The Dwindling: How the Workforce Hollows Out

The Last Shift — Column 3

There’s a town in northeastern Ohio — one of dozens like it — where the biggest employer used to be a stamping plant that made parts for cars. At its peak, 1,400 people worked there. It was the kind of job you didn’t need a degree for but could raise a family on. Union wages, pension, health insurance, a sense that if you showed up and worked hard, the future was reasonably predictable.

The plant didn’t close. It automated.

Today it employs 340 people. It produces more parts than it ever did. The town, meanwhile, has been slowly hollowing out for fifteen years. The hardware store is gone. The diner is gone. The high school graduated 60 kids last year, down from 230 in 2005. The ones who could leave, left. The ones who stayed are figuring out what a town does when its reason for existing has been handed to a machine.

This is what displacement looks like when it isn’t on the news.

The Numbers Nobody Celebrates

Between 2000 and 2024, the United States lost more than 4.5 million manufacturing jobs — a 26 percent decline in the workforce — while manufacturing output actually grew by 45 percent. Read that twice. The factories are producing more than ever, with roughly a quarter fewer workers. The gap between what the machines make and what the people earn has never been wider, and it is widening.

In just the twelve months ending in late 2025, the manufacturing sector shed nearly 94,000 jobs. September alone was the fifth straight month of decline. This was happening, notably, while the broader economy was adding jobs in other sectors — healthcare, construction, some corners of tech. The stock market was fine. GDP was fine. The macroeconomic picture looked, depending on who was drawing it, mostly okay.

The people in that Ohio town were not fine.

This is the central deception of the current moment: the aggregate numbers absorb the damage. When a thousand workers lose jobs in one county, and two hundred coding jobs appear in a city three states away, the net number looks manageable. The economy “added” jobs. The system is “working.” What the numbers don’t show is that the thousand displaced workers cannot easily become the two hundred coders, and that even if they could, two hundred doesn’t equal a thousand.

The Reskilling Myth

Whenever this topic comes up in policy circles, someone says the word “reskilling,” and everyone nods. The idea is intuitive and appealing: workers displaced by automation simply learn new skills, pivot to new industries, and re-enter the workforce in roles that machines haven’t taken yet. Problem solved. Next question.

The Brookings Institution, which has studied this carefully, puts it plainly: we should stop treating reskilling as a reliable answer. The evidence that public retraining programs significantly improve outcomes for displaced workers is, at best, inconclusive. Workers who retrain frequently end up in occupations that are equally exposed to automation — moving from one vulnerable job to another. The skills gap they’re trying to close is a moving target, and the programs training them often can’t predict fast enough where the safe harbors actually are.

Then there’s the age problem. The workers most likely to be displaced are not fresh out of college. They are 45, 52, 58 years old — people who spent decades getting good at something specific and have neither the time nor, in many cases, the financial runway to start over from scratch. Older workers, when they lose jobs, stay unemployed more than twice as long as younger workers. When they do find new work, it typically pays less than what they lost. The math of reskilling works a lot better on a whiteboard than it does in a 54-year-old’s life.

And the United States, compared to its peer nations, invests almost nothing in workforce development. The country ranks near the bottom of OECD nations in active labor market policy spending — about 0.1 percent of GDP, putting it second to last, just ahead of Mexico. Germany, Denmark, and France spend many times that. The gap between what the reskilling rhetoric promises and what the system actually delivers is vast, and workers are falling into it every day.

Automation doesn’t spread evenly. It concentrates—geographically, economically, and demographically—hollowing out entire communities at once.

Where It Hits Hardest

Automation does not land evenly. It concentrates.

It concentrates geographically — in manufacturing-dependent towns in Ohio, Michigan, Pennsylvania, and Indiana, where entire local economies were built around a handful of large employers. When those employers automate, the ripple effects are total. The tax base contracts. Schools cut programs. Local businesses lose their customer base. Property values fall. Young people leave for cities with more opportunities, and the towns that remain skew older, poorer, and increasingly disconnected from the economic mainstream.

It concentrates demographically — falling hardest on workers without college degrees, on women (who hold a disproportionate share of the clerical, retail, and service jobs most exposed to automation), and on communities of color, who are overrepresented in the sectors being hollowed out first.

And here is what makes the current moment different from previous waves of automation: it is no longer concentrating only in blue-collar work. The junior analyst, the paralegal, the insurance adjuster, the entry-level software developer — all of these are now in the same economic position as the factory worker was twenty years ago. The credentials changed. The vulnerability didn’t.

The Tax Base Nobody’s Talking About

There is a fiscal crisis hiding inside this story that almost no one in public life wants to address directly.

Robots don’t pay income tax. They don’t pay into Social Security or Medicare. They don’t contribute to the local property tax base. When a company replaces a hundred workers with an automated system, it captures enormous productivity gains while the public sector absorbs the cost — unemployment insurance, retraining programs, increased demand for social services, the long-term economic depression of the communities left behind.

The companies doing the automating are not villains. They are following the logic of the market, and the market is rewarding them handsomely for it. But the mismatch between who benefits and who pays is not sustainable indefinitely, and the political consequences of ignoring it are already visible in the kind of grievance-driven politics that flourishes when people feel that the system has stopped working for them — because, in material terms, it has.

Automation once replaced jobs slowly. This time it’s hollowing out the middle of the workforce within a single lifetime.

The Part Nobody Wants to Say Out Loud

The optimists have a reply to all of this, and it deserves to be taken seriously. Every previous wave of automation eventually created new jobs — jobs that nobody predicted and that couldn’t have existed before the technology made them possible. The elevator operator became the computer programmer. The buggy whip maker became the auto mechanic. New industries absorbed the displaced workforce, and the overall standard of living rose. This happened. It’s real. History is genuinely on the optimists’ side.

But the pace is different now, and the breadth is different now, and the workers being displaced are different now. Previous disruptions moved through one sector at a time, across generations. This one is moving through all sectors simultaneously, within a single working lifetime, and it is moving into cognitive work — the kind of work that used to be the safe destination for people escaping the automated physical jobs.

The workforce isn’t just dwindling. It’s hollowing out from the middle — the stable, skilled, mid-wage jobs that held communities together and gave ordinary people a sense of economic dignity. What’s left, increasingly, is a thin layer of highly-paid work at the top, a thick layer of low-wage service work at the bottom, and a growing silence in between.

The town in Ohio is quiet in a way that towns aren’t supposed to be quiet on a Tuesday afternoon.

Next column: “After Work — What Happens to a Planet That Doesn’t Need Most of Us”