On Nov 6, 2007, at 4:02 PM, ira wrote:
> http://www.rgemonitor.com/blog/roubini/224473/
>
> Nouriel Roubini | Nov 02, 2007
>
> The U.S. October employment report came better than expected. But the
> details of this report are much weaker than the headline figure of
> 166K
> non-farm jobs created in October.
>
> First, the establishment survey says that 166K jobs were created in
> October. But the household survey says that 250K jobs were lost in
> October. Which survey is more correct?
The two surveys aren't strictly comparable. The household measure includes agricultural workers and the self-employed; the establishment survey doesn't. The household count, when adjusted to match the coverage of the establishment survey, was down by 55,000.
Most of the time, the household survey is considered to be less reliable than the establishment survey. The household survey covers about 60,000 people (huge compared to your average poll, whose sample size is around 1,200, but small for estimating monthly changes); the establishment survey (which, as its name suggests, asks employers how many workers they've got on the payroll and how much they're being paid) has a sample size of around 300,000. But the establishment survey is often unreliable at turning points, because of this:
> Second, the bias of the establishment survey in periods of growth
> slowdown is exacerbated by the birth/death model used by the BLS that
> creates statistical jobs that do not exists based on their statistical
> model of new born firms and dying firms.
They do not "create statistical jobs that do not exists." The birth/ death model attempts to account for the creation of new firms and the death of existing ones that might not be covered in the monthly sample (which, at 300,000, covers something like a quarter of all U.S. employers). Most of the time it works very well, and only know- nothings like John Crudele of the New York Post dismiss it with a sneer. But when the economy is slowing or accelerating, the b/d model is often off. So the establishment survey undercounts employment in the early stages of recovery and overcounts at the onset of slowdowns or recessions. Since there's a good chance we're early in a slowdown, if not an outright recession, that's a real issue now. About 80% of the new jobs in the establishment survey so far this year have come from the b/d model. It's quite possible that those jobs don't really exist. Over the same period, the adjusted household survey reports about half as many new jobs as the establishment survey. And, the employment/population ratio (the share of the non-jailed adult population working, another useful check on the establishment survey) has fallen from 63.4% in Dec 2006 to 62.7% in Oct 2007. And the growth in the adjusted household measure of employment was 0.7% for the year ending in October, vs. 1.2% for the establishment survey. All this suggests a weak job market - which is confirmed by the consumer sentiment surveys.
> Third, the unemployment rate remained stable at 4.7% because in
> October
> 211K workers left the labor force, most likely discouraged workers who
> cannot find jobs. Without this reduction in the labor force the
> unemployment rate would have been higher.
Weirdly, the labor force shrank in October, but the number of people who are classed as "not in labor force - want job now" (meaning they want work, but aren't actively looking, which means they can't be counted as unemployed) also fell. So that doesn't suggest withdrawal from discouragement.
Here's a fun fact: almost half the growth in jobs over the last year has been in health care and and bars and restaurants.
> Seventh, employment is a lagging indicator of the economy:
No it's not. Historically, employment peaks and troughs have been almost identical to cyclical peaks and troughs. The last two cycles - the early 90s and early 00s - did see very slow recoveries in the job market. But that's unprecedented.
Doug