Three functions and an interest bill carry three-quarters of it
Welfare, health and education account for $98.7 billion — 70% of everything the core Crown spends. Add the interest on past deficits and you reach 76%. Everything else the state does — police, courts, defence, roads, environment, housing, culture — fits in the remaining quarter.
Two lines deserve a hard look. NZ Superannuation alone costs $23.2 billion — more than the entire education system — and is forecast to reach $30.9 billion by 2030. And finance costs of $8.9 billion are pure deadweight: interest on money already spent, now the fourth-largest function of the New Zealand state.
This page is the itemised bill; our companion site taxdollars.nz lets you explore the same $141.7 billion interactively — every line, what it costs per person, and where your own tax goes.
Data table & source
| Function | FY2025 | Share |
|---|
Source: Financial Statements of the Government of New Zealand, year ended 30 June 2025, Fiscal Indicator Analysis — Expenses by Functional Classification (p. 160); Treasury's classification labels the top line "social security and welfare". NZ Superannuation from Note 8, Transfer Payments and Subsidies.
Open the three biggest boxes
"Welfare", "health" and "education" are labels on boxes, and labels hide things. Opened up, the audited accounts tell a sharper story: half of all welfare transfers is NZ Superannuation — $23.2 billion before the first Jobseeker dollar is counted. Hospitals absorb over half of health delivery. And the third-biggest line in schooling isn't teachers or ECE — it's the school property portfolio, at $3.4 billion a year. This is the level of detail a visible tax bill invites voters to read.
Source: FSGNZ 2025, Note 8 — Transfer Payments and Subsidies (p. 71), audited actuals, total Crown; rows sum exactly to the note's $44,679m. The note spans all functions, so it includes international development co-operation and student allowances (paid via Vote Social Development). The $47.5b welfare function total additionally carries non-transfer spend — MSD operating costs, disability support services ($2.6b, delivered by MSD since Sept 2024), employment programmes and similar.
Source: Vote Health, Estimates of Appropriations 2025/26 — 2024/25 estimated actuals, expense-type appropriations (the audited accounts don't publish a health line-item split). Mental health is not separately appropriated — it sits inside both delivery lines. On top of these sit ~$3.2b of capital-type items (hospital builds, the $1.3b Holidays Act remediation, pay-equity injections), which with accrual differences bridge most of the gap to the $30.3b health function total. Disability support services ($2.6b) are classified under welfare, not health.
Sources: Vote Education and Vote Tertiary Education, Estimates of Appropriations 2025/26 — 2024/25 estimated actuals, expense-type appropriations. Student allowances ($574m) are paid through Vote Social Development and appear in the welfare table above, not here. The two votes' expense lines total ~$21.4b against the $20.9b education function — the ~$0.5b difference is mostly the school-property capital charge, an intra-Crown payment eliminated in the audited consolidation.
The state's size is a choice. The interest bill isn't.
Over 25 years, core Crown spending has moved between roughly 28% and 34% of GDP — rising in crises (the GFC, COVID), and settling back only partly afterwards. The architecture takes no side on where in that band New Zealand should sit: that is Parliament's argument to have, visibly, at elections. What the architecture insists on is that the argument be honest — paid for with visible rates on durable bases, not with bracket creep or with borrowing whose costs land on the next generation's ledger as "finance costs".
Because that last line is the one nobody votes for. The interest bill collapsed to $1.9 billion in FY2021, then more than quadrupled in four years as deficits and rates returned. Treasury forecasts it at $13.4 billion by FY2030 — by then bigger than defence, law & order and transport combined. Every dollar of it buys nothing.
Sources: Treasury Fiscal Time Series 1972–2025 (history) · HYEFU 2025 (forecasts). OBEGALx is the Government's headline balance measure, excluding ACC.
The architecture finds savings the scalpel can't
Every government hunts line-item savings, and the recent record shows exactly what the scalpel yields. Budget 2024 set every agency a 6.5–7.5% baseline target and found $1.5 billion a year; the consultant-and-contractor purge cut $915 million over two years — double its target. Real money, hard won — and ultimately about 1% of spending, with public-service headcount already drifting back up by March 2026.
The deeper savings sit in the structure: money spent collecting, complying with, and churning a tax system far more complicated than the state it funds. None of it appears as a Budget line called "waste" — it's spread across IRD's running costs, 400,000 businesses' accounting bills, and a transfer system that taxes the same families it pays. This is the layer the architecture removes by design rather than by review.
And the incentive to stop buying growth
One more line deserves scrutiny: the state paid out about $2.3 billion in subsidies in 2024 (Stats NZ), including $180.8m of film rebates in a single year — $1.5 billion over the past decade, by the Government's own count. Some of it evaluates well, some doesn't (Motu found Callaghan growth grants returned $0.83 per dollar). The architecture's incentive clause cuts at the root: a government whose revenue rises automatically with spending, land value and profits doesn't need to purchase growth stories with grants — it gets paid by growth itself, wherever growth happens. And every dollar of genuine saving compounds: the interest bill on past deficits is already $8.9b and heading for $12.8b by 2029.
Sources: IRD Annual Report 2024–25 (departmental expenses $752.7m; 4,526 FTEs; cost of collection 46¢ vs 80¢ in 2015) · IRD SME compliance cost survey 2024 (median 32 hrs / $5,749; PAYE 26 hrs; 91% use external professionals) · Sandford & Hasseldine (1992), business tax compliance ≈ 2.5% of GDP (1990) · IRD WfF statistics ($3,043m, 328,400 families) · IRD WfF consultation 2025 (~80% IR-paid; $273.5m debt) · Treasury AN 23/03 & AN 25/01 (net-tax and EMTR analysis) · IRD OIA on FamilyBoost · Treasury, Budget 2024 Summary of Initiatives ($1.5b/yr baseline savings) · Beehive/PSC (consultant spend −$915m; workforce partially rebounding by Mar 2026) · Stats NZ GFS (subsidies $2.3b, 2024) · NZ Film Commission 2024–25 ($180.8m rebates) · Motu WP 25-11 (growth grants bang-for-buck 0.83) · BEFU 2026 (finance costs to $12.8b by FY2029). The $2.3b SME aggregate is our own multiplication of IRD's median by the SME count — a floor, since means exceed medians.