Australia promised 1.2 million new homes by June 2029 — 21 months in, 307,635 were built against a 420,000 benchmark and the national housing council forecasts a 220,000 shortfall

What is happening?

In August 2023 the National Cabinet of Australia raised an earlier target of one million new homes to 1.2 million new well-located homes over the five years beginning 2024-07-01. The agreement was signed by the Commonwealth, the eight states and territories, the national association of local government, nine superannuation funds, two institutional investors and three industry bodies. The target divides into 240,000 homes a year and 60,000 a quarter, and that division is the benchmark every later assessment of progress is measured against.

Twenty-one months in, the cumulative benchmark stood at 420,000 homes. Completions recorded to 2026-03-31 were 307,635, which is 112,365 short. Measured on commencements instead of completions, the same period recorded 331,903 against the same benchmark, a shortfall of 88,097. At the earlier eighteen-month mark, to the end of the December quarter of 2025, cumulative completions were 262,592 against a benchmark of 360,000, short by 97,408.

The quarterly figures behind those totals are not moving at the rate the benchmark requires. In the December quarter of 2025 the Australian Bureau of Statistics recorded 43,536 completions on a seasonally adjusted basis, down 1.7 percent on the previous quarter, against a quarterly benchmark of 60,000. Commencements in that same quarter were 53,567, up 8.0 percent.

On 2026-04-30 the National Housing Supply and Affordability Council published its assessment of the housing system for 2026 and forecast 980,000 homes across the accord period — 220,000 short of the target — with 1.2 million not reached until about September 2030, more than a year after the accord ends. That forecast is 42,000 higher than the 938,000 the council forecast in 2025. The same release named a rise in oil prices arising from events in the Middle East as a new downside risk to it.

Whose problem is this?

The body that set the target does not build the homes, and the parties that do build them signed an agreement that asks for a quantity without saying what happens if the quantity does not arrive.

Where does this problem end?

The boundary here runs between an agreed quantity of housing and the quantity actually delivered. It does not extend to whether the target was set at the right level, and it does not extend to the demand side of the housing market, which enters this document only where a source names it as a cause of the shortfall.

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

Now

What it should be

The agreement states the target itself, so no inference is needed here. It is 1.2 million new well-located homes completed in the five years to 2029-06-30, at a rate of 240,000 a year and 60,000 a quarter. What the agreement does not state is anything below that figure that would still count as compliance, and it names nothing that follows if the figure is not reached.

How big is it?

Measured in homes the gap is 112,365 as at 2026-03-31, and it is forecast to reach about 220,000 by 2029-06-30. The first of those is arithmetic, the difference between completions recorded and the benchmark the target implies. The second is a forecast published on 2026-04-30 and carries the uncertainty of a forecast rather than of a count.

Measured in people it is not derivable, and the population file for this document records that rather than a number. Every count opened here is a count of homes or of households, and no source supplies the term that would convert either one into people. The households that can be counted are the 254,571 on social housing waiting lists in 2024-25, of which 122,457 were recorded as being in greatest need, a figure 12 percent above the year before. Those are households with an open application in the one part of the housing system that keeps a queue, which is narrower than everyone affected by a shortfall in market supply, and the greatest-need figure reaches this document through a secondary report rather than from the commission that publishes the series.

The national population was 27,801,023 at 2025-12-31. That figure bounds any count of people affected and does not produce one, and dividing it by anything above would be invention.

The two quantities that can be counted answer different questions and are not added together anywhere in this document. The homes figure says how far behind the agreed rate of building is running. The households figure says how many applications are open in the segment served by the fund that had completed 1,432 homes of a 40,000 target as at 2026-05.

Under what conditions does it arise?

1. The capacity to build sits below the agreed rate. A national residential building industry body forecast a shortfall of 15 to 20 percent in key trades across the major cities by the end of 2026. Average development approval times in those cities run 90 to 120 days. Building materials costs in the first quarter of 2026 were 5 to 8 percent above the same quarter a year earlier on figures from a commercial data provider. A target can raise the priority of building but it cannot by itself supply a carpenter or shorten an assessment queue.

