labor statistics
When the payroll numbers come out tomorrow, BLS will also release a revision of data from April of 2005 to last March, and is expected to raise payroll numbers by 800,000 jobs or so.

Economists widely consider the payroll survey to be significantly more accurate of the two readings. But [Bernard Baumohl, managing director of the Economic Outlook Group,] and other economists say it's not accurate enough to justify the attention paid to the monthly net change in payrolls.

"The BLS says that the payroll estimate has a margin of error of 150,000 jobs, while the household survey is plus or minus 300,000 jobs," he said. "So when you see a gain of 150,000 in the payroll number, it could be zero, or 300,000. It's tough to draw any conclusions about the state of the economy from that."

Well, okay, but why do I care about the number of jobs produced? As a practical matter, there are two components to "number of people employed": number of people in the labor force, and the fraction of them with jobs. A lot of the uncertainty in the household number is due to uncertainty in that first figure; if it were all in the second, that would still amount to 0.2 percentage points on the unemployment rate, but in fact the uncertainty there is significantly less than that. The best measure of the tightness of the labor market is the unemployment rate, which comes reasonably directly from the household survey; if an exogenous event were to dramatically and suddenly change the size of the labor market without changing the number of jobs, there is no doubt that wage pressures would be affected. The payroll figures are trivia.
- posted by dWj @ 9:40 PM