I just finished a maddening, circular discussion with my friend Drew regarding the Bureau of Labor Statistics employment numbers. Regular readers know that I take issue with the way my friend argues a point. For one, he posts on a personal journal page that provides no avenue for rebuttal, and that I think he is intellectually dishonest in his arguments, relentlessly panning for the anti-American, anti-Bush-Rove-Ashcroft nugget. Regarding unemployment, he posits that since the BLS changed its methodology 8 months ago, something he calls the CNES Birth-Death model, the vast majority of new jobs being created are not comparable to past, pre-change numbers. Effectively, he believes that the recent positive jobs numbers are an anomaly. I think that argument is without merit, because we should have seen an immediate aberration 8 months ago if the model alone was responsible for the numbers. He claims that the BLS phased in the implementation of the model. I am throwing this out to you, since I am sure there is a labor statistician out there somewhere. Is it at all plausible that this change in calculation method could have been:
1. Implemented by a phased in approach?
2. Solely responsible for the improving employment numbers?
3. Not picked and drilled by the Democrats and Kerry’s economic advisor as a sham?
I may be completely off-base here, but I am willing to take my lumps if that is so. I await your responses.
Update: Upon further research, I found it’s the CES Birth/Death model.
I am not a labor statistician, but I do sample surveys and estimation. I will address only your first issue … is a phase in possible.
Yes. Suppose you have an estimate based on the old technique, Yo, and another on the new procedure, Yn, and that you wish to phase it in over some number of months, m. Your final, phased-in estimate, Y, would be:
Y = (t/m)*Yn + (1-(t/M))*Yo.
Thus, in the first month (t = 1), if you were phasing the estimate in over 6 months (m = 6) only one sixth of the estimate would arise from the new model, and 5/6ths from the old.
Note, however, that the Bureau of Labor Statistics models such changes for prolonged periods before implementation, and they do not change willy nilly. Those people are professionals, not political hacks.
To say that the unemployment numbers aren’t real would then lead one to call into question all the other economic data coming out as well. Taken together, they are consistent with one another and the delayed boost in employment is typical for a recovery. As far as the BLS models, I don’t have a clue.
Query: why are we speculating? Doesn’t BLS publish their models, and figures showing numbers with both the old and the new model? Their webpage seems to indicate the latter; there are plenty of technical notes, there’s raw data, etc etc.
Because I am a trader, not an actuary. I want to hear from someone apolitical who can speak with authority on the subject rather than fumbling through the bls haystack.