New Zealand reported meeting its 75 percent emergency-housing reduction target in December 2024, five years early — the target counts households in emergency housing, and the Ministry of Social Development said no agency had to tell it where households go after they leave

What is happening?

New Zealand provides emergency housing through a discretionary welfare payment, the Emergency Housing Special Needs Grant, administered by the Ministry of Social Development. The number of households receiving it is also a Cabinet-level numeric target. Government Target 8 is a 75 percent reduction from a baseline of 4,983 households in November 2021, to be reached by 2030, and the page of the Ministry of Housing and Urban Development that explains the target states it in those terms.

In December 2024 the government announced 591 households in emergency housing and said the target had been met five years early. Subtracting the two published figures gives a fall of about 88 percent from the baseline. A report in May 2026 put the national figure at about 471 households.

From August 2024 the Ministry of Social Development began applying an eligibility test that asks whether an applicant caused or contributed to the immediate need. Over the same period the share of applications declined rose sharply. One opened report gives a monthly decline rate of about 3 percent in March 2023 and 507 declines out of 1,602 applications in February 2025, about 31 percent. Another gives 402 declines out of 642 applications in urban Auckland in August 2025, about 63 percent. Whether the fall in the count was produced by the narrower gateway is disputed and no source opened here settles it — community organisations assert the connection, the report carrying that assertion records that the point is contested, and no opened government document addresses it.

Before the change took effect the Ministry of Social Development advised the responsible minister that narrowing access before housing supply increased carried a risk of more homelessness, more people living in vehicles and more crowded housing; that about half of emergency housing applicants are Māori and could be affected disproportionately; and that the change raised questions under the United Nations Convention on the Rights of the Child.

In June 2026 it became public that the reduction target had been written into performance agreements for regional managers at the Ministry of Social Development. A regional social-services provider criticised the arrangement in public. The ministry responded that the target had been given to it by the government and was accordingly part of how performance is measured. On 2026-08-06 the ministry issued a new performance letter with the regional reduction targets removed, directing managers instead toward sustainable housing for people currently in emergency housing.

Asked where households go once emergency housing ends, the ministry said about 80 percent move into social, transitional or private housing with government support, and that for the remainder no agency had an obligation to tell it where they go.

Whose problem is this?

The body that reports progress against the target is not the body that records what happens to a household after it is declined. The count travels upward on one line and the household leaves on another, and no opened document joins the two.

Where does this problem end?

Three adjacent value questions sit outside this frame and are named here so that they are not answered by implication — whether the overall level of housing assistance should rise or fall, whether reducing the emergency housing count is a desirable goal in itself, and whether the eligibility test is fair to the people it refuses. This document takes the target as the government stated it and asks only what that target measures and what it leaves unmeasured.

The boundary here separates the count that the target measures from the housing outcome that no published measure records.

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

Now

Should be

The government has stated one target and only one. Government Target 8 sets a 75 percent reduction from the November 2021 baseline by 2030, and the ministry page describing it frames the goal as fewer people in emergency housing. Measured against that stated target, the state now is ahead of schedule by the figures the government published.

No opened source states a target for what should happen to a household that is declined or that leaves emergency housing. There is no published figure the destination of departing households is meant to reach, no stated share of departures that should end in permanent housing, and no date by which such a measure is meant to exist. The gap in this block is not that the government set no target. It is that the target it set counts one thing, and nothing beside it counts the other.

How big is it?

A single figure for how many households this problem reaches cannot be derived from the sources opened here.

The counts that exist measure different things. 4,983 households were in emergency housing at the November 2021 baseline and 591 in December 2024, a fall of about 88 percent by subtraction of those two published figures, with about 471 reported for May 2026. Those are stocks at three moments rather than a count of the households that passed through, and no opened source publishes turnover, so the stock cannot be converted into a number of households affected.

The decline figures are monthly, and they count applications rather than households. February 2025 records 507 declines out of 1,602 applications nationally, and August 2025 records 402 out of 642 in urban Auckland. A household may apply more than once, no opened source states how many distinct households those monthly totals represent, and no opened source gives the months in between, so the monthly figures cannot be added across the period.

The destination figure has no denominator. The ministry said about 80 percent of departing households move into supported housing and that no agency had to report where the rest go, without stating the cohort or the period that share describes.

The one national measure of the wider outcome is a census estimate — 112,496 people, or 2.3 percent of the usually resident population, at 2023-03-07, against 99,462 and 2.1 percent in 2018. That count sits about 17 months before the August 2024 eligibility change, so it describes the period before the change rather than after it.

Under what conditions does it arise?

1. The target counts a category, not an outcome. Government Target 8 is expressed as a reduction in the number of households in emergency housing. A target that counts households in a category can be met by any route that lowers the count, and the target carries no companion measure that would tell the routes apart.

2. The eligibility test turns on a judgement made at the counter. From August 2024 an applicant may be refused on the ground of having caused or contributed to the immediate need. No source opened here identifies the statute, regulation or published guideline that carries the test, so the level at which it could be reviewed or amended is not established in this document.

3. The measure reached the people who apply it. As at June 2026 reduction targets appeared in performance agreements for regional managers, which placed the count in the same document as the assessment of the staff who decide applications. That arrangement was removed on 2026-08-06. The target it referred to was not.

