Urbanization across eras
For most of history the city was a magnificent dead end — capped by how much food its hinterland could spare. Then something underground broke the ceiling, and the city became the most productive thing humans had ever built. This walkthrough follows one pattern across four eras: economic life keeps concentrating into cities, and the constraint that holds the city back keeps changing — from food, to disease, to the price of a place to live. The agglomeration force is constant. The ceiling moves. That is the whole story, and it is the journey, not the recap.
The city under the food ceiling
This page traces the city as a recurring pattern — how the urban form kept reinventing itself as production technology changed. It is not the housing-affordability debate (for why housing is expensive now, see Why is housing so expensive?), not the question of why the industrial takeoff happened where it did (for that, see the Great Divergence walkthroughs), and not the story of who migrated and what happened to their wages. It is the story of the cities themselves.
“The city of Constantinople is great and populous… In greatness and beauty no city in the world is to be compared with it… The amount of business which is daily transacted is beyond belief.”
— Pero Tafur, Travels and Adventures, on Constantinople, c.1437
Every age has its wonder-of-the-world capital — Constantinople, Beijing, Delhi, Istanbul, the city the traveler cannot believe is real. And yet here is the fact that should stop you: even these capitals were home to a small fraction of their societies. Nearly everyone, in every pre-modern civilization, lived on the land and worked it. The greatest cities the world had ever seen sat under an invisible ceiling, and that ceiling was made of food.
The ceiling has a simple cause. A pre-industrial economy runs on current solar income — food, fodder, firewood, the wind and the water. A city is, by definition, a population that does not grow its own food. So the size of any city is capped by two things: the agricultural surplus the surrounding countryside can spare, and how much of that surplus can reach the city before it spoils. Better farming or better transport — a canal, a coastal sea-lane — raises the ceiling. Nothing in the organic economy removes it. This is why pre-modern urbanization was not a story of steady climb but of a plateau: the city as a noun, not a verb.
Economists frame this as the very first stage of structural transformation — the slow shift of labor out of agriculture and into cities that defines a country’s development path. The pre-modern world is the world stuck at the starting line: agriculture is so labor-hungry that there is almost no one left over to put in a town. The development apparatus that names this transition — the Lewis model of surplus rural labor, the Kuznets fact that urbanization rises with income — is the right lens, but it is a lens built to watch the ceiling lift. In Stage 1 it has not lifted yet.
What did an “urbanization rate” even measure in such a world? Just the urban population as a share of the whole — and across the great pre-modern cores it barely moved for centuries. The cleanest way to see the ceiling is to put four of those cores side by side.
Take the pre-modern understanding of the city at full strength, because it was correct for its world. In that understanding the city is a consumption center, not a production engine. Wealth flows to the city — as tax, as rent, as tribute, as the proceeds of long-distance trade — and the city spends it. The court, the cathedral, the garrison, the great market: these are where a society’s surplus is gathered and consumed, not where it is made. Manufacturing, such as it was, was dispersed across the countryside in small workshops and households. The city was magnificent precisely because it was the place the surrounding economy poured its surplus into.
This is not a naive view to be corrected; it is an accurate description of an organic economy. Constantinople, Beijing, Delhi, the Italian and Flemish trade cities — their grandeur was real and their dependence on the hinterland was real, and the two facts are the same fact. A great city was the visible interest paid on a vast agricultural countryside. When the harvest failed, the city starved first and worst, because it had no fields of its own. The consumption-city framing held because, for the organic economy, it was true.
The pre-modern city was real, large by the standards of its world, and capped. The pull of agglomeration was already there — people clustered for exchange, administration, safety, worship, and the simple density of opportunity — but the food constraint held the ceiling down, and the consumption-city was the most a society running on current solar income could build. What is about to change is not the agglomeration pull. That force is a constant in this story, present in every era. What changes is the ceiling. And the ceiling is about to be broken not by anything happening in the city, but by something dug out of the ground beneath it.
When a city no longer has to eat from its own hinterland — when grain arrives by rail from a thousand miles away and the energy to run the place comes from coal seams instead of acreage — the consumption-city becomes something the world has never seen. A city that produces. The first society to build one would also be the first to discover what it cost.
