss down to 24,000.  October started at 151,000 new jobs and ended up at 210,000: November went from 39,000 to 93,000; December from 103,000 to 152,000.  And given that errors usually move in the direction of the trend, there’s every reason to believe that the next annual revision will move the numbers yet higher from where they started.

This doesn’t mean that BLS is half-assed or sloppy.  They understand this pattern of revision and report it candidly.  But their job is not to guess what employment did – it’s to report what employers tell them it was, and then let opinionated blowhards offer their views about why BLS is wrong.

So why are they wrong?  There’s always a little error in the data – that’s why they call them “data.”  BLS surveys over 100,000 business each month to get a basis for estimating job growth.  Sometimes the businesses don’t report on time, sometimes they change their internal numbers, and so on.  But the real catch has to do with how BLS knows that a business exists – how does BLS know whether to put a business on the list from which it draws this sample of 100,000+ companies (which represents over 400,000 “establishments,” meaning an individual factory or store). 

The answer is that they learn of a business’s existence when that company pays state unemployment taxes or otherwise reports itself to the government.  In some cases, that lag can be short, in others, it might go on for a while.  For example, my son Nick is a personal trainer and boxing instructor in Baltimore.  He draws no salary – he’s paid by clients or by gyms or other facilities that want him to attract clients for them.   The government “learns” of his employment at the end of the year, when he files a Schedule C on his tax return.  That’s one reason why there are after-the-fact annual revisions – because there are millions of “proprietors” who work that way. 

That’s an extreme example of the problem.  When the economy grows, new businesses are formed, and it takes the system a while to learn that they’re there.  BLS understands that and, using past experience, derives a statistical correction to the employment report for new businesses that they think are out there but haven’t reported yet, but I think they’re undercounting that.  For example, ADP – the largest company that processes payrolls for companies - using their own data, think that companies with 49 or fewer employees have added 600,000 jobs over the last year, from February 2010 to February 2011.  Half of that total occurred in the first nine months and the other total in the last three.  So there’s good evidence that the pace of job creation by small businesses is accelerating, and those are the ones BLS’s process has the hardest time capturing.

Moreover, like ADP’s, other measures of employment are showing bigger gains than BLS’s survey of establishments.  The most important of these is the Current Population Survey, which the Census Bureau undertakes on behalf of BLS in order to measure the unemployment rate.

You can’t estimate the unemployment rate except by asking people if they’re unemployed.  You can track the number of people filing unemployment claims, but not all unemployed people do – some aren’t eligible, some have benefits that have run out, and so on.  Many critics point to people who decide the-heck-with-it and stop looking for work (“discouraged workers”) but, once again, BLS’s job is not to guess what these folks would do if a job suddenly jumped out of a tree and bit them on the neck – it’s to report how many of them there are and let the opinionated blowhards do the rest.

So BLS hires the Census Bureau to interview 60,000 families every month, from which it estimates the unemployment rate.  In some ways, it’s a little dicier an estimate, because 400,000 business establishments is a bigger and better chunk of all the economy’s business establishments than 60,000 families is a chunk of all the nation’s families.  So the sampling errors are larger for the survey of households.  Moreover, just as with businesses, new households are harder to catch – for example, a young person or family leaves their parents’ house and goes out or their own, or immigrants arrive, or people move.  But young families usually split off when they have a job, and people move to go to new jobs, so it’s possible the survey undercounts the employed.  On the other hand, if my trainer son (wait, that doesn’t sound right…) were to be in the “household survey,” he’d be counted as employed, as opposed to not being included in the “establishment survey.”

Regardless, over the past three months – that is, from November 2010 to February 2011 – the survey of business establishments says the economy’s added slightly over 400,000 new jobs.  But the survey of households reports that it’s added 1,000,000 in the same three months.  The ADP survey is closer to 600,000, but also north of the BLS establishment survey.

And other economic evidence, however, circumstantial, support a higher number.  Initial claims for unemployment insurance are dropping and are about 25 percent lower than a year ago.  Retail sales continue to grow, and new orders and shipments of manufactured goods are 9 percent and 8 percent, respectively, over their levels of last June, the nadir following the Spring, 2010 “false dawn.”

The economy faces serious headwinds, among them the fiscal retrenchment of the states and the rising (although, in my view, temporary) price of oil.  And there’s the risk of federal budget cuts at level above the economy’s ability to withstand their impact.  But right now the economy is beginning to grow and employment isn’t tracking it.  The numbers we’re seeing, then, suggest that output is being produced the way Mickey Mouse produced brooms in The Sorcerer’s Apprentice segment of Fantasia.    I don’t buy it.  We’re going to find out soon that there’s more employment in the economy than we now think.