Still confused why you continue to emphasise estimate when ADP is also an estimate.

Because BLS is a much shittier estimate.

revisions are not shocking, but perfectly normal.

Lol no, not to this degree, they absolutely are not, and its been getting worse for 2 years.

I don't think anyone is that surprised (though we can agree that there are a lot of overconfident and less statistically literate professionals)

Well the DJIA drop on the print, and every single news outlet's reaction indicates that a lof of people were surprised, but ok then.

Literally, expectations were +110k to +120k.

Which is pathetic and anemic even as an expectation, should be more like +220k for the economy to just be treading water basically, but nope, missed even that sad expectation by almost half.

ADP was +104k in July, so by your previous logic we should expect upwards revisions in the BLS July number as the year goes on right?

My... logic points to this?

I don't see how you draw that conclusion but ok.

Lets compare June ADP and BLS:

ADP: -28k / not revised because they don't need to.

BLS: +147k / revised down to +14k

ADP measures private payrolls, and then basically just statistically broadens their sample set to an estimated size of entire private economy by sectors and business sizes, for the same month.

BLS on the other hand does this for the entire economy, and has as fundamental parts of their modelling both an attempt to normalize for seasonality and also tries to figure in estimated population growth at the same time.

Meanwhile, the BLS counts of just pure private sector jobs that are almost exactly the same as what ADP counts... they are actually pretty close.

Post revision BLS has July at +83k private sector jobs. Thats not that far from ADP's +104k July private sector.

Roughly, to get the ADP number from BLS, for private sector, you just undo their seasonal normalization, then you're quite close to the ADP number most of the time.

BLS just royally fucked up everything not core private sector.

So no, I would not expect much of an upward revision in July BLS numbers, at best, they'll find they slightly undercounted private sector jobs, but overcounted a bunch of other categories, that would be my expectation.

...

BLS has a giant tome of complexity that goes into the totality of their big number print, with many, many factors that can be (and have been) wrong.

ADP's is less comprehensive over the whole economy, but is also much less complex and direct.

Less failure points = More reliable.

Its kinda like how in the past year or so, roughly 1/3 of the reported prices that actually make up the inflation number, CPI... well they are not actually reported prices.

They are model derived. Not actual reported data.

Lies, damned lies and statistics.

They more and more stats you pile onto something means you have more and more assumptions and reliance on models, the fact that BLS keeps having to revise shit means their models and assumptions are inadequate.

The BLS should probably be reporting error bars and ranges for their sector specific and the compounded error bars for their total metrics, I can tell you that the stuff they've been putting out would not make it through a peer reviewed academic journal as being statistically valid to any serious degree.

While I agree that Trump just firing the BLS head and likely replacing them with a stooge is even worse, the chuckleheads at BLS have been incompetent for a while, as has been everyone else who hasn't noticed this.