The payroll survey, which is based on responses from 400,000 companies, shows a loss of 2.2 million jobs since early 2001.

But the payroll survey indicated 164,000 more people were on the job.

When this is done the gains in both the household and payroll surveys average about 300,000 a month.

As a result, 1 million jobs have been artificially "lost" in the payroll survey since 2001.

An even bigger problem with the payroll survey is the evolution of what constitutes work.

We should be prepared for the job numbers in the payroll survey to permanently look anemic compared to other measures.

But the payroll survey has limitations, and they may be important ones right now.

Measured the same way, the payroll survey has shown annual declines for 22 months.

As far as the payroll survey goes, there weren't any big surprises.

The numbers these models produce have been lagging behind the payroll survey for years.