Anatomie de la densité

Retranscription de la conférence du 19 janvier 2024 à Sciences Po (Paris)
Colloque Organic Cities


UN-Habitat & New York University

Date de publication

19 janvier 2024


20 mars 2024

I have been thinking and writing about urban density for 20 years now. I confess that one thing I like about urban density is that it is measurable and, like Lord Kelvin said, when something is measurable you can do something about it.

Figure 1

I would like to focus my talk today on some of the findings associated with density with a special emphasis on density and climate change and on the issue of densification as an urban planning strategy. Let me start by noting that, when speaking of density, we need to look at the city or metropolitan area, in their entirety.

Figure 2

Cities are not, as some people will tell us, mosaics of self-contained live-work communities, where people walk or bicycle to work in 15 minutes. They derive their great productivity and their economic resilience by forming integrated metropolitan labor markets, large single markets where every worker has access to all jobs and every firm has access to all workers and to all other firms. And the easier it is for all workers to get to their jobs, wherever they may be, the more productive the city will be. We can therefore focus on gross measure of average urban density as the ratio of the total population of a city or metropolitan area and the total area of its urban footprint.

Figure 3

This way of defining density already tells us something important about the city. It tells us first how much land, on average, each person in the city uses or consumes, because the reciprocal of the urban density is land consumption per capita. Cities with lower density consume more land per person than cities with higher density, and if the policy issue at hand is land conservation, for example, then average urban density measures the efficiency of land consumption in the city. The second thing that average urban density tells us is that, in two cities with footprints of the same shape, the city with the lower density will have higher average distances between its locations than the city with a higher density. That means that cities with higher densities have, on average, shorter distances between locations, requiring shorter infrastructure lines and shorter trips and making it easier for more trips to be taken by walking and biking than, say, by driving. Moreover, it has been shown that public transport becomes feasible above a certain density threshold—some say 30 people per hectare, some say 50 people per hectare—and therefore cities with higher densities can better accommodate public transport.

Figure 4

Atlanta and Barcelona that in 1990 had roughly the same population, while the density of Barcelona was 26 times that of Atlanta. You understand then that you can walk, bicycle, and use public transport in Barcelona to get to a lot more places than you can in Atlanta, where pretty much every trip requires a car.

This connection between average urban density and energy use in transport or greenhouse gas emissions from urban transport has been front and center in connecting urban density with climate change. It is generally agreed that higher average densities in cities contribute to the mitigation of climate change the IPCC Fifth Assessment report, for example, suggests that doubling the average density of a city, coupled with other interventions—like mixed land use, improved public transport, and behavior change—could be associated with up to a 25% reduction in greenhouse gas emissions.

Figure 5

This connection, which is typically viewed as a causal connection between urban density and greenhouse gas emissions, has energized urban planning agendas the world over to call for urban densification in the name of mitigating climate change. This is in addition to other good reasons for densification such as the conservation of prime agricultural lands on the peripheries of cities, shorter infrastructure lines, and a host of other benefits from higher densities that have been measured and presented in the literature over the last couple of decades. This causal connection between density and emissions is surely an important insight but to make sense of it we must subject it to some quantitative analysis. Because even if a city, any city, could retain its present urban footprint and do not expand at al — a highly unlikely scenario — doubling densities would require the city to double its population. Let us look at two sets of data, the projections for urban population growth in different regions between now and 2050 and the distribution of greenhouse gas emissions in different regions in 2020. In general, most urban growth is now projected to be in the global South.

Figure 6

In fact, for every added person to a city in the global north between now and 2015 eighteen persons will be added to cities in the global South. More specifically, one-third of the increase in the urban population will be in sub-Saharan Africa and one-quarter will be in the Indian subcontinent.

