Australia set a target to cut the over-representation of Aboriginal and Torres Strait Islander children in out-of-home care by 45 percent by 2031 — the rate rose to 52.2 per 1,000 while 63.6 percent of child protection spending goes to care after removal

What is happening?

In July 2020 the Australian government and all eight state and territory governments signed the National Agreement on Closing the Gap. Target 12 commits every one of them to reduce the over-representation of Aboriginal and Torres Strait Islander children aged 0 to 17 in out-of-home care by 45 percent by 2031.

The measure has moved the other way. The Productivity Commission, which maintains the official data repository for the agreement, classifies Target 12 as worsening. The measure stood at 47.3 per 1,000 children in the 2019 baseline year and at 52.2 per 1,000 in 2025.

The distance from other children has widened over the same period. At 30 June 2024 there were 19,987 Aboriginal and Torres Strait Islander children in out-of-home care nationally, about 50 per 1,000 against 4.6 per 1,000 for non-Indigenous children. That ratio of 11.0 to one was 9.6 to one in 2020. At 30 June 2025 Aboriginal and Torres Strait Islander children were 44 percent of all children in out-of-home care in Australia, up from 42 percent in 2020-21, while making up about 7 percent of the population aged 0 to 17.

The money runs in the opposite direction to the target. In 2024-25 Australian governments spent about 11.3 billion dollars across family support, intensive family support, protective intervention and care services. 7.2 billion dollars of that, or 63.6 percent, went to care services — the cost that begins once a child has been removed. Family support and prevention received the remaining 36.4 percent.

The clause written into the same agreement to correct this carries no number. Priority Reform Two commits the signatories to shift government funding towards Aboriginal and Torres Strait Islander community controlled organisations, but it sets no share, no dollar figure and no deadline, and delivery is established through voluntary annual reporting that carries no penalty.

One jurisdiction has been audited against the placement rules in detail. In June 2024 the Audit Office of New South Wales found that the state child protection department cannot demonstrate compliance with the Aboriginal and Torres Strait Islander Child Placement Principle. Of the 105 recommendations from the 2019 Family Is Culture review that fell to that department, 12 were complete four years later. The audit estimated that at the observed rate, transferring case management for Aboriginal children to community controlled organisations would take about 57 years. It also recorded that 65 percent of removals of Aboriginal children between 2018 and 2023 used emergency powers that bypass a court process.

Whose problem is this?

The governments that set the target are the same governments that write the budgets running against it. Nobody outside them can change the ratio, and the annual report that would show whether they are changing it is written by those same governments and carries no consequence either way.

Where does this problem end?

The boundary matters here because nothing in it is disputed at the level of intent. Every government in the country signed the target, and the same governments publish the spending figures that run against it, so the open question is not what ought to happen but why both the money and the outcome keep moving the other way.

What is the state now, and what should it be?

Now

Should be

The target state is written down and dated for one half of this problem and absent for the other. Target 12 requires the measure it tracks to sit 45 percent below its 2019 baseline of 47.3 per 1,000 by 2031, which is a number, a direction and a year. Priority Reform Two requires funding to move towards community controlled organisations and stops there.

No source opened here gives a target share of child protection funding that should sit with community controlled organisations, and none gives a target split between prevention and care after removal. The spending half of this problem has a direction but no destination, which is why nothing about the 63.6 percent recorded for 2024-25 can be called ahead of schedule or behind it.

How big is it?

19,987 children. That is the national count of Aboriginal and Torres Strait Islander children aged 0 to 17 in out-of-home care at 30 June 2024, published in the Indigenous health performance framework. It is about 50 per 1,000 Aboriginal and Torres Strait Islander children, against 4.6 per 1,000 non-Indigenous children.

A second national series does not agree with it. A figure reported from the 2026 government services series gives 15,340 at 30 June 2025, about a quarter lower one year later. Nothing opened here states that the count fell, every other indicator in this document moves the other way, and no document reconciles the two series or states what each one covers. The lower figure is therefore recorded here as an unexplained disagreement rather than used as a bound, because a count that may cover fewer jurisdictions is not a smaller measurement of the same thing.

This counts children already removed on one day. It is not a count of children at risk of removal, and it is not a count of the families and communities on the other side of each removal. The population living under the conditions this document describes is larger than the population counted in care, and no source opened here measures that larger group.

The share is the part that has moved fastest. Aboriginal and Torres Strait Islander children were 42 percent of all children in out-of-home care in 2020-21 and 44 percent at 30 June 2025, against about 7 percent of all children in the country.

Under what conditions does it arise?

