Earlier this week Gallup reported that their unemployment rate had risen to 9.1 percent and predicted that the government's unemployment rate would rise to 8.6 percent. That didn't happen. Today, the Department of Labor's Bureau of Labor and Statistics (BLS) reported that the U.S. economy added 227,000 jobs and unemployment was steady at 8.3 percent.

Why the huge difference between the two numbers?

Both Gallup and the BLS use randomized surveys to produce estimates of the current state of the labor market. Gallup calls 30,000 people every month over the span of the entire month. BLS conducts 60,000 interviews a month (both face-to-face and over the phone), but conducts them all in one week. More importantly, however, the BLS uses a model to smooth their raw numbers out to account for seasonal swings in the labor market. Gallup does not.

As the chart above shows, every January the U.S. economy sheds more than million a jobs as retailers let people go after the Christmas shopping season. There is another smaller drop off in the summer as kids leave their summer jobs and return to school. 

Once seasonal patterns are accounted for, the BLS and Gallup numbers generally track each other. But not always. And not right now. Gallup explains:

Last February, the U.S. Bureau of Labor Statistics applied a seasonal adjustment factor of 0.5 points to its unadjusted unemployment rate for the month. If that same seasonal adjustment is applied to Gallup's mid-month unemployment rate of 9.0%, it would produce a seasonally adjusted unemployment rate of 8.5%. Alternatively, if it was applied to Gallup's full-month unemployment rate of 9.1%, it would produce a seasonally adjusted rate of 8.6%. Gallup therefore forecasts an increase in the unemployment rate.

The BLS does include a seasonally adjusted unemployment number in each report. For February it is 8.7 percent, which is still below Gallup's 9.1 percent number. Why?

Every month reporters usually mention two numbers from the BLS: the number of jobs created/lost and the unemployment rate. Most people assume that the unemployment rate is a function of the jobs number. It's not. The BLS creates both numbers from completely different surveys.

The 227,000 jobs number you heard about this month comes from the Current Employment Statistics (CES) survey. The 8.3 percent unemployment rate comes from the Current Population Survey (CPS). The CES is a survey of employers. The CPS is a survey of workers.

The CPS jobs total is always higher than the CES total, but their gains and losses usually do track each other. But not this month. In February, the CES reported a seasonally adjusted jobs gain of 428,000. That is almost double the reported 227,000 CES jobs number. And that 428,000 is the number the BLS used to come up with 8.3 percent unemployment.

Will the CPS adjusted job gains continue to outpace the CES numbers? Probably not. Just look at the unadjusted CPS job numbers which were actually smaller than the CES numbers were. While employers reported 851,000 unadjusted new jobs last month, workers reported just 740,000 unadjusted new jobs. In other words, while the seasonally adjusted CPS job gains were double the CES numbers, the unadjusted CPS job gains were actually smaller than the CES ones.

Unless the unadjusted CPS numbers begin to rise as fast adjusted numbers do, the unemployment rate will go up.