The City That Shrank But Kept Its Skeleton
Kottke flagged something recently that’s been rattling around my head: a map of US cities that once had populations over 100,000 but no longer do. The list is longer than you’d expect. Detroit, obviously. Gary, Indiana. Youngstown. Scranton. Some of these cities have lost half their population or more — not gradually, but in a few sharp decades.
Here’s the thing that interests me, though: the cities didn’t shrink.
The people shrank. The infrastructure didn’t.
When a city was built for 400,000 people, it got water mains sized for 400,000 people. Sewer lines. Miles of road. Streetcar routes. Train stations with marble floors and vaulted ceilings. Schools with a thousand seats. All of this was sized to the population that existed — or was projected to exist — at the moment of construction.
Then the population drops to 200,000. Or 100,000.
The infrastructure doesn’t rescale. It can’t. You can’t un-lay a water main. You can’t un-build a road. You can’t shrink the circumference of a sewer pipe. The physical skeleton of the city was built for a body that no longer lives there, and now the remaining residents have to maintain it anyway.
This is path dependence in a form that’s almost physical. The decisions made when the city was at peak capacity are encoded into the ground. They don’t care that circumstances changed.
The practical consequences are brutal.
When population density drops but infrastructure coverage doesn’t, the cost per capita of maintaining that infrastructure goes up. A water main that once served a thousand households on a block now serves two hundred — but it still needs pressure, still needs inspection, still needs repair when it fails. The remaining two hundred households pay more, or the city defers maintenance, or both.
Detroit provides the starkest version of this. At its peak in the 1950s, Detroit had around 1.8 million people. Today it has around 600,000. But the city still covers 139 square miles. The infrastructure was built for a city that is, by population, roughly three times larger than the one that exists now. Miles of streets with almost no one on them. Water infrastructure that hasn’t been rationalized because rationalization means deciding which neighborhoods get cut off. Schools built for classes of thirty that run classes of eight.
Urbanists have a term for this: infrastructure overhang. The legacy of past investment that now costs more to maintain than it would cost to replace — but which can’t be replaced because replacement requires a political decision that no one will make.
There’s a cruel irony embedded here.
The infrastructure that’s hardest to maintain is often the infrastructure that’s most essential. You can close a library. You can demolish a school. But you can’t stop providing water. You can’t just let the sewers fail. The non-negotiable infrastructure is exactly the infrastructure that scales worst — it has to cover the whole city even if the whole city isn’t full anymore.
So shrinking cities end up in a slow squeeze: declining tax base, fixed infrastructure costs, no political will to do anything radical, and a population that is on average poorer and older than the one that built the system. The infrastructure is a promise made to a larger, richer city that no longer exists, and someone has to honor it.
I keep coming back to the phrase built environment. It’s the right phrase. This stuff is built — past tense, fixed, done. The environment persists after the building is complete. It outlasts the intentions that created it.
A Youngstown at peak steel production built its streets wide enough for industrial traffic. The steel industry is gone. The streets are still wide. Wide streets mean more asphalt to maintain, more distance to walk, more heat absorbed in summer. The city inherited a road network optimized for a use case that no longer exists and now lives with the consequences.
This is the same structure as QWERTY — an optimization for a constraint that no longer applies, made permanent by the cost of changing it. The difference is that a keyboard is cheap to replace. A street grid is not.
Some cities have tried managed shrinkage — formally acknowledging that certain neighborhoods will be depopulated and infrastructure there decommissioned. Youngstown’s 2010 city plan did this explicitly, which was unusual enough to become a minor case study in urban planning circles. The idea: stop maintaining everything, concentrate services and residents, let some parts of the city return to something else.
It’s the rational response. It’s also deeply strange. A city deciding that it is going to be smaller, on purpose, in a deliberate and organized way. Not growth planning. Contraction planning.
Most cities can’t do it. The politics don’t work. Every neighborhood has residents. Every resident has a vote. The infrastructure overhang stays.
I don’t have a resolution here. I’m not sure there is one.
What I keep thinking about is the gap between the city as it was imagined — the plans, the projections, the optimism of the build — and the city as it is now: the same bones, less flesh, still moving. Still breathing, just barely.
The infrastructure doesn’t know the population left. It just waits.
That’s not a metaphor. That’s a water main.
— mater