This publish is the third in a (considerably interrupted) sequence on inhabitants points within the Pacific, re-generating the charts I utilized in a keynote speech earlier than the November 2025 assembly of the Pacific Heads of Planning and Statistics in Wellington, New Zealand. To this point we’ve:
We frequently hear that Auckland is the world’s largest Polynesian metropolis, and even the world’s largest Pacific Islander metropolis; however which is the second or third largest?
This might be a brief publish. The tip level is that this single chart:
Port Moresby (the capital of Papua New Guinea) is the second largest city assortment of Pacific Islanders, and actually it isn’t far behind Auckland. Subsequent come the biggest cities of Indonesian Western New Guinea. I’m not effectively accustomed to that a part of the world and I may need missed some additional cities of comparable dimension, however am assured I obtained the primary two. Coming in at numbers 5 and 6 we’ve Suva in Fiji, and Papeete in French Polynesia (on the island of Tahiti). Then we see that Sydney and Brisbane in Australia in all probability have extra Pacific Islanders than do most of the well-known cities of the Pacific, comparable to Lae, Honiara, Noumea and Port Vila. Samoa’s Apia doesn’t even make it on to the chart as a result of it’s restricted to the highest 24 cities.
I couldn’t get information on French cities within the mainland ‘hexagon’, for which ethnicity data is troublesome to acquire for deliberate choices on the a part of the statistical authorities. There are good causes for this based in historical past. However the quantity might be too small to make it to the chart. Los Angeles might possibly be there if a broad sufficient geography is included, however the metropolis definitions had been a bit robust for me to take care of and in the long run I opted to depart it out.
I’m positive there’s some omissions or errors right here so would welcome corrections and feedback, as normal. However the principle level was illustrative, and geared toward declaring the significance of some cities maybe not typically considered Pacific Islander city concentrations, and I’m completely satisfied that it does that fairly precisely.
There have been a number of selections right here, comparable to whether or not to incorporate West Papuans, Māori (who’re fairly quite a few in Australian cities in addition to in New Zealand) and Hawaiians (and fewer materially as there are much less of them, Torres Strait Islanders) as Pacific Islanders. I’m fairly completely satisfied that the reply is “sure” to incorporate all of them, for our functions. Be aware that if we excluded Māori from the Auckland depend, it could not be the world’s largest Pacific Islander metropolis.
The actual issue, and one I’m assured my resolution for which might be improved, was in getting constant definitions of “metropolis” and good estimates of what quantity of that metropolis are Pacific Islanders. The latter can come from census information, however I didn’t have time to go to every nation’s newest census and guarantee a comparable quantity, so needed to resort to Wikipedia in some instances.
For instance of the issue of a definition of ‘metropolis’, Honolulu itself has a inhabitants of round 350,000, however the City Honolulu metropolitan space is round 1 million (solely a small proportion of whom are Pacific Islanders). Suva’s inhabitants is round 100,000; its metropolitan space brings this as much as 185,000; and should you embrace Lami, Nasinu and Nausori (the place the airport is) this turns into 330,000. In each these instances I used the higher metropolitan space, however not Nausori, and so forth. for Suva.
For Australia and New Zealand I used the “Better Capital Metropolis Statistical Areas” and “Territorial Authorities” respectively. This implies I miss out on non-capital cities, like Gold Coast (inhabitants round 600,000 and round 1 per cent Pacific Islander) however I feel that’s okay. It means we’re under-counting Wellington by the usual I used for Suva and Honolulu (Decrease Hutt and Higher Hutt ought to in all probability be included, however they’re their very own TAs). Once more, I feel that’s in all probability okay.
There’s a minimum of one different extra controversial downside I’ve skimmed over and gained’t point out.
For cities outdoors Australia and New Zealand I didn’t have time to get definitive estimates instantly from every census and relied on Wikipedia and different secondary sources. This bit is very error-prone, and will do with a extra cautious method! General I’ve obtained a considerably dim view of the tossed-together code under, which was an actual compromise between time and thoroughness. However hopefully the outcomes are ok for our illustrative functions! Wherever, right here’s the code:
# this can be a crude exploration of the query:
# "What are the biggest Pacific islander cities on this planet?"
# It's probably incomplete and there are a bunch of extra detailed
# points to enter if we wished to do that definitively.
#
# Peter Ellis 2025-11
library(tidyverse)
library(scales)
#--------------------New Zealand census data----------
# Massive file of Stats NZ census information to obtain. apparently the Census 2023
# equal isn't but availalbe, so we simply use the 2018 model:
dir.create("raw-data")
fn <- "raw-data/nz_census_2018.zip"
if(!file.exists(fn)){
obtain.file("https://www3.stats.govt.nz/2018census/8317_Agepercent20andpercent20sexpercent20bypercent20ethnicpercent20grouppercent20(groupedpercent20totalpercent20responses),%20forpercent20censuspercent20nightpercent20populationpercent20counts,%202006,%202013,%20andpercent202018percent20Censusespercent20(RC,%20TA,%20SA2,%20DHB).zip",
destfile = fn, mode = "wb")
}
# the file is a zipped assortment of lengthy skinny coded information desk and
# dimension lookup tables explaining what every of the codes imply:
unzip(fn, exdir = "raw-data")
ethnic <- read_csv("raw-data/DimenLookupEthnic8317.csv")
space <- read_csv("raw-data/DimenLookupArea8317.csv")
