On Friday, February 2, 2007 the BLS released the Employment Report for January 2007.
Nonfarm payroll employment rose by 111,000 in January, and the unemployment rate was essentially unchanged at 4.6 percent, the Bureau of Labor Statistics of the U.S. Department of Labor reported today. Job growth continued in several service-providing industries over the month, and construction employment also rose. The number of manufacturing jobs continued to decline. The number of unemployed persons (7.0 million) and the unemployment rate (4.6 percent) were about unchanged in January.
The expectation was for 150,000 to 170,000 jobs. The number came in at at 111,000. So was the raw number a bad showing? To help answer that question let's take a look at birth/death adjustments.
January and July are typically the revision months for the BLS. Given that many (myself included) harp about all the positive assumptions, this month we see a negative adjustment of 175,000 jobs. In spite of that significant downward adjustment, jobs still came in at +111,000. All in all that is a respectable showing. But what does it say about September through December? Hmm. The Birth/Death assumptions for those months were +221,000. In effect the BLS assumed 221,000 jobs were created in those four months but now it is saying that 79.2% of those jobs were fictional. But rather than telling us what months they were wrong about, we are left guessing.
The BLS website has this to say about Birth/Death assumptions
- The net birth/death model component figures are unique to each month and exhibit a seasonal pattern that can result in negative adjustments in some months. These models do not attempt to correct for any other potential error sources in the CES estimates such as sampling error or design limitations.
- Note that the the net birth/death figures are not seasonally adjusted, and are applied to non-seasonally adjusted monthly employment links to determine the final estimate.
- The most significant potential drawback to this or any model-based approach is that time series modeling assumes a predictable continuation of historical patterns and relationships and therefore is likely to have some difficulty producing reliable estimates at economic turning points or during periods when there are sudden changes in trend.
Given the BLS does not disclose what they are actually basing their assumptions on, what their methodology is, or what months they admit their assumptions were wrong, it is tough to draw real conclusions based on these revisions. We do know they will be wrong at the turn, that sampling errors may be significant, and the whole game of adding jobs Feb-June, and August-December then attempting to correct those mistakes in January and July makes it tough to do much other than look at overall trends. The overall trend was towards weakening numbers throughout 2006. Let's turn to the establishment data for additional clues.
Establishment Data
Although the raw number of 111,000 was respectable given the massive downward adjustment, a closer look at the establishment data raises additional questions and comments.
For starters note that 14,000 of the 111,000 jobs created were government jobs. That is a fairly robust 12.6%, and not a good sign of anything but waste. Note also that manufacturing and durable goods both shed a lot of jobs. This is not actually a sign of a healthy economy. Most suspicious, however, is the 22,000 construction jobs that were created in the midst of a huge housing slump. Remember that those numbers are seasonally adjusted. So what happens to normal seasonal adjustments when weather was unseasonably warm? Hopefully the answer is obvious.
Thus we have the odd anomaly of a weak 111,000 overall number being stronger than it looks but the makeup of that 111,000 being weaker than it looks. Sheeesh.
Are we done yet? Unfortunately not. The BLS issued this notice in the January report.
Establishment and Household Data Changes
The establishment survey data in this release have been revised as a result of the annual benchmarking process and the updating of seasonal adjustment factors. See the note beginning on page 5 for more information on the revisions. In addition, household survey data for January 2007 reflect updated population controls. See the note on page 6 for more information. ....
Revisions to Establishment Survey Data
In accordance with annual practice, the establishment survey data have been revised to reflect comprehensive universe counts of payroll jobs, or benchmarks. These counts are derived principally from unemployment insurance tax records for March 2006. As a result of the benchmark process, all not seasonally adjusted data series were subject to revision from April 2005 forward, the time period since the last benchmark was established. In addition, with this release, the seasonally adjusted establishment survey data from January 2002 forward were subject to revision due to the introduction of updated seasonal adjustment factors. Table B presents revised total nonfarm employment data on a seasonally adjusted basis for January through December 2006.
The revised data for April 2006 forward incorporate the effect of applying the rate of change measured by the sample to the new benchmark level, as well as updated net business birth/ death model adjustments and new seasonal adjustment factors. The November and December 2006 revisions also reflect the routine incorporation of additional sample receipts into the November final and December second preliminary estimates.
The total nonfarm employment level for March 2006 was revised upward by 752,000 (754,000 on a seasonally adjusted basis). The previously published level for December 2006 was revised upward by 981,000 (933,00 on a seasonally adjusted basis).
As you can see from the above table there was a massive revision based on a methodology change. Did those jobs really occur? In the months stated? Or in 2005 or 2004? If those jobs were indeed created in the stated months, then 2006 was a bit stronger than generally reported.
The Ugly
When it comes to jobs data we are seeing revision after revision and some of the revisions and adjustments can not even be added together because they are not compatible. For example the birth/death figures are not seasonally adjusted, but everything else is, and in some sort of magic those unadjusted numbers are applied to seasonally adjusted monthly employment numbers to determine the final estimate. If that was not enough in and of itself, we have new seasonally adjusted methodology changes, new birth/death model adjustments, and to top it all off there are post-benchmark revisions. Does anyone have any confidence in this?
"The February 2007 issue of Employment and Earnings will contain an article that discusses the benchmark and post-benchmark revisions."
I can hardly wait. In the meantime, let's call this report what it really is: ugly.