Figure 7

The population of the Global North, for example, is expected to grow by 12% between now and 2050, and so if cities in the Global North did not expand at all, perfect densification will increase average densities by 12% and reduce greenhouse gas emissions by one quarter of that, namely by 3%. The population of China, to take another example, is expected to grow by 25% during this period. Perfect densification there will decrease greenhouse gas emissions by one-quarter of that, i.e. by 6%. Now let us look at the data about the distribution of greenhouse gas emissions in different regions.

Figure 8

We see that one-third is produced in the Global North, one-third in China, and one-third in all other countries together. We can also see that a 3% reduction in greenhouse gas emissions in the Global North will result in a 1% reduction in global emissions. A 6% reduction in China will result in a 2% reduction in global emissions. Africa, as a whole, to take another example, produced 4% of total greenhouse gas emissions in 2020. Its population is projected to grow by an additional 150% by 2050. So, with perfect densification it could reduce greenhouse gas emissions by 25% times 150% or by 37.5%. That would amount to reducing global emissions by 1.5%. In other words, the places that are producing a lot of greenhouse gas emissions, like the Global North and China, are not growing in population and therefore cannot really densify and the places that are producing very little greenhouse gas emissions are growing in population and can, in theory, densify, but even if they do they will contribute little to the global reduction in greenhouse gas emissions. Let us now ask ourselves what would doubling urban densities in 30 years would mean and how are cities performing now in comparison to this imaginary target. On average, it would mean that densities would have to increase at an average compound rate of 2.3% per year. Recent data from our study of 3’800+ cities and metropolitan areas that had 100’000 people or more in 2020 show that, on average, global urban densities declined, rather than increased, by 0.75% per year during the 2000-2020 period.

Figure 9

In regions where urban densities increased, say in Sub-Saharan Africa, they increase by an average of less than 1% per annum. In fact, only 121 cities, or 3.2% of all the cities studied densified at a rate equal or higher than 2.3%, the rate required for densities to double in 30 years. In short, given current trends, a business-as-usual scenario is unlikely to result in a significant increase in average urban densities at the required scale to affect greenhouse gas emissions in a meaningful way. What is more, the estimates we derived earlier assumed a perfect densification scenario, where all population growth is accommodated within existing footprints, with perfect containment of outward expansion and thus zero growth of a city’s urban footprint. How realistic is that assumption? As cities grew and developed, they needed more room or, more precisely, more floor space. They could create more floor space in three ways: By building upwards, by infilling the vacant open spaces between buildings, or by expanding outwards, and they typically grew in all three ways together.

Figure 10

We can distinguish between these three ways by referring to building upwards and infill within existing urban footprints as densification and to building outwards as expansion. To an important extent, densification and expansion are substitutes: Typically, when not enough room can be made available to meet the demand for floor space through densification, then room is inevitably made available through expansion. Conversely, when there are barriers to expansion, densification creates more room than it would create in the absence of such barriers. Together with colleagues, I studied a global sample of 200 cities, a stratified sample of all cities and metropolitan areas that had 100’000 people or more in 2010. A map of the sample cities is shown here.

Figure 11

Using satellite imagery, we than identified their urban footprints in 1990 and 2014. The urban footprints of Addis Ababa, Ethiopia, in 1990 and 2014 are shown here.

Figure 12

We then focused on a simple question: What share of the population added to these cities between 1990 and 2014 densified their existing urban footprints and what share was accommodated in the expansion areas built between 1990 and 2014? We found that, on average, only one-quarter of the population added to the 200 cities in the sample in the 1990-2014 period was accommodated within their 1990 urban footprints, while three-quarters was accommodated within their newly built expansion areas. We conducted a similar study in 2022 in 69 cities in the Latin America and Caribbean region for the 1990-2020 period. Lima, Peru, shown here, increased its population by a factor of 1.9 and its area by a factor of 1.8 during this period. Its 1990 footprint (shown in yellow and blue) densified, from 99 persons per hectare in 1990 to 152 in 2020. It accommodated 57% of the added population during the 1990-2020 period, while its expansion area (shown in red) accommodated 43%.

