Revised Unemployment Rates Analyzed

Former General Electric CEO Jack Welch is questioning the sudden drop in the unemployment rate claimed by the Bureau of Labor Statistics (BLS). Many pundits are wondering whether this drop is politically motivated. When arguing with MSNBC Host Chris Matthews this is what Mr. Welch had to say:

“We had 600,000 government jobs added in the last two months. We had 873,00 jobs by a household survey — which is a total estimate — from 50,000 phone calls. Of those, 600,000 were temporary workers. Chris, these numbers are all a series of assumptions. Tons of assumptions. And it just seems somewhat coincidental that the month before the election, the numbers go one-tenth of a point below where the president started. Although, I don’t see anything in the economy that says these surges are true.”

At one point Chris Matthews interjected with, “You’re talking about the President of the United States playing with the Bureau of Labor Statistics’ numbers.” Welch did not back down.

“You don’t think it’s coincidental that we’ve got the biggest surge since 1983 in the jobs surge? Come on, Chris! It’s a six percent improvement in employment in two months…The numbers don’t jibe. These numbers defy logic. They defy logic. We do not have a 4 to 5 percent booming economy with 873,000 people added. I mean, stop it, Chris. On the face of it, we don’t have this GDP. I love you, but you can’t get there.”

I decided to take a look and see what explains the sudden drop in the unemployment rate. Most likely it’s just an anomaly from the statistics-gathering process. The Bureau of Labor Statistics (BLS) is calling 60,000 households, not 50,000 as Mr. Welch said. But he is right about the 873,000 jobs supposedly being added. You can take a look at the raw survey numbers on the BLS site (keep in mind the numbers are in thousands).

Between August and September the number of unemployed is estimated to be 12,544,000 and 12,088,000 respectively. So that makes an difference of 456,000 people who fit the requirements for being considered unemployed. This is extrapolated from calling 60,000 households and may greatly be affected by the margin of error in the surveying process. The BLS explains:

A sample is not a total count, and the survey may not produce the same results that would be obtained from interviewing the entire population. But the chances are 90 out of 100 that the monthly estimate of unemployment from the sample is within about 290,000 of the figure obtainable from a total census.

But Mr. Welch is right about there being a jump in temporary seasonal or part-timers, which is 582,000 of the estimated 873,000 jobs. Teenagers (16-19) and Black people had the largest sudden decreases in unemployment. Then of course there’s the people that are considered to be “not in labor force” which is estimated to have gone down by 211,000 people. This is a sudden reversal since the trend before then was for the number of those not being counted in the labor force to be going up. This number includes “unemployed workers who have exhausted their benefits” as an example.

I would suggest reading the BLS page to see how they calculate. If there’s any funny business it would probably be with the “seasonal adjustment factor” or in the way they “weight” each response:

“the total numbers are “weighted,” or adjusted to independent population estimates (based on updated decennial census results). The weighting takes into account the age, sex, race, Hispanic ethnicity, and State of residence of the person, so that these characteristics are reflected in the proper proportions in the final estimates.”

and

“To deal with such problems, a statistical technique called seasonal adjustment is used. This technique uses the past history of the series to identify the seasonal movements and to calculate the size and direction of these movements. A seasonal adjustment factor is then developed and applied to the estimates to eliminate the effects of regular seasonal fluctuations on the data. When a statistical series has been seasonally adjusted, the normal seasonal fluctuations are smoothed out and data for any month can be more meaningfully compared with data from any other month or with an annual average. Many time series that are based on monthly data are seasonally adjusted.”

By every four months the households being called get changed. Thus the sample must be designed and selected. For those looking for a conspiracy this process could be suspect:

The CPS sample is selected so as to be representative of the entire population of the United States. In order to select the sample, all of the counties and county-equivalent cities in the country first are grouped into 2,025 geographic areas (sampling units). The Census Bureau then designs and selects a sample consisting of 824 of these geographic areas to represent each State and the District of Columbia. The sample is a State-based design and reflects urban and rural areas, different types of industrial and farming areas, and the major geographic divisions of each State.

So is there a conspiracy with the unemployment rate? Who knows. The holiday/seasonal work combined with the margin of error seems to cover the majority of jobs. Either way, it’s not the recovery anyone wants but a reprieve is (or should be) welcomed by all.