On September 5, the latest “Employment Situation” release for August became available from the U.S. Bureau of Labor Statistics (BLS). One of the closely watched numbers in that release is the latest month-to-month (July to August) change in nonfarm payroll employment. According to the release, the preliminary seasonally adjusted estimate is that there was a net increase of 142,000 jobs in August. Commentators immediately pronounced this number to be disappointing because it was lower than (someone) expected and lower than recent employment gains.
Download this week's Musing.
Some readers may recall the Al Jolson’s line from The Jazz Singer, “Wait a minute; wait a minute, you ain’t heard nothing yet.” (If not, go to https://www.youtube.com/watch?v=22NQuPrwbHA or watch here.)
The problem with getting excited or disappointed as a result of the reported monthly change in employment is that the preliminary data, which already reflect the vagaries of seasonal adjustment, are subject to major revision. So we need to stand back and “wait a minute” before reading the monthly figures as guides to policy or to the general state of the economy. Maybe we haven’t heard “nothing” when we get the monthly number, but the “something” we have heard is pretty fuzzy.
The chart below shows the reported preliminary monthly changes for 2013 versus what now (as of September 2014) appears on the BLS website:
As it turned out for 2013, not only are the deviations of the preliminary figures relative to the current figures large, but they don’t even average out over that year. That is, except for July 2013, the preliminary numbers consistently understated what is now taken to be the true figure. Only in July 2013 was there an overestimate. Almost all of 2013, as seen initially, was disappointing relative to 2013 as seen today. The average error was a monthly understatement of about 30,000 jobs, i.e., something like 360,000 jobs over the year.
In past musings, I have made the heretical statement that maybe we don’t really need monthly releases and maybe, say, quarterly releases would be better. Of course, we will continue to do monthly releases because that’s how it has always been. Moreover, any change in frequency would become the subject of conspiracy theories about manipulated figures. But it would be nice if commentators just focused on changes over longer periods than just one month. Why not use August 2013 to August 2014 (a full year)? If you do it that way – and use just the numbers that are not seasonally adjusted since you are covering a full year – it turns out that we are creating net jobs per month at a rate of around 209,000. You can then decide whether that number is disappointing or not and what policy response, if any, should be taken. You may be right or wrong in your thinking but at least you won’t be basing it on statistical noise.