New Zealand has no law saying who acquires a home in a hazard zone or who pays for it — the Climate Change Response Amendment Bill introduced on 2026-07-15 does not address funding

What is happening?

New Zealand has no statutory framework for managed retreat — no law that says who decides a home can no longer be lived in, who acquires it, and who pays. The Climate Change Response (Zero Carbon) Amendment Act of 2019 made a national climate change risk assessment a duty every six years. It does not deal with retreat and it does not deal with funding.

A separate Climate Adaptation Act was signalled in 2020, with introduction targeted for the end of 2023. It was not introduced. After the change of government in December 2023 the approach changed, and in April 2024 the government asked the Finance and Expenditure Committee of Parliament to run an inquiry into climate adaptation on narrowed terms of reference.

That committee reported on 2024-10-01. It described its own recommendations as vague, open to differing interpretations and at times contradictory, and it recorded a lack of visibility over the investment plans of government agencies and of local government. Who pays was not settled.

The second national climate change risk assessment, released 2026-05-07, records that there is no legal framework for managed retreat and no national approach to displacement and relocation.

On 2026-07-15 the government introduced the Climate Change Response Amendment Bill. It would require councils to produce climate adaptation plans looking thirty years ahead — within five years of the adoption of a regional spatial plan for priority areas, within ten years for the rest, reviewed every ten years. A legal summary of the bill records that it does not address how adaptation work is to be funded, so the uncertainty continues. The government has said the cost-sharing decision falls to the next term of government. The bill was expected to reach select committee before Parliament rises on 2026-09-24, and passage before the November 2026 election is not expected.

Meanwhile the retreat has already happened, without a law. After Cyclone Gabrielle in February 2023 about 3,000 properties in Hawke's Bay alone were placed in risk categories, and the most severe class covered homes assessed as unsafe to occupy. That process ran on Cabinet decisions and on agreements negotiated with two district councils, and about NZD 100 million of public money had gone into buyouts, relocation grants and demolition by 2025-11-27.

Whose problem is this?

The body that would have to write the plans is not the body that decides the money, and the bill introduced in 2026 keeps that split in place. Every party in the table above can act correctly under the rules that exist and still leave a house in a hazard zone with nobody obliged to buy it.

Where does this problem end?

The boundary matters because the hazard maps already exist and the money question does not have an owner. What is missing here is not a map and not a warning but a rule that says who buys the house.

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

Now

Needs a new measurementthe target state for funding: no source opened here gives a rule for who pays for a retreat, a formula for sharing that cost, or a date by which either will exist. The 2026 bill sets deadlines for plans and no deadline for money, and the reported government position is that the cost-sharing decision falls to the next term of government, with no date attached.

How big is it?

About 219,000 residential properties. That is the count of dwellings standing inside mapped coastal inundation or inland flood zones, published by the Ministry for the Environment and Stats NZ in Our Marine Environment 2025 on a 2023 basis, with a combined value of about NZD 180 billion. The same publication projects that about 1,300 homes in the coastal inundation zone will take more than 20 percent damage in an extreme weather event between 2026 and 2060, at a value of about NZD 900 million.

The two numbers answer different questions and neither is a retreat count. The first says how many dwellings stand in a zone that a hazard map already marks. The second says how many are expected to be hit hard enough within the next thirty-five years for the question of moving to arise at all.

No source opened here converts either figure into a number of people. Household size is not given anywhere in the opened sources, so the unit stays as dwellings and no multiplication is performed.

In one region the counting has already been done a different way, after the event rather than from a map. About 3,000 properties in Hawke's Bay were placed in risk categories following Cyclone Gabrielle, the most severe class was reported as 287 in the final 2023 count and as 327 in 2026 reporting, and 167 properties were eligible for a voluntary buyout.

Under what conditions does it arise?

1. The risk is assessed on a statutory clock and the response is not. The 2019 amendment requires a national risk assessment every six years. Nothing in that law requires a plan for moving people, and nothing attaches money to a finding of risk. The assessment released in 2026 records the same absence that the first one had no power to close.

2. The 2026 bill separates the duty to plan from the power to pay. Councils would have to write thirty-year adaptation plans on a five-year or ten-year clock depending on priority, and the legal summary of the bill records that funding is not addressed. A council that identifies a street to retreat from would hold a document and no budget line.

