Anthony Fox

Elbow room

November 2026

Key points

A growing city has two places to put its new people. It can build at its edge, on farmland, desert or land reclaimed from the sea, or it can crowd more people onto the land it has already built. The split decides how far people travel to work, how much road and pipe the city must lay and how much land it takes.

Three fast-growing cities split their growth very differently between 1975 and 2020. Shenzhen and Hong Kong went from 5.7 million residents to 26.7 million, the Lagos area from 3.3 million to 13.7 million and Cairo from 9.4 million to 27.0 million. I used satellite maps of their buildings since 1975 to find where each one put its new people.

How can a map tell growing out from filling in?

I dated the land under each new resident. The European Commission's satellite record maps buildings every five years from 1975, so I gave each square kilometre of a city the year buildings first covered a good part of its ground. I then grouped the squares by that year, which splits the residents a city added into those living on land built by 1975 and those living on land built since.

The three cities split almost opposite ways. Lagos put 64% of its added residents on land already built by 1975, Cairo 49%, Shenzhen and Hong Kong only 18%. I reran the split nine ways, with squares from 90 metres to a kilometre across and with a tenth to a third of the ground covered, to check the order holds. It does. Lagos keeps 53 to 67% of its growth on old land, Cairo 32 to 56% and Shenzhen and Hong Kong 10 to 27%.

Lagos grew on old land and Shenzhen on new

Residents added from 1975 to 2020, a hundred squares a city, each coloured by when the land under them was first built
Built by 19751980 to 19901995 to 20002005 to 20102015 to 2020Still thinly built
Squares of about 1 km; a square counts as built once a third of its 90-metre cells are at least a fifth covered by buildings. Each area is the square on the maps below, 80 to 90 km across. Source: European Commission, Global Human Settlement Layer, GHS-POP and GHS-BUILT-S, release 2023A.

Did the state open the land in Shenzhen?

The state drew the first line. Shenzhen became a special economic zone in 1980, a strip along the Hong Kong border. The zone's built-up area grew from under 3 km² to 136 km² by 2000, according to the geographer Jianfa Shen. The growth soon ran past its fence. From 1988 the districts outside the zone drew more migrant workers than the zone itself, as factories set up on village land in Bao'an county.

When I split Shenzhen by district, most of its growth sits in that old county. Of the 11.9 million residents Shenzhen added between 1975 and 2020, 68% live in its outer districts, four in five of them on land first built after 1975. The zone was widened to take in the whole city only in July 2010.

Hong Kong grew outward by plan. Its new towns, begun in 1973 on land partly reclaimed from the sea, house about 3.6 million people today. Hong Kong took 15% of the residents added across the map, against 25% for Dongguan, the factory city to the north.

Two thirds of Shenzhen's new residents live outside the 1980 zone

Left, land under buildings by the year it was first built. Right, residents, one column for each square kilometre. 1975 to 2030, the last two steps projected
Source: European Commission, Global Human Settlement Layer, release 2023A; Jianfa Shen (2008) for the zone. Coastline, rivers and borders © OpenStreetMap contributors.

Why did Lagos fill in?

Water hems in the old city, with the lagoon on one side and the Atlantic on the other. The banks, head offices and big firms stayed on Lagos Island, Victoria Island and Ikoyi. New residents crowded onto the mainland within reach of the bridges.

Land had to be made from water. Reclamation took Lagos State from 3,577 km² to 4,050 km² before the state government suspended it in September 2025. Within Lagos State, 79% of the residents added between 1975 and 2020 live on land already built by 1975. The number of people on each km² of built-up land rose by almost a third after 1990.

The overflow crossed the state line to the north. Ogun State took 22% of the wider city's added residents, three in four of them on new land.

Four in five of Lagos State's new residents live on land built by 1975

Left, land under buildings by the year it was first built. Right, residents, one column for each square kilometre. 1975 to 2030, the last two steps projected
Source: European Commission, Global Human Settlement Layer, release 2023A. Coastline, lagoon and the Lagos and Ogun line © OpenStreetMap contributors.

Did the desert towns take Cairo's growth?

Egypt has spent five decades building towns in the desert to draw people out of the crowded Nile valley. The first, such as 10th of Ramadan and 6th of October, date from the late 1970s. New Cairo followed in 2000.

I marked as desert any land with almost no green cover around it. The buildings followed the plan. Desert land took 44% of the ground Cairo built on between 1975 and 2020. The people mostly stayed behind. Desert land holds about one in ten of the residents Cairo added, while the old city and the farmland around it took the rest.

The desert took 44% of Cairo's new building land and one in ten of its new residents

Left, land under buildings by the year it was first built. Right, residents, one column for each square kilometre. 1975 to 2030, the last two steps projected
Desert: land with under a tenth of green cover within about 5 km in 2021 (ESA WorldCover). Source: European Commission, Global Human Settlement Layer, release 2023A. Nile, canals and the Cairo and Giza line © OpenStreetMap contributors.

