Why Cities Stop Growing

Detroit peaked in 1950. At that moment it was the fourth-largest city in the United States, home to roughly 1.85 million people, producing more automobiles per capita than anywhere else on earth, and generating a standard of living for blue-collar workers that would have seemed fantastical to their parents. By 2013 it had filed for the largest municipal bankruptcy in American history. The population had fallen to around 680,000. The tax base had collapsed. Thousands of acres within the city limits were functionally abandoned.

The standard explanation attributes this to the decline of the American automobile industry, and the standard explanation is not wrong exactly, but it is radically incomplete. Other cities built around single industries have declined and then rebuilt themselves around something new. Pittsburgh lost its steel industry and eventually became a mid-sized city with a diversified economy anchored in healthcare, education, and technology. Cleveland lost manufacturing and did something similar, though with more difficulty and less success. Detroit lost its main industry and did not recover in the same way. The question worth asking is not what happened to Detroit’s cars, but why Detroit could not do what Pittsburgh did.

The answer, studied carefully by economists like Ed Glaeser and historians like Thomas Sugrue, is that Detroit’s institutions were specifically structured around the automobile industry in ways that made reorientation unusually difficult. The tax structure depended on manufacturing employment. The physical layout of the city was optimized for factory workers who drove, not for the dense walkable development that tends to attract the knowledge workers who replace manufacturing workers in successful urban transitions. The political coalitions that governed the city had interests tied to the old economy. The educational institutions were calibrated to produce workers for a type of work that was disappearing. None of this was anyone’s fault in a simple sense. All of it combined to make adaptation much harder than it needed to be.

The deeper pattern here is one of institutional lag. Cities are collections of physical infrastructure, social relationships, political coalitions, and legal arrangements that take decades to build and resist rapid change. When the economic environment shifts, the city’s institutions keep doing what they were doing. Sometimes the environment shifts back, or the new environment is close enough to the old one that the mismatch doesn’t matter. When neither of those things happens, you get urban stagnation.

This is not a uniquely American story. Venice reached its commercial peak in the late fifteenth century, when it controlled a significant fraction of Mediterranean trade and was by some measures the wealthiest city on earth relative to its size. After Vasco da Gama’s voyage around Africa opened a direct route from Portugal to India in 1498, Venice’s position as the indispensable intermediary in the spice trade began to erode. The city’s institutions, which had been optimized over centuries for a specific kind of long-distance trade finance, could not reorganize fast enough to compete in a world where the trade routes had literally moved. Venice did not die — it became a tourist economy, which is a kind of survival — but it never recovered its commercial position. The physical city remains. The economic city of 1490 does not.

What makes this episode instructive is that Venice had enormous advantages: a sophisticated legal system for commercial contracts, deep capital markets, experienced merchants, and a population highly skilled in the relevant trades. These advantages were real and substantial. They were also specifically adapted to a world that stopped existing. The skills and institutions that made Venice dominant in 1490 were, in the changed world of 1510, partially a liability, because they directed attention and capital toward things that were no longer valuable rather than toward the new opportunities.

The concept that economists use for this dynamic is “path dependence,” and it is one of the genuinely useful ideas in economic history even though it gets used as a catch-all in ways that sometimes obscure more than they illuminate. The basic claim is that the choices a city or firm or country made in the past constrain the choices available to it in the present, not because of any physical necessity but because of sunk costs, coordination requirements, and the way expectations shape investments.

Consider the development of Silicon Valley. By the 1970s, the concentration of semiconductor firms in the Santa Clara Valley had created a labor market, a legal environment, a social culture, and a capital infrastructure that made it easier to start a technology company there than anywhere else in the world. This was not inevitable — the early semiconductor industry could have concentrated in Boston, or around the government defense contractors in the Washington corridor, or in Texas. It happened to concentrate in California partly for reasons that had more to do with Stanford’s particular culture in the 1950s and the personalities of a handful of engineers than with any systematic advantage.

