Why You Can't Trust The New Job-Loss Numbers

Enlarge

The pink line is misleading. Click for a larger version.

The pink line is misleading. Click for a larger version.

The Department of Labor released its weekly unemployment insurance claims report today, which says there were 565,000 new jobless claims the week ending July 4. But that's not the actual number of jobless claims — it's the seasonally adjusted number. Actual jobless claims for that week were 577,506.

Welcome to one of the weirder corners of labor statistics.

There are really two ways of looking at today's report. Everyday folks can look at it to try and find out how many new people claimed joblessness in the past week. But what Labor Department researchers want to see is whether the economy is going along as they predict it should in any given month. Their predictions follow seasonal trends. For example, the unadjusted number of claims always spikes in January, since temporary holiday workers are laid off at the end of each holiday shopping binge.

The Labor Department adjusts the jobless claim numbers based on those patterns. Any variation on top of that is what's unpredictable. The more the jobless numbers change week-to-week after seasonal trends are factored in, the more unpredictable the job market. In a perfect world, the pink line on the chart above, for the adjusted number, would be flat — the Labor Department would be able to predict any factors affecting claims and account for them.

Of course, we're not in that world, and the recession has made the job market markedly unpredictable. As you can see from the chart below, adjusted jobless claims are way up from last year.

Note the gap between last year and this year. (Numbers are seasonally adjusted.)

Here's the deal with today's numbers: the Department of Labor is still using the old formula. It factors in the typical pattern for July, when auto companies lay off workers and retool their factories to build cars for the next model year. This year, so many auto plants closed earlier that there are far fewer July auto layoffs. But the department's formula still factors them in — it expects those layoffs to happen. Department researchers know the problem exists, and they've been telling reporters not to make too much of today's figures.

Economists generally use the weekly numbers to get a bead on the monthly jobs report — the one that just clocked in at 9.5 percent unemployment. Now, writes economist Ian Shepherdson, the weekly data will be "meaningless" for the next few weeks. As a result, it will be hard to know if jobless claims are actually going down because the recession is easing or because the formulas can't account for the recession.