Town by town, the gap is wide. On Egypt's official estimates for 2024, the six desert towns with published plans held about 960,000 people, 11% of the 8.6 million they were built for. Most of Cairo's growth went instead into informal districts built without permits, often on farmland, which the urban planner David Sims puts at about 70% of the city.

Cairo's desert towns hold a tenth of the people they were built for

Planned population and official estimate, 1 January 2024
TownPlannedEstimate, 2024Share of plan
New Cairo4,000,000321,0918%
10th of Ramadan2,100,000267,01513%
El Obour1,054,500144,07414%
15th of May500,000101,01420%
El Shorouk500,00094,22819%
Badr450,00033,7688%
Six towns8,604,500961,19011%
Source: Egyptian Center for Economic Studies, Economic Lens 20 (August 2026), from the New Urban Communities Authority (plans) and CAPMAS (estimates). 6th of October and Sheikh Zayed publish no plan figure.

Is Lagos unusual?

To place the three among the world's big cities, I used a simpler measure that runs on every one of them. It counts the residents on each km² of land under buildings inside a city's boundary, in 2020 against 1990. Of 499 cities of a million people or more, 53% grew more crowded on this measure.

Residents per km² in Lagos rose by a third, more than in 77% of the others. Cairo, up 8%, sits near the middle and Hong Kong fell by a tenth. The cities that spread out fastest are Chinese and Indian, led by Chengdu, where residents per km² of building land fell 60%. Most of those that filled in fastest are in Africa and the Gulf. Shenzhen has no entry of its own, because the database folds it into one city of 41 million around Guangzhou.

Lagos grew more crowded than three in four of the world's big cities

Residents per km² of land under buildings, 2020 against 1990. One dot for each urban centre of a million people or more
Each centre's boundary is fixed at its 2025 extent. Five centres above 4 times, among them Luanda, are drawn at the right edge. Source: European Commission, GHSL Urban Centre Database, release 2024A.

So why did they grow out or fill in?

Shenzhen grew out because the land beyond its zone lay open to factories and to the villages that housed their workers. Lagos filled in because its jobs stayed on the islands and water closed off the land around them. Cairo built outward by plan, but its people stayed close to the old city and its farmland edge. The split now sets how far each city's people travel to work.

Method and sources

Residents and buildings: European Commission Joint Research Centre, Global Human Settlement Layer, GHS-POP (about 1 km) and GHS-BUILT-S (about 90 m), release 2023A, CC BY 4.0. Both are modelled. Residents are census counts spread onto buildings, so read a single square as an estimate. The layer counts 12.0 million people in Shenzhen in 2020, well short of the 17.56 million in the census. 2025 and 2030 are the layer's projections.

Dating the land: I count a square of about 1 km as built once a third of its 90-metre cells are at least a fifth covered by buildings. I grouped each square's change in residents from 1975 to 2020 by the five-year step in which it was first built. I reran the split at 10, 20 and 30% cover and at about 1 km, 450 m and 90 m; below 1 km, I shared residents by building cover. The 1975 map reads more building than the later ones in Lagos and Shenzhen, so I also checked the Lagos State share from 1990 (80%). I split each map into regions along OpenStreetMap's administrative boundaries.

A second measure, residents per km² of built-up land, answers a different question. On it Shenzhen alone grew more crowded, from about 1,800 to 18,500 between 1975 and 2020, because the farmland it built on held few people. The outward spread across the Shenzhen and Hong Kong map comes from Hong Kong and Dongguan, where that measure fell by about a third and a half after 1990.

Cairo's desert: land with under a tenth of green cover within about 5 km, from ESA WorldCover 2021; the share of growth on it runs from 8 to 11% under other cut-offs. On the satellite layer the named desert towns held about 900,000 people in 2020, below the official estimates. Egypt's prime minister put New Cairo alone near 3 million in October 2026, a figure no published count supports.

World comparison: I used the GHSL Urban Centre Database R2024A, 499 centres with a million residents or more in 2020, land under buildings (building footprint) inside each centre's fixed boundary. I start in 1990 because several 1975 maps, Luanda's above all, read too little building.

Background: Jianfa Shen, The Open Environmental Journal, 2008, on Shenzhen's zone and its migrants; China Briefing, 11 June 2010, on the zone's extension; Hong Kong Civil Engineering and Development Department, new towns factsheet, 2021; Nairametrics, 14 September 2026, on Lagos reclamation; Peoples Gazette, 5 April 2024, on Lagos's commercial districts; Egyptian Center for Economic Studies, Economic Lens 20, 2026; David Sims in The National Interest, 13 March 2014. Maps: coastline, water and borders © OpenStreetMap contributors (ODbL); locator land from Natural Earth.