1. The budget is built around the event the target exists to prevent. In 2024-25, 63.6 percent of the 11.3 billion dollars spent went to care services, which begin at the moment of removal, and 36.4 percent went to family support and prevention, which exist to stop it. Removal is the funded pathway and prevention competes for what is left over.

2. The clause meant to correct this carries no number. Priority Reform Two commits every signatory to move funding towards Aboriginal and Torres Strait Islander community controlled organisations, without a share, a dollar figure or a date, and delivery is reported annually by the governments bound by it with no penalty for reporting nothing.

3. Compliance with the placement principle cannot be shown even where it is law. The placement principle is enacted in every state and territory, and the one audit that tested it found the responsible department unable to demonstrate compliance. Over the same period 65 percent of removals of Aboriginal children in that state used emergency powers that bypass a court process, which is the pathway with the least external checking of all.

4. The transfer to community control runs slower than the caseload it is meant to take on. An audit projection put completion at about 57 years at the observed rate, and 12 of the 105 review recommendations falling to that department were complete after four years.

A structure like this does not need anyone to decide against the target. It only needs each part of it to keep working exactly as it was funded to work, and the aggregate then moves in the direction the money points rather than the direction the agreement points.

What has been tried?

The results are recorded in the same places as the attempts. The target is assessed as worsening, the funding commitment has no figure it could fall short of, the placement principle is law everywhere and demonstrable nowhere that was audited, and the transfer of case management was projected at about 57 years to completion. Every one of these was designed, delivered and scored by the parties that hold the power to finish it.

What was found?

Why is it still unsolved?

Incentive inversion — the budget is largest at the point the target is trying to reach least, and the clause written to fix that is the one part of the agreement with no number in it.

The first part is arithmetic rather than intent. A child protection system that spends 63.6 percent of 11.3 billion dollars on what happens after a child is removed has built its capacity, its staffing and its contracts around removal, and the 36.4 percent left over is what prevention has to work with. Money spent after removal is a legal obligation attached to a child already in care, while money spent before removal is discretionary and competes every year against that obligation. The obligation always wins, and it wins without anyone arguing for it.

The second part is that the correction has nothing to be short of. Priority Reform Two commits every signatory to move funding towards community controlled organisations and then declines to say how much or by when, so no year can be assessed as a shortfall and no government can be shown to have missed anything. Delivery is reported by the parties bound by it and no penalty attaches to reporting nothing, which turns the commitment into a record of effort rather than a record of results. The one place an external body did apply a measure, a state audit office, the finding was not that the target was missed by some margin but that compliance could not be demonstrated at all.

The third part is that the failure never becomes an event. Each removal is decided one at a time, on its own file, by people acting inside their own authority, and 65 percent of them in the audited state went through emergency powers that do not pass a court on the way. There is no moment at which the aggregate is decided, so there is no moment at which it can be refused. The rate is published once a year, the assessment says worsening, the agreement stays in force, and nothing in the process has changed by the time the next figure arrives.

What observation would mean it is solved?

Candidates — (a) the measure Target 12 tracks falls 45 percent below its 2019 baseline of 47.3 per 1,000, by 2031 (b) the share of child protection spending going to family support and prevention rises well above the 36.4 percent recorded for 2024-25 (c) Priority Reform Two acquires a published share and a published date, with delivery reported by somebody other than the governments bound by it.

(a) alone cannot be read on its own. The two national series already differ by about a quarter with no reconciliation, so a fall of that size would not be distinguishable from a change in counting basis or in jurisdictional coverage. The measure also has a denominator that moves for reasons unrelated to removal, including changes in how many people are identified as Aboriginal or Torres Strait Islander, so a ratio can improve while the number of children removed does not fall at all.

(b) alone counts money and not children. A spending share can shift because services are reclassified between the four groups rather than because fewer children are removed, and a prevention budget can grow while reaching nobody new. Spending is the input this document says is misaligned, so it is the right thing to watch, but it is not the outcome.

(c) alone counts paper. A published share and date would remove the specific defect described here, that the correcting clause has no number, and it would still say nothing about whether the money arrived, whether the organisations receiving it had the capacity to take on case management, or whether a single removal was avoided. The three have to be read together, and (a) has to be read against a stated counting basis before any movement in it means anything.

What is it connected to?

Fills with researchhousing and income support as upstream drivers of the neglect findings that lead to removal, the youth justice pathway that a separate document covers, the funding and capacity of the community controlled sector as a whole, and comparable child removal patterns for Indigenous children in other countries. Relation type and evidence grade were not established in this round.

