The plant that never reopened
Picture the road out of a town built around a single plant. For two or three generations the plant set the rhythm of the place: it filled the shifts, the diner at shift change, the union hall, the high-school lot on a Friday night. Then one year the plant closed. Not for a strike or a slow season but for good, the machines auctioned off or crated for a port, the gates chained. The town waited for something to take its place, because something always had before, and this time nothing came. What came instead was a slow leaving. The people with the most schooling and the fewest ties went first, then the younger families, until the ones left on the street were mostly those who could not easily go: the old, the sick, the owner of a house now worth less than its mortgage. Standing on that road a decade on, past the dark plant and the houses with plywood where the windows were, a person is right to ask the question the rest of the country asks about places like this, usually with a shrug. Why did it never come back, and whose fault was that?
This chapter is about the places that fall and do not get up: the mill town, the mining valley, the one-industry city whose industry died. It asks why recovery, which the textbooks promise will arrive as workers and capital flow toward better uses, so often does not arrive at all, and whether the usual verdict on these towns, that they simply failed to adapt, survives contact with the evidence. The answer runs through two things a town does not have that a country does, and one thing it has that a country does not.
A town is not a country
The chapters just before this one followed decline at the scale of a whole nation: the economy that stalls after a long boom, and the one whose population ages. A nation in trouble, for all its trouble, is still sovereign. It has a central bank that can cut interest rates or print money, a treasury that can tax and spend and borrow, a border it controls, an exchange rate that can fall and make its goods cheap again. It can be wrong about which lever to pull, and often is, but it has the levers. The one thing it cannot do is leave. A country cannot pick itself up and relocate to a part of the world with better prospects; it has to solve its problem where it stands.
A region is the mirror image of that. A town or a county has almost none of the tools. It cannot set an interest rate, cannot run a currency, cannot devalue to make its shuttered factory competitive again, cannot borrow at will or print a dollar. Its budget rises and falls with a tax base it does not control, and it competes with every other town for whatever investment is going. What it has instead, the one power a country lacks, is that its people can move. A worker in a dying town can do the thing the textbooks recommend and go to where the work is, in a way a whole nation never can.
That freedom looks like the region’s single advantage, and it is the reason so many economists long expected these shocks to heal themselves. If a factory closes and the local wage falls, workers should move out until the ones who remain are scarce enough to be worth the old wage again, and capital, chasing that cheap labor and empty land, should move in. The town should re-equilibrate. The trap is that the same freedom, worked through real people rather than the smooth averages of a model, is exactly what turns a one-time shock into a spiral that feeds on itself.
Falling, and staying fallen
Detroit is the case everyone reaches for. In the first third of the twentieth century the automobile turned it into a boomtown of a kind the country had rarely seen: its population more than quintupled between 1900 and 1930, and by 1950 it peaked at just over 1.85 million people, the fifth-largest city in the United States, a place whose assembly-line wages had built a broad blue-collar middle class. Then the line turned over and never came back up. By 2020 the city held about 639,000 people, roughly a third of its peak and a smaller number than it had in 1920. Over the same century the country it sits in more than quadrupled in size.
What the line does not show, and what the word decline can hide, is who was left standing in it. The people who leave a failing place are not a random sample of it. They are the younger, the better schooled, the ones with a trade that travels and savings enough to move, and they go where wages are higher, which is to say out. What remains behind is older, poorer, and sicker than what left. Detroit today has a median household income a little under half the national figure and roughly a third of its people below the poverty line, against about one in eight across the country. The metropolitan area that contains it, counting the suburbs many of those leavers moved to, sits far closer to the national average. The collapse is a fact about the city inside the ring, and the ring is where a good deal of the money went.
Who paid for the cheap goods
A single plant closing for local reasons is one kind of story. What made regional decline a national subject, rather than a scattering of local misfortunes, is that a great many towns lost their factories to the same force at the same time, and economists have measured that force directly, town by town. When China entered world manufacturing in the 1990s and 2000s, the flood of cheap imports did not fall evenly. It fell hardest on the particular American places that made the competing goods: the furniture towns, the textile counties, the makers of toys and shoes and small metal parts. David Autor, David Dorn, and Gordon Hanson, comparing local labor markets by how exposed each was to Chinese competition, found that this one force explained about a quarter of the fall in American manufacturing employment between 1990 and 2007. A later study put the loss at roughly a million manufacturing jobs, and close to two million once the suppliers and the shops those workers spent their wages in were counted, over the years to 2011.