The mineral city
“The view from this bridge… is characteristic for the whole district. At the bottom flows the Irk, a narrow, coal-black, foul-smelling stream… Above the bridge are tanneries, bonemills, and gasworks, from which all drains and refuse find their way into the Irk… In dry weather, a long string of the most disgusting, blackish-green, slime pools are left standing on this bank.”
— Friedrich Engels, The Condition of the Working Class in England, on Manchester, 1845
Engels is describing the most productive place on earth. Manchester multiplied roughly twelvefold across the Industrial Revolution; Britain crossed the threshold to become the first society in human history where most people lived in towns. And it was a graveyard: life expectancy in Manchester ran around 26 years against 38 in the rural counties. The mineral city arrives leading with its cost, and any honest verdict has to earn the agglomeration benefit against that ledger.
Two decouplings broke the food ceiling. Coal is stored solar income, not current — millions of years of it, dug out of the ground, releasing the city from the acreage that had always bound it for fuel. And rail and the steamship decoupled the city’s food from its local hinterland: Chicago could eat the Great Plains, London could eat the world. With both constraints lifted at once, the city stopped being a place where surplus was merely spent and became a place where it was made. The factory concentrated labor, capital, and power at a single site; the wage pulled people off the land faster than any city had ever grown. For the first time the urbanization rate did not plateau — it climbed, and kept climbing.
On the production side this is the economy stepping off the Malthusian treadmill — the negative feedback loop where every gain in population or living standards pushed harder against the land constraint and was eventually clawed back. The industrial city is what that escape looks like in human geography. The detailed record of the explosion — Manchester’s growth, the crossing into a majority-urban society, the brutal life-expectancy gap — is the spine of History Ch 7 (The Industrial Revolution), and the spread of the mineral city beyond Britain runs through Ch 8 (Industrialization beyond Britain). One caution this walkthrough will keep: why the takeoff happened in Britain first is a separate question, fought out in the Great Divergence debate — here we take the takeoff as the fact that broke the urban ceiling, not as the thing to be explained.
The anti-urban tradition was right about the costs, and it deserves to be argued in its own voice before any verdict answers it. Engels and the sanitary reformers — and Dickens, who fictionalized Coketown into a place where the chimneys smoked “serpents of smoke” forever — were not exaggerating. The early industrial city was lethal. Urban mortality outran rural; cities only grew at all because in-migration outpaced the deaths, the grim arithmetic historians call the “urban graveyard effect.” Cholera moved through neighborhoods with no sewers; children worked fifteen-hour days; the rents on a cellar dwelling with contaminated water bought less life than a rural cottage with a garden had. The first generations of the mineral city paid a real and disproportionate price for a productivity dividend most of them never lived to collect.
Behind the sanitary critique sat a deeper one: the Jeffersonian agrarian ideal, and it is a serious political economy, not nostalgia. Its claim is that landholding independence is the material basis of civic virtue and political freedom — a republic of farmers who own their own ground cannot be easily coerced, because each household commands its own subsistence. The city, on this view, is a place of dependence: the wage-worker owns nothing, can be turned out at will, and is therefore a poor foundation for a free polity. Faction, crowding, and corruption follow. This is an argument about the relationship between property, autonomy, and the citizen, and it has to be met on those terms, not waved away as fear of the new.
The agrarian ideal was right about the cost and wrong about the verdict
The intuition that cities are unnatural, corrupting, and unsustainable never died — it runs from Jefferson through the back-to-the-land movements to today’s “cities are over” discourse. It picks up real moral force from the genuine costs of the early industrial city. It loses the argument because the agrarian republic could not match the production-city’s output, and the costs it named were eventually engineered away by sewers and clean water — not by abandoning the city.
Both things are true: the costs were real, and the agglomeration benefits eventually dominated. The cities that were death-traps in 1845 became, within two generations, the most productive places on the planet — and the great Victorian sanitation projects, clean water, and public-health investment eventually flipped the urban-mortality penalty into an urban-mortality advantage. The Jeffersonian ideal lost the argument, but not because it was wrong about the price of dependence. It lost because the production-city’s output was decisive: an agrarian republic simply could not match what the industrial metropolis could make. The anti-urban tradition named the costs honestly. What it underweighted was the agglomeration dividend — the force that the next stage finally gives a name and a model.
