Recent Employment Data Didn't Capture the Negative Impacts

The employment report for October was extremely negative, consistent with recession levels.  Here is our assessment written on Friday for RealMoney:

Today's employment report had a number of interesting features. The large downward revision to last month brought the total net losses close to what we expected. Modelers (like me) use the final revised data. (There is no advantage to using the unrevised data -- a subject for another day). Revisions come because the "early returns" only include those responding on time, perhaps 65% of the sample. We get big revisions when the "late responders" are significantly different from the early responders.

There was a dramatic spike in other unemployment measures. Involuntary part-time workers rose 645,000 in the month, to a total of 6.7 million. These are people who would prefer full-time work. There was also an increase of 164,000 in "discouraged workers" to a total of 484,000. These people are not counted as official unemployed since they did not look for work in the last four weeks.

The payroll data come from the week including October 12th, and therefore may not capture events since then.

As to the future, there are even more big variables than normal, so forecasting is even more difficult. Here is what to look for. Obama economic policy, the transition, the use of TARP funds, the timing and nature of additional stimulus will be important both in the direct effects, and in whether they succeed in improving consumer and business confidence.

In short, the headline numbers did not really capture the negative impacts.

The Birth/Death Adjustment

Some of those commenting continue to cite the BLS birth/death adjustment as a source of error.  These observers depict a bad situation as even worse, continuing their multi-year theme of attacking the BLS methodology.

This viewpoint has been so pervasive that it is now accepted as the conventional wisdom.  Nearly everyone on the NYSE floor, as reported by our favorite source, Art Cashin, thinks that the BLS overstates employment.  The mainstream media has extensively reported this viewpoint, and the biggest Internet bloggers push the story.

They are all wrong -- completely wrong.

This is easy to check.  On a regular basis the BLS compares their forecasts with the actual results.  If the critics were correct, the results from state employment data would eventually reveal this.

Last month we highlighted the results from an extensive study by the BLS where they compared their forecasts with actual results.  It was a slam-dunk victory for the BLS.  We urge readers to review this article for a factual assessment of BLS methods.

Despite this evidence, many critics repeat the old lies.  Here is a typical comment from Alan Abelson, who says almost the same thing about the payroll report each month:

All told, 240,000 payroll slots went up in smoke, and that doesn't count those mythical jobs that padded the total by 71,000 via the infamous birth/death model. And the unemployment rate jumped to 6.5%, from 6.1%, the highest level since March 1994.  

We are determined to correct this inaccuracy.

The Challenge for Critics

Anyone who maintains that the BLS is inventing jobs or padding the total should be able to meet this simple challenge:

Find a year where the payroll employment report would have been more accurate without the BLS adjustment.

We eventually have the data from state sources, so if the critic is correct this should be easy to do.  Critics, put up or shut up.

Comments

[Steven Hansen:]

you have to select a census year - only these years have real data for comparison. any other year we are playing liars poker.

try census year 1990. i ran numbers a few days ago. that year the employment numbers (total population, employment, unemployment, etc) between the BLS and US Census showed a statistical anomaly. i have tried to find my worksheet but i must have deleted it as what i was trying to verify turned out to be non-conclusive.

actually, i am not trying to rain on your parade. i understand the need to create statistical baseline material that is useful in year-to-year analysis. however, they could lead with a number which is virgin, and follow it with a homogenized number.

nobody trusts the government any more to tell the truth. it really is to bad if we discover they tell the truth some times.

2008 Nov 10 02:43 AM

[rdr4:]

The possibility they have modified the count exists. The government also can modify the input numbers so it is possible they are simply ahead of your math. Perversion of statistics is one of the oldest scams.

During the gas crunch and 55 mph speed limit days of the seventies, the government, to justify the 55 mph speed limit, changed the way they counted traffic deaths. Before the change, a person who died in a traffic accident was a traffic fatality, however, after the change, only those who die within 30 days of an accident are a traffic fatality. Presto, the government is saving lives by changing the speed limit to 55.
Perhaps the government would modify numbers about such a trivial topic, but I am sure they would never alter numbers related to unemployment, WMD, or anything important.

2008 Nov 10 08:49 AM