The unemployment figure: error ranges cut both ways.
You surely know about the right-wing-fueled concept that Friday's drop in the unemployment rate was so large that the numbers either a) had to be cooked, or, failing that, b) were untrustworthy. The first part is silly; as Ezra Klein pointed out, an administration set on cooking the figures would also have cooked up better hiring figures in the establishment survey.
But the concept that the drop in the unemployment percent is statistical noise bears more merit. Every competent statistician, and every educated news reader knows there are plus or minuses to all these figures. Meaning that the "real" unemployment figure may be greater than 7.8%. Or perhaps even lower still.
And it's also possible that prior months' unemployment figures were also overly negative. (Kevin Drum wrote about this yesterday. The same thought had occurred to me on Friday; they revise prior hiring figures, but never the reported unemployment rate. But Drum wrote about it first.)
I do like the idea, though, that confidence interval on the unemployment figure could also go lower than 7.8%.
