Your Progress 0%

How to Read Suburb Housing Figures

Census housing data is the best information New Zealand has about what living somewhere costs, and it is misread more often than almost any other official statistic. Not because it is difficult, but because four things about it are not obvious, and each one produces a confident wrong conclusion.

This guide covers those four. What a statistical area actually is. Why two figures that sit next to each other must not be divided. What random rounding does to small numbers. And the large thing the Census simply never asks.

Get those right and the figures are excellent: comprehensive, free, collected from every household rather than a sample, and available for about 2,200 areas across the country.

A statistical area is not a suburb

Census figures are published by statistical area 2, usually shortened to SA2. An SA2 holds roughly 2,000 to 4,000 people and is drawn to make statistics work: reasonably even populations, sensible boundaries, stable over time. It is not drawn to match how anybody names where they live.

Sometimes an SA2 lines up with a suburb you would recognise. Often it splits a large suburb in two, and gives the halves names with "North" and "South" appended that nobody local uses. Sometimes it bundles several small suburbs under the name of whichever is largest. Occasionally its name refers to a geographic feature rather than a settlement at all.

The practical consequence: when a figure is published for an area with a familiar name, check what the area actually covers before concluding anything about the suburb you have in mind. The two overlap; they are not the same thing.

Why this matters more than it sounds

Property marketing frequently quotes Census figures against suburb names, because doing so is easy and the numbers are free. If the SA2 covers a wider area than the suburb, the figure will be pulled toward whatever else is inside it, which in a mixed area can move a median a long way.

Two numbers that must not be divided

On any area page you will find a median weekly rent and a median household income, sitting near each other. Dividing one by the other looks like it would give you a rent-to-income ratio. It does not.

Median weekly rent is measured over households in rented dwellings. Median household income is measured over all households, including those who own their home outright and pay no housing cost at all.

Divide one by the other and you have a ratio describing a household that rents at the median rent and earns the median income of a different population. In an area with high ownership, those two groups can be quite different: owner-occupiers are on average older and better off, so the income figure is pulled up by people who are not paying the rent.

This is why the affordability tool on each area page asks for your income rather than using the area's. Your income is a real number attached to a real household. The area's median income is not the income of the households paying the area's median rent.

What you can do instead

Compare like with like. Set the area's median rent against your own household income, which is what the tool does. Or compare the rent in one area against the rent in another, which is a valid comparison because both are measured the same way over the same kind of population.

Random rounding, and what the Census never asks

Stats NZ randomly rounds Census counts to a multiple of three. This is a privacy measure: in a small area, an exact count can identify people. A published 3 might really be 1, 2, 3, 4 or 5.

Two consequences. First, components need not sum exactly to their stated total; a difference of one or two is the rounding, not an error. Second, the smaller the area, the more the rounding matters in proportion. A count of 900 is barely affected; a count of 6 could be half or double.

That is why areas with fewer than 100 households are not published on this site at all. A median rent for an area with nine renting households is a number, and it is not information.

The thing the Census does not collect

The Census asks renting households what rent they pay. It does not ask owning households what their mortgage costs.

So there is no median mortgage payment for any New Zealand area, anywhere, and no way to derive one from Census data. Which also means there is no single housing cost figure covering everyone in an area: only what renters pay, and separately how many people own.

If a source offers you a suburb-level mortgage figure, it is coming from somewhere else, most likely from sale prices and an assumed deposit and interest rate. That may be a reasonable estimate, but it is a model rather than a measurement, and it should be labelled as one.

Test Your Knowledge

Quiz on reading Census housing figures

1. Is a statistical area 2 the same thing as a suburb?
Yes, they are two names for the same boundary
No, it is a Census geography of roughly 2,000 to 4,000 people that sometimes matches a suburb and often does not
Yes, except in rural areas
No, it is always smaller than a suburb
2. Why can median weekly rent not be divided by median household income?
Because one is weekly and the other annual
Because they are measured over different populations: rent over renting households only, income over all households
Because Stats NZ does not permit it
Because rent is rounded and income is not
3. What does random rounding to base three do?
Rounds every figure down to the nearest three
Randomly adjusts counts to a multiple of three to protect privacy, so components need not sum exactly to totals
Applies only to figures over 1,000
Rounds dollar amounts to the nearest three dollars
4. What does the Census ask owning households about their mortgage?
The weekly repayment amount
The outstanding balance
Nothing, which is why no median mortgage payment exists for any New Zealand area
The interest rate and term
5. Why are areas with very few households not published?
Because Stats NZ withholds them
Because random rounding makes small counts proportionally unreliable, so the figure is a number rather than information
Because they are always rural
Because the data costs more to obtain
6. An area shows a low median household income and a high ownership rate. What is the most likely explanation?
The figures are wrong
An older or retired population, who have lower incomes and own their homes outright
Widespread housing stress
A large student population

Frequently Asked Questions

Is a statistical area the same as a suburb?

No. It is a Census geography of roughly 2,000 to 4,000 people. Sometimes it matches a suburb, often it splits one or covers several.

Why can I not divide median rent by median household income?

Rent is measured over renting households only; income over all households including outright owners. The ratio would describe a population that does not exist.

What is random rounding to base three?

A privacy measure that adjusts counts to a multiple of three. Components need not sum exactly to totals, and small counts are proportionally unreliable.

Does the Census record mortgage payments?

No. It asks renters what they pay and asks owners nothing about their mortgage, so no median mortgage payment exists for any New Zealand area.

Why do some areas have no figures?

Too few households, or none. Many statistical areas are harbours, inlets or offshore islands.

Related pages