362 | Luis Bettencourt on the Universal Properties of Self-Organizing Cities
What this covers
Sean Carroll talks with Luis Bettencourt, a theoretical physicist at the University of Chicago, about whether the reasoning behind biological scaling laws can be applied to cities. The argument is built in steps. In a circular city, travel cost proportional to radius is set equal to an interaction rate proportional to density. That gives area growing as population to the two-thirds power, density to the one-third, and total interactions to the four-thirds. Hierarchical road networks, which work like a parallel circuit, yield the one-sixth exponent. Bettencourt says this value depends on space being two-dimensional and on travel tracing a line, so a three-dimensional city would give a different number.
The empirical range is wide. Per-capita wages and expenditures rise about 16 percent per doubling of population, while per-capita road and pipe infrastructure falls by a similar amount. Patents, crime, and early COVID-19 spread also accelerate with size. Bettencourt predicts that people walk faster in big cities and that highways are especially prone to jamming. He cites the pattern in China, Brazil, and South Africa, treats the UK as less certain, and uses pre-Columbian Mexican cities as a natural experiment, with monument construction standing in for output.
The later turns are about people. Bettencourt criticizes the mean field approximation because no individual resembles the city average. He ties interaction counts to wealth and isolation to poverty, and discusses specialization as a division of knowledge. He also covers why young migrants arrive, a Kenyan cell-phone study he describes as anecdotal, East Asian urbanization and poverty reduction, and the claim that urban need gave rise to social democracy.
Bettencourt argues that cities are self-organizing complex systems whose size-dependent outputs, interactions, and infrastructure follow universal scaling exponents derivable from simple geometric and interaction arguments, so urban dynamics are more fundamental than any particular political or economic system.
- Per-capita outputs such as wages, patents, and interaction-driven outcomes rise by roughly 15-16% when city population doubles, while per-capita infrastructure falls by about the same amount
- The exponents follow from balancing travel cost against interaction rate in two-dimensional space, with the one-sixth value arising from hierarchical infrastructure, and they recur across nations, history, and the pre-Columbian Americas
- Because these regularities hold across very different political and economic systems, cities are a more fundamental unit of analysis than capitalism, socialism, or democracy
Dense cities follow shared scaling dynamics beyond local politics, which supports complex systems as a field.
- Will Wilkinson's density divide research found that city residents score higher on the Big Five trait of openness, meaning people who want to explore new things sort into cities, and Carroll suggests dense environments provide more opportunity for this exploration.
“the people who tend to go to cities have higher openness”
- Carroll's paraphrase that the history and local politics of a city matter little for these scaling quantities is accepted by Bettencourt, who argues that the regularities show cities have shared dynamics that transcend specific political choices.
“the local politics at the moment, Democrats, Republicans, or whatever your local political parties are, kind of don't matter that much for a certain set of really important quantities”
- Cities show that very similar reasoning about scaling applies to biological organisms and cities, two very different kinds of things, which justifies complex systems as a field of study in its own right.
“very similar kinds of reasoning lead you to insight about both biological organisms and about cities, which are completely different kinds of things.”