The always-colorful Jack Rasmus:

> The current population survey (the 873,000) represents a statistical 
> operation on raw jobs data that adjusts that raw (i.e. actual) jobs data 
> by means of several statistical operations–i.e. seasonality, etc.—to get 
> to the 873,000.  But before the raw data is statistically adjusted, 
> another source of raw data is added to the initial data and only after 
> that is the statistical adjustment carried out. This second source is 
> jobs data estimated from assumptions about New Business Formation that 
> are lagged up to nine months.
> 
> Here’s how it works. The labor department assumes net new businesses are 
> formed nine months previous. That would be last November-December 2011. 
> These data on new business formation are very inexact. It’s not actual 
> new businesses but an assumed historical average of new businesses. So 
> the past years in which new business formation was high is substitute 
> for the more recent period when new business formation is in fact low, 
> or even negative. It’s really a  shaky estimation process.

This is completely wrong. The business formation estimates figure into the 
establishment survey, the monthly survey of almost 500,000 worksites - more 
info here:

  http://www.bls.gov/ces/. 

That survey is the principal source of the headline job gain/loss figures. The 
household survey - of 60,000 households - is independent of the establishment 
survey. The unemployment figures and this CPS 873,000 figure come from the HH 
survey. It's the result of asking people if they've been working, looking for 
work, neither, etc. More detail on the HH survey here:

  http://www.bls.gov/cps/faq.htm#Ques1

The HH survey has nothing to do with the establishment survey. There are no 
assumptions about new business formation at all. It just comes from asking 
people questions and adding up their answers. It's very volatile - you need a 
change of 400,000 to reach statistical significance. The employment numbers 
bounce around a lot. You can get +800,000 one month and -500,000 the next. The 
raggedness evens out over the course of a year, but monthly changes should be 
taken with a grain of salt, or three.

In any case, the so-called birth/death model, the estimate of new business 
formations, actually subtracted 9,000 jobs in September from the establishment 
count. Over time, the model has actually proven to be very accurate - it's not 
shaky at all. There are yearly corrections to the monthly estimated data, based 
on the near-complete coverage of the employment universe offered by the 
unemployment insurance system. These benchmark revisions are typically in the 
range of 0.3% or less. There are sometimes larger misses at business cycle 
turning points - the monthly numbers underreport reality early in expansions, 
and overreport it at the beginning of recession. But the long-term record is 
very good. We just had a preliminary announcement of the benchmark revisions to 
be applied to the March 2012 data:

  http://www.bls.gov/ces/cesprelbmk.htm

and it's an upward revision of 0.3%, or 386,000 jobs. In other words, the 
monthly numbers had been slightly too low.

The birth/death model is described here:

  http://www.bls.gov/web/empsit/cesbd.htm

Doug
_______________________________________________
pen-l mailing list
[email protected]
https://lists.csuchico.edu/mailman/listinfo/pen-l

Reply via email to