3. A disaster forces the decision before the law arrives. After February 2023 the categorisation and buyout in Hawke's Bay ran on Cabinet decisions and on agreements negotiated with two district councils, with the cost split evenly between central government and those councils. It worked once, for one region, and it left no rule behind for the next region.

4. The temporary arrangement is being withdrawn without a successor. From December 2025 the lowest category was discontinued and the most severe class is being phased out as flood risk modelling moves into district plans. The regional body running that system said that after the phase-out the decision falls to each council on its own.

What has been tried?

Two routes were tried in sequence and neither reached the money. The first was a commitment to legislate, which ran from 2020 to 2023 and produced no bill. The second was an inquiry, which produced a report whose recommendations the committee itself described as open to differing interpretations. The third route, the one that actually moved households, was not a route at all but a negotiation held after the flood water had gone down.

What was found?

Why is it still unsolved?

Institutional gap — the duty to assess the risk exists, the duty to plan is arriving, and the duty to pay has never been written.

A risk assessment can name a hazard but it cannot buy a house. That is the difference between the instrument that exists and the instrument that does not.

The 2019 law put the assessment on a six-year clock, so the finding of risk arrives on schedule. Nothing arrives with it. The 2026 bill adds a second scheduled document, the council adaptation plan, and leaves the transaction that such a plan would trigger without a payer. A plan that identifies a street to move, and cannot fund the move, is a restatement of the problem in an official format.

The second part is that a deferral looks the same from outside as work in progress. The act signalled in 2020 for the end of 2023 was never introduced, and no source opened here carries an on-the-record reason. The 2024 inquiry reported, and the committee wrote that its own recommendations were open to differing interpretations. The 2026 bill arrived seven years after the assessment duty and states that funding is out of scope. Each step is real, none of them closes the question, and the sequence reads as progress.

The third part, which is the one that keeps the pattern stable, is that the improvised route worked well enough to remove the pressure. After February 2023 homes were acquired, money moved and households left, all without a statute. An outcome that arrives anyway makes the missing law look optional. It is not repeatable at national scale: the arrangement covered one region, ran on negotiated agreements between three parties, cost about NZD 100 million for 167 eligible properties, and is now being wound down with each council left to decide on its own.

What observation would mean it is solved?

Candidates — (a) an act in force stating who decides that a property must be vacated, who acquires it, and how the cost is shared (b) a published cost-sharing rule with a commencement date (c) council adaptation plans in force whose first steps are funded rather than only scheduled.

(a) alone is weaker than it looks. An act can commence with thresholds high enough, or an acquisition power discretionary enough, that no household reaches it. It can also be amended or repealed by a later Parliament, which turns a solved observation back into an unsolved one with nothing having happened on any street.

(b) alone counts a formula, not a payment. A rule that assigns councils a share larger than the rates base can carry is a rule that will not be used. The even split that ran after 2023 covered two districts with a single declared event behind it, and no source opened here says whether that ratio is affordable anywhere else.

(c) alone counts documents. Plans on a five-year and ten-year clock will exist because the bill would require them. Whether a plan moves a house is a separate observation, and the one region that has already moved houses did it without a plan of that kind.

The three have to be read together, and none of them can be observed before the next Parliament, because the reported government position is that the cost-sharing decision falls to the next term.

What is it connected to?

