On Feb 19, 2007, at 11:50 AM, Michael Perelman wrote:
Is this work any good? http://www.shadowstats.com/cgi-bin/sgs/
Not really. E.g.:
Conventional wisdom in the financial community is that the payroll survey is more accurate, given its larger sampling base. To the contrary, the household is scientifically designed, and the error can be estimated to any degree desired. The payroll data are haphazard at best, and the BLS has no idea of potential reporting error.
There are several problems with the household survey, starting with the sample size. (Dean Baker & CEPR did a good piece a few months ago on undercoverage of darker, poorer populations.) But there are also difficulties associated with inflating the sample into a national estimate - e.g., there's only spotty info on immigration, and population estimates can be rigorously adjusted only every 10 years, with the decennial census. The payroll survey, however, is benchmarked every year with records from the unemployment insurance (UI) system, which covers 98-99% of the employment universe. Recent benchmark adjustments have been very small - except for the most recent one, which was quite large, but appears to be the result of a bunch of weird stuff that happened in late 2005/early 2006. Based on the quarterly UI records, it appears to be back in whack. And it *was* benchmarked, which you can't say about the household survey. Doug
