Last month, ADP projected a net loss of private-sector jobs in May of about 2.7 million, part of the expected damage from the COVID-19 shutdowns. The Bureau of Labor Statistics’ official jobs report, however, showed a net gain of 2.7 million, a figure so large and so far off from ADP’s report that a few thought BLS had cooked the books. As more economic measures emerged, however, it became clear that the economy had actually rebounded in May.

Did it continue its rebound in June, however? ADP’s report for June shows a net gain of 2.37 million, a vast change from its pessimistic take on May. More accurately, it’s a big change from its original May report. ADP announced today that they have revised that report by, oh … nearly six million jobs.

Perhaps they can supply some Dramamine with it to deal with turbulence:

And maybe some Dramamine for the explanation, too:

Companies in June continued to bring workers back from their pandemic furlough as the national economy slowly came back to life.

Private payrolls grew by 2.369 million for the month, a bit lower than the 2.5 million expectation from economists surveyed by Dow Jones, according to a report Wednesday from ADP and Moody’s Analytics.

The total actually represented a decline from the previous month, which saw a dramatic upward revision to 3.065 million. ADP initially said May saw a loss of 2.76 million. However, the Labor Department two days later reported a gain of 2.5 million for May, a number that itself was far better than the Wall Street estimate of an 8 million loss.

“There is no information in that revision. It is simply the result of the fact that our objective here is to predict the [Bureau of Labor Statistics] number with the ADP data and to do that as accurately as possible,” said Mark Zandi, chief economist at Moody’s Analytics. “You can’t glean from that that something positive is happening in the labor market.”

Is that really the main objective of the ADP report? That seems strange, especially since it usually precedes the BLS report by only a day or two, which makes its value extremely short-lived. Zandi got brought in by ADP several years ago when people complained that their report had gotten less predictive, so it’s not entirely fair to say that predictive ability has no value. However, ADP’s position as the nation’s largest payroll service company gives it more direct insight into private-sector trends than the BLS Establishment survey does, or it should. If there’s a disconnect between the two, it might not be an issue with ADP.

Liesman’s correct that the instability in the jobs market makes all of these metrics somewhat unreliable. It’s not exactly a surprise that we’re getting surprised, in other words. The scale of the misses make it hard to take any of these reports too seriously, though. Revisions of the five-figure range are understandable, but revisions in the seven figure range make these metrics all but useless, especially with regard to public policy.

The BLS jobs report will come out tomorrow, by the way, rather than Friday, thanks to the Independence Day holiday. That shortens the value of the ADP objective by 24 hours, but it also gives us a quicker look at whether BLS has any corrections of its own. The new ADP number will likely stoke expectations of an over/under of 2.5 million upside in the jobs numbers, which other economic indicators also suggest.

One in particular has not, however — the weekly initial jobless claims figures. Those have remained sky-high at 1.5 million or greater for more than three months now. How can we be adding jobs while seeing that kind of weekly meltdown? The answer, Bloomberg says, is another metric failure:

The weekly U.S. jobless claims report shows that 11 million people are on a new federal aid program geared toward the self-employed, accounting for more than a third of total claims across the country.

But that figure — cited by economists alongside traditional claims in regular state programs as helping provide a more complete picture of unemployment — likely overstates the true count. In some cases, by a lot.

Eight states and Puerto Rico have reported Pandemic Unemployment Assistance claims higher than those for regular state programs, according to figures released Thursday. On the surface, this suggests there are many more self-employed and gig workers losing their jobs.

In reality, states contacted by Bloomberg News explained that the totals reflect backlogged claims from prior weeks finally making their way through the system. In addition, with each retroactive week counted as a separate claim, one person can account for multiple claims in the tally.

In other words, it was a lot worse than we thought in March and April, and much better than the numbers indicate now. Or at least did at the time BLS conducted its June surveys; by now, those numbers might start showing new damage from cities and states reimposing restrictions on commerce as the COVID-19 case numbers spike upward again.

The metrics are interesting, and perhaps even entertaining. Reliable? Well, YMMV.