Jobs Overstated: The BLS Comes Clean

After nearly two years of pointing out how the Birth-Death model by the Bureau of Labor Statistics (BLS) has been badly overstating job creation, I want to publicly recognize and commend the BLS for publishing the massive benchmark revision showing that more than 800,000 jobs that had been modeled into existence never really existed.

I'll bet that revelation would have been swept under the rug in more than one country.

Over time, my main complaint about the model was that it didn't make sense.  It added jobs when times were good and even more when times were bad.  It seemed completely disconnected from the world around it, and therefore the main source of my ire was that it simply didn't make any sense.

In particular, I railed against continued increases in construction jobs being modeled in, even as retail construction investment was falling off a cliff.  I wondered at the continual creation of financial services jobs, even as mass layoffs were being announced across the industry.

Further, I have this point of view that attempting to pilot a nation (or an economy or a portfolio…) using bad instrumentation is a bad idea. Good decisions made with bad data usually become bad decisions.

Now a lot has already been written on this subject, and I don't want to cover all the territory.

The basics are that from March 2008 to March 2009, the BLS overestimated payroll employment by some 824,000 jobs, or nearly 70,000 jobs per month.

Between March of 2008 and March of 2009, the BLS Birth-Death model added 717,000 jobs.  Without these additions, the benchmark revision would have come in with a very acceptable error of slightly more than 100,000 jobs too many.

The part of the story that I want you to focus on is this:  Since March of 2009, the Birth-Death model has added an additional 815,000 jobs, which, I am quite sure, will also have to be removed from the rolls when next year's revisions are made.

As Floyd Norris wrote in the NY Times:

The so-called “benchmark revision” that was announced today will not formally be incorporated into the job figures until February, and could be revised. But the figures indicate that last March the government overestimated the total number of jobs by 824,000, or 0.6 percent. Its overestimate of private-sector employment was even greater — 855,000 jobs, or 0.8 percent.

It is the largest benchmark revision in at least the past dozen years. In 2006, when the economy was booming, it underestimated the total number of jobs by 752,000 jobs, or 0.6 percent. The private sector accounted for nearly all of that, causing an 0.7 percent upward revision.

In each of the three years since then, the benchmark revisions have indicated the government overestimated the number of jobs in the economy. But the 2007 and 2008 revisions were relatively small. The culprit is probably the much maligned birth-death model, although Victoria Battista, an economist at the Bureau of Labor Statistics, said the bureau was looking into other possible issues, such as changing response rates to the questionnaire sent out by the bureau to employers each month.

So the government has been systematically underestimating job losses for three years running.  And, surprise of all surprises, the worse the economy got, the worse a model that only ever seemed capable of adding jobs performed.

I did love that possible explanation offered by the BLS economist that it might have been "changing response rates" to questionnaires sent out to employers.

Memo to BLS:  Bankrupt companies probably have low response rates.  You might want to begin there.

The critical message here is that even though the last employment report was bad, reality was certainly worse.

My best guess is that the Birth-Death model is overestimating by nearly 100,000 jobs per month right now.  I base this on past performance (the benchmark revision), tax receipts, retail data, unemployment benefits, initial claims, and employment index data.  Admittedly, this is not scientific, but at least it makes some sense.

But kudos to the BLS for owning up to the error.

There are many observers watching to see what the response will be.  Let's hope that a desire to repair a broken instrument exists within the BLS, because we're going to need the best information we can get, as we navigate these unknown economic canyons.

This is a companion discussion topic for the original entry at

One of the posters on Nate's site, Joe, noticed this update to the Civilian Unemployment Numbers this morning,

The federal reserve of St Louis has updated their household data charts with last week's jobs data. Attached is the YoY change in civilian employment in the US. The economy lost 785,000 jobs last month, not the 250,000 that was reported. All told, we've lost 7.8 MILLION jobs since the peak in November 2007. The government LIES to us.

Chart 1

Chart 2

Chart 3

Holy Cow, thats a huge difference! Nate is going to verify the information and post it later.


So they have come to the conclusion that the jobs numbers were overstated, but if I’m understanding it correctly the use of ‘corrected’ adjustments won’t go into effect until February of next year and in the meantime we’ll still be using the previous adjustments?  Is this normal procedure, or do they normally implement new revisions immediately (i.e. the next month)?  I hate to be a cynic (even though it’s served me well in this environment thus far!), but this seems another example of kicking the can down the road for another several months, perhaps in the hope that the economy really does start turning around by then.

  • Nickbert

But what is the net effect of this revision?
When unemployment numbers are released to the market, there is an official announcement and the market responds to the report. When the benchmark revision is stated, is there an official announcement? Does CNBC do a countdown to the BLS Revision like they do for the jobless claims? How will an admission to faulty reporting effect the markets?

Or does this all just filed under the Errors and Omissions portion of the newpaper?


These numbers would really bother me
If it wasn’t a “jobless recovery”

Graphs from the St. Louis Fed:
Civilians Unemployed - 15 Weeks & Over

Civilians Unemployed for 27 Weeks and Over


Employees on Nonfarm Payrolls: Manufacturing



Now, how does that economy look?

At least on this board "Everybody knows "


Our career community is dedicated to helping New Mexico Jobs Professionals,like you, connect with local employers and find great career opportunities in the Bay Area. We even provide email alerts to inform you as soon as new opportunities become available and offer the latest information and resources to help you land a great job and advance in your career.

"Our career community is dedicated to helping  palo alto jobs Professionals,like you, connect with local employers and find great career opportunities in the Bay Area. We even provide email alerts to inform you as soon as new opportunities become available and offer the latest information and resources to help you land a great job and advance in your career

Has been created keeping in view of the IT-professionals.This site will help you ““Professionals””“” to quickly and efficiently locate many opportunities that exist.
It’s user friendly tool to help you match your own Specifications, Qualifications and Requirements.“”“”“” Banbury Jobs