I no longer bother with the Over/Under bet we created so many years ago. The data from BLS has been too darn funky — noisy, subject to revision, overly massaged– to bother with that bet any longer.
But the spate of upgrades to the estimates is why I am tempted to take the under:
• The ADP data for the month of November has been consistently too bullish. Likely suspect: difficulty with the seasonal adjustment.
• The Household survey (which determines Unemployment) tends to lead the Establishment survey at the beginning and end of cycles.
• The WSJ’s Justin Lahart points out in this morning WSJ that Household survey gave a better depiction of the 2001 deterioration in jobs than Establishment:
"Now the household survey is flashing red again. For
October, it showed there were just 0.5% more workers than there were a
year ago, compared to 1.2% gain shown in the payroll figures. If the
household survey is giving the right read, the payroll figures are
overstated to the tune of one million jobs."
Consensus was for 85,000 jobs, but following ADP report, many economists raised their forecasts.
Note: The Household Survey and the Establishment Survey each measure two different, though overlapping, things (See Self-Employed Work-at-Home Contractors?).
However, if you adjust them so they count the same things, their differences mostly go away. For the wonkier amongst you, we addressed this way back in 2004’s BLS on Payroll vs. Household Survey.
I am out of pocket for a few hours, but I will follow up later this morn. Please update the 8:30am release and reaction in comments . . .
Survey to Trust: Is Household Poll The Real Deal?
WSJ, December 7, 2007; Page C1