The number matters, but the shape of it matters more. The gains from cheap imports were spread across every household in the country, a few dollars off the price of a sofa or a set of tools. The losses were concentrated on the towns that had made those things, and they did not disperse. Autor and his co-authors found that adjustment in the hit places was, in their word, remarkably slow: a full decade after the shock arrived, wages and the share of people working were still depressed and unemployment still raised in the exposed towns. And the escape valve the models counted on, people moving out, barely opened; the net loss of jobs was not made up by out-migration or by new work arriving. The distress stuck to the place because most people did not leave, and absorbed the blow instead as lower pay, lost jobs, and rising rolls of disability and assistance. The cheap goods everyone bought were real, and so was the bill; it was simply handed to a small number of particular towns rather than split among the many who benefited.
The imports, or the robots?
Trade is not the only suspect for the missing factory jobs, and the evidence does not cleanly convict a single culprit. Over the same decades that imports rose, factories were also automating, replacing workers with machines that did their tasks faster and without pay. Daron Acemoglu and Pascual Restrepo, tracking industrial robots across American local labor markets from 1990 to 2007, found that each additional robot per thousand workers measurably lowered local employment and wages, with a single new industrial robot associated with several fewer jobs in the area around it. On this reading the towns lost their work to technology as much as to trade.
A competing view says the automation story is weaker than it looks. American manufacturing output kept rising even as its workforce shrank, which is usually taken as proof that machines were producing more with fewer hands. Susan Houseman has shown that most of that measured productivity miracle comes from a single corner of manufacturing, computers and electronics, and largely from the way statisticians account for ever-faster chips rather than from factories actually shedding labor to machines. Strip out that one industry and the evidence that automation emptied the factories thins considerably. Where the trade shock leaves a clear statistical fingerprint on specific exposed towns, the automation shock is harder to pin to a place and a year.
The two forces moved together, through the same decades and often the same towns, and the studies built to separate them disagree about how the blame divides. That disagreement is genuine and it is not resolved here. For the worker standing on the road, it is also beside the point. Whether the job left for a factory in Guangdong or for a robot two counties over, it left for reasons settled far from the town and past its power to refuse. The cause was not a local failure of effort. It was a change in the world economy that happened to break on this particular shore.
The spiral
What turns a hard shock into a permanent one is the loop it sets spinning, and the engine of that loop is the very mobility that was supposed to cure it. When the anchor employer goes, the people best able to leave leave, because they are the ones with the skills and savings to start again elsewhere. Their departure does two things at once. It removes from the town the workers and entrepreneurs a recovery would have been built on, and it shrinks the tax base and the customer base that pay for everything shared. A city’s costs are largely fixed in its pipes, roads, streetlights, and school buildings, sized for the population it used to have, so when the people thin out, the cost of keeping the place running is spread over fewer shoulders. Services get worse or taxes go up, usually both. A town with failing schools, thin policing, and boarded storefronts is a worse place to stay, which gives the next tranche of people who can leave their own reason to go. Each turn of the loop strips out more of exactly what a revival would need.
This is where the region’s one power, the freedom to move, turns against the place. A country cannot lose its most capable people to the country next door in quite this way; its talent is more or less stuck inside its borders, available to be taxed and put to work on a recovery. A town’s talent can and does walk to the next town, and the walking is individually sensible every time. Nobody in the story does anything foolish. The young engineer who takes the job three states away is making the right call for herself. It is the sum of all those right calls, with no central bank or treasury to offset them, that leaves the town with the fewest tools exactly when it needs the most. What becomes of the workers who leave, and whether retraining and relocation policy could do more than it has to cushion the landing, is the subject of a later volume (Volume IV, on what happens to the workers automation displaces).
The places that climbed back out
Not every fallen region stays down, and the exceptions matter here, because the case for blaming the towns themselves rests on them. Pittsburgh is the one usually named. When its steel industry collapsed in the 1980s the region lost well over a hundred thousand manufacturing jobs and more than a hundred and fifty thousand residents, a blow as heavy as any mill town’s. Yet over the following decades Pittsburgh built a different economy on top of the wreckage, one of universities, hospitals, and later software and robotics, and its unemployment and incomes recovered where a Youngstown’s or a Flint’s did not. Adjustment, the hopeful theory that regions eventually reallocate toward new work, is not pure fantasy. It happened here.
Two things keep Pittsburgh from being a template the others simply failed to copy. The first is that its recovery ran on assets most single-industry towns never had: two major research universities, a large medical complex, and a base of corporate headquarters that outlived the mills, none of which a mill town can conjure by wanting one. The second is that even Pittsburgh did not refill. Its economy came back; its city population did not, sliding further even as the new jobs arrived. Reinventing a regional economy is possible, on the right foundations and over a generation or more. Refilling the town that emptied is rarer still. The reinventions are real, and they are few, and both facts have to sit in the verdict at once.
