The director of the National Economic Council has found a version of the July jobs report he likes better than the one the Bureau of Labor Statistics published. Kevin Hassett, the White House's top economist, said of July's numbers that "if you throw out the World Cup and the government workers, we actually had a number that was about plus 100,000." [1]

That sentence deserves a label before it deserves a response. It is a framing exercise: a subtraction performed on a published statistic, with the remainder presented as the figure the economy actually produced. The technique underneath it is real. The use it is being put to is not a measurement, and the difference between those two things is the story.

Start with what the report says, because the report is not ambiguous. The BLS recorded a net loss of 23,000 jobs in July, the first negative monthly print since February. [1] The same release revised the two prior months down hard: May fell from 129,000 to 63,000, a cut of 66,000, and June fell from 57,000 to 20,000, a cut of 37,000. Between the two revisions, 103,000 jobs that had been reported as created were removed from the record. The unemployment rate fell to 4.1 percent, which reads as good news only until you see the mechanism: labor force participation dropped to 61.4 percent, its lowest level since February 2021. Fewer people counted as unemployed in part because fewer people were counted at all. We walked through all of this when the report landed two days before Hassett's reframe of it.

Here is what Hassett gets right, because the technique he is reaching for is a legitimate one. Economists strip volatile or one-off components out of a data series all the time to look for an underlying trend. It is why the most watched inflation measure excludes food and energy, and why a single month distorted by a strike or a storm gets an asterisk in any careful analysis. If Hassett had said that two unusual sectors dragged the headline down and the rest of the economy added jobs, that would be an analytical claim about composition, the kind of thing forecasters argue about in good faith.

That is not what he said. He said the excluded version was the number "we actually had." [1] The subtraction stops being analysis and becomes spin at exactly that word, actually. A sector-excluded subcalculation is a lens on the report. It is not the report. The economy Americans work in includes government payrolls and includes whatever one-off effects came and went with the World Cup, and in that economy, the one the BLS measures, July produced 23,000 fewer jobs than June. The distance between the number Hassett offered and the number the government published is 123,000 jobs, and every one of them belongs to the record.

An underlying-trend argument also has to survive contact with the trend, and this one does not. The case for looking past a single bad month is that the months around it are healthy. The July release said the opposite: the two prior months were substantially weaker than first reported, by a combined 103,000 jobs. Strip out July entirely and the revisions alone describe a labor market losing momentum. The participation rate tells the same story from a different angle. A falling unemployment rate driven by people leaving the labor force is not a statistic anyone brags about once the denominator is in view.

The beneficiary of the reframe is not hard to name. An administration facing its first negative jobs print since February would prefer the public remember plus 100,000, and its top economist supplied that number by deleting the parts of the economy that shrank. The cost lands on everyone who has to read the labor market as it is rather than as narrated: workers deciding whether this is a safe moment to change jobs, households deciding whether to take on debt, policymakers deciding what the data licenses them to do. Those decisions run on the whole report.

There is a reason the jobs report includes the sectors that contracted. That is what makes it the report.