Earlier, I wrote a post that said the unemployment rate dropped from 9.4% to 9.0% due in large part to unemployed Americans no longer looking for work and exiting the workforce. The Bureau of Labor Statistics contacted me to disagree with this assertion. They say their population revision skews the data to make it appear this way, but in reality 589,000 more people found jobs. And zero stopped looking for work. If you weren't confused before, you will be now.

This information was included in "Table C" of the BLS report:

unemp Table C 2011-01.png

It turns out that because BLS does not update their prior months data for the Household Survey, so you can't accurately compare the January data to the December data. That means using their database isn't a useful exercise for this comparison, because the data is bad. The chart above is all we get. This makes it virtually impossible to make many meaningful claims about changes in the Household Survey data from December to January beyond the few they show in the chart.

So what really happened? Well, according to BLS, the unemployment rate declined in January because people actually found jobs -- a lot of them. To put 589,000 new jobs into perspective, from October though December just 385,000 new jobs were created in all three months combined.

If the Household Survey is accurate, then this is great news. It would easily be a high high for the recovery for monthly jobs added. But is it accurate? Remember, the Establishment Survey said only 36,000 jobs were created. This would mean that the two surveys disagree by a factor of 15. That's a huge difference.

I spoke with Betsey Stevenson, Chief Economist at the Labor Department. She explained that this discrepancy could be due to the shortcomings of the Establishment Survey. As explained here, it doesn't count all types of workers, leaving out self-employed and a few other categories. She also stated that new businesses are harder for the Establishment Survey to find and account for, so there could be some delay in its recognition of these jobs.

But it just seems hard to believe that these effects could account for the huge discrepancy. The biggest overestimate by the Household Survey throughout 2010 between the two surveys was just 335,000 in August. January's is 75% larger.

So what does this say about unemployment in January? There's very little certainty surrounding the data in the report. Some jobs were certainly added -- both surveys agree on that. But the difference between 36,000 and 589,000 is enormous. Hopefully we'll know more in coming months, as this sort of difference between the two surveys isn't common.