January Employment Report Preview

As usual, I have a monthly comment on the employment situation.  This month's report is especially tricky, since the BLS will incorporate "benchmark revisions" to prior data.

Anything about revisions causes most observers to cringe.  The average person just wonders why they couldn't get it right the first time.  The conspiracy buffs see revisions as evidence of data manipulation.  Since many people are ready to believe that the US government is in an Orwellian mode, this is a message that resonates.  Those who know nothing about the actual operation of government seem to think that the BLS switches allegiance whenever there is an election.  George W's picture is taken down, and Obama's is put up.  I have been trying for several years to educate readers about which agencies spin, and which do not.  The BLS is among those least subject to political pressure.

I recently participated in a new survey of economic bloggers.  I have high hopes for this approach or I would not spend time on it.  My own take on the first results is that it tells us more about the economic  bloggers than it does about the economy.  They are extremely pessimistic on economic prospects, much more so than surveys of economists.  They think that (for some reason) things are much worse than official statistics indicate.  The background info seems to suggest that Republican, Democratic, and Libertarian parties are all about equally represented, with a huge number of "independents."  That is an eyebrow raiser!

Why Revisions?  Why Benchmarking?

The BLS attempts to report timely data about the employment situation, even though many businesses do not respond to the survey.  As more data comes in, they add the information to the results.  They also do a standard adjustment for seasonality.  It is a system, not discretionary.  When new data comes in, the seasonal results are recalculated.  This is another adjustment.

It is a very standard and mechanical process.  The commentators who impute some dark motivation to this are just ignorant about government.  Those who think they can "guess" the pattern of revisions need a model relating to non-respondents or changes in seasonality.  Most are just guessing, or making an inference from the last couple of months.  They need to read a book covering the "law of small numbers"!

It is Keeping Score

Measuring employment changes may be guesswork at the time of the original estimate, but eventually we know the truth.  We know it about nine months later when the Feds reconcile the state unemployment data.  No business pays unemployment insurance on phantom jobs, so we can expect this to be a good count.

Why can't the BLS do this faster?  I do not know.  At least they do it eventually.  Each year the BLS makes a "benchmark adjustment" by comparing actual counts to their original forecasts and reconciling discrepancies.

BLS Critics

It is only fair to note that the vocal BLS critics — people like Ritholtz, Abelson, and Mish — do not do any  benchmarking of their own.  They criticized the BLS and their methods for many years.  When the actual data proved them to be completely wrong there was no 'fessing up.  They were wrong for many years, but still pretend otherwise.

In the last year the BLS noted that their method (finally) broke down.  Here at "A Dash" we go by data, not by any pre-conceived opinion.  When the BLS was correct, we noted that.  More recently we have emphasized the breakdown in their method of estimating job creation.

The Benchmark Revisions

I am going to state this a simply as possible.  Most critics focus upon the birth/death adjustment aspect of the BLS process.  They look at raw data, and ignore all of the warnings and explanations about seasonality, not adding these numbers to the monthly result, etc.

This is a big mistake — a blunder.  It would not pass muster in a peer-reviewed journal or a court case.

So Bloomberg's article – -nice graphics, wrong data — misses the point.

Barry Ritholtz's assertion about the B/D model and the business cycle is a mistake because he is asking the wrong question.  Most of the implied job creation comes from the imputation step, something that he never discusses.

Both sources err by looking only at only one part of the process — and the part that is less important — the birth/death adjustment.  They are missing the "imputation step" which is much more important.

The birth/death adjustment was never intended to capture cyclical behavior.  It was a time series geared to the residual job creation after the "imputation step" did the heavy lifting, and has historically captured

For emphasis – -and I have tried to alert many sources — 

Anyone who does not understand and discuss the "imputation step" as part of the BLS job creation process is not a true expert.  You should ignore that source.

Our Own Estimates

As regular readers know, I am not a fan of the current BLS estimates.  While it may seem like hairsplitting to disagree with other critics for a different reason, I think it is an important distinction.

Each month we ask the question, "What change in payroll employment
would be consistent with other economic data from the same time period
(the middle of the prior month)?

This is not a forecast, per se, since we do not posit any
causal relationship among these variables.  They are all concomitant
indicators of economic activity. 

  • We use the four-week moving average of initial unemployment
    culminating in the week of the employment survey.  This is the best
    direct indicator of new lob losses.  This has improved in the last
    three months to a loss of 447K.  Note that today's increase to 480K is not within the survey period for the monthly report.  Last week we noted a blip in the other direction.
  • We look at the University of Michigan sentiment survey,
    which we find to be more useful than the Conference Board's sentiment
    index.  Michigan uses a panel, where some families are carried over
    from month to month.  This is a good technique.    Sentiment is
    influenced by employment.  When people have lost jobs, or are worried
    about losing jobs, it shows up in sentiment.  It is a good concurrent
    indicator.  The Michigan index is now at 74.4, continuing a rebound.
  • We us the ISM manufacturing index, which surged to 58.4 from 55.9 last month.  This is strongly bullish for the overall

Our long-term record has been pretty good, especially when compared
to the final revised data.  This makes sense because our model was
derived from the final data.  In recent months we have been too
bearish.  The BLS benchmark revisions suggest that we have been much
better than first thought.  I am working on a comparison with the final

This month, our estimate is for a slight job gain of about 5000.


