Why Cities Become More Productive: The Economic Drivers

Cities: Geoffrey West spent most of his career as a theoretical physicist, studying the behavior of subatomic particles at the energy frontier. Then, in the mid-1990s, a collaborator asked him a question about biology: why do metabolic rates across species follow such precise mathematical relationships with body mass?

A mouse burns energy at a rate per kilogram roughly a hundred times higher than an elephant, yet both obey the same underlying power law. West and his colleagues worked out the mathematical theory of these biological scaling laws and published their findings in 1997.

The discovery prompted a natural extension. If biological organisms obey universal scaling laws rooted in the geometry of their nutrient distribution networks, might cities do the same?

The answer, which West, Luis Bettencourt, and colleagues at the Santa Fe Institute published in the Proceedings of the National Academy of Sciences in 2007, was more striking than almost anyone anticipated.^[1]

Cities do obey precise mathematical scaling laws, and these laws hold with remarkable consistency across countries, cultures, and time periods.

When the researchers plotted urban metrics against population size on logarithmic axes, the data organized itself into straight lines.

Infrastructure such as road lengths and electrical cable scales sublinearly at roughly the 0.85 power: a city ten times larger needs only about seven times as much road. But socioeconomic outputs - wages, patents, GDP, number of new companies - scale superlinearly at roughly the 1.15 power.

A city ten times larger produces approximately fourteen times the total economic output, or 40 percent more output per capita.

The practical implication is direct: every time a city doubles in population, per-capita socioeconomic output increases by approximately 15 percent. This is not a selection effect, not an artifact of which skilled people happen to choose cities, and not specific to any particular country or historical moment.

West's team found the same exponents in medieval European cities reconstructed from historical records, in Brazilian and Chinese cities as well as American ones, in cities spanning the full range of modern economic development.

Something in the nature of dense human aggregation itself - something mathematical and structural - continuously amplifies human productive capacity in ways that no other arrangement of people has ever matched.

"Cities are the great innovation machines of our species. But they are also the great engines of inequality, crime, and disease. The scaling laws don't discriminate. They amplify everything." - Geoffrey West, Scale (2017)^[9]

Key Definitions

Urban scaling laws - Mathematical power-law relationships between city population size and urban metrics, in which the exponent of the relationship determines whether the metric scales sublinearly (economies of scale) or superlinearly (accelerating returns). First formally documented by Bettencourt, West, and colleagues in 2007.

Superlinear scaling - A scaling relationship in which a metric grows faster than proportionately with population. At an exponent of 1.15, a city twice as large produces approximately 2.22 times the per-capita economic output, not merely twice as much. Over large population differences, this compounds dramatically.

Sublinear scaling - A scaling relationship in which a metric grows slower than proportionately with population. Infrastructure that scales at the 0.85 power embodies genuine economies of scale: the per-capita material cost of a city decreases as it grows larger.

Agglomeration economics - The study of the economic benefits that firms and workers gain from geographic clustering. Agglomeration economics identifies the mechanisms by which proximity generates productivity and explains why economic activity does not distribute evenly across space even when digital communication is widely available.

Knowledge spillovers - The transfer of knowledge from one economic actor to others, generating productive benefits for recipients who did not pay for the knowledge. Knowledge spillovers are highly localized geographically and are considered a primary mechanism of the urban productivity premium.^[4]

Tacit knowledge - Knowledge that cannot be fully articulated in explicit form, conveyed instead through demonstration, apprenticeship, informal conversation, and direct observation. Tacit knowledge is a major reason why proximity continues to matter for knowledge-intensive work even in an era of advanced communication technology.

Urban wage premium - The consistently documented finding that workers in larger cities earn substantially more than otherwise identical workers in smaller places, even after controlling for education, occupation, age, and individual skill levels.

Urban Scaling: What Doubles When a City Doubles in Size

Marshall's Three Mechanisms: The Foundation of Agglomeration Theory

The observation that economic activity clusters geographically is not new. In 1890, the British economist Alfred Marshall identified three mechanisms by which geographic concentration of similar industries creates value for all firms in a cluster.^[6]

His framework, developed in the "Principles of Economics," remains the foundation of agglomeration economics more than a century later.

