There’s a line that President Trump uses a lot when he’s bragging about how good the economy is. He trotted it out last week at a campaign rally in Manchester, N.H.

“We’ve created over 6 million new jobs since the election,” Trump said. “If I would have said that during the campaign, those people back there would have been headlines, headlines.”

One of the things we’ve learned about Trump over time is that his rhetorical points often devolve into a sort of insidery shorthand. Trump has said this thing about the number of jobs that have been created any number of times and followed it up any number of times with a claim that the media would have been flabbergasted/furious had he predicted such job growth during the campaign. His supporters know that’s what he means, since he says it all the time, so Trump just sort of waves at the punchline.

As has been noted scores of times, though, it’s not true. Over the course of 2016, the economy was adding jobs at a rate of about 193,000 a month. Maintaining that clip over 31 months in office, we would expect Trump to have added about 5,983,000 jobs — or, to round up, 6 million.

Notice that Trump doesn’t measure from the moment he took office, though. He measures from the election, giving himself credit for creating jobs simply by virtue of not losing the electoral college vote. Since the election, the country has added about 6.2 million jobs. Since Trump took office, though, it’s added about 5.7 million.

Those figures are from the Bureau of Labor Statistics’ monthly reports. On the first Friday of every month, the BLS releases new data on job growth in the United States, focused on non-farm employment. Last month, for example, the country added 164,000 jobs, about in line with the monthly change over the course of 2019.

But there’s another important part of the monthly data. Each report also includes revisions to past jobs numbers, correcting the released figure up or down as more accurate data comes in. The initial release about May’s employment, for example, indicated that 75,000 jobs had been added. In June’s report, that number was revised downward: Actually, 72,000 jobs were added. In the July report, it was revised down even more to 62,000. That’s a change of 13,000 — about 17 percent lower than originally indicated.

This happens every month, and the swings can be significant. At the end of each fiscal year, final numbers are released that further adjust the data. The result is that each month’s figures waver several times before being considered complete.

In practice, that looks something like this. Since January 2016, the monthly numbers have swung by as many as 114,000 jobs (as happened in September 2016). The revisions can result in final numbers that are several times larger or smaller than originally thought. For example, the initial 20,000 figure for this February was updated to 56,000, nearly triple the original value.

(Philip Bump/The Washington Post)

Over the course of Trump’s presidency, most of the original monthly numbers have ended up higher than originally reported. From March through June, though, they were all revised downward — to the tune of more than 130,000 jobs. (July’s numbers haven’t been revised yet, since the month was the subject of the most recent jobs report.)

In other words, 130,000 fewer jobs have been added since March than were originally reported. Compared with a revision released Wednesday, though, that change is relatively modest.

So that it can confirm the accuracy of its data, the BLS also conducts an annual benchmark revision looking at state unemployment records. A preliminary estimate of the benchmark revision is released in August and is finalized at the beginning of the following calendar year.

The benchmark released Wednesday estimates that more than half a million fewer jobs were added between April 2018 and March 2019 than originally estimated. In other words, the government now estimates not that 6.2 million jobs were added since the election but that about 5.2 million have been added since January 2017.

(Philip Bump/The Washington Post)

The downward revisions affected most industries. Retail added 146,000 fewer jobs than originally thought. Leisure and hospitality added 175,000 fewer. Manufacturing added 3,000 fewer. The overall effect is that instead of 210,000 jobs being added per month during that period, the figure was instead closer to 170,000 a month — below the rate in 2016 and 2017.

These revisions are preliminary and, as with the jobs numbers themselves, could move up or down again when finalized. But it does have an immediate effect — making Trump’s rhetoric about jobs being added even more wrong.

“We’ve created over 6 million new jobs since the election,” he said in Manchester, which is true if you measure from the election (which doesn’t make sense) and if you ignore that the number of jobs created appears to have been off by half a million, or about 9 percent of the total originally reported to have been added since Trump took office.

The paragraph above would, in fact, have made for an interesting headline in 2016.