The Shifting Sands of Employment Data
|By Patrick M. Barkey
June 19, 2006
Americans like to keep score on things. And in the realm of economics, there are plenty of things to keep score on. But the economy is a huge, often unwieldy beast, and the data we use to track it are often quite a bit fuzzier than the rows of hard-looking numbers in the graphs and statistical reports we digest would make it seem. In fact, as the old joke goes, we economists like to present growth rates out to two decimal points just to show we have a sense of humor.
That unpleasant reality is fully exposed in an area of economic performance that has increasingly occupied center stage in the eyes of the general public – job growth. Employment, you see, is not measured once – it’s actually measured four different times by various federal and state agencies. And it is an unavoidable truth that these estimates produce different outcomes. In fact, sometimes they even tell very different stories.
The Bureau of Labor Statistics, in its closely watched monthly employment report, released data showing disappointing job growth in the national economy of just 75,000 net new jobs created during the month of May. Most major business media picked up that story, and many market traders, analysts, and policymakers voiced a reaction.
Flip to a different page of that same BLS report, however, and you’ll see that a different measure of employment, one based on a survey of individual households instead of businesses, showed a job increase of 288,000 jobs for the same month. And as if that discrepancy weren’t confusing enough, this household-based employment measure says that there are nearly 9 million more jobs in the national economy than what the business-based measure says.
It’s a situation that even the professionals who design and implement the data collection and analysis don’t fully understand, even after years of study.
But at least part of it is easy to explain, because the two surveys are actually measuring two different things. The survey of business and government establishments asks for payroll information, pure and simple. How many folks are on your payroll? Tally them up and you get a job count. But if you start with people instead of businesses, what you get is the number of folks who are working. And when people are self-employed, unpaid family workers, or hold multiple jobs, there’s some slippage between the two concepts.
Another part of the explanation is much less comforting, at least to a lay audience. That is the unpleasant fact that both of these employment estimates come from surveys. And the thing about surveys is that no two give the same exact result, even when they ask the same question. The “facts” on job growth are something we can never exactly know, although we’d expect a well designed survey to come acceptably close.
Since the employer-based estimates of employment are based on a much larger sample of workers, in normal times they receive the lion’s share of attention from most analysts, keeping confusion to a minimum. But by anyone’s assessment, the years just before, and just after, the recession of 2001 have not been normal.
Remember the jobless recovery? With the eyes of the policymakers and politicians impatiently trained on the labor market for signs of recovery in the wake of the last recession, the employment data coming from BLS refused to cooperate. In 2003, the U.S. economy lost another 347,000 jobs, on top of almost 1.5 million jobs lost the year before.
But the household-based estimates told a completely different story. Those reports said the economy actually gained 1.25 million new jobs in 2003, with job losses in 2002 fully two-thirds lower than business-based results. The temptation was too strong for some bullish analysts to resist, and the household-based job estimates found their way into the spotlight as the “best” measure of new hiring.
Anyone will tell you that it’s always best to argue from facts. But what do you do when the facts argue with each other?
Business Research | Phone: (765) 285 5926 | Fax: (765) 285 8024 |E-mail |Website: