Misleading Commentary on Job Creation

How easily can you be fooled?  See if you can answer this question:

How many jobs were created by the U.S. economy last month?

If your answer was something like 100K to 150K, you are failing your course in Labor Economics! You had better read on to learn the correct answer.

Economic Pundits and the BLS Birth/Death Adjustment

Certain pundits are out to convince you that job creation is mostly a result of statistical maneuvering by the Bureau of Labor Statistics.  According to some the BLS is creating "fictional jobs" while others just question their statistical techniques.  Let us turn to the observations of two leading commentators.

Barry Ritholtz writes as follows:

I've mentioned the B/D adjustment over the years, and how its become an increasingly large portion of the reported BLS new jobs.

What I haven't previously mentioned is that over the past year, it is accelerating: the Birth/Death Adjustment has become an ever-large portion of the reported NFP payrolls.

How much larger? Well, consider the following data points: Over the five month period ending in June, BLS B/D adds was a total of 922,000 new jobs. During the same period, the actually head counted Non-farm Payrolls (NFP) job creation was 709,000.

That's right, fictional Birth/Death job adds have been outpacing actually measured job creation by some 30%.

As they do every year, BLS Net Business Birth/Death Model deleted jobs in January -- in 2007, it was 175k.  That means that year-to-date, the net fabricated BLS new jobs was 747k -- versus NFP growth of 871k -- that's 85.58% of NFP job growth.

The second comment is from Paul Kasriel, an economist with Northern Trust.  Kasriel is cited in the Wall Street Journal Economics blog.  You can find his entire statement here.

The “birth/death” adjustment contributed 1.11 million out of the 1.98 million jobs added to nonfarm payrolls in the 12 months ending in June, he says in a research note. That is 57% of the increase. In the 12 months through March 2006, those “phantom” workers accounted for 31% of the increase. It moved higher since then even as economic growth slowed.

Kasriel goes on to argue that the Birth/Death adjustment should be expected to have a reduced impact in a slowing economy.

Analysis

We wonder how many readers spotted the error on  their own.  The key is in knowing how many jobs are actually created each month.  The correct answer is about 2.5 million. Readers used to thinking in quantitative terms should realize that the Ritholtz statement, "the actually head counted Non-farm Payrolls (NFP) job creation was 709,000" cannot possibly be correct.  (It is actually a double error, since these were all survey results.  There was no counting going on.)

Anyone who thinks that job creation is only 150K per month is not well-informed about the dynamics of the labor market.  That would be job creation of about 0.1%.

The Kasriel commentary carefully avoids using the term job creation, instead referring to jobs "added to non-farm payrolls" but the impression is the same.

Contrast the misleading statements with the following:

During the first six months of 2007, private job creation was about 15 million.  The birth/death adjustment of 750,000 total jobs represents about 5% of the total.  This percentage has changed only slightly over the last few years.

Interpreted in this way, the Birth/Death adjustment seems to have a more reasonable scale.

Why the Confusion?

The problem results because most pundits -- even those like Ritholtz and Kasriel who know better -- focus on the monthly change in non-farm payrolls and call that job creation.  There are two problems with that approach.

1. The BLS is not directly measuring the monthly change.  In fact, it has no good way to do so.  It is estimating the number of jobs through a survey technique with a significant sampling error, a process we described carefully last year.
2. The monthly difference does not represent "job creation" or even "jobs added."  It is simply the difference between one month's estimate and the next.

Actual job creation is much higher, about 2.5 million jobs monthly.  But so is job destruction!  The labor force dynamics are much larger and more powerful than most understand, even though the BLS puts out a quarterly report showing the effects.

Our Take

The Birth/Death adjustment was introduced to correct a persistent underestimate in job change from the monthly survey methods.  Since the sampling frame only captured existing businesses, it failed to generate good estimates of job creation.  So far, the Birth/Death adjustment has improved estimation but has not gone far enough in adding jobs.  We discussed this and entertained comments from Barry Ritholtz in a prior article.

