BLS puts out two surveys in its monthly reports. They are the Establishment Survey and the Household Survey. Unemployment is calculated from the Household Survey.
Links:
http://www.bls.gov/news.release/empsit.nr0.htmhttp://www.bls.gov/news.release/empsit.a.htmhttp://www.bls.gov/news.release/empsit.t17.htmWhat you see there is that the two surveys are disagreeing with each other right now. Since April, the household survey shows over 300,000 jobs lost, and the establishment survey shows over 40,000 jobs gained.
But it is important to realize that the establishment survey is adjusted each month by the Birth/Death model numbers, which can be found here:
http://www.bls.gov/web/empsit/cesbd.htmNote that there were big net positives in May and June.
The B/D adjustment is figured from the B.E.D. survey (Business Employment Dynamics). However that survey is always at least 3/4s of a year out of date. This is the most recent, and you can see it is old - this is data from the third quarter of 2010.
http://www.bls.gov/news.release/cewbd.toc.htmSo what they do is to look at how many new businesses and jobs were created by sector (basically in relationship to the sampled jobs), and then adjust the reports from the businesses included in the establishment survey. Usually this works quite well, and it is necessary, because for a while these new businesses aren't known to BLS so they couldn't possibly sample them. However they do see businesses drop out of their establishment sample, so if they didn't adjust for new business creation, almost always the establishment survey figures would be too low.
However, in times of rapid economic change (positive or negative), the B/D adjustment is out of sync with reality because it is always at least three-quarters of a year behind. BLS always goes back and adjusts its establishment survey numbers as they get more current data, but those adjustments occur long after the month is reported. So for historical research the establishment survey is very good, but it isn't always accurate in the first few months.
The correct way to read the monthly employment report is to look at both surveys. When you see a lot of change and they suddenly diverge, check the B/D adjustment and glance through the other indicators, and they usually will give you a strong inkling as to whether the establishment survey numbers are off.
In this case, due to the rise in initial claims, due to high BKs and non-payments reported in some business surveys, due to the much lower estimate in the household survey, and due to the drop in temporary employment reported by the establishment survey, it appears that the household survey numbers are far more accurate at this time.
So I conclude that we are losing jobs, and the final backup to that comes from the Treasury hospital insurance monthly data, which showed a trend change over the last few months indicating that we are losing jobs. June Treasury receipts will be published next week and we will see what they say.