# we're going to use the Territorial Authority stage so we will choose up
# Christchurch, Wellington that are TAs. Be aware this implies we're
# not counting eg Decrease Hutt as a part of Wellington. An interpretation of 'higher Wellington'
# in all probability would come with this. However that is an okay compromise for our functions, I feel?
# Takes some time as a result of there's a mass of very detailed information right here
# however we're solely utilizing a tiny little bit of it - second largest regional teams
# and only a small subset of the ethnic teams
nz2018 <- read_csv("raw-data/Data8317.csv") |>
filter(12 months == 2018) |>
left_join(ethnic, by = c("Ethnic" = "Code")) |>
rename(ethnic_name = Description) |>
left_join(space, by = c("Space" = "Code")) |>
rename(area_name = Description) |>
filter(ethnic_name %in% c("Maori", "Pacific Peoples")) |>
# solely Territorial Authority stage:
filter(str_length(Space) %in% 3) |>
filter(!area_name %in% c("Complete - Territorial Authority areas")) |>
# complete all folks:
filter(Age == "999999") |>
# complete all sexes:
filter(Intercourse == 9) |>
# simply cities (not districts) |>
filter(grepl("Metropolis", area_name) | area_name == "Auckland") |>
mutate(worth = as.numeric(depend)) |>
choose(ethnic_name, area_name, worth) |>
mutate(nation = "New Zealand")
# fast actuality test - print to console the largest TAs with Pacific peoples:
nz2018 |>
group_by(area_name) |>
summarise(worth = sum(worth)) |>
organize(desc(worth))
nz2018 |>
choose(ethnic_name, worth, area_name) |>
unfold(ethnic_name, worth) |>
organize(desc(`Pacific Peoples`))
#--------------Australian census data--------------
# Initially downloaded from australian tablebuilder,
# file is small so is dedicated to this repo:
# `/raw-data/ancestry pacific by higher metropolis 2021 australia census.csv`
aus2021 <- read_csv("https://uncooked.githubusercontent.com/ellisp/blog-source/refs/heads/grasp/information/ancestrypercent20pacificpercent20bypercent20greaterpercent20citypercent202021percent20australiapercent20census.csv",
skip = 9, n_max = 26) |>
choose(-Complete, -...11) |>
rename(ethnic_name = `GCCSA (UR)`) |>
filter(!is.na(`Better Sydney`)) |>
collect(area_name, worth, -ethnic_name) |>
filter(!grepl("Complete", ethnic_name)) |>
mutate(worth = as.numeric(worth)) |>
mutate(ethnic_name = if_else(
ethnic_name == "Maori", "Maori", "Pacific Peoples"
)) |>
group_by(ethnic_name, area_name) |>
summarise(worth = sum(worth)) |>
mutate(nation = "Australia")
#--------------Different--------------
# these estimates from numerous advert hoc sources, largely
# Wikipedia. Remembering we would like variety of pacific islanders,
# not complete ppulation. Which implies we've two troublesome numbers
# to pay money for. So this bit is definitely incorrect! - simply the
# finest estimate I might do in a rush.
different <- tribble(~area_name, ~worth, ~nation,
"Port Moresby", 400000, "PNG",
"Lae", 100000, "PNG",
"Mount Hagen", 50000, "PNG",
# pop is 400k+ however what quantity is pacific? - typically west papua about 75% papuans:
"Jayapura", 320000, "Indonesia",
"Sorong", .75 * 300000, "Indonesia",
"Better Suva", 185000, "Fiji", # not counting nausori
"Lautoka", 75000, "Fiji",
"Nasinu", 74000, "Fiji",
# Solely about 9% of higher honolulu establish as pacific islander:
"Honolulu city space", 0.09 * 1e6, "USA",
"Better Noumea", 0.26 * 200000, "New Caledonia",
"Papeete", 137000, "French Polynesia",
"Honiara", 80000, "Solomon Islands",
"South Tarawa", 70000, "Kiribati",
"Majuro", 20000, "Marshall Islands",
"Apia", 30000, "Samoa",
"Port Vila", 50000, "Vanuatu"
) |>
mutate(ethnic_name = "Pacific Peoples")
#----------------draw bar chart--------------
nz2018 |>
rbind(aus2021) |>
rbind(different) |>
group_by(area_name) |>
mutate(complete = sum(worth)) |>
ungroup() |>
organize(desc(complete)) |>
slice(1:24) |>
mutate(area_name = fct_reorder(area_name, -worth, .enjoyable = sum)) |>
mutate(country_type = case_when(
nation %in% c("Australia", "New Zealand", "France", "USA") ~ "Metropolitan SPC member",
nation %in% c("Indonesia") ~ "Non-SPC member" ,
TRUE ~ "Pacific island SPC member")) |>
ggplot(aes(y = worth, x = area_name, fill = country_type)) +
geom_col(place = "stack") +
scale_y_continuous(label = comma) +
scale_fill_manual(values = c("darkgreen", "brown", "steelblue")) +
labs(fill = "", x = "", y = "Variety of Pacific Islanders
(together with Māori, Papuans and Hawaiians)",
title = "The world's largest Pacific Islander cities",
subtitle = "Deal with these estimates with some warning... corrections are welcomed!",
caption = "Supply: Australia Census 2021, New Zealand Census 2018, Wikipedia and creator estimates ") +
theme(axis.textual content.x = element_text(angle = 45, hjust = 1),
legend.place = c(0.8, 0.7),
plot.caption = element_text(color = "grey50"))That’s all for now. Arising we take a look at how a lot of Pacific Islander populations are within the “dwelling” nation and the way a lot elsewhere (e.g. New Zealand); some extra on inhabitants profiles; remittances information; and a abstract publish the place I’ll tie issues along with the messaging I used within the precise speak.