Figure 13

57% of the added population during the 1990-2020 in Lima, Peru, was accommodated through densification of its 1990 footprint. Yet, on average, the 1990 footprints of the 69 Latin American and Caribbean cities we studied accommodated only 18.5%, less than one-fifth, of the population added during the 1990-2020 period, while their expansion areas accommodated 82.5%, more than four-fifths of this population. That said, however, the share of the population accommodated by the densification of existing urban footprints increased significantly over time. It increased from an average of 25% during the 1990-2000 period to almost 40% during the 2010-2020 period.

Figure 14

Still, these findings pertaining to Latin American and Caribbean cities, as well as the earlier finding in the global sample of 200 cities, suggest that, as things stand, we cannot expect business-as-usual densification to accommodate the lion share of the population added to cities in coming decades. Both trends I outlined here—declining average densities and the relatively small share of the added population to cities accommodated by the densification of their existing footprints—are anathema to the great majority of urban planners. For decades now, the prevailing paradigm in the profession has been the compact city paradigm, a paradigm or worldview that is now, and I quote, “enshrined in land use planning policy in many countries”. The essence of this paradigm is the call for a halt to the excessive sprawl of cities into the countryside.
This has led many urban planners the world over to advocate for the containment of urban expansion—by green belts, urban growth boundaries, or by other restrictions on the conversion of land from rural to urban use.

Figure 15

This image shows the edge of the Urban Growth Boundary in Portland, Oregon, one of the few examples of the successful containment of urban expansion. The London green belt shown here is another example of successful containment.

Figure 16

And where strict containment has been largely successful, as in Portland, London or Seoul, Korea, we see three problematic outcomes. First, once the contained area is largely filled, we typically witness a rapid increase in land prices and a subsequent affordability crisis. This has been well documented in all three cities. Containment is now seen by more and more people as excluding them from the cities that they would like to live and work in but cannot afford to, because of housing supply bottlenecks that result in exorbitant house price inflation. Second, containment policies have often been pursued without a comprehensive strategy for densification. Namely, planners that have advocated containment have relied on the private sector to densify the existing footprints of cities—be it through the infill of vacant plots or through the more intensive use of existing plots—without much concern of whether existing regulations, existing infrastructure, or existing transportation modes support or hinder such densification and whether it can be implemented in a way that does not compromise affordability, does not generate implacable neighborhood resistance, does not compromise access to open space, and does not reduce the city’s resilience in the face of climate change. Third, once the area within greenbelts has been largely built-up, development leaps across the greenbelt while still connected to their metropolitan labor markets. This results in longer commutes across the greenbelt and, of course, in larger than necessary greenhouse gas emissions. This has been the fate of the Frankfurt metropolitan area shown here.

Figure 17

This image shows that Frankfurt has a green belt that encircles the city core, which has 750 thousand people. But the metropolitan area of Frankfurt, including outlying cities with strong commuting relations to the city core, is more than three times larger and has expanded beyond the greenbelt: The greenbelt was incapable of containing its outward expansion because it did not allow for a sufficient area for the city to grow within it. Broadly speaking, containment may be successfully enforced when a well-managed public authority—with jurisdiction over the entire extent of the city and beyond - manages to withstand the pressure to expand over a long period of time. Except for a few cities like Portland, London and Seoul, effective containment is the exception rather than the rule. It may be on the books in many cities in official vision statements or master plans that have no statutory authority and are not enforced. In many other cities, good intentions coupled with weak efforts at enforcement have, until now, simply failed. I focus here briefly on two examples of failed containment: Tokyo, Japan, and Beijing, China.

Figure 18

This image shows the abandoned 1941 green belt in Tokyo, superimposed on its built-up urban extent in 2014. The area within this abandoned green belt is 450 square kilometers. The area of Tokyo in 2014 was 14 times larger, 6’400+ square kilometers. Again, Tokyo’s abandoned green belt would not have been able to contain its outward expansion.

