Rebel Traders – Analysis Without The Hype

Unemployment statistics are flawed.

What have I been saying ever sine the recession began, the Bureau of Labor Statistics use of the “birth/death” model is flawed. To adjust the monthly jobs data by a number which is computed from a ‘model’ of how many businesses were created and how many closed up during any given month is simply asking for trouble in maintaining any sense of accuracy.

The main purpose of the birth/death model is an attempt to offset the jobs data by what the BLS believes to be happening in the real world, the only problem is that the model relies on assumptions. And assumptions are worthless in an economic collapse.

When the Government releases Friday’s employment report, nearly a million jobs could be erased. The change won’t show up in the monthly report. Rather, the expected drop will show up in the government’s revised job losses from April 2008 to March 2009, showing the labor market was in much worse shape than we knew at the time.[…]

Blame the birth/death model for the revision. Built on years of research, the model’s key assumption is fairly simple: most of the time jobs created at new companies make up for losses at companies that close. In October, Keith Hall, Commissioner of the Bureau of Labor Statistics, said most of the coming revision “appears to be due in part to an increase in the number of business closings,” short-circuiting the model’s assumption that deaths are offset by births.[…]

The Labor Department says there are flaws in its models and its monthly employment survey but it defends the process, saying the annual employment revisions are small and stable. […] (Source: Bloomberg)

Small and stable? I don’t think a calculation error of 800,000 to 1 million additional job losses is ‘small and stable’.

The birth/death model should be shelved, permanently.

[...]If numbers are correct, N.J. job losses in November made up 60% of U.S. rate [...]

[...] The release of November’s payroll employment by the New Jersey Department of Labor and Workforce Development continues the odyssey of strange monthly numbers that have accompanied the recession. New Jersey lost 10,900 private-sector jobs in November, which in itself is not unusual — during the 22-month recession (January 2008 to November 2009), the state has lost, on average, 8,232 private-sector jobs per month. Thus, New Jersey’s November’s loss, while higher than the average recessionary monthly loss, is not significantly out of line with it.

What is strange is that the United States as a whole lost just 18,000 private-sector jobs in November. If both the nation and state numbers are to be believed, New Jersey by itself accounted for more than three-fifths (60.6 percent, or 10,900 out of 18,000) of the nation’s total November private-sector job loss. That appears to be unreasonable. So, as we have suggested before, caution must be used in analyzing monthly numbers, particularly since annual benchmark revisions, which will be released in the first quarter of 2010, promise to alter them substantially. [...] (Source: NJ Biz)

I’ll take the bet that the New Jersey data is more accurate than the U.S. data.

BTE (Better Than Expected) has been the norm these last times… Pushing the market always higher. Now, how about some WTE (Worse Than Expected)?

The earnings of BAC and of GE this morning have left some bitter taste for the bulls as BAC pulled up a huge loss and GE did not perform as well as could have been expected. Now it would appear that the “real” unemployment numbers have been actually worse than reported as the BLS (Bureau of Labor Statistics) announced it would have to revise the numbers downward because of issues with the “birth/death model”…

It is not useless to recall that Chuck pointed out several times in his videos the issues with this “model” to explain those BTE employment numbers.

The Bureau of Labor Statistics (BLS) recently announced that they will be making downward benchmark revisions to past monthly nonfarm employment data that casts doubt on the validity of the recent figures as well. As we will explain, it is highly likely that substantially more jobs are now being lost than is currently reported.

The BLS makes annual revisions to the previously announced payroll reports to account for job increases or decreases that were not picked up in the initial data.  The main reason for the differences in the preliminary and final reports is the difficulty in getting numbers from the many small and medium sized business and accounting for new startups and firms going out of business.  To make an educated guess at the data that they are missing, the BLS uses something called the ARIMA time series model (commonly called the birth/death model) to estimate employment changes resulting from business births and deaths that are not accounted for by other methods.  The model is based on the actual births and deaths over the five prior years.

As you can imagine, when the prior five years encompassed a period of economic expansion, the application of these numbers to a period of recession can result in a substantial overestimation of job changes, and that is evidently what happened recently.  The BLS candidly acknowledges this and states that it is “likely to have some difficulty producing reliable estimates at economic turning points or during periods when there are sudden changes in trend.”

In the current instance the BLS announced that preliminary tabulations indicated they would have to reduce the estimate of total nonfarm employment by about 824,000 for the year ended March 31, 2009.  On average, therefore, the net change in payrolls for the period was overstated by about 68,000 per month.  Interestingly enough, the birth/death adjustment had added about 717,000 jobs during the same period.  So it’s apparent that the benchmark revision will more than wipe out the entire amount added by the model.

What does this mean for the period following March 31, 2009, which will not be revised until next October?  For the six months since March 31^st the birth/death adjustment has added 815,000 jobs, an average of 135,000 per month.  Since small and medium sized firms are suffering from severe credit restrictions, they are much more likely to have reduced employment significantly rather to have added that many jobs.  That means current monthly job losses may be running as much as 135,000 higher than is currently being reported.  While we won’t know the true number for another year, those being laid off will know, and they will be reducing their spending accordingly.  The Fed certainly knows what’s happening and that’s one reason they are promising near-zero interest rates indefinitely.

When we combine the weak job numbers with declining wages, tight credit, record household debt and the impending explosion of home foreclosures, the chances of a sustainable economic recovery looks exceedingly slim.  Yet, for the third time in this decade the stock market is off on another binge not based on reality.  Such flights into fantasy always end badly.