Fills with researchwithdrawal of flood insurance cover from hazard zones, the rating capacity of small councils, the resource management reform programme that produces the regional spatial plans this bill keys its deadlines to, Māori land and settlement land inside hazard zones, and managed retreat law in other jurisdictions. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• Why the promised act was never introduced. No source opened here carries an on-the-record reason from the government for the absence of a Climate Adaptation Act by the end of 2023, or for the decision to replace it with a narrowed committee inquiry.
• Why the cost-sharing decision was deferred. Reporting records the position that it falls to the next term of government. Nothing opened here gives a reason for the deferral or a date on which it ends.
• How many properties were in the most severe category. One opened source gives 287 as the final 2023 count for the region and another gives 327 in 2026 reporting. Neither acknowledges the other, and neither states its counting basis or cut-off date, so it is not known whether the difference is a later addition, a different set of councils, or a correction.
• What happened in the other two regions. Auckland and Tairāwhiti also ran category-based buyouts after the same 2023 storms. This round confirmed only that those programmes existed and did not open their pages or obtain any counts, so no national buyout total appears anywhere in this document.
• How far the Hawke's Bay buyout has been completed. The council page that carries a running completion count returned HTTP 403 to automated fetching, and no completion rate for 2025 or 2026 was obtained.
• What happens to a property that was categorised but not acquired. The categories are being withdrawn from December 2025 and no opened source says how a household mid-process is treated after that.
• The primary legislation. No bill text or statute was read in the original. The New Zealand legislation site did not open to automated fetching in this round, so the description of the bill rests on a law firm summary and on news reporting, and the 2026 risk assessment is likewise read through a law firm summary rather than through the assessment itself. No section or clause number is cited anywhere in this document, and none should be inferred from it.
• The proposed funding models. A cost-sharing approach based on who benefits was mentioned in search results, attributed to an independent expert group. No opened source sets out its terms, so nothing is said here about what it would require of anyone.
• The position of the institutions named. This one is unusually well covered and the exception is worth stating precisely. The bill, the committee report, the risk assessment and the ministry publication are all documents produced by the government and by Parliament, so the position of those institutions is carried in this document in their own words rather than through a critic. What could not be found is a stated reason on the two points where the record simply stops: the missing act and the deferred cost-sharing decision.

See the evidence

No primary legal document was read in this round. The two government sources opened directly are the Ministry for the Environment page describing the 2019 assessment duty and the Our Marine Environment 2025 chapter carrying the exposure figures; everything about the 2026 bill and the 2026 risk assessment comes from law firm summaries and from news reporting that cite them. The New Zealand legislation site did not open to automated fetching, which is why no section number appears anywhere above. Where opened sources overlap they agree: the NZD 180 billion exposure value appears in two, and the finding that the bill leaves funding unaddressed appears in a legal summary and in reporting of the deferral to the next term of government. Where they disagree the disagreement is left visible rather than resolved — the most severe category is 287 in the 2023 final count and 327 in 2026 reporting, with neither source stating its basis. Four rows carry no URL because automated fetching returned HTTP 403 or an empty body, and those rows are kept rather than deleted so that the gap is visible; no claim in this document rests on any of them. This is a Path A output, research-based rather than observation-linked, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 1,300–219,000 As of exposure count on a 2023 basis; damage projection for 2026 to 2060

Sensitivity This is not a confidence interval and the two ends are not two estimates of one quantity. The ceiling is a stock of dwellings inside a mapped zone at a point in time; the floor is a projection of dwellings expected to be damaged past a severity threshold over a thirty-five year window. They are not nested either, because the floor is restricted to the coastal inundation zone while the ceiling also includes inland flood zones, so a home can sit in the ceiling and be structurally ineligible for the floor. No arithmetic between the two ends is meaningful and none is performed here. The unit is dwellings, not people. A person count was deliberately not derived: no source opened here gives an occupancy figure, a household size, or an owner-occupier versus tenant split for either set, so multiplying by any average would invent precision that no opened source supports. Two corrections would move the interval in opposite directions and neither can be computed from the opened sources: downward, because being inside a hazard zone is not the same as being a retreat candidate and most of the 219,000 will never be acquired under any framework; upward, because the floor covers only one hazard type and only one damage threshold, and because the projection window ends in 2060 while sea level exposure continues past it. A third quantity exists at a different scale and is not part of this chain: about 3,000 properties were categorised in one region after Cyclone Gabrielle in 2023, of which the most severe class was reported as 287 in the final 2023 count and as 327 in 2026 reporting, and 167 were eligible for a voluntary buyout. Those figures are post-event assessments of a single region rather than hazard-map exposure, so they measure a different thing and are excluded from the chain rather than averaged into it.

Regional breakdown No source opened here decomposes the national count of 219,000 properties by region, territorial authority or coastline segment. The only sub-national figures available are the post-cyclone categorisation counts for one region, and those rest on a different basis entirely: they come from case-by-case damage assessment after a declared event rather than from hazard-zone mapping, so they cannot be subtracted from or fitted into the national total. Counts for the other two regions that ran category-based buyouts after the same 2023 storms were not obtained in this round. Splitting the national figure by population would be proportional allocation, and it would be wrong in a known direction here, because flood and coastal inundation exposure follows river plains and shorelines rather than where people live.