On the first Friday of every month, the Labor Department releases a detailed summary of the previous month’s employment picture. It includes several numbers that quickly spur news coverage, such as initial estimates of the number of jobs added or lost and the unemployment rate. Digging deeper, there are data related to employment shifts by industry, wage data and unemployment figures for various demographic groups.

The report released Friday, covering August, offered a mixed bag of data. Some 130,000 more Americans are working, and wages increased. But that 130,000 figure was lower than expected, and the Labor Department revised its estimates for June and July downward by a combined 20,000 employees.

Those revisions don’t often make headlines, but they’re significant. The initial June jobs report, for example, indicated that the economy had added 224,000 jobs. President Trump retweeted a celebratory tweet from Rep. Mark Meadows (R-N.C.), amplifying that number — particularly in contrast with the 160,000 jobs the economy was expected to add.

After two months of revisions, though, the current estimate for the number of jobs added in June is 178,000, much closer to expectations than the initial report.

For five of the seven months of 2019 that have seen revisions, those revisions have been downward. That’s unusual in recent years; in 2016 and 2017, the revisions were about evenly split, while in 2018, they were mostly revised higher.

(The figures above are seasonally adjusted nonfarm employment, for those curious.)

Overall, jobs numbers in 2016 ended up 168,000 higher than initial reports. In 2017, it was 53,000, and in 2018, 254,000. This year, jobs numbers are 110,000 lower than Labor’s initial reports.

The implication is probably obvious. Looking just at initial jobs report numbers, the three-month average of jobs added would be 173,000 per month at this point. After revisions, though, the figure is 156,000 — about 10 percent lower.

Notice how the opposite was true in 2018, the year in which revisions regularly pushed initial estimates higher. There’s an important footnote to those numbers, though.

Jobs numbers go through three revisions. After the initial announcement, they are revised in each of the next two months. At the end of each fiscal year, there’s an additional final revision, which is reflected in the 2016 and 2017 numbers above.

Last month, the department announced its benchmark estimate for the period from April 2018 to March 2019. According to the initial jobs reports, some 2.3 million jobs were added over that period, a total later revised upward to about 2.5 million. But those numbers were significantly too high, according to the department’s benchmark estimate — by about half a million jobs.

(Since those revisions haven’t been finalized, which they will be in the spring, the numbers for those months on the first chart in this article don’t reflect the new estimates.)

This is an important asterisk that should be applied to the initial jobs numbers. There’s no guarantee that the current numbers will be revised downward, of course, but it’s worth remembering that they are not static — particularly given the frequency with which Trump points to employment numbers as a marker of his success as president.

The other factor that’s important to note in the numbers released Friday is that we’re starting to see hiring for the 2020 Census. Every 10 years, federal government hiring spikes as the Census Bureau hires part-time employees to conduct its survey. The hiring generally peaks in the spring of the census year, but hiring increases in the months earlier.

The latest jobs report indicates that 34,000 of the 130,000 jobs added were census-related. Those are real jobs, of course — but they are also the front end of a surge that will recede in the middle of next year.

There’s a subtext to this article that we might as well make explicit. Relying on initial jobs numbers as a metric of political success is often misleading, if not intentionally disingenuous. On average, the final jobs numbers since Trump has been president have been 27 percent higher or lower than the initial report.

With major revisions to the 2018 numbers looming.