Once that concentration happened, it became self-reinforcing. Each new company that located in Silicon Valley made it marginally easier for the next company to locate there, because each new company brought workers, suppliers, investors, and knowledge. The path became the path not because it was optimal in some abstract sense but because it was established. Other regions that wanted to replicate Silicon Valley’s success found that the most important element of Silicon Valley — the thick web of relationships, knowledge, and norms that had built up over decades — was exactly the element that could not be copied.

This helps explain why urban planning is so often disappointing. Planners can change the physical environment. They can build infrastructure, zone land, offer tax incentives, construct buildings. What they cannot do is directly create the social and institutional density that makes cities economically productive. They can create conditions in which that density might develop. The development itself requires time and the emergent behavior of many independent actors, which means it cannot be planned into existence on a schedule.

The cities that have successfully reinvented themselves share certain characteristics that are less obvious than the usual list of amenities and infrastructure. They tend to have universities that were already engaged with the local economy before the decline arrived — not as causes of recovery but as anchors that survived the contraction and could serve as nuclei for regrowth. Pittsburgh has Carnegie Mellon and the University of Pittsburgh. Cleveland has Case Western. Detroit, notably, has Wayne State, but it is located in the city itself rather than in the suburban ring, which matters for the density effect. Universities that are physically embedded in cities contribute to urban recovery in a way that universities that are spatially isolated from the urban core often don’t, even when they’re intellectually excellent.

The cities that have failed to reinvent themselves tend to lack those anchors, or to have anchors that were culturally disconnected from the city around them. They also tend to have particularly rigid political institutions — strong mayors with patronage networks, city councils representing constituencies with concentrated interests in the old economy, public employee unions whose members’ jobs are tied to the tax base generated by the disappearing industries. None of these features is unique to declining cities, but their combination in the absence of the countervailing institutions that anchor recovery tends to produce stagnation.

There is an uncomfortable implication here for contemporary urban policy, which is that the features of cities most associated with decline are also the features most resistant to policy intervention. You cannot mandate university-industry relationships. You cannot dissolve entrenched political coalitions by passing legislation. You cannot create overnight the risk tolerance and network density that make entrepreneurial ecosystems work. The things that can be done — physical planning, tax incentives, infrastructure investment — are the things that matter least. The things that matter most are the hardest to change.

This is not a counsel of despair. Cities do recover. Pittsburgh is a real example, not a fairy tale. Austin grew from a minor Texas capital to a major technology hub within two decades, largely because of the University of Texas and a particular policy decision (the decision by Michael Dell to locate his company there in 1984, which was itself partly accidental). Salt Lake City built an economy around technology and logistics that few would have predicted forty years ago. The recoveries happen. They just don’t happen on the schedule that local politicians promise or that federal funding programs assume.

The implication for thinking about urban stagnation is that the relevant time horizon is very long. The institutions that make a city’s economy work took decades to build. The institutions that will enable recovery, where recovery is possible, will also take decades to build. The honest thing to say to a city in the early stages of economic decline is not that the right policy package will turn things around in five years, but that the right investments made now will bear fruit in fifteen or twenty years, if sustained, and if lucky. This is not a message that wins elections. It is, however, the message that the historical record supports.

The cities that stopped growing are not warnings about the limits of urban life. They are warnings about the brittleness of institutions that optimize too completely for a single economic role. The most resilient cities in economic history — London, New York, Amsterdam — are the ones that managed to be useful in multiple ways simultaneously, so that when one source of economic advantage disappeared, others remained. The lesson is not to avoid specialization entirely. It is to never let specialization calcify into the only thing a city knows how to do.

Detroit in 1950 was a city that was very good at exactly one thing, and it was extremely good at it. The problem was not the excellence. The problem was the singularity. When the one thing changed, there was nothing left to fall back on — not because the people were less capable, but because the institutions, the physical layout, the political coalitions, and the mental models of what the city was for had all converged on a single point. The contraction that followed was not punishment for success. It was the consequence of building success in a way that left no redundancy.

Understanding that distinction is the beginning of sensible urban policy, even if sensible urban policy remains, as it always has been, very difficult to actually practice.