----- Original Message ----- From: "Erik Reuter" <[EMAIL PROTECTED]> To: "Killer Bs Discussion" <[EMAIL PROTECTED]> Sent: Wednesday, March 24, 2004 8:35 PM Subject: Re: Household vs. payroll employment
> On Wed, Mar 24, 2004 at 08:14:04PM -0600, Dan Minette wrote: > > > Right, but there is a seperate catagory for them. > > I don't see your point. The categories are not well-defined, so they > cannot be fully reconciled between the two surveys. > > > The difference between payroll figures are still substantial. > > Yes, that is rather the point, don't you think? If they weren't > different, then we wouldn't be having this discussion. > > The important question is which is a better representation of the > employment situation? > > > > There are a few logica possibilities that I can think of for the > > discrepency to exist. > > > > 1) There has been a great growth in workers in off the books jobs. > > > > 2) There has been a great growth in people reporting their off the books > > jobs. > > > > 3) There has been a great drop in second jobs, which affect the payroll > > figures, but not the household figures. > > > > 4) The normalization > > B.L.S. lists several possibilities for the differences in the article I > referenced: > > - sampling error > > - payroll survey benchmark > > - new business births in the payroll survey > > - population controls in the household survey > > - worker classification in the household survey > > - "off-the-books" employment I appreciate those catagories and they have validity. One thing that is worth thinking about any catagories is why they would change enough in 4 years to account for the differences. The biggest factor is the population control, so that is one good reason to focus on that factor. For example, why would the worker classification errors show a ~3 million shift towards reporting self employed as payroll employed between 2001 and 2004? > > But, if you look at the graph you reference, it has not been 7> > renormalized between the start and end of the period that is covered. > > So, as far as I can tell, it would not have corrected for the recent > > effect that Greenspan commented on. > > What do you think they mean by "smoothed for population control > revisions"? I think that they did something this last January, but not nearly enough. The latest revision downward only affected data after Jan 04...and was only about 400k It might be worthwhile to trace the normalization over the last few years: 1994 -7.2% 1995 -6.8% 1995 -6.8% 1997 -6.7% 1998 -6.0% 1999 -5.7% 2000 -5.7% 2001 -5.6% 2002 -5.2% 2003 -5.4% 2004 -5.7% So, the normalization for errors in the population estimate between 2000 and 2003 is actually in the opposition direction from what Greenspan indicates is needed. After the correction of ~400k in Jan, 2004, its close to flat. It actually was renormalized after all, but just not by much...and not over the time frame I was looking at most closely. That's why I missed it when I concentrated on the graph. But, you are right, they are trying to correct for the population. One way to look at it is that they got the normalization factor wrong...in the opposite direction of where is should be. Its an interesting mix, one survey is much more accurate but doesn't cover the whole field, the other is broader, but with greater statistical and systematic errors. One might argue that one should use the gross numbers from payroll and look for trends in things like self employment in the household numbers. Dan M. _______________________________________________ http://www.mccmedia.com/mailman/listinfo/brin-l
