Someone on the internet (me, today) asked which great cities allocate the most vertical metres of supertall to each resident. Not total floorspace, not skyline mass, not aesthetic charm: the blunt ratio of architectural height of the tallest completed high-rise divided by headcount, for a fixed list of mega-agglomerations.
The toy version is obvious. A hundred people sharing one 100 m tower get 1 metre of pinnacle per person. Double the population without raising the spire and you halve the ratio. The question is what happens when you run the same arithmetic on real municipalities at nine-figure scale, where politics, geology, airport flight paths, and CTBUH definitions all interfere with the clean parable.
The metric
For each city i, let Hi be the tallest completed building in the urban jurisdiction tracked by the CTBUH Skyscraper Center (municipal population in their city facts panel when present), and Pi residential population from Wikidata property P1082 using the March 2026 snapshot returned by their public SPARQL endpoint. The ratio is Ri = Hi / Pi.
When Skyscraper Center had no city page, I took the tallest figure from curated national tables on Wikipedia (India, Bangladesh, Pakistan summaries) or a short list of hand-checked municipal sources documented in the build script. Where both existed, I used the maximum of the Center height and the national-table height (mainly to correct under-reported figures in some South Asian extracts), never a fuzzy cross-match across unrelated cities.
Ri is tiny, so the tables below multiply by 106 and report micrometres of pinnacle per resident (µm / person). Dubai at roughly 210 µm is handing each inhabitant about 0.21 millimetres of Burj Khalifa if you insist on visualising it that way. The point of the scaling is comparative: Dhaka and Lagos sit in the low tens of micrometres because a tower that looks heroic in silhouette still disappears inside a demographic ocean.
Who made the cut
Wikidata typing is messier than it has any right to be. I took the hundred most populous distinct English labels among entities whose instance of is exactly Q515 (city). That captures Beijing and omits Shanghai, because Shanghai’s ontology uses more specific “city in China” classes instead. Delhi and Guangzhou fall out for similar reasons. The ranking is therefore not “the world population top hundred by any single census definition”; it is “the hundred largest things Wikidata confidently calls city in the narrow sense.” Treat the set as a large stratified sample of the urban human experiment, not as a claim to completeness.
What the numbers do
The distribution is right-skewed but not lawless. After the micrometre conversion the mean is about 54 µm, the median near 48 µm, and the 90th percentile lands just under 100 µm. Below that mass you find city-states, affluent second cities, and petro-capital skylines. Above it you find the ultra-dense South Asian and West African basins where even a respectable tower leaves almost no per-capita trace.
Upper tiers (selected)
| Rank | City | Tallest (m) | Pop (Wikidata) | µm / resident |
|---|---|---|---|---|
| 1 | Dubai | 828 | 3,944,751 | 209.9 |
| 2 | Taipei | 508 | 2,442,991 | 207.9 |
| 3 | Kuwait City | 413 | 2,989,000 | 138.2 |
| 4 | Kaohsiung | 348 | 2,733,964 | 127.3 |
| 5 | Busan | 412 | 3,453,198 | 119.3 |
| 6 | Perth | 253 | 2,141,834 | 118.1 |
| 7 | Algiers | 264 | 2,364,230 | 111.7 |
| 8 | Phnom Penh | 228 | 2,129,371 | 107.1 |
| 9 | Toronto | 298 | 2,794,356 | 106.6 |
| 10 | Caracas | 225 | 2,245,744 | 100.2 |
| 11 | Brisbane | 270 | 2,706,966 | 99.7 |
| 12 | Pyongyang | 274 | 2,863,000 | 95.7 |
| 13 | Tashkent | 267 | 2,956,384 | 90.3 |
| 14 | Xining | 219 | 2,467,965 | 88.7 |
| 15 | Los Angeles | 335 | 3,898,747 | 85.9 |
Dubai and Taipei are effectively tied once you account for rounding: different politics, similar order-of-magnitude story about small denominator populations carrying very tall symbolic needles. American coastal hubs (Los Angeles, later New York in the full table) land mid-pack not because their towers are short (they are not) but because Wikidata populations for consolidated cities swallow a lot of people.
Chinese administrative cities split into surreal land-area units; several megacity prefectures occupy the lower third despite skyscrapers exceeding 400 m, because Pi includes millions of exurban residents who will never set foot in the central business district tall zone. That is not a bug in the algebra; it is the algebra punishing inclusive municipal boundaries.
Basement floor (selected)
| Rank | City | Tallest (m) | Pop (Wikidata) | µm / resident |
|---|---|---|---|---|
| 89 | Bamako | 80 | 4,227,569 | 18.9 |
| 90 | Prayagraj | 110 | 5,954,391 | 18.5 |
| 91 | Yangon | 122 | 6,874,000 | 17.7 |
| 92 | Kumasi | 65 | 3,903,480 | 16.7 |
| 93 | Surat | 93 | 5,935,000 | 15.6 |
| 94 | São Paulo | 172 | 11,451,999 | 15.0 |
| 95 | Kano | 62 | 4,348,000 | 14.3 |
| 96 | Lima | 140 | 9,943,800 | 14.1 |
| 97 | Chengdu | 284 | 20,937,757 | 13.6 |
| 98 | Lahore | 150 | 11,126,285 | 13.5 |
| 99 | Lagos | 160 | 15,070,000 | 10.6 |
| 100 | Dhaka | 171 | 16,800,000 | 10.2 |
Dhaka finishes last with ~10.2 µm despite an earnest 171 m roof: the numerator is fine; the denominator is the demographic equivalent of a neutron star. If you wanted a policy reading, it is not “build tall”; it is “tall is a feeble lever when housing demand scales like compound interest.”
How seriously to take this
Not very, and also quite a lot. The ranking is fragile to boundary choice: swap city-proper for metropolitan definitions and Taipei, Toronto, and Los Angeles dance several slots without changing a single steel beam. It ignores slums, informal floors, and anything below high-rise prestige thresholds. It treats one spire as the stand-in for an entire vertical program. It cannot know whether you wanted skyline ego or flood-safe mass timber.
What survives those caveats is the reminder that extensive growth (more bodies) and intensive growth (higher steel) are logged on different ledgers. Most public conversations fuse them into a single aesthetic judgement. Separating them with a brutish ratio does not solve urban planning. It does make the trade-off legible: you can always win the height arms race in the photograph; you cannot automatically win it in the denominator.
Full machine-readable rankings (all 100 rows) live alongside the Jekyll tree in misc/skyline_per_capita_ranked.json. The reproducible merge logic is misc/build_skyline_rankings.py, with raw Skyscraper Center passes cached by misc/skyline_per_capita_data.py.
If you redo the scrape in five years, check whether Dubai still anchors the top: the exciting scientific result would not be a reshuffle, but a systematic decline in the ratio across emerging markets as populations outrun even aggressive vertical construction. That would tell you something about asymptotics that photographs never will.