The massive downward revisions (-258k) are puzzling. Even as someone who long believes payrolls are overstated by massive amount, these downward revisions are anormay large outside of benchmark month or weather related events. The downward revision are largest for state and local government, but also broad based. So far we can rule out the following explanations: 1. Seasonal adjustment. The revisions are mostly NSA. 2. Collection rate of the survey. The collection rate not abnormally low and in fact higher than last year’s may to july. 3. Birth and death model revisions. But BD is for private payrolls not government. Plus the published page got BD shows estimates were unchanged. Notably government payrolls are not connected from survey but from administrative data. Why the administrative data is missing by so much is a quandary.

Today’s payrolls numbers appear to corroborate what one expects should have happened since April (ie our payrolls “bloodbath” warning back then)…logistics down, construction down, all levels of government down…. But what left me wondering is why the contemporaneous administrative data and macro data, plus the normal wedge between BLS reported data and underlying, are all shifting.

I will also add that a stretched BLS with staff shortage can indeed bring about all sorts of new and unpredictable methodological changes in statistics.