The United States publishes two employment numbers: one based on a survey of employers ("the payroll survey") and one based on a survey of households. The household survey gets more ink, but its proven itself unreliable.
As I've discussed repeatedly, unlike most other countries, the United States publishes two employment numbers: one based on a survey of employers ("the payroll survey") and one based on a survey of households. Since they in principle describe the same real world phenomenon, the number of people employed in the U.S., they should have the same result. As it happens, they usually don't meaning that at least one of them must be wrong.
Because the payroll survey result is less volatile on a monthly basis, economists usually consider it more reliable. I basically agree with that, with the reservation that if the household survey systematically (not necessarily all individual months, but on average) diverge in one direction then this is a hint that the payroll survey number likely underestimates or overestimates (depending on which direction the divergence is) employment growth.
And lately, we have seen that the household survey has in fact been systematically weaker. During the latest six months, between February 2012 and August 2012, payroll survey employment saw a gain of 580,000 jobs or about 0.45%. That is not exactly what one would call a vigorous boom, but at least it's in line with or even slightly above population growth. By contrast, household survey employment saw a gain of merely 36,000 jobs, less than 0.03%, far below population growth.
This means that total unemployment, counting not just those officially unemployed, but also discouraged job seekers, has in fact probably increased this year, despite the fact that the headline numbers cited by the financial press might give you the opposite impression