Is it possible that benchmark revisions will show that the adjustment has been too great?  Of course.  The evidence from the unemployment survey supports this idea.  The constant effort to compare the non-farm payroll report, the employment survey, and the final benchmarks from state data, provide a good historical series.  As so often happens, the real-time information is not nearly as informative as the final counts.

Kasriel is correct in asserting that the BLS model is not "adjusted for the phase of the business cycle."  It is a statistical time series method, using prior data.  As such, it is data-based and objective.

We wonder how Kasriel would suggest making a business cycle adjustment?  He currently has his own (minority) opinion about where we are in the business cycle.  Must we all agree?  The employment data are supposed to provide evidence on the business cycle, not just someone's assumption.

Your Takeaway

Here is a summary of what you should remember from this article.

1. The BLS Birth/Death model represents only 5% of job creation.
2. Pundits who compare the adjustment to the monthly change either have an agenda or do not understand labor force dynamics.
3. The Birth/Death model has improved estimates for several years, but has usually been too low in adding jobs.  This year might be different.  We shall see.
4. If you are reading "A Dash" then you have an advantage over the tens of thousands reading the other sources we cited.  That understanding is a contrarian advantage.

You can review all of the extensive methodological work we have done on interpreting government employment reports.

 

Comments

[Bill aka NO DooDahs!]

If the pundits believed their own bear scat, the level of cognitive dissonance necessary to insulate their repeated statements from the outcomes in the market would be pathological. I see these alternatives: either they believe what they write, making them both incompetent as market analysts and greatly capable of self-delusion about their abilities, or they are intentionally striking a pose for the purpose of deluding others. At the moment, I'm inclined to believe it's an act.

[oldprof]

Marc -

Thanks for the kind words, and for your comment.

I may have done a poor job of trying to explain a difficult subject. The main idea is that what the critics are attacking, the B/D adjustment, is not very large or very important. It does not change that much. Most of the job creation is already imputed from job "deaths." Why not question that model? That's the big one since it accounts for many of the 2.5 million jobs or so created each month. Of course without these adjustments there would be no meaningful payroll job estimate. We would be waiting many months for the actual count.

Thanks again for your comment. Maybe someday I will be inspired to explain this better, since I think it is important.

Jeff

[Marc]

Thanks for the website. It's one of my daily bookmarks.

I understand your comment about "job creation" as being 2.5M/month but what does this have to do with anything? I think it's pretty obvious that the two authors you mentioned were talking about "net jobs added" and not "jobs created exclusive of jobs lost".

For instance, if the net jobs added for a month is 175k and the adjustment is 100k of that, then you have a 57% adjustment. I don't see how you can calculate a % change with a "net" BLS adjustment applied to only the jobs created number without factoring in the jobs lost number. The BLS adjustment affects the net directly and as we've been seeing is a very substantial adjustment to this net.

Thanks,
Marc

[me]

Thank you Jeff,for this blog. I subscribe and never miss it, and always value your insight as another piece in the giant puzzle of investing.

[oldprof]

Welcome, "me" and thanks for your comment.

While I am a strong interest in political issues and politics, the purpose of "A Dash" is investments. To do that, one must look objectively, regardless of who was in office.

Market historians like to compare cycles and recoveries, and tops, and bottoms, and some use these for prediction. Usually they do not have enough cases to do so.

To take the current recovery, it has generated less overall job growth than those in the past because the preceding recession was not very severe. The size of a recovery depends upon where you start. The fact that this argument is often made by Republicans does not necessarily mean that it is false.

Thanks for making a point that many folks probably think about.

Jeff

[me]

You may or may not be correct in joining the argument, but the fact remains that 8 years under Clinton, job creation or whatever you choose to call it was 225,000 a month and now under Bush about half that.

I don't recall economic anxiety under Clinton but those that chose to get off wall street or out of DC will find people are pissed. They may blame illegals or Indians, but the fact is, the great American jobs machine is dead.

And you forgot to mention no wage pressure.