When schools start asking parents to keep sick children at home, as RNZ reported this week, the households that feel it hardest are the ones juggling the most kids. Three children, one household, one working parent suddenly on unpaid leave. That's the arithmetic of a big family in winter.

Here's the tension. In 2019, Stats NZ counted 92,300 couple households raising three or more dependent children. The most recent reading in the same series, for 2026, is 74,900 (Source: Stats NZ, household-income-by-region). That is a fall of 17,400 households, close to one in five, in seven years. It's also the lowest figure in a series that stretches back to 1998.

Two numbers that shouldn't sit side by side: 2015 and 2026. Back in 2015, the count dipped to 75,700 and looked like a floor. Within two years it had bounced back to 89,900, and by 2019 it was above 92,000. The 2026 figure of 74,900 sits below that supposed floor. What once looked like a one-off dip now looks like the shape of the trend.

The longer view is starker. In 1999, New Zealand had 98,700 of these households. That's roughly 23,800 more than today, or about a quarter of the total gone in a generation. Large families haven't vanished, but as a share of the country's household mix they are a shrinking category.

The path since 2019 has been choppy rather than a clean slide. The count fell in 2020, 2021 and 2022, ticked up to 84,800 in 2023, dropped hard to 79,100 in 2024, nudged up to 79,900 in 2025, then fell again to 74,900. Five declines and two rises across seven years of movement. The direction is down; the ride is bumpy.

Now the honest caveat, because it matters here more than usual. This is survey-based data, and the year-to-year wobbles are big enough that any single reading should be treated with caution. The 2015 dip and the 2017 rebound are the proof: a nearly 14,000-household swing in two years is more likely sampling variation than 14,000 families genuinely materialising. The most defensible reading is the medium-term one, not the single latest print. Household definitions and dependency ages also shape who lands in this bucket, so a level shift is not automatically a behavioural shift.

Even allowing all that, the 1999-to-2026 gap is far too wide to be noise. And here is my view: we build policy, housing and school planning around a family shape that fewer and fewer people are living in. Three-bedroom-plus stock, family tax settings, the assumption that a sick child can be absorbed by a stay-at-home parent. The households that fit that picture are now a minority within a minority.

If you are one of the 74,900, you already know this. You are the household that needs the seven-seater, the four school lunches, the two bedrooms doubled up. And you are, statistically speaking, rarer than at any point this century.