Figure 19

The fate of the green belt of Beijing, shown here, is illustrative of the failure of Chinese authorities, powerful as they may be, to limit urban expansion. The red area in this map is the area within Beijing’s 2003 officially designated green belt (light blue boundary) that has been built upon by 2016. Much of the remaining area of the original green belt has been fragmented, and much of it now stands as a barrier between the older city and its newer suburbs, rendering the city less compact than before, and increasing the average travel distance in the city unnecessarily. It is surely possible to contain urban expansion in cities in the Global North that are not growing or shrinking. Yet for better or worse, the containment of urban expansion is not likely to succeeed in cities in the Global South — most notably in Sub-Saharan Africa and the Indian Subcontinent — that are still growing in leaps and bounds. Many city governments in these regions are simply too weak to practice effective containment in the face of overwhelming demand for land on the urban fringe of cities, by both formal and informal developers. Containment there is most likely to fail. To conclude, the continued commitment of urban planners to containment presents a real intellectual barrier to accommodating urban growth in cities that are still growing. In most of these cities, there are good reasons to believe that it will fail, and in a few where it may succeed, its adverse consequences may outweigh its benefits. This raises an important question for urban planners, especially in the rapidly growing cities in the Global South: Is there a ‘green’ alternative to containment? There are good reasons to believe there is.
Making cities more compact — and greener as well — does not call for strict containment but rather for a proper balance between densification and expansion. This proper balance can be attained, in my view, through the regular operations of land and housing markets if, and only if, pervasive organized resistance to both densification and expansion is overcome, the distortions created by inappropriate regulations and incentives are removed, and making room through densification and expansion takes place on a level playing field. The pragmatic alternative to containment is a two-pronged approach to making room in cities: The densification of existing footprints using a comprehensive framework which I call The Anatomy of Density on the one hand, and the ‘greening’ of urban expansion on the other. In the remaining minutes of my talk today, I will elaborate on the Anatomy of Density framework. Unfortunately, I do not have the time to elaborate on the ‘greening’ of urban expansion, a subject close to my heart. In short, it involves ensuring that the rural periphery is converted to urban use in a way that conserves areas of high environmental risk or high amenity value, in a way that connects the periphery to the urban job market with a grid arterial roads that carry public transport, in a way that enables the plots adjacent to these arterial roads to develop as high density areas of ‘transit-oriented development’, in a way that makes small blocks in local neighborhoods easier for walking and biking, and—most importantly—in a way that reduces informal development and keeps housing accessible and affordable to all inhabitants. What do I mean by the anatomy of density? We defined urban density at the outset: Urban Density is simply the ratio of the total population of a city and the total footprint the city occupies. We measure Urban Density in, say, persons per hectare, where a hectare is one percent of a square kilometer.

Figure 20

Dhaka, the capital of Bangladesh, had an average of 372 persons per hectare in 2014; Minneapolis in the United States had an average of 10. Urban Density in Dhaka was 37 times that of Minneapolis. But just knowing these facts does not tell us much. They hide more than they reveal. In what sense in Dhaka’s density high? In what sense in Minneapolis’s density low? The key insight I present to you here is in exposing the anatomy of density, in showing that Urban Density can be factored into separate metrics that expose its internal structure. What do we mean by ‘factoring’? We can factor the number 12, for example, into two factors, 3 and 4, that when multiplied together yield 12. 3 and 4 are factors of 12. So are 6 and 2. In a similar manner, we can factor Urban Density into two, three, four, and then seven metrics. In so doing, we build a structure that incorporates and integrates most of the familiar density metrics used by urban planners. What we discovered — and I can report here — is that the factors that constitute Urban Density are measurable as well. In fact, we measured them in a rigorous manner in ten representative cities from different world regions and in 30 cities in Latin America and the Caribbean. An independent research team unknown to us used our methodology to measured them in Japanese cities too. The results in the first ten cities we analyzed surprised us. The density of each of these ten cities depended on quite different factors. We will get into that soon. Let me start by showing what it means to break urban density into two factors, factors that when multiplied together reconstitute urban density.