The city was now a production engine. But that raised a question no one had quite answered: why? Why did the factories cluster in the same towns, the same districts, the same handful of streets, instead of spreading out evenly across the cheap and empty countryside? Land was dearer in the city, not cheaper. Something kept pulling activity together against the cost of the ground it stood on. A Cambridge economist was about to give that something a name.
Why activity clusters
“When an industry has thus chosen a locality for itself, it is likely to stay there long… The mysteries of the trade become no mysteries; but are as it were in the air, and children learn many of them unconsciously.”
— Alfred Marshall, Principles of Economics, 1890
“In the air.” In one phrase Marshall named the force that had pulled the factories together against the price of urban land. Knowledge that is not written down anywhere — the feel of a trade, the trick that works, the supplier who can be trusted — sits in the atmosphere of a place where the trade is concentrated, and you absorb it by being there. This is the agglomeration force, and the rest of the stage is the discipline catching up to it.
Marshall’s answer in 1890 was three external economies — benefits that come from the cluster as a whole, not from any one firm. The first is the knowledge spillover he made famous: in a dense industrial district the secrets of the trade are “in the air,” learned by proximity. The second is labor-market pooling: a thick local market for specialized skills means a firm can find the worker it needs and a worker can find the next job without leaving town, so both sides take the risk of specializing. The third is input-sharing: specialized suppliers and services that no single firm could keep busy become viable when many firms share them. All three are external to the individual firm but internal to the city — which is exactly why activity piles up where land is dear instead of spreading to where it is cheap.
Jane Jacobs, in The Economy of Cities (1969), pushed the claim further and inverted the old order of causation. The orthodox view held that the countryside produces and the city is where the surplus is spent — agriculture first, cities afterward. Jacobs argued the reverse: cities are the source of new economic work, and the new work then transforms the countryside, not the other way around. Her emphasis was on cross-industry diversity — novelty comes from the recombination of unlike trades crowded together, not just from the deepening of one. That set up a debate the field still runs — whether the gains come from same-industry specialization (the “localization” economies, later tagged Marshall–Arrow–Romer) or from Jacobs’s cross-industry diversity (“urbanization” economies).
Then Paul Krugman, in 1991, gave the cluster a formal model — the new economic geography — and showed that agglomeration need not be assumed; it can emerge. Take increasing returns at the firm level, add transport costs and mobile workers, and the model produces a core and a periphery on its own: a small initial advantage tips into self-reinforcing concentration, because firms want to locate near the big market and the big market is wherever firms have already located.
The mechanism rests on monopolistic competition with scale economies (the Dixit–Stiglitz form). A firm enjoys internal increasing returns — lower average cost at higher output — so it wants one large plant, not many small ones, and it wants that plant near the largest market. But the market is large wherever workers are, and workers go where the firms (and the wages and the variety of goods) are. Writing real wages as a function of access to varieties net of transport cost, the symmetric “spread” equilibrium becomes unstable below a critical transport cost: any small concentration raises local real wages, drawing in more workers and firms in a cumulative loop — the home-market effect. Above the critical cost activity disperses; below it, it agglomerates into a core. The location of the core is not pinned down by fundamentals — history and accident choose it, and then increasing returns lock it in.
Firms want to be where the customers are; customers (who are also workers) want to be where the jobs and the goods are. Each pulls the other, so once a place gets a little ahead it gets further ahead on its own — a snowball. Which place wins is mostly an accident of history; what is not an accident is that some place wins big and the rest stay thin. That is why economic activity clumps instead of spreading evenly, even when the empty land is cheaper.
The intellectual lineage runs Marshall → Jacobs → Krugman, and it sits across two clusters of the history of economic thought: Marshall in the marginalist formalizers (Ch 5 §5.3, Marshall), and Krugman’s new economic geography together with Jacobs’s heterodox-but-absorbed voice in the modern-pluralism frontier (Ch 17). The industrial-district evidence Marshall was generalizing from — the Lancashire cotton towns, the Sheffield cutlery trades — is in the Industrial-Revolution record itself.