The first mechanism is labor market pooling. When many firms in the same industry locate near each other, they collectively create a labor market large enough to support specialists.

A single manufacturing plant in an isolated location cannot support a professional class of component engineers; an automotive cluster in Stuttgart or a technology cluster in Silicon Valley can.

Workers benefit from multiple potential employers; firms benefit from access to an appropriately skilled workforce without bearing the cost of developing it exclusively.

The second mechanism is specialized input suppliers. A dense cluster of firms generates demand for specialized intermediate goods and services that would not be economically viable to provide to a dispersed market.

A fashion district supports specialized textile printers, button makers, and pattern drafters whose existence depends entirely on cluster density. Each firm in the cluster gains access to capabilities that none could justify maintaining internally.

The third mechanism, and the most important for understanding urban scaling, is knowledge spillovers. Knowledge generated within one firm or research group leaks to nearby firms through employee movement, informal conversation, observation, and the general circulation of ideas through a shared community.

This is what Marshall described as ideas being "in the air." The cluster collectively learns faster than any of its members could learn in isolation, because knowledge travels through the thick informal networks that dense proximity creates.

Measuring Knowledge Spillovers: The Patent Citation Evidence

For decades, the evidence for knowledge spillovers was suggestive but difficult to quantify. Adam Jaffe's 1989 study in the American Economic Review provided the first rigorous measurement.^[5] Patents are required to cite prior patents on which they build, creating a traceable record of how knowledge propagates across space and industry.

Jaffe found that patents cite other patents from the same metropolitan area at rates far above what the geographic distribution of technical expertise would predict.

Even after controlling for industry concentration, geographic proximity independently predicted citation probability. Knowledge travels, but it travels faster and farther within a few miles of its origin.

Stuart Rosenthal and William Strange extended this finding with greater spatial precision, demonstrating that the productivity benefits of proximity decay sharply over short distances - substantially within a single mile.^[7] The spillover premium is concentrated within urban cores and becomes negligible in suburban and rural contexts.

This sharp spatial decay has a critical implication: proximity cannot be replaced by communication infrastructure, because the spillovers are not primarily of the kind that can be transmitted through cables and screens.

Gaspar and Glaeser addressed the relationship between communication technology and cities directly in their 1996 analysis.^[8] Their finding was counterintuitive. Rather than being substitutes, communication technology and physical proximity are complements for knowledge workers.

The telephone did not empty cities; the internet did not empty cities. The reason appears to be that the most valuable knowledge - the kind that creates the largest productivity gains - is tacit rather than explicit.

Tacit knowledge requires observation, demonstration, and the informal conversation that happens when people share physical space over time. Video calls transmit information; they do not transmit the accumulated contextual understanding that comes from years of working in proximity.

Jane Jacobs and the Mechanism of Urban Life

Decades before the Santa Fe researchers formalized these ideas in mathematical terms, Jane Jacobs described the same mechanism through close observation of urban street life.

Her 1961 book "The Death and Life of Great American Cities" is one of the most important works of urban theory ever written - remarkable partly because Jacobs was a journalist and community activist, not an academic economist, and partly because the quantitative research of the following half-century has largely confirmed what she saw.

Jacobs argued that urban economic vitality depends on four physical conditions.^[3] First, mixed primary uses: residential, commercial, and civic activity interleaved in the same blocks, ensuring that streets are populated at different hours by different kinds of people with different purposes.

A street that is purely residential empties during working hours; a street that is purely commercial empties on weekends. Only mixed use generates the consistent pedestrian density that enables the casual encounters and informal interactions that constitute the productive social fabric of a city.

Second, short city blocks: small blocks create more corner intersections and more possible routes between any two points, forcing people through shared paths and maximizing the frequency of unexpected encounters.

The long superblocks favored by mid-century urban planners reduced interaction geometry, reducing the probability that any two people would cross paths.

Third, buildings of varying age: old buildings, with low rents and flexible layouts, provide the cheap space where new enterprises form. A neighborhood of gleaming new construction is a neighborhood of high rents accessible only to established businesses.

The bookshop, the experimental restaurant, the small design studio - the kinds of enterprises where innovation and novelty disproportionately arise - require inexpensive space. Old buildings provide this.

