Two Jobs Data: Can Both Be Right?
Here's a guest blog from regular HobbsOnline reader and email correspondent Stan Brown, who writes:

Real Clear Politics has two articles about the economy today. The first is by Mort Zuckerman and the second by Robert Samuelson. Two points on jobs/unemployment data occurred to me while reading the articles:

1) If Congress extends benefits and makes it easier and easier to stay on unemployment, any comparison between the present number of people drawing benefits and some number in the past is bogus. Zuckerman's statement that we have the most drawing unemployment in 20 years is meaningless - we are comparing apples and oranges.

2) Samuelson tells us that the household data and the payroll data are incredibly divergent. Apparently there are two ways we measure the number of employees. We ask employers about their payrolls and we ask households (employees) whether they have jobs. The employees tell us that unemployment is much, much lower. Samuelson says that the payroll data shows a loss of 2.6 million jobs since early 2001. The household data shows a loss of only 108,000 (and a gain of 1.2 million this year).

This kind of disparity is enormous.

He goes on to cite a Standard and Poor's analyst who thinks that the disparity is due to businesses hiring "contract" workers which they don't count on their payroll, but who considered themselves employed since they get paid for working.

The quality of journalism on this issue has been abominable. Why haven't we heard anything about the incredibly good news from the household data? Why haven't we at least heard of the dispute? The administration should have made sure this news was out there. And why doesn't an administration defender point out the apples/oranges nature of the unemployment comparisons?

Good questions, Stan. I'd tend to believe the jobs data gathered from households more than the data gathered from employers. It would go a long way to explaining why the "recession" never seemed to show up in the consistently-strong consumer spending data. I'll shoot a link to this post over to Donald Luskin and the "EconoPundit" Steven Antler and see what they think.

UPDATE: Luskin responds:

I don’t know the precise answers, and I’m not going to pretend I do. My approach to economics is not very much about all these bogus statistics. But what I do know is the following:

1) The "jobless claims" numbers are compiled from actual claims. However, as you point out, all kinds of factors can affect why people might make these claims or not.

2) The "unemployment rate" is derived from a survey of, I think, 60,000 households. It's basically an opinion poll. Because it is a "rate," the number can fluctuate in relation to changes both in the numerator and the denominator. For example, suppose 1 million women suddenly decide to go from being moms to being babysitters. Suddenly you have a million unemployed people.

3) I'm not sure how the "payrolls" data is compiled. Let's assume it is a perfect compilation of actual payrolls. In that case, I think (but I'm not dead sure), that temporary workers might show up on the payrolls of their agencies. That, plus the assumption that it's an actual compilation rather than a poll, suggests that it would be much more accurate - if all you are trying to do is know the "number of jobs." You still have to put that number in relation to something to make it meaningful.

The lesson in this is that these statistics are stupid and dangerous.

And Antler responds:

Answered most of it at EconoPundit. I think your reader is a little off base about the change in unemployment benefits, because these get changed around every recession and even then some.