Argue the view Jacobs overturned at full strength, because it was the orthodoxy and it was not foolish. The standard development sequence put agriculture first and the city second: a society raises its farm productivity, the surplus frees hands and mouths, and only then can a town grow on the back of that surplus. On this telling the city is downstream — a consequence of the countryside’s success, the place the agricultural surplus is gathered and spent. It is exactly the consumption-city picture from Stage 1, promoted into a theory of growth: the land produces, the city follows. For the organic economy this was simply true, and it remained the natural assumption long after the mineral city had quietly falsified it.
Jacobs’s inversion is the surprise. She argued — controversially, on archaeological grounds — that cities can come first: that the dense crowding of unlike trades is where genuinely new work is invented, and that the new work then radiates outward to transform agriculture, rather than waiting on it. Whether or not her “cities first” archaeology holds, the deeper point survived and reshaped the field: the city is not merely where surplus is spent but where novelty is generated. The agglomeration that Marshall named as a force, and Krugman modeled as an equilibrium, is on Jacobs’s reading the engine of growth itself — which is why the orthodox “land produces, city follows” sequence no longer describes the modern economy.
Once the food ceiling is lifted, agglomeration economics is the settled mainstream answer to why cities exist: density generates external economies, and Marshall’s three channels — knowledge spillovers, labor-market pooling, input-sharing — are the mechanism, formalized by Krugman and measured by the modern urban-economics literature. That much is not in dispute at the frame level. What is genuinely open lies one layer down: the relative weight of the three channels, and the Jacobs-versus-Marshall question of whether the gains come mostly from same-industry specialization or cross-industry diversity. Those are real parameter-and-method disputes inside a locked frame — not reasons to doubt that density is productive, but the live edges of how and how much. The force the Industrial Revolution unleashed now has a name and a model. What it does not yet have, in this story, is a new ceiling.
If density is this productive, the modern city should be an unambiguous triumph. It is — and it has a new problem. The food ceiling is gone for good, but another ceiling has quietly taken its place, and this one is not made of acreage or harvests. It is made of zoning maps, scarce land, and the price of a place to live.
The modern urban economy
“Cities, the dense agglomerations that dot the globe, have been engines of innovation since Plato and Socrates bickered in an Athenian marketplace… The city has triumphed.”
— Edward Glaeser, Triumph of the City, 2011
Glaeser’s claim is that the city is humanity’s greatest invention — the place where proximity magnifies what people can do together. The modern urban-economics literature backs the boast: the densest metropolitan areas really are more productive, and a meaningful share of that is causal, not just a matter of talented people sorting into them. And yet the most productive cities have become the most ruinously expensive places to live in. Both facts are true at once, and the tension between them is the modern city’s defining problem.
The first move is the density premium. Output per worker in the densest metros runs well above the national average, and the agglomeration-elasticity literature finds that a non-trivial part of the gap is genuinely causal — in the consensus range, productivity rises on the order of a few percent for each doubling of density, after stripping out the selection of skilled workers into big cities. The mechanism is exactly Stage 3’s: spillovers, pooling, sharing, now measured rather than asserted. The density premium is real and it is large.
The second move is the one this thread has been building toward: the binding constraint has shifted from food to housing. The pre-modern city was capped by the surplus its hinterland could spare. The modern city is capped by how much housing its zoning, its land, and its politics will let it build. When the most productive cities cannot add homes fast enough, the density premium does not flow to workers as higher real wages — it is captured by landowners as rent. The standard supply-and-cap diagram tells the story: a regulatory cap to the left of where supply would otherwise stretch turns the gap into a wedge of pure rent, and the city’s success shows up as a housing-cost crisis. This is the “superstar cities” dynamic — the most productive metros pull away in wages and in housing cost together, and the second cancels much of the first for anyone who has to move there to earn it.
There is a long-tail of lineage here too: the idea that the value created by a city’s density flows to whoever owns the land under it is the late-Ricardian theory of locational rent — Henry George’s argument that the unearned increment belongs to the community that created it. That land-as-institution framing has its intellectual home in the institutional tradition (History of Economic Thought Ch 15). The historical rise of the superstar and “global” cities is the spine of History Ch 18 §18.6 (within-country inequality across the rich world), and the post-2008, post-COVID urban moment runs through Ch 19 §19.6 (the unresolved frontier).
