What's Wrong with the November Employment Numbers

Regular readers of "A Dash" may be surprised to see that I have objections to the most recent payroll employment report results from the Bureau of Labor Statistics.  Ironically, my objections come at a time when many critics say it is a "clean" report.  In addition, I think that the problem relates to the measurement of job creation.

Some Background

I hope that I have established some credibility on this subject.  The BLS has a method for estimating the monthly job change, including job creation.  For several years I have insisted that the right way to keep score was to look at the final results, which we eventually know from state employment data, and test the estimates against those results.

Until recently, the results were excellent.

Something happened.  It did not happen at the onset of the recession, as many critics predicted.  In fact, the BLS method had worked through the 2001 recession, something that everyone ignored.

It did not even happen in Q408, at least not very dramatically.  The problem showed up in Q109, as I reported and discussed (here) and in my November preview.

Let's repeat my recent review of the BLS method for estimating job creation.  If you take a moment to read this carefully, you will see why the critics were wrong before, and are also missing the problem now.

How the BLS Handles Job Creation

The BLS approach is to make an estimate of the total payroll jobs in one month, make another estimate for the next month, and subtract the two to determine the change.  They use an excellent and sophisticated survey technique to do this.  Their historical record, judged by the eventual count from the states, has been very good -- until quite recently.

The Survey Problem.  Any time you do a survey, there will be non-respondents.  When the question is something like "How many people favor health care with a public option?" the non-respondent problem takes a simple form.  You need only ask whether the non-respondents are similar to those who actually answered.  Most polls make this assumption.

The employment question is qualitatively different.  We are not asking the opinions of non-respondents.  We are asking whether they are even still in business.  If the BLS were to assume that non-respondents had all ceased operations, they would seriously underestimate total employment.  Historical data conclusively show that the non-respondents are split between those who did not answer and those who are out of business.  The data also show that new job creation, running at about 2 million jobs per month even in recessions, are a predictable function of dying businesses.

Let me emphasize the difficulty.  There are always non-respondents to the voluntary survey, despite the best efforts to get everyone.  If the BLS assumed that the non respondents were all lost jobs, and that the impact was proportional, we would see a loss of 13 million jobs per month, a silly result.  Instead they attempt to impute business deaths and births.  At one point, they assumed a business birth for every death.  This is the natural result from extrapolating the sample to the entire population.

This is not the +/- 100K jobs from sampling error; it is non-sampling error.  This means that the non-respondents are different in an important way from those who answer the survey.  We know this to be true, so the problem is how to compensate.

The Job Creation Estimation.  Because of this, the BLS employs a two-step process.  The imputation step forecasts job creation from job destruction, and includes a cyclical component..  The Birth/Death adjustment, (the only thing cited by most critics, who ignore the more important imputation step), is a residual.  For many years this residual was stable.  The most recent test against the state data indicated a significant error, showing that the BLS estimates have been wrong for nearly a year, especially since Q1 09.

The Result

The preliminary benchmark revisions show that as of March, 2009, the number of jobs was over-estimated by 824K jobs.  When the official revisions are announced in February, for the January report, there will be three important effects:

1. These job losses will be apportioned to the prior eleven months, lowering each by about 75K per month.  (The actual adjustment may vary for technical reasons, but this is a good starting point).
2. The months after March, 2009, will also be adjusted to conform to a new set of calculations.
3. The Birth/Death adjustment, the calculation of the "residual effect" will also be adjusted.  We may see dramatic downward adjustments for most of 2009.

Two years ago I asked BLS experts if the Birth/Death adjustment could ever be a negative number.  The answer was that while it was theoretically possible, it had never occurred in the recession periods during the development of the model.  It is possible that this adjustment will now become neutral or negative, assuming that the BLS maintains the current methodology.

Conclusions

There are several key conclusions.

1. The universal focus on the Birth/Death adjustment is a blunder.  The critics think that because the B/D adjustment added only 30K jobs (not seasonally adjusted) in November, that the problem does not lie with job creation.  The problem lies in the imputation step -- far more important than the B/D adjustment.
2. Something important happened at the start of the year - probably the loss of credit available to new businesses.  The strong historical relationship used by the BLS finally broke down.  Without a good estimate of job creation, the BLS monthly change is suspect.
3. Private estimates are important.  For many months, preceding the identification of the breakdown in the BLS method, I have emphasized the need to look at other approaches.  This should now be clear to everyone.

There was a general sense of surprise at the November results, but no one has a clear concept of what went wrong.  TrimTabs has entered an objection, and I agree.  The estimates of job change from our model, and the other approaches that I report each month (including TrimTabs), will prove to be better estimates than recent BLS reports.

It will take some months before we see the actual data to prove this, but I intend to follow up with some estimates.  Meanwhile, I doubt that employment has improved as much as the current report indicates.  It is not consistent with other economic data.

And finally, readers should note that this had nothing to do with BLS bias, manipulating the numbers, or creating "phantom jobs" on demand for President Obama.  It is all about methodology, and the inherent limitations on the survey approach.  The BLS team devised a good approach and implemented it in consistent fashion.  The change in the credit markets - not a normal recession -- seems to have undermined their empirical models.

I am reporting about data.  My conclusions are based completely upon where the data leads me.  For many years, the BLS method worked extremely well.  We should now use a variety of methods to assess job changes.

I have a continuing concern about concurrent seasonal adjustment.  More to come....

Comments

[RAN]

Thank you for a thoughtful explanation and for debunking the conspiracy theorists' response to all surprizes, particularly upside surprizes. I also agree with your hypothesis in "Dumbing Down the News" and believe it has a major affect on the conspirators' thought processes.

[Gary]

As usual, insightful and informative. You deserve a gold star.

[oldprof]

Grasp - The trend is for better (less bad) numbers, in line with what we are seeing from other indicators like ISM and jobless claims. The improvement is coming more slowly and from a lower level than the recent reports would seem to indicate.

Thanks for helping me clarify this.

Jeff

[graspthemarket]

Thank you for you insight, I look foward to reading the "more to come" part. I understand your points, and I will have to study your work more. If you get a chance, could you explain where you believe the trend is headed. I understand that you disagree with the number, but where, in your best estimation are we headed? Thanks for you time.

Jason