The Bureau of Labor Statistics released preliminary July job numbers this morning and they are not good. “Job growth totaled a seasonally adjusted 73,000 for the month… below even the meager Dow Jones estimate for a gain of 100,000, the Bureau of Labor Statistics reported Friday,” CNBC reported Friday. But even 73,000 added jobs seems optimistic given that the report also contained the second and third updates for May and June’s job numbers, respectively:

So they were off by an order of magnitude for job growth numbers in May and June. Is that… normal? I don’t know, because like most of you (I hope), I never studied economics. If you are looking for an economics degree I can tell you I don't have one, but what I do have are a very particular set of skills. Skills I have acquired over a very long career. Skills that make me a nightmare for people like your dumbass brother in law, who you might run into this weekend at a BBQ, loudly talking his usual line of bullshit based on the one Yahoo! News headline he half-glimpsed in a new Chrome tab when he was pulling up his fantasy baseball stats. Skills like “Googling to find out what the news means.” Come with me and let’s become an…

So what happened?

Every month the U.S. government’s Bureau of Labor Statistics publishes what it thinks is the total number of “nonfarm” jobs in the American economy. Why do we exclude farm jobs? I don’t know, it doesn’t matter. They count farm jobs too but… listen we’re already in the weeds here which is what happens when farming gets involved.

Today was July 2025 job numbers day and they came in quite low. Generally everyone focuses not on the total number of jobs but on how much the number grew or declined since last month, compared to how much economic forecasters predicted it would grow or decline since last month. This second-derivative number allows people like Joe Weisenthal to post “BREAKING: SLIGHT MISS” and be understood by other economy-knowers, and now by you and I as well. The opposite of a “miss” is a “beat,” where numbers beat the estimates. So now you can roll up to the drinks cooler and say something like “You see those jobs numbers? Huge miss, huh!” with verve and panache.

The total job growth reported today was +73,000, a “SLIGHT MISS.” But these numbers are revised one month after they’re initially announced and then again two months after, and this report contained the second June revision and the third May revision, which changed May’s estimate from +144,000 to a paltry +19,000 and June’s from +147,000 to an even paltrier +14,000.

Wow, those are big revisions for May and June. Is that normal?

Not really! But kinda? Even the B.L.S. says these revisions are “larger than normal,” and the B.L.S. also publishes lots of nice tables of numbers if you want to investigate how much larger than normal they are exactly. Here’s the table for 2025 so far, which you can read and understand now that you know what the columns for “1st,” “2nd,” and “3rd” mean:

For the visual learners, here’s what has everyone agog:

And given Fig. 1 and Fig. 2, many people are asking:

For comparison, here’s the same table for 2024:

As you can see, the numbers always do change, sometimes by a little and sometimes by a lot. But even a very quick Instant Expert™ glance shows us that the largest changes don’t normally affect the order of magnitude of the original guess. “We guessed 147,000 and the actual number is 14,000” is pretty wild.

For a different example of what was considered remarkable in more normal times, here’s the New York Times’ Ben Casselman a year ago, reporting on a downward revision of 818,000 jobs in the twelve months from March 2023 to March 2024, an average error of 68,000 jobs per month, or “a downward revision of about 28 percent.” The downward revisions to May and June 2025 total almost 89 percent.

So this proves Trump is cooking the books right?!

Of course not, if anything this is evidence for the opposite. These revisions didn’t come from some intrepid muckraking journalist turning up secret government memos—they’re normal, scheduled information releases from the U.S. government bureau in charge of measuring jobs. We should probably be glad that someone in the Trump administration is still releasing economic information that tells us the Trump economy is not good. Once again, we don’t need conspiracy theories to observe what is openly happening around us!

Why the big adjustments, then?

To try to answer that question, first I had to find out where the numbers come from because like, was someone lying or what? This was extremely boring so without getting into a lot of detail, the jobs numbers come from a survey called “Current Employment Statistics,” or C.E.S. C.E.S. info is collected from, according to the B.L.S., “a randomly selected proportion of businesses and most government agencies, representing their individual worksites” and updated after one month and two months “because not all respondents report their payroll data by the initial release of employment, hours, and earnings.”

So the initial number is an extrapolation based on a relatively small quantity of survey responses fed into a statistical model based on past results. And statistical models based on past results work great, until they don’t. If economic conditions change unexpectedly between the beginning of a month and the end of the month in a way that severely impacts employment, that might very well weaken the effectiveness of a model that (rightly) assumes that kind of thing usually doesn’t happen. For example, some of the largest past adjustments in initial job estimates were in March and April of 2020 and September and October of 2008.

Wow that was boring, right up until the part about 2020 and 2008. Hey—is the economy crashing?

Maybe! One reasonable conclusion you could draw from these jobs numbers is that the economy might be starting to behave in a way that breaks with normal expectation. Here’s New York Times job numbers correspondent Ben Casselman again: “When hiring is consistently revised down, it often means the economy is in, or headed for, a recession.” And whether the adjustments themselves mean anything or not in what is at best a noisy data series, the actual job numbers suck. +19,000, +14,000, and (if it holds up) +73,000 would be the weakest three-month run in job growth since 2010, when we first started to claw our way out of the 2008 recession. That seems suboptimal.

I think what we can safely conclude is that the job reports are bad, the early reports that Trump’s idiotic monkeying with the levers of the economy wasn’t doing any harm were probably wrong, and the U.S. economy is possibly teetering on the brink of recession. It’s also still possible, of course, that the June numbers will get adjusted back up next month, and so will July’s, and all of this will look like nothing much in the end. We just don’t know yet.

That’s not a very satisfying conclusion.

I know, so this is the point in your weekend BBQ chat where you need to be able to shift from data to ✨vibes✨ in order to keep the conversation flowing. Here’s Christopher Mims in the Wall St. Journal reporting that our economy might just be three AI data centers in a trench coat:

And the Washington Post’s Meryl Kornfield, Hannah Natanson and Laura Meckler reported that “[t]he government is paying more than 154,000 federal employees not to work as part of the Trump administration’s deferred resignation program, according to two administration officials.” Thanks, DOGE! All of the cost, none of the service—a classic Elon Musk business strat. Meanwhile tariffs are higher than they’ve been since 1935 and rather than making America great again, they appear to be driving more U.S. manufacturing offshore. Also the Corporation for Public Broadcasting is dead.

If you’re as tired of talking about the economy by now as I am, this is a good time to pivot to “Hey, did you know Trump paved over the Rose Garden?”

Illiquid Swords:

Talia Lavin expanded on the Great Sword Heist of 2025 in a post. Police have issued the following sketch of the perpetrator, who should be considered armed but heavily encumbered (movement speed reduced by ten):

For Gothamist Clayton Guse, David Brand and Brigid Bergin tracked down 52 fraudulent signatures submitted by Eric Adams’ increasingly criminal campaign for reëlection. This follows last week’s THE CITY report that Adams received money from straw donors and applied it toward public matching funds requests, again.

Today in Cursed Phrases: Labubu BBL, Italian Brainrot Slopjects, and radioactive wasp nest.

And Finally: “Why Did National Geographic Disappear Its Own Documentary About A Queer Climate Scientist?” by Eva Holland for Defector. This is baffling and infuriating and also now I really want to see the film.

Today’s Song: Hayley Williams, “True Believer”

Hayley Williams is doing Liz Phair’s career in reverse, and I’m here for it. Here’s a playlist of all the new surprise singles. Maybe you should play them at the BBQ!