A Look Into Employment Revision Numbers — a.k.a. Why I Like 12-Month Moving Averages

Wow. The payroll report for February jobs this morning was well received by Wall Street as the S&P 500 promptly hit an all-time record high. That’s a lot of faith in a single number, and you would hope that number is sound, solid and truly representative of what is going on in the economy.

As I was writing the blog this morning on the just-released nonfarm payroll numbers, I again questioned how valid these original data are given significant revisions which often occur multiple times on the same data. To see how large these swings in data are, I went back and gathered the original nonfarm, seasonally-adjusted estimates of payroll jobs commencing January 2013. The following table shows the original estimate of total, nonfarm payroll jobs and tracks the changes over time.

3-7-14a graph1

The next table examines the original monthly job gain and compares it against the latest revision. Since February 2013, the original job estimates have increased by 331,000 jobs – a gain of 17.1 percent. In all but one month, the original estimates have increased. Some of the revisions are minor, while others are truly significant. The change from the original estimate of job gains for March 2013 has been revised, increasing that number by 60 percent.

3-7-14a graph2

The previous table might lead one to assume that the government’s methodology systematically underestimates employment numbers initially, but catches up on revisions. I can assure you that is not the case. Just take a look at the original job estimate for January 2009, of 134.58 million jobs. The Department of Labor now has restated that estimate to 133.976 million jobs. Hence they are now telling us that we did not lose as many jobs as thought (604,000 fewer lost jobs), and we have gained more jobs than we have realized (331,000 since January 2013 as restated). That’s a combined swing of almost one million jobs. To me that is material.

If one had a conspiracy outlook, you might claim that the government is manipulating data to look better. I am not a member of the conspiracy camp. Rather, I believe it is extremely difficult to estimate most national economic data on a monthly basis given the scope and variability of this majestic country. It also says to me that Wall Street looks for whatever factoid it can hang it’s hat on for the moment, regardless of how stable that point is.

What is my takeaway on this? Recall that in my analyses and econometric models, I prefer to use a 12-month moving average of whatever datum is under the microscope – whether employment, housing sales or home prices. Stick with the 12-month moving average for a trend analysis that remains relatively stable—even after revisions. And never bet the bank on one month of data of any type.


Leave a Reply