The towns that lost their factories just failed to adapt.
Backwards Moderate confidence
The claim puts the cause of the ruin inside the town, in a failure of effort, nerve, or imagination on the part of the people who lived there. Turn the arrow around and the evidence fits far better. The shock that emptied these places, a global trade realignment or a wave of factory automation, was decided far away and was no town’s to refuse. The one adaptation the textbooks prescribe, moving to where the work is, is precisely what the people who could adapt did, and their leaving is what drained the tax base and the services and sealed the fate of those who stayed. Where people did not leave, the measured record of the China trade shock shows that adaptation mostly failed for a decade and more. The failure was not one of effort; a laid-off machinist of fifty in a town with no other employer has almost nowhere to move within it. So the successful adaptation, the leaving, deepened the collapse, and the collective adaptation, reinvention, was blocked by structure rather than sloth. That is close to the reverse of a town that simply would not adjust. What keeps the ruling short of a clean reversal, and holds the confidence to moderate, is that adaptation is not always impossible: a few regions with the right foundations, Pittsburgh above all, did rebuild an economy, which means the towns that could not were not every one of them doomed from the start. The claim mistakes the direction of cause; it treats a wound delivered from outside, and made permanent by rational individual flight, as a local failure of will.
Sources
- The concentrated, lasting incidence of the trade shock: D. H. Autor, D. Dorn & G. H. Hanson, “The China Syndrome: Local Labor Market Effects of Import Competition in the United States,” American Economic Review 103(6) (2013), pp. 2121–2168 — import competition explains about a quarter of the 1990–2007 decline in US manufacturing employment; and D. Acemoglu, D. Autor, D. Dorn, G. H. Hanson & B. Price, “Import Competition and the Great US Employment Sag of the 2000s,” Journal of Labor Economics 34(S1) (2016), pp. S141–S198 — roughly one million manufacturing jobs, and close to two million economy-wide, lost to 2011.
- Slow adjustment and limited out-migration: D. H. Autor, D. Dorn & G. H. Hanson, “The China Shock: Learning from Labor-Market Adjustment to Large Changes in Trade,” Annual Review of Economics 8 (2016), pp. 205–240 — adjustment in exposed local labor markets is “remarkably slow,” with wages and labor-force participation depressed for at least a decade, and the net job loss not offset by out-migration.
- The competing automation account, and the reason it is contested: D. Acemoglu & P. Restrepo, “Robots and Jobs: Evidence from US Labor Markets,” Journal of Political Economy 128(6) (2020), pp. 2188–2244 — each robot per thousand workers lowered local employment and wages; against S. N. Houseman, “Understanding the Decline of US Manufacturing Employment” (Upjohn Institute, 2018) — most measured manufacturing productivity growth is concentrated in computers and reflects statistical quality-adjustment, so the automation-did-it story is weaker than the aggregate output suggests.
- Falling and staying fallen: U.S. Census Bureau decennial census (population) and American Community Survey, 2024 one-year estimates (income and poverty) — the city of Detroit peaked at 1,849,568 in 1950 and held about 639,000 in 2020, roughly a third of its peak; its median household income is a little under half the national figure and about a third of its residents live below the poverty line, against roughly one in eight nationally.
- Reinvention is possible but rare and asset-dependent: Pittsburgh lost well over 100,000 manufacturing jobs and more than 150,000 residents when steel collapsed in the 1980s, then rebuilt an economy on its research universities and medical centers even as its city population kept falling — a recovery of the regional economy, not a refilling of the town.
- Confidence is moderate under the rubric: the pip scores the weaker of evidence directness and construct match, and construct match binds here. The evidence is direct — the incidence and the decade-long persistence are measured, and the population collapse is observed — but “failed to adapt” runs together two different things: the individual adaptation that did happen, moving away, which hollowed the town, and the collective revival that a few places managed and most could not. Because the reverse is reliably true for the typical single-industry town but not universal, the claim is a reversal of cause rather than a clean inversion of fact.
Where the argument goes next
The scale ladder has one rung left. A region loses when its people leave; a city can fail in the opposite direction, by succeeding so well that the people who keep it running can no longer afford to live in it. The same divergence that pulls nations and regions apart has a version at the scale of a single expensive city, and it turns on who owns the ground, not on who leaves. The last chapter goes to the richest cities in the world and asks why, in the very places with the most work and the highest pay, a paycheck buys less and less of a home.