Few agenda items rank higher on the MAGA priority list than boosting manufacturing employment. “Jobs and factories,” President Donald Trump touted in his Rose Garden tariffs announcement on April 2, “will come roaring back into our country and you see it happening already.” But shouldn’t we first know how many manufacturing jobs we have? Turns out, getting to a solid number isn’t easy.
As my Economic Innovation Group colleagues Adam Ozimek, Ben Glasner, and Jason He detailed in a recent analysis, asking businesses and asking workers yield wildly different figures. A leading business survey recorded 12.3 million manufacturing jobs in 2023; a leading worker survey recorded 15.1 million. To put that difference in context, 2.8 million is greater than the number of manufacturing jobs that were lost to either the “China Shock” or the Great Recession.
Coming from a family with multigenerational ties to one of the country’s biggest manufacturers, I’ve had a front-row seat for observing not only the ructions of industrial change that have, in part, given rise to the MAGA movement, but also some of the complications that come with employment measurements and cloud the public policy picture. My grandfather, Pat McGillis, started at Boeing in 1957, working in the security division. Though my dad didn’t follow in Grandpa Pat’s footsteps, six of his siblings—Steve, Mary, Kevin, Mark, Patty, and Annie—did.
Kevin got his foot in the door at Boeing in 1977, checking, cleaning, and replacing light fixtures in a plant at Boeing Field south of Seattle. He took advantage of the company’s educational programs and worked his way through college in the 1980s, with Boeing covering tuition. After that, he moved into facilities management.
Mark, Kevin’s twin, started around the same time and would stay throughout his career on the blue-collar side, working as an electrician in a Boeing General Construction Unit, one of a few dozen in the Seattle metro area. Mark’s team spent a lot of its time revamping buildings for new uses, such as converting early-vintage manufacturing-specific buildings into office buildings as the company’s workforce expanded and production was centralized at the giant newer plants further out in the suburbs.
Kevin and Mark are perfect examples of how the manufacturing jobs data can get out of whack. As my colleagues explain in their analysis, a big reason that business-reported numbers and worker-reported numbers diverge is that businesses are directed to categorize their workers’ fields based on the majority activity at the particular facility where they’re based—i.e., rather than by the workers’ particular jobs or by the firm’s overall business. So, while Kevin was working behind a desk as a manager at a manufacturing site, he would probably have been included by Boeing in its reporting of manufacturing workers. Meanwhile, Mark, the electrician, probably would not have been counted when his crew was working on plant-to-office building conversions.
Workers in the aforementioned second survey were also instructed to base their answers on the “activity, product, or service provided at the location where employed.” But many workers are apt to respond with how they think of their own job or with what the company does as a whole, not to scrutinize the survey’s fine print. If both Kevin and Mark were to fill out the survey and record themselves as manufacturing workers, could you really blame them?
While economists usually defer to business-reported numbers, workers are worth listening to as well. Given the high social status that comes with manufacturing—compared to other fields that are open to workers with similar educational attainment, like retail—I think it’s only natural that workers will tend to check that box, even if the location where they work is mostly focused on a different part of the business, like research, training, or sales.
Subfields like automobile and aerospace manufacturing display some of the largest worker-reported versus business-reported gaps. Per the 2023 worker survey, auto manufacturing employment is near an all-time high, and manufacturing employment in the subfield of “ships, planes, trains, and military vehicles” is almost 70 percent higher than in the business survey.
Far from being erroneous, the worker reporting likely communicates that the workers think they’ve got pretty good jobs. And good jobs, obviously, are a part of what the Trump administration is after. But if the number of manufacturing jobs in this country is understood to be 20 percent higher than commonly thought, our policy conversations should take that into account.
Importantly, the gaps between worker-reported and business-reported surveys have widened over time, due in part to definitional changes. In 1997 some statistical agencies, including the Census Bureau and Bureau of Labor Statistics, switched from using the Standard Industrial Classification System, which had been in place since the 1930s, to the North American Industry Classification System. Not coincidentally, it was in the years after 1997 that large deltas emerged between some of the worker-reported and business-reported datasets. Research from economists Teresa Fort and Shawn Klimek suggests that the change in industry definitions led to 1.4 million jobs being reclassified from manufacturing to services in the Census Bureau’s Longitudinal Business Database, for example. As my colleagues write, “what is counted as a manufacturing job is a statistical choice, and one that government statisticians have made differently in the past.”
My uncles Kevin and Mark neither worked directly on aircraft assembly, nor, for long periods, at manufacturing-specific facilities. Yet their work contributed to Boeing’s manufacturing enterprise. Should a worker like Kevin be counted toward manufacturing if he has taken off his toolbelt and stepped behind a desk? Should one like Mark be counted toward manufacturing totals even if his in-house construction unit is working on a hangar-to-office conversion?
The answers here aren’t obvious. But Kevin’s and Mark’s paths show how worker surveys might better capture the health of our country’s industries than the oft-preferred—and more pessimistic—business surveys. Indeed, a manufacturing enterprise can grow its overall domestic employment—i.e., add good American jobs—and expand its productive output even while reducing the number of workers that would be categorized as manufacturing in a business survey.
For the economy at large, there’s an important point here: Our business-reported manufacturing numbers are missing a big chunk of workers who have family-supporting, middle-class jobs and who are engaged in work that bolsters physical production. That these workers often think of themselves as having manufacturing jobs is worth keeping in mind when we’re told American industry has been hollowed out. If we take the worker perspective seriously, the familiar narrative of manufacturing’s decline doesn’t land quite so forcefully.