Figure 21

Floorspace Occupancy, on the left, is the ratio of the population of the city and its total residential floor area. It is a measure of how many people occupy a unit of residential floor area. The more there are, the less floor area per person. It is a gross measure of how crowded dwelling units in the city are. Floor Area Density, on the right, is the ratio of the total residential floor area of the city and the total area of the city’s urban footprint. It is a measure of the density of residential floor area in the city; or, more generally, of how common it is to find residential floor space in the city. The more of it there is, the more people the city accommodates. You can see that, when we multiply one by the other, ‘floor area’ cancels out and we get people divided by the urban footprint, which is our definition of Urban Density. So by factoring Urban Density into two factors, we gain some insight about its anatomy: The more people per unit floor area, the higher the Urban Density. And the more floor area per areal unit of the city’s urban footprint, the higher the Urban Density. We can go further. We can also break down Floor Area Density into two factors, thus decomposing Urban Density into three factors.

Figure 22

The Floor Area Ratio — the yellow box — is a common density metric. It is defined here as the ratio of the total residential floor area in the city and the total residential area of the city, net of streets and other land uses. The Residential Share — the blue box on the right — is the share of the city’s urban footprint occupied by residential areas. Again, when we multiply the Floor Area Ratio by Residential Share, the residential area in the city cancels out and we get Floor Area Density. This means that we can represent Urban Density as a product of three factors: Floorspace Occupancy, Floor Area Ratio, and Residential Share. When we represent Urban Density as a product of three factors, we can think about it as a box in three-dimensional space, where Floorspace Occupancy is its width, its X-axis, Residential Share is its length, its Y-axis, and Floor Area Ratio is its height, its Z-axis. We obtained values for these three factors for ten cities in different world regions: Dhaka, Hong Kong, Kinshasa, Bogotá, Cairo, Baku, Madrid, Bangkok, Wuhan, and Minneapolis. The Urban Density boxes for these cities are shown here in grey in order of declining density from top left to bottom right. The colored cube on the top left represents the ten-city average value for each factor.

Figure 23

We can now see that different cities get their Urban Density from quite different factors. Dhaka, the city with the highest density in the group gets its density from above-average Floorspace Occupancy and Residential Share, while its Floor Area Ratio is below average. Hong Kong, with the second highest Urban Density, gets its high density from its above average Floor Area Ratio, even though it has a low Residential Share. Kinshasa, with the third highest Urban Density, gets its high density from its very high Floorspace Occupancy — it is extremely overcrowded — and above average Residential Share, but its Floor Area Ratio is way below average. In fact, every city has a unique combination of factors that give it its Urban Density. And we cannot guess the values of these factors from information about Urban Density alone. We need to look for them and measure them one by one. And the Anatomy of Density can go even further. We can factor Urban Density into four factors. We do so by factoring the Floor Area Ratio into two factors: Building Height and Plot Coverage.

Building Height is the ratio of the residential floor area and the area of all residential building footprints in the city. It is a measure of the average number of residential floors on a given residential building footprint. Plot Coverage is the share of the residential area in the city that is covered by residential building footprints. Again, when we multiply them together ‘residential building footprints’ cancels out and we get ‘Floor Area’ divided by ‘Residential Area’, which is the Floor Area Ratio. We now know that Urban Density is a product of four factors: Floorspace Occupancy, Building Height, Plot Coverage, and Residential Share. It increases when either one increases and declines when either one declines. Finally, we can go even further. We can also decompose Floorspace Occupancy into four factors as well.