Cities that gentrify completely and replace their aging building stock lose the capacity for the productive churn that drives innovation.

Fourth, sufficient density: enough people per acre to sustain diverse businesses, fill streets with pedestrian life, and support the thick informal networks through which ideas circulate.

The Critique of Robert Moses

Jacobs' particular target was Robert Moses, the enormously powerful New York City planning commissioner who oversaw the demolition of dense working-class neighborhoods to build highways and superblock housing projects.

Moses was not ignorant or malicious - he was working from a coherent (if misguided) theory of urban improvement that prioritized automobile access, sanitary living conditions, and the replacement of "slums" with planned housing.

But Jacobs argued that he was systematically destroying the interaction fabric that made cities productive, replacing the fine-grained texture of mixed-use street life with monofunctional zones optimized for cars and separated from pedestrian life.

The subsequent decades largely vindicated Jacobs. The neighborhoods Moses demolished that survived in comparable form elsewhere showed continued economic vitality.

The housing projects he built have largely failed as communities, generating the concentrated poverty and social pathology that the urban renewal movement claimed to be solving.

And the econometric literature on agglomeration has quantified exactly the mechanisms that Jacobs identified qualitatively: mixed use, density, and the preservation of old buildings for cheap startup space are not aesthetic preferences but descriptions of the physical conditions under which human interaction generates economic and cultural value.

Why Innovation Requires Recombination

The urban scaling model treats the city not as a collection of buildings and infrastructure but as a network of human interactions.

In this framing, the productive unit of a city is not the firm or the individual but the encounter: the conversation at a conference, the chance meeting over coffee, the argument that reshapes a research direction.

Innovation is fundamentally a recombination process. New ideas emerge from connecting existing pieces of knowledge in novel configurations.

A materials engineer who encounters a molecular biologist in the lunch line of a dense urban research district is more likely to generate genuinely novel combinations than the same engineer working in an isolated campus surrounded only by other materials engineers.

Not because the individual is more talented, but because the combinatorial space available to them is larger.

The mathematics is straightforward. In a city of N people, the number of possible pairwise connections is N(N-1)/2, which scales as N squared. Density means that more of these potential connections are realized in practice.

Doubling the population more than doubles the rate of productive interaction, because each person has access to a larger network and can encounter a wider range of knowledge types.

The 15 percent per-capita productivity gain from population doubling is the empirical signature of this combinatorial logic.

Edward Glaeser's documentation of the urban wage premium reinforced this picture from a different angle. Workers in cities of one million earn approximately 30 percent more per hour than otherwise identical workers in cities of 100,000, even after controlling for education, occupation, and individual skill.

This residual premium implies that cities raise the productivity of workers beyond what their observable characteristics would predict - either by accelerating skill acquisition through tacit knowledge transfer, by enabling more productive matches between workers and firms, or by providing the knowledge spillovers that make the whole network smarter than the sum of its parts.

The Dark Side: Superlinear Scaling Amplifies Everything

Bettencourt's 2013 paper in Science was careful to present both sides of the scaling ledger.^[2] The same 1.15 superlinear exponent that governs wages and patents governs with equal mathematical regularity crime rates, disease transmission, traffic congestion, and income inequality.

A city twice as large does not simply have twice the crime - it has approximately 2.15 times the crime. The interaction density that spreads ideas and raises wages also spreads pathogens, social tensions, and every other contagion of human contact.

This is not a finding about bad urban policy or inadequate governance. It is mathematically universal. The urban premium comes bundled with urban pathologies because they arise from the same source: the density of human interaction. Managing this tradeoff is the fundamental challenge of urban governance.

Policies that increase density to capture agglomeration benefits must simultaneously address the superlinear scaling of negative outcomes, not as an afterthought but as an inherent feature of the same dynamic they are trying to exploit.

Housing costs represent the most visible manifestation of this tension. Superstar cities generate extraordinary wage premiums, but their housing markets have absorbed much of those premiums through price increases, limiting the real income gains available to workers who must actually live there.

A city that produces superlinear productivity gains while blocking the residential density that would allow more workers to access those gains is failing on both sides of the scaling equation simultaneously.

