Sister communes
The four communes that most resemble each one, across the 28 figures PuroChile publishes for the 342 communes with complete data. Nothing in the computation knows where any commune is: the kinship is statistical, not geographic, and 52% of the pairs cross into another region.
The country's tree
12 branchesThe same distances, read all at once. Each commune joins the one it most resembles, those pairs join each other, and so on up to the whole country; cut at 12 branches, this is what there is. No branch carries an invented name: each is named by its three largest communes and described by the three published figures on which it sits furthest from the median commune.
The tree's first cut is not north against south: it sets 79 dense, schooled, apartment-dwelling communes apart from the other 266. Each junction's height is Ward's — an increase in variance, not a distance — which is why it is not read on the same scale as the 1.48 that separates any two communes. Ward was chosen over average linkage, which is easier to explain, because on this data average linkage chains: cut at twelve it leaves one branch of 318 communes and eleven of a single one.
How it is computed
- Census population
- Population density
- Average age
- 65 and over
- Under 15
- People per household
- One-person households
- Female-headed households
- Indigenous peoples
- Foreign-born share
- Arrived since 2020
- Average years of schooling
- Reached higher education
- Illiteracy
- Disability prevalence
- Disability among 65+
- Average monthly pay
- Employment rate
- Self-employed
- Firms per 1,000 residents
- Contractors as share of workers
- Overcrowded dwellings
- Internet access
- Apartments
- Households heating with wood
- Renting households
- Precarious wall materials
- Public water network
Each dimension carries a seventh of the weight, whether it holds two variables or seven. It is the decision that moves the result most, and the reason it exists: half a dozen of PuroChile's indicators — internet, computers, schooling, higher education, pay — rise and fall almost together, and added in equally they would make two communes sisters for being equally well off.
This is not a measurement. No authority publishes a list of similar communes: it is arithmetic over the same figures PuroChile already publishes on every profile.
Each variable is standardised against its own median and clipped at four robust deviations, so that neither Santiago's density nor Ollagüe's population takes over its axis. Distances are averaged within each dimension and then across the seven.
Whether the lists hang on any single variable is checkable, so it is checked: recomputed 28 times, leaving one variable out each time, 89.4% of the 1,368 pairings hold on average, and never fewer than 74.6%.
Antártica is left out: the census's wide comuna table does not cover it and 1 of the 28 variables are missing. It has no sisters and none are invented for it.
The ones that pick each other
70Kinship is not symmetric: the commune most like A may have someone else as its own closest match. These are the pairs where each is the other's first, closest first — the strongest thing this construction can say.
The 24 closest of 70 mutual pairs are shown.
The ones with no sisters
35Every commune has one that resembles it more than the rest; not every commune has one that resembles it. These 35 are the ones whose best match sits above the 90th percentile of the country's 342 best matches. Their profiles publish the four least unlike them, labelled as that.