Some loyal readers have suggested that I get off this case … That I’ve made my point.
I promise that I’ll stop writing about BLS reporting bias when the streak ends.
Now we’re up to 77 out of 78 weeks — and, at least 18 weeks in a row — that the BLS’s “headline number” has under-reported the number of initial unemployment claims … and cast the jobs situation as brighter than it really is.
Based on yesterday’s BLS report, the number for the week ending August 25 was revised upward from 374,000 to 377,000.
In itself, the 3,000 isn’t a big deal.
But, in context it is
Again, I ask: statistical bias or political bias?
If the former: fix it already, BLS.
Hint to BLS: just add 2k or 3k … or .8% to your prelim forecast !
Tags: BLS, employment, unemployment
September 7, 2012 at 10:08 am |
Does the BLS calculate numbers differently when the Republicans are in the White House? If not, why is this a political issue?
September 10, 2012 at 4:24 pm |
Troll troll troll troll troll troll
Anyways, since you asked: The process for measuring unemployment is not as simple as summing up all of the claims. So, the Bureau of Labor Statistics uses a survey of 60,000 households. In this setup, there are approximately five thousand ways to introduce a bias, including these three:
1. The sample changes. 25 percent from month to month and 50 percent from year to year. Guess who selects the new sample?
2. The regional distribution changes: BLS splits the nation into 2,025 geographic units and then samples respondents in 824 of those 2,025 units. Guess who selects the chosen units?
3. Respondent “weighting”: Following activity classification, each individual response is “weighted” to align with the current national demograpic. Some guys count more than once, other count less. Guess who weights the responses?
You could also add that the current uneployment statistical model (according to BLS) assumes 6 percent unemployment rate to ensure a stable coefficient of variability, but who is really in the mood for stat porn?
The point is to use the various survey levers to increase accuracy. Instead, it seems like they are being used to manipulate the numbers. So, well, that is why it is a political issue.
FYI, I had Guerrero-Cusumano.