> BLS numbers are routinely revised in retrospect to be worse a quarter or two later and with a friday release so that no one notices. As in, more than one in t

> BLS numbers are routinely revised in retrospect to be worse a quarter or two later and with a friday release so that no one notices. As in, more than one in two BLS unemployment reports is retroactively revised, and I don't remember a single "it's actually better than we previously reported" revision. That's how it is supposed to work, they do the best with the data they have and then update as more data comes in. They are trying to gauge hard to measure things using samples, it's a tricky business. As for revising to the upside, it happens all the time. December and January were both revised upwards: http://www.bls.gov/news.release/archives/empsit_03042016.htm http://www.bls.gov/news.release/archives/empsit_03042016.htm > The change in total nonfarm payroll employment for December was revised from +262,000 to +271,000, and the change for January was revised from +151,000 to +172,000. With these revisions, employment gains in December and January combined were 30,000 more than previously reported.