Bigger Cities Build Lighter: Global Study Finds Built Mass Grows Slower Than Population

For more than a century, urban planners and scientists have largely assumed a simple rule: every additional resident of a city requires roughly the same amount of concrete, asphalt and steel as the last one. A new analysis of more than 3,000 cities worldwide, published in Nature Cities by Kangning Huang and Mingzhen Lu, overturns that linear assumption. The study finds that the mass of the built environment grows more slowly than population, both when comparing cities with one another and when comparing neighborhoods within a single city. Larger cities and denser neighborhoods, in other words, achieve economies of scale in physical construction, requiring less built mass per person than their smaller or sparser counterparts.

The researchers quantified this pattern with power-law scaling exponents. Across cities, the exponent β was estimated at 0.90, meaning that a tenfold increase in population is associated with roughly a 7.9-fold increase in built mass rather than a tenfold one. Within cities, across neighborhoods, the exponent δ was even lower, at 0.75. Because both values fall below 1, the relationship is described as sublinear scaling, a signature of economies of scale. In practical terms, the built environment per capita shrinks systematically as urban systems grow larger and denser, a pattern that simple per capita accounting conceals.

Measuring the physical mass of cities on a global scale is a formidable technical challenge, and the study’s methods are among its most notable contributions. The team built their estimates from the ground up, combining raster-based estimates of building volume from three independent satellite-derived datasets by Esch and colleagues, Li and colleagues, and Liu and colleagues. Averaging across the three datasets helped minimize bias specific to any single product. Building volumes were then converted to mass using a material-intensity framework that accounts for both building type and regional construction practice, drawing on the work of Haberl and colleagues, which distinguishes categories such as OECD countries, North America, Japan and China.

Roads and pavements posed a separate problem. The researchers started with the Global Impervious Surface Area product at 10-meter resolution, subtracted estimated building footprints to isolate pavement, and then separated five grades of road network, from highways down to local streets, because each grade demands different structural depth and therefore different material intensity. Other pavements, such as sidewalks and parking lots, were assigned a material intensity of 570 kilograms per square meter, equivalent to roughly 20 centimeters of concrete or asphalt. Subway infrastructure, which evades remote sensing, was estimated from compiled network geometries using per-kilometer tunnel and per-station coefficients, and sensitivity analysis showed that including it shifted the scaling exponents negligibly.

Cities were defined using the Urban Cores concept from the Global Human Settlement Layer developed by the European Commission’s Joint Research Centre, which delineates contiguous built-up areas of high population density from satellite imagery and census data. Only cities with at least 50,000 residents were included, a threshold consistent with other global urban studies, and the authors report that their main finding is robust to the choice of cut-off. Neighborhoods were delineated with Uber’s H3 hexagonal grid at resolution 6, where each cell spans roughly 36 square kilometers with an average edge length of about 3.7 kilometers. The authors chose hexagons because they minimize perimeter-to-area ratio and avoid the directional bias of square grids, and because the cell size captures the scale of daily human activity, with residents spending approximately 89 percent of their trips within the focal hexagon.

Population figures came from the WorldPop dataset at 100-meter resolution, summed within city and neighborhood boundaries. To handle the statistical complications of scaling analysis, the team used a centered-scaling approach developed by Bettencourt and Lobo, subtracting group-level means of log population and log mass before regression. At the city level, country means were removed to strip away differences in construction norms, material availability and development history; at the neighborhood level, city means were removed to isolate within-city variation. This design yields a universal elasticity of built mass with respect to population, net of baseline differences between countries or between cities.

Perhaps the study’s deepest result is theoretical: the across-city scaling pattern emerges from the uneven within-city distribution of people. Neighborhood populations follow Zipf-like heavy-tailed distributions, which formal testing following Clauset, Shalizi and Newman strongly preferred over log-normal alternatives. The authors modeled built mass as a log-normal conditional expectation given population, following the statistical framework of Gomez-Lievano and colleagues, and derived an exact constraint linking the neighborhood scaling exponent to the relative shapes of the population and mass distributions. Because the distribution of neighborhood built mass is more homogeneous than the distribution of neighborhood population, δ falls below 1. Aggregating neighborhoods into cities then bounds the city-level exponent β between δ and 1: in a highly disparate regime where the largest neighborhood dominates, β equals δ, while in a perfectly homogeneous regime β equals 1. A simulation sweeping the Zipf shape parameter reproduced the observed relationship between within-city disparity and across-city scaling.

The robustness of these conclusions was tested from four directions. The analyses were repeated with each building-volume dataset individually, in random combinations, and weighted by regional validation accuracy. Material-intensity assumptions were compared across a single global average, a five-region classification and a fine-grained 32-region classification accounting for climate and development stage. Underground infrastructure was bounded using subway data and high-resolution United States material-stock assessments. And the hexagonal grid was analyzed at three resolutions and displaced in cardinal directions. Across all tests, the exponents remained robustly sublinear, with variation from any single methodological choice roughly an order of magnitude smaller than the deviation of β and δ from unity.

The study situates its analysis against historical datasets of global built structure mass and human biomass from 1900 to 2015, with projections to 2037, using a coarse estimate of average human body mass of about 50 kilograms derived from global age distributions and age-specific body masses. The authors note that age-specific body mass varies across countries and diets, but that their qualitative conclusions are not sensitive to this estimation. These historical and projected mass comparisons frame the scale of the global built stock that the scaling analysis addresses, underscoring why the efficiency of material use in growing cities matters for understanding long-term material demand.

The policy implications are pointed. The authors caution against well-meaning interventions, such as height and building-size limits, that artificially flatten cities’ natural density gradients, since such rules push urban form away from the material-efficient configurations that sublinear scaling reflects. They argue instead that effective urban policy should move beyond per capita linear thinking, adopting scale-adjusted metrics and managing cities as nested complex systems in which neighborhood-level disparities generate city-level regularities. All data and code supporting the study are publicly available, along with an interactive visualization of the global mass patterns, allowing other researchers to scrutinize and extend a result that reframes how the material weight of urban life should be counted.

Subject of Research: Sublinear scaling of urban built mass with population across global cities and neighborhoods

Article Title: Nested economies of scale in global city mass

Article References: Huang, K., & Lu, M. (2026). Nested economies of scale in global city mass. Nature Cities. https://doi.org/10.1038/s44284-026-00532-x

Image Credits: AI Generated

DOI: 10.1038/s44284-026-00532-x

Keywords: urban scaling, built mass, cities, economies of scale, sublinear scaling, urban density, material stocks, Zipf's law, remote sensing, infrastructure, urban policy, complex systems