April 10, 2007

Inside the Labor Report

Last Friday's labor report was notable not only for the overall count of 180,000 new jobs, but also for the rebound in construction hiring that resulted in 56,000 new positions and the disappointing loss of 7,000 Professional and Business Services spots.

A closer look at some of the data provides some intrigue and raises a few questions.

First, the overall total looks quite good when viewed in the context of the last two years or so - a moderation in job growth after a stellar period in most of 2005 and early 2006. Note the huge 300,000+ gains per month back in 2005, presumably the result of the recent annual benchmark revision by the Bureau of Labor Statistics.

There have been no downward revisions to the monthly data since early 2005, so, if that trend stays intact, look for an improvement on the 180,000 total with next month's release.

On the surface, the most recent seasonally adjusted total appears to be part of a gentle slowdown in job creation, but upon closer inspection, the seasonal adjustments occurring around this time of year are having an outsized impact.

Construction

As would be expected, construction hiring is very seasonal, even more so in colder parts of the country where snow and ice linger for many months during winter. The normal pattern of hiring can be seen clearly in the raw data below (blue bars), a good portion of which would be better termed "raw estimates" as some of this data is derived from the birth/death model.

So, in the March report, a total of 181,000 new construction jobs were reported/estimated and of that total 27,000 were derived from the birth/death model, meaning that 154,000 new construction jobs were reported by state unemployment insurance agencies.

The seasonally adjusted total, however, only comes to 56,000 and herein lies the intrigue. As shown in the chart above, March is the first month of the year when the seasonally adjusted number is less than the reported/estimated number, the prior six months always showing the opposite effect in order to help "smooth" out the data (e.g., January 2007 saw 289,000 construction jobs lost, but the seasonally adjusted total was +34,000).

What does this mean?

This is a particularly strange time of the year as seasonal adjustments combined with erratic weather patterns can play havoc with the data. The next three months of construction hiring will provide a much clearer indication of what is in store for the rest of the year since, understandably, April through June are the peak months for construction hiring and are seasonally adjusted downward more than any other months.

If the recent cold snap depresses construction hiring in parts of the country for the next month's report, look for a truly horrible construction number in the April data.

Professional and Business Services

A similar effect can be seen in the professional and business services employment category, where, last month over 100,000 new positions were created, but the seasonally adjusted total came in at -7,000.

Astonishingly, the 378,000 jobs lost in this category in January showed up as a gain of 12,000 after the seasonal adjustment and you can look back at some of the late 2006 gains posted for this category only to see actual losses or only piddling improvement.

The Bureau of Labor Statistics has been doing seasonal adjustments for many, many years so they surely know what they are doing and the "adjusted data" is, without a doubt, more meaningful than the raw data simply because of the convoluted explanation that would be required to properly describe the meaning of the reported data.

Can you imagine the headlines in February if 378,000 professional and business services jobs were reported to be lost in January?

Well, it happens like that every year.

This time of year, at least for construction payrolls, the raw data may be more meaningful than the seasonally adjusted data.