What these sources do not say

• Why the two national counts disagree. The health and welfare series gives 19,987 at 30 June 2024 and a figure reported from the 2026 government services series gives 15,340 at 30 June 2025. No source opened here places the two side by side, states what each covers, or explains the difference. Detail tables in the second series are understood to exclude some jurisdictions, which would explain a gap of this size, but nothing opened here confirms that for these particular figures. Part of this is a limit of the research time available and part of it is that no reconciliation document appears to exist.
• Whether the placement principle is measured across all of its elements. The placement principle is described as having five elements, and a statement that several of them are not measured nationally appeared in search results, but the page carrying it refused automated retrieval and was never opened. The claim is recorded here as unverified and is used nowhere in this document.
• How far Priority Reform Two has actually got. No national aggregate of the share of Indigenous programme funding flowing through community controlled organisations was found, either as a current value or as a baseline. The commitment is worded qualitatively, so it is possible that no measurement basis was ever defined.
• What share of the 11.3 billion dollars is spent on Aboriginal and Torres Strait Islander children. No national figure was found. The 45 percent recorded by the New South Wales audit is a share of children in care, not a share of money, and the two must not be read as the same quantity.
• Whether 65 percent of removals under emergency powers is high or low. No source opened here gives the equivalent figure for any other jurisdiction, so that number has no comparison and cannot yet be read as an outlier or as normal practice.
• What the New South Wales figures of 6,500 and 6,563 have in common. They come from different bodies on different dates and neither states its counting basis, so the difference between them is not treated here as a change in the number of children.

See the evidence

No primary statistical table was opened in full. The agreement pages carrying Target 12 and Priority Reform Two are primary documents and were opened directly, and they are the basis for every statement here about what the two commitments do and do not contain. The Productivity Commission dashboard and the government services chapter page are the sources for the trajectory and for the spending split. The state audit report is the only document here that tested delivery against the placement principle and it covers one jurisdiction of eight, so every finding drawn from it is stated as a New South Wales finding and not as a national one. Where sources overlap they agree on direction: the dashboard, the health performance framework and the secretariat statement all record over-representation rising rather than falling, and the 2020 and 2025 comparisons in each move the same way. Where they disagree the disagreement is left visible rather than resolved — two national series give 19,987 at 30 June 2024 and 15,340 at 30 June 2025 with no document reconciling them, and two sources give 6,500 and 6,563 for New South Wales on different dates with no stated common basis. Four pages that would most likely carry the placement principle indicators and the campaign figures refused automated retrieval, and no claim in this document rests on any of them. This is a research-based definition, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 19,987–19,987 As of 2024-06-30

Sensitivity The interval has zero width and is not a confidence interval. It is a single published count on one date, and no source opened here restates it, projects it forward or attaches a margin to it. A second national count exists and was deliberately not used as a bound: a figure reported from the Report on Government Services 2026 series gives 15,340 at 30 June 2025, about a quarter lower one year later, and no document opened here places the two series side by side, states what each one covers, or explains the difference. Detail tables in that second series are understood to exclude some jurisdictions, which would explain a gap of this size, but that was not confirmed for these particular figures. Using 15,340 as a lower bound would therefore risk treating a count of fewer jurisdictions as a smaller measurement of the whole country, which is a category error rather than a conservative estimate, and it would run against every other indicator here, all of which move the other way. The direction of the correction is unknown until the two series are reconciled. Two further corrections run upward and neither could be computed from the sources opened. Children at risk of removal but not yet removed are excluded by construction, and no source opened here counts them, although the prevention spending this document describes is aimed at exactly that group. The families and communities on the other side of each removal are excluded as well. The count is also rising rather than stable on the measure that is tracked: the rate stood at 47.3 per 1,000 in the 2019 baseline year and 52.2 per 1,000 in 2025, and the ratio against non-Indigenous children widened from 9.6 to one in 2020 to 11.0 to one in 2024, so a count for any later date should be expected to be higher rather than lower. Against all of that the count measures exposure and not harm, and this document does not attempt to quantify harm.

Regional breakdown No source opened here gives a state and territory split of the national count of Aboriginal and Torres Strait Islander children in out-of-home care. Only one jurisdiction was measured in detail, New South Wales, where a state audit recorded 6,500 Aboriginal children in care at June 2023 and a peak body cited 6,563 in February 2026, on different dates and with no stated common basis. Splitting the national figure by population would be proportional allocation, and it would be wrong by a wide margin here, because removal rates differ enormously between Australian jurisdictions while the Aboriginal and Torres Strait Islander share of the child population differs on a different pattern again. The New South Wales figures are also not safe to subtract from the national total, because they sit on dates that match neither national series and because the two national series themselves disagree by about a quarter.