Frank de Libero on Putting the Jobs Numbers in Context
This post is by reader Frank de Libero.
————————————————
The Jobs Numbers in Context
Toward the beginning of each month the Bureau of Labor Statistics presents its estimate of the number of jobs created during the previous month. These estimates are usually large numbers, in the tens or hundreds of thousands, and are given with little context. Thus they often elicit lame, misleading or even fantastical statements about their meaning.
Here are some examples of attributed meaning to the March 9 estimate of 97,000 new jobs in February 2007:
• “Nonfarm payroll employment continued to trend up (+97,000)…”
BLS News Release, March 9, 2007
• “Job growth slowed last month but still remained firm…97,000 — slightly more than analysts expected.”
Jeremy W. Peters, NYT, March 10, 2007
• “…economists were generally happy with the latest employment data and said it shows that the labor market remains healthy.”
Conor Dougherty, WSJ, March 10-11, 2007
• “The latest jobs report is further evidence that the doomsayers aren’t right about the state of the U.S. economy. There were 97,000 new jobs last month…”
Michael Darda, WSJ, March 13, 2007
• “The U.S. job-creating machine is firing on all cylinders.”
Larry Kudlow, NRO, March 13, 2007
Whatever the motivation behind such assertions, sincere or propagandistic, they misrepresent what is actually happening. Furthermore, how does the average citizen relate to an estimate like 97,000? On the face of it, it is just a big number.
To better understand job growth, common analytic methods applied are to: display the raw job numbers over time; use moving averages; chart a function of the number of months after an event like a recession; look at period-to-period percent changes; graph employment-to-population ratios.
What is variously missing in these approaches is, for example, being able to: identify specific months such as September 2001; make meaningful comparisons over time or across administrations; evaluate against some objective criterion; directly take into account US population growth; clearly represent the data so that it can tell its own story.
Interestingly, working-age population data can supply context and further our understanding of the jobs data. The BLS has monthly payroll and population data from January 1948 on. The Bureau’s default working-age population is defined as ages 16 and up. More realistic is to use ages 18-64. It turns out that these data, ages 18-64 beginning with January 1948, correlate highly with the payroll data (simple correlation coefficient > .99). Indeed, a strong correlation makes sense: more people mean more consumers mean more jobs.
As an improvement then, besides the raw or nominal count, the monthly job creation numbers can be expressed on a more human scale, as a rate that adjusts for population growth. This can easily be accomplished by reporting the monthly estimates as the number of jobs created per 1,000 adults (JPT) aged 18-64. That is: {total monthly change in the number of jobs created} divided by {total working-age population in 1,000’s for that month}. This simple metric would enhance understanding. Furthermore, it would put job-creation numbers in real terms so reasonable comparisons can be made, just as we use the CPI to make dollars comparable over time.
Explicitly recognizing the important role population growth plays in job creation, at least for developed countries, shows that economic policy often affects only the margins of job growth (recall, one JPT is one new job per 1,000 people per month), and can help belie incredible claims about job creation.
For example, Allan Hubbard and Edward P. Lazear, WSJ, October 2, 2006 wrote “In the past three years [we have added more jobs] than all the jobs added in the European Union and Japan combined.” Such claims continued to appear on the Whitehouse website at least into February 2007, e.g., here. Assuming a similar relationship exists between job and population growth with the EU and Japan as with the US, a quick check in the CIA World Factbook and a little arithmetic show that for every one person added to the EU and Japan combined, we put four new people into our population. Of course our job creation is higher. One might just as well claim that “Under our administration we have fewer people per square mile than in the EU or Japan.”
With the JPT index we can put the payroll numbers in perspective. For instance, the average JPT since January 1948 is 1. Since the overall average is 1 and the working-age population is now about 186 million, the March payroll numbers would have to be 186,000 just to be average. Individual data points can be identified and put in context, such as September 2001. Beginning with Truman, the average Democratic JPT is more than double the average Republican JPT. Bush I and Bush II are virtually tied for the lowest average JPT of all administrations, at 0.35 and 0.36 respectively.
Here is a chart to illustrate the value of the index. I believe it speaks for itself, and shows how the example assertions given at the beginning misrepresent.
In addition to displaying the general value of the index, this chart also shows that approximately since June 2005 the aggregate job situation is getting worse. And that is just the aggregate numbers, not about the kinds of jobs.
—————————————————
Update from cactus: Apologies. As posted earlier, the links did not work.