4. Nothing downstream reports back. The ministry said no agency had an obligation to tell it where households go once emergency housing ends, so the system that produces the count receives no information from the far side of the decision.

5. The national measure of the wider outcome arrives once a census. The severe housing deprivation estimate is census-based and the most recent edition predates the change, so for the period since August 2024 there is no national figure against which the count can be read.

What has been tried?

The sequence is part of the finding. The narrower gate came first, the target was reported met four months later, the arrangement tying the count to manager performance became public after that, and the withdrawal of that arrangement came after the target had already been reported met. Removing a performance measure in August 2026 does not create a record of where the households that left or were refused before then went.

What was found?

Why is it still unsolved?

Incentive inversion — the quantity the system is steered by is the count of households in emergency housing, and the quantity that count stands in for is not measured at all.

A target that counts households in a category can be met by any route that lowers the count. Government Target 8 names a number of households and a date. Nothing in it distinguishes a household that reached settled housing from a household refused at the counter, or from a household that left with no recorded destination. A count of who is inside a category is a proxy for whether a need was met, and a proxy becomes the goal when it is the only thing reported upward.

The second part is that the proxy reached the decision. For a period that ended on 2026-08-06 the reduction target sat in performance agreements for regional managers, which put the count in the same document as the assessment of the people who decide applications. That arrangement is gone. The target is not, no published measure of destinations replaced it, and the structure that allowed the arrangement in the first place is intact — a single reported number, and no second number that could contradict it.

The third part is the missing return path. The ministry said no agency had an obligation to tell it where households go once emergency housing ends. A system that reports a count and receives nothing back cannot be corrected by its own data, because the data that would correct it is the data it does not collect. The measure that might serve instead is a census estimate whose most recent edition predates the change by about 17 months.

The fourth part is why the question stays open rather than being settled either way. Whether the fall in the count followed from the narrower gate is asserted by some and disputed in the same reports that carry the assertion, and it cannot be answered from published figures, because the figures that would answer it are the destination records that are not kept. A dispute that no dataset can close is a dispute that continues.

What observation would mean it is solved?

Candidates — (a) a published figure for where households go once emergency housing ends, covering all departures rather than only those placed with government support (b) a national measure of severe housing deprivation covering the period after August 2024, readable against the count (c) a stated target that names an outcome rather than a category count, with an instrument that measures it.

(a) alone is weaker than it looks. A destination field can be filled in on the day the grant ends and say nothing about where the household is a month later. It can also be completed for the placements the government makes and left blank for the rest, which is close to the position described now.

(b) alone is slow and coarse. Census-based estimates arrive years apart and move with definitions as well as with conditions. No opened source states whether the 2018 and 2023 estimates rest on the same method, so a difference between two editions cannot yet be read as a change in circumstances.

(c) alone counts paper. A target can name an outcome and still have no instrument behind it, in which case the reported number goes on being the only number. The three have to be read together, and the count by itself says nothing about any of them.

What is it connected to?

Fills with researchtransitional housing and the public housing register as the next steps for households that leave or are refused, the use of numeric reduction targets across other public services, and the design of destination reporting in comparable jurisdictions. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• Where declined or departing households actually went. No opened source carries a government figure. The ministry stated a share moving into supported housing and said no agency had to report the rest, and nothing opened here follows any household beyond the day the grant ended.
• Whether the fall in the count followed from the narrower gate. No opened source establishes the link, and the report carrying the assertion records that the point is contested. No opened government document addresses it either way.
• What legal instrument carries the eligibility test. No opened source states whether the test of causing or contributing to the immediate need sits in statute, in regulation or in internal guidance, so the level at which it could be reviewed is unknown to this document.
• The denominator behind the destination share. No opened source gives the cohort, the period or the total that the reported share of departing households describes.
• When the share of applicants who are Māori was measured. No opened source dates that figure or defines the applicant population it counts.
• Whether the risks named in advance occurred. The ministry named homelessness, vehicle living and crowding as risks before the change, and no opened source reports a government assessment of whether those risks materialised.
• What the withdrawal of the performance measure changed. No opened source reports what regional managers are measured on after 2026-08-06, or whether decline rates moved afterward.
• Whether the two severe housing deprivation estimates are comparable. No opened source states whether the method changed between the 2018 and 2023 editions.
• The months between the two decline figures. No opened source gives a continuous series of monthly applications and declines, so the two months reported here cannot be read as a trend on their own.

See the evidence

No primary document was read in full. The page of the housing ministry describing Government Target 8 and the monthly housing reporting page of the social development ministry were opened directly and are the only first-party government pages read here; the media release announcing that the target had been met was read in redistribution, because the page of the issuing government returned an empty body. The supplementary analysis report behind the 2024 eligibility change was not retrieved, so everything said here about the advice that preceded the change comes from a report describing that document. Where sources overlap they agree — the announcement that the target was met five years early appears in the redistributed release and in a separate news report, and the presence of reduction targets in performance agreements for regional managers appears both in the report that made it public and in the later report of its removal. Where they leave a gap the gap is left visible rather than filled: the share of applicants who are Māori carries no date, the share of departing households whose destination is unrecorded carries no denominator, and the national decline rate rests on two single months, March 2023 and February 2025, with nothing published in between. This is a Path A output, a definition built from public reporting rather than from collected observations, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range Not derivable

The reason and what is missing are listed under “What is missing” below