Two questions about the modern city are genuinely open, and the honest move is to name the live positions without pretending to adjudicate them. The first is remote work. If the knowledge spillovers that make density productive can route through a screen, does the density premium survive the pandemic’s great experiment in working from home? There is a sharp “cities are over” reading — the “donut effect,” offices and high-end demand hollowing out of city centers — and an equally serious counter-reading that agglomeration is resilient, that hybrid work concentrates the in-person premium into fewer, denser days rather than erasing it. The data is recent and the verdict is not in.
“Remote work killed the city” mistakes a re-pricing for a death
The pandemic-era claim that knowledge work can route through a screen, so the expensive city is obsolete, has real evidence behind it — the donut effect is measurable. But the density premium it would have to erase is large and old, and the early data looks more like a shift in where and how often people cluster than like the end of clustering. The honest read: the premium is being re-priced, not abolished — though this is a live question, not a settled one.
The second open question is the developing-world megacity. Lagos, Dhaka, Kinshasa are urbanizing at a scale and speed the rich world never saw — but is agglomeration delivering the productivity premium, or does the housing-and-infrastructure constraint bind before the dividend arrives? There is a real worry, sometimes called “urbanization without growth” or the rise of “consumption cities,” that some megacities are filling up without the industrial or tradable-services base that made the historical city productive — agglomerating the congestion without the premium. The structural-transformation apparatus that frames this is the same development economics Stage 1 opened with, now asking whether the modern megacity is an engine or a trap. The superstar-inequality dynamic, where the most productive cities pull away and the gains concentrate in the people and landowners already inside them, runs underneath both questions.
Two things are settled and one frontier is open. The density premium is real and large: cities are productivity engines, exactly as the agglomeration apparatus says. And the binding constraint has shifted from food to housing — the modern city’s ceiling is built of zoning and land, not harvests. What to do about that housing constraint is a genuine controversy with live policy stakes, and this thread does not run it — for the dispute over supply, homevoter politics, and decommodification, the full treatment is Why is housing so expensive?. Here the claim is only the historical one: this is the modern era’s binding constraint, the way food was the pre-modern era’s.
On the frontier the honest answer is to name what is open and lean where the evidence leans — which is a position, not a shrug. The density premium probably survives remote work in modified, hybrid form — the spillovers concentrate into fewer in-person days rather than vanishing — and the developing-world megacity can be an engine, but only where the housing-and-infrastructure constraint is solved before congestion eats the dividend. Both are live and unresolved, and the responsible thing is to say exactly that: here are the questions, here are their coordinates, and here is where the evidence currently leans — not a verdict dressed up as one. The thread’s job was to lay the apparatus and the shifting constraint end to end; the frontier is where it points, not where it pretends to close.
The shifting ceiling
- The consumption-city under the food ceiling. The pre-modern great city was magnificent and capped — its size held down by the agricultural surplus its hinterland could spare. The city consumed; the countryside produced.
- The mineral production-city. Coal and rail broke the food ceiling, decoupling the city from its local energy and its local food. The city became a production engine for the first time — at a brutal cost the first generations paid disproportionately, before sanitation flipped the urban-mortality penalty.
- The agglomeration force, named. Marshall named the spillovers, Jacobs inverted the city-and-countryside order, and Krugman modeled the cluster as something that emerges on its own — the discipline catching up to the force industrialization had unleashed.
- The modern density premium under the housing ceiling. The premium is real and large; the binding constraint has shifted from food to housing supply. The same agglomeration force, now bottlenecked by the built environment instead of the food supply.
Lay the four eras end to end and one structure does all the work. The pull of agglomeration — the productivity of density — is a constant, present and operating in every age. What changes is the ceiling that binds it. Food capped the organic-economy city; the costs and diseases of the mineral city were the next era’s ceiling, engineered down by public health; housing supply caps the modern superstar city. An era-organized history that walks one period at a time cannot show this, because the pattern is not in any one period — it is in the sequence of constraints, and that sequence is the thread.
Which is also why the frontier is honestly open and worth pointing at rather than papering over. The current ceiling is housing, and whether the next shift comes from technology loosening the agglomeration tie (remote work routing spillovers through a screen) or from scale pushing the megacity past where its infrastructure can follow, the question has the same shape it has always had: not whether economic life will keep concentrating into cities — it will — but what new ceiling the next kind of city will run into. The city has reinvented itself in every era. The only safe prediction is that it will do so again.