Bob O has pointed out more on the auto employment weirdness to me. The mess that Greenspan made has some great writing on how the early shutdown of much of the auto industry due to the bankruptcies as GM and Chrysler is skewing the unemployment numbers.

Up until recently there was a very regular pattern to the late July shutdown in which a number of people would be laid off and then rehired. That didn't happen at the same time this year.

It's even clearer looking at the raw numbers.

Between 1999-2008, on average the auto shutdowns lead to layoffs which put 47,630 people out of work. This year those normal layoffs came in April and May. And GM and Chrysler brought people back in in July. The long term trajectory of the sector is reduced employment, but the because of this early shut off as TMTGM noted:

Instead of 50,000 or more layoffs in July, there were just 8,600 actual job losses which then turned into a seasonally adjusted gain of 28,200 jobs, the biggest July increase in decades.

At GM I know for a fact that the people that they've brought back on at the factories are working 10 hour shifts and six hour days. Basically, they are working a shrunk workforce harder, so increased production isn't going to necessarily mean increased employment. Also, the company is planning a 3 week shutdown in October to do the retooling for the next year's models. Which means that the same statistical weirdness that made it look like there was a drop in job losses in July is going to make it look like we are on our ways to a double dip recession when the seasonal adjusting of numbers turns a modest shutdown into something that looks much worse.