It is always interesting to compare the job forecasts from different
sources.  We follow several because of the interesting and widely
varying methods they use.  A wise interpretation would be to consider
all of  these disparate sources of information.

ADP has proprietary data because of its payroll management business. 
ADP sees losses of 22K.  This estimate does not
include government jobs.

WANTED Technologies refuses to estimate this month because of the benchmark revisions.  I disagree.  This is the wrong attitude.  The benchmark revision should be encouraged for reasons noted above.  The estimate of the monthly change is still very relevant.

Briefing.com cites the consensus as a gain of 15K
and their own forecast is a loss of 25K. 

All of these sources are valuable.  The 90% confidence interval on
the BLS estimate, something that no mainstream media sources report, is
+/- 100K or so.  And that is after revisions and benchmarking.  It is a
survey — a good one — but it has an error band.

Final Interpretation

The report will include a massive benchmark revision for the prior year — a reduction of jobs of over 800,000.  There will be new birth/death factors and new seasonal adjustments.  Since the revisions only go back to March of 2009 (the last available data), there may be additional revisions for recent months.  This would bring the entire series much closer to the monthly forecasts which I have published.

It will be a field day for pundits, most of whom will focus on the benchmark revision — something everyone has known and talked about for months.  It is "old news" but will probably not be treated that way.

Meanwhile, our assessment is that the jobs picture continues to improve, but there is a long road ahead.  The evidence is that the engine of job creation was broken during the credit crunch.  As yet, there is no evidence of a change.

We will get some read data in a few weeks with the next round of the actual state employment data.  That information is not a forecast.  It is not a survey.  It is not a model.

The data will be old, but relevant.  Stay tuned.

You may also like


  • Barry Ritholtz February 5, 2010  

    Wrong for years? In what way?
    My position on B/D has been consistent for a long time: Ever since a major modification was made to the Birth Death adjustment (proposed in 2001, and implemented a few years after), the NFP data has become increasingly misleading.
    The attempt was to capture early improvements in employment at the start of a recovery was the goal. But the trade off was it wildly overstated economic strength at the end of a cycle.
    For example, in 2007, approximately 75% of reported new jobs were due to this adjustment.
    In 2008, the BD adjustment inexplicably showed lots of job creation in construction and finance.
    The people who ignored the below the headline changes missed the warning signs. Lots of economists blindly cheerleaded their way right into the recession — never saw it coming. The worst offenders were the ones who dismissed the effects of the B/D changes. This cohort of economists totally missed the worst recession in generations — they were utterly blindsided, and the people who followed their advice saw their 401ks and retirement accounts mangled.
    Jeff, with all due respect, I do not recall where you stood pre-recession (Q4 2007) — were you expecting an economic contraction? Were you misled by the BLS data? Did you have any clue of the coming storm?
    My criticism of the BLS data — I am NOT one of the a conspiracy folks — is that it needs to be carefully parsed and dissected to get at the worthwhile data hidden within. The BD data was only part of it.

  • Jeff Miller February 6, 2010  

    Barry — You are certainly not one of the conspiracy folks, although you think that government agencies have a bias. You are trying to reach a valid conclusion, and I agree with the need to parse the data effectively. I also agree that your position has been consistent.
    I think you are mistaken, and I would like to try to convince you with data. Perhaps I’ll write a longer piece and lay out a few points. For now, here are two thoughts:
    1)You state that “For example, in 2007, approximately 75% of reported new jobs were due to this adjustment.” This is a factual error, and a big one. There were about 13 million new jobs in 2007, 4 million from new establishments. This is what I mean about benchmarking. You never look at the actual count (nine months later) to see if you were right.
    2) What you are really doing is comparing the birth/death adjustment to the net job change. This has an important sound to it, but it is actually meaningless. Let us suppose, for example, that there was a net job loss in the month, but an actual gain in jobs at new establishments. (This is quite common). Your 75% measure makes the error of starting with the conclusion — your pre-conceived notion that there are no new jobs. This leads your readers to believe that any such adjustment is an important error.
    If we really want to parse the data, as you suggest, why not begin with data instead of an opinion? Go to the business dynamics series and see what really happened. Then see what the BLS forecast. Compare it to your own comments. If you do that for the entire period of time you have been criticizing them, you will see that you were wrong in 2006, 2007, and most of 2008. Only in 2009 did thing change, and it was not because of the b/d adjustment. It was due to the “imputation step” something you never analyze at all.
    Thanks for taking the time to comment. Let’s stick to the BLS point. We can discuss the recession and data another time, perhaps when I finally finish my review of your book — which I enjoyed and have recommended to friends.