Remote Work and the Resilience of Urban Agglomeration

The COVID-19 pandemic produced the largest involuntary natural experiment in urban economics in history. Within weeks in early 2020, tens of millions of knowledge workers transitioned to full-time remote work. Office districts emptied. Commuter networks collapsed.

Some workers relocated from major metropolitan areas to smaller cities and rural areas with lower costs of living. Commentators widely predicted either a permanent restructuring of urban geography or a rapid snap-back. The actual outcome, which became clear by 2024-2025, was more nuanced than either prediction.

Some secondary and mid-sized cities did experience genuine population and economic growth. Geographic flexibility increased for workers in roles with limited tacit knowledge requirements.

Commercial real estate in some office-dependent urban cores remained structurally weakened, reflecting a real shift in the geography of where knowledge work gets done on a day-to-day basis.

But major knowledge hubs - New York, San Francisco, London, Singapore - largely recovered in population and economic vitality. The agglomeration premium had been compressed but not eliminated.

The persistence of the premium, even after workers experienced years of full-time remote work, provides powerful evidence about what cities actually provide. The productivity gains of urban agglomeration are not primarily about being physically present in a specific building on specific days.

They arise from long-term immersion in the social networks, informal knowledge flows, and tacit learning opportunities that dense urban environments generate over months and years.

Workers who relocated during COVID did not immediately lose these advantages because they had already accumulated them. Whether their children, starting careers in less dense environments, will accumulate them at the same rate is a question that urban economists will be studying for decades.

For a broader perspective on how spatial and environmental context shapes cognition and culture, see why cities and rural areas think differently about the world.

Implications for Urban Policy

The urban scaling evidence carries concrete policy implications. Restrictions on urban density - height limits, minimum lot sizes, excessive parking requirements, slow permitting processes - impose costs that go beyond housing prices.

If the 15 percent per-capita productivity gain from population doubling is real and generalizable, then cities that prevent their own growth are suppressing the rate of innovation and economic dynamism that their residents and the broader economy would otherwise generate.

The costs of restrictive land use policy are borne not only by those who cannot afford to live in a productive city, but by everyone who would benefit from the innovation those missing residents would have produced.

This argument does not imply that all growth is automatically beneficial or that density alone is sufficient. The superlinear scaling of urban pathologies means that growth without the institutional capacity to manage crime, disease transmission, traffic congestion, and inequality is not a net improvement.

The scaling laws describe tendencies of urban systems; governance determines which side of the ledger those tendencies are allowed to dominate.

Jacobs' architectural prescriptions remain as relevant in 2025 as they were in 1961. The fine-grained mixed-use urbanism she described - dense, walkable, old buildings alongside new, short blocks, street life at human scale - is not an aesthetic preference.

It is a description of the physical conditions under which the interaction rates that drive superlinear scaling actually materialize. Urban design that destroys these conditions, however well-intentioned, destroys the mechanism by which cities produce their advantages.

Sources & Further Reading

1. Bettencourt, L. M. A., Lobo, J., Helbing, D., Kuhnert, C., & West, G. B. (2007). Growth, innovation, scaling, and the pace of life in cities. PNAS, 104(17), 7301-7306. DOI: 10.1073/pnas.0610172104
2. Bettencourt, L. M. A. (2013). The origins of scaling in cities. Science, 340(6139), 1438-1441. DOI: 10.1126/science.1235823
3. Jacobs, J. (1961). The Death and Life of Great American Cities. Random House.
4. Glaeser, E. L., Kallal, H. D., Scheinkman, J. A., & Shleifer, A. (1992). Growth in cities. Journal of Political Economy, 100(6), 1126-1152.
5. Jaffe, A. B. (1989). Real effects of academic research. American Economic Review, 79(5), 957-970.
6. Marshall, A. (1890). Principles of Economics. Macmillan.
7. Rosenthal, S. S., & Strange, W. C. (2004). Evidence on the nature and sources of agglomeration economies. Handbook of Regional and Urban Economics, 4, 2119-2171.
8. Gaspar, J., & Glaeser, E. L. (1998). Information technology and the future of cities. Journal of Urban Economics, 43(1), 136-156.
9. West, G. (2017). Scale: The Universal Laws of Growth, Innovation, Sustainability, and the Pace of Life. Penguin Press.