Figure 24

The first is Dwelling Unit Occupancy, which measures the number of persons in an occupied dwelling unit. The second is the Occupancy Rate, which measures the share of the total number of dwelling units in the city that are occupied, i.e. not vacant. The third is Dwelling Unit Packing which measures the number of dwelling units in a hectare of residential living area. It is the reciprocal of the average size of a dwelling unit in the city. The fourth is Floor Plan Efficiency, which measures the share of the floor area of residential buildings in the city that is devoted to living areas, net of common spaces. Again, when we multiply all four together everything cancels out and we get Floorspace Occupancy. This essentially means that we can represent Urban Density as a product of seven measurable factors: Dwelling Unit Occupancy, the Occupancy Rate, Dwelling Unit Packing, Floor plan Efficiency, Building Height, Plot Coverage, and Residential Share. You can see now that in order to densify a city we must increase one or more of these factors while ensuring that others do not decline. This kind of analysis has useful policy implications. We can increase Dwelling Unit Occupancy, for example, by increasing the number of people that occupy a single dwelling unit.

Figure 25

In times of crisis, we can do so by forcing families to share dwelling units; otherwise, we can encourage communal living and larger households; or we can encourage lowering dwelling unit size. We can increase the Occupancy Rate by removing barriers and incentives that keep a larger-than-necessary share of the housing stock vacant.

Figure 26

We can do so by introducing a vacancy tax on empty units; by permitting the leasing of vacant homes; by overcoming the resistance of owners to rent vacant units; by discouraging the supply of units not in demand; and by reversing central city abandonment. We can increase Dwelling Unit Packing by encouraging people to reduce their consumption of floor area per person and to refrain from increasing it as they become better off.

Figure 27

We can do this by allowing the subdivision of dwelling units; by removing minimum size limits on apartments; by eliminating the lending bias against condos; by avoiding displacement through urban renewal; and by encouraging the residential mobility of older people. We can increase Floor Plan Efficiency by removing barriers and incentives that lower the share of living areas in the total residential floor area under construction.

Figure 28

We can do that by encouraging lightweight construction; by improving high-rise building design; and by reducing parking requirements. We can increase Building Height by revising regulatory barriers and incentives that prevent the increase of residential floor space with taller multi-story buildings.

Figure 29

We can do that by relaxing building height restrictions; by allowing the addition of floors on existing roofs; by increasing access to construction finance; by increasing allowable floor area ratios; and by expanding zoning for multi-family dwellings. We can increase Plot Coverage by removing or redesigning regulations that limit the share of the area of residential plots that building footprints can occupy.

Figure 30

We can do that by reducing minimum lot size for single family homes; by relaxing setback regulations; by allowing multiple units on single family plots; and by increasing allowable floor area ratios. Finally, we can increase Residential Share by increasing the share of residential and mixed-use areas in the city, while limiting the loss of residential area to other land uses.

Figure 31

We can do so by accelerating conversion of commercial and industrial areas to residential land use; by encouraging mixed use; by avoiding mass evictions for urban highways; and by avoiding the over-allocation of lands for industrial use. To conclude, we can decompose Urban Density into measurable factors that allow us to explore its anatomy. They allow us to understand how different cities acquire their density and to see that they acquire it in different ways, something we could not see by simply looking at variations in Urban Density. Decomposing Urban Density into its factors also allows us to design comprehensive urban densification strategies, strategies that look at all the possible ways to increase Urban Density in a systematic way. The Anatomy of Density thus provides urban planners and policy makers with a tool that allows them both to implement complex densification plans and to measure how increases in each of the factors contribute to progress in densification, now a key goal in our global urban agenda.

Figure 32



  author = {Angel, Shlomo},
  publisher = {Sciences Po \& Villes Vivantes},
  title = {Anatomie de la densité},
  date = {2024-01-19},
  url = {},
  langid = {fr}
Veuillez citer ce travail comme suit :
Angel, S. (2024, January 19). Anatomie de la densité. Organic Cities, Paris. Sciences Po & Villes Vivantes.