2. The cause named by the responsible minister sits years upstream. The federal housing minister attributed the delay to a fall in approvals several years earlier compounding with construction cost rises and labour shortages after the pandemic. Approvals granted before the accord began determine how much can be completed inside it, so the first years of a five-year target inherit a pipeline nobody can change during the period.

3. The financial lever runs one way only. The New Home Bonus of 3 billion Australian dollars pays jurisdictions that exceed their allocation under the target. It carries no matching provision for jurisdictions that fall short, and the agreement itself sets out no consequence for missing the target. With no jurisdiction rated on track at 2026-03, an instrument that rewards outperformance has nothing to operate on.

4. Delivery is split across governments that did not each sign the same obligation. The target was set by the Commonwealth and the eight states and territories. Planning and zoning powers sit with the states and territories, most development approvals sit with local government, and the building itself is done by private industry and community housing providers. Each layer can be doing what it agreed to do while the national total still lands 220,000 homes short.

What has been tried?

Three of these are money and one is a target, and the money is small beside the quantity being asked for. The fund covers 40,000 homes of the 1.2 million, which is about one in thirty, and its own completion count stands at 1,432. Everything else in the target has to be delivered by parties the agreement encourages rather than obliges.

What was found?

Why is it still unsolved?

Supply shortage — the target names a quantity of homes that the industry asked to build them has not been able to produce at that rate, and the agreement that names the quantity does not say what happens when it is not produced.

The first part is a capacity constraint and it is visible in three separate measures at once. Trades are forecast 15 to 20 percent short across the major cities by the end of 2026, approvals in those cities take 90 to 120 days on average, and materials cost 5 to 8 percent more in early 2026 than a year earlier. None of the three is something a signature can change inside the five years. The responsible minister points further upstream still, to a fall in approvals granted before the accord period began, which fixes part of the completion count for the first years of the target before the target starts running.

The second part is that this is an agreement to try. The one financial lever attached to the target, the New Home Bonus, pays jurisdictions that beat their allocation and says nothing about jurisdictions that miss it, and the agreement itself carries no penalty, no correction step and no review that triggers on a shortfall. In a period when no jurisdiction is rated on track, an instrument that only rewards outperformance has nothing at all to operate on. The absence here is deliberate in form rather than accidental in drafting, and it was read directly on the agreement page rather than inferred from silence in reporting.

The third part, which is what keeps the pattern stable, is that the shortfall arrives as a forecast and never as an event. The 2025 forecast of 938,000 implied a gap of 262,000 against the target; the 2026 forecast of 980,000 implies 220,000. The number improved by 42,000 in a year and stayed far below the line, and an improving number produces no moment at which anyone has to act. The target itself has not moved and the date has not moved, so the only thing that changes each year is an estimate of how far short the finish will be.

What observation would mean it is solved?

Candidates — (a) cumulative completions reach 1.2 million within the five years to 2029-06-30 (b) the council forecast for the accord period converges on the target rather than sitting below it (c) every state and territory is rated as on track against its own allocation.

(a) is the only one that counts homes, and it still counts the wrong thing on its own. A completion is a dwelling finished. The target says well-located, and no source opened here states how that condition is tested or whether any completion has ever failed it. A target met by 1.2 million dwellings in places nobody needed them would satisfy the arithmetic exactly.

(b) alone is weaker still. A forecast is a projection over inputs that move, and the same release carrying the 980,000 figure named an oil price shock as a risk to it. A forecast that rises to meet the target is a change in an estimate, not a change in the housing stock, and this document has already recorded one such rise of 42,000 homes that left the shortfall at 220,000.

(c) counts jurisdictions rather than homes. Allocations are set per jurisdiction, so every jurisdiction can be on track against its own allocation while the national total still misses, and the reverse is possible too. The queue that the fund was built to serve needs its own observation, and at 1,432 homes of 40,000 as at 2026-05 that observation is not close to being made either.

What is it connected to?

Fills with researchplanning and zoning law across the states and territories, migration and household formation on the demand side, insolvency and training pipelines in the residential construction industry, and comparable national housing targets in other countries. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• Which jurisdiction sits where in the forecast range. The council states that expected target dates run from September 2029 to beyond 2034, and the table naming which jurisdiction falls where was not readable. The report was received as a file but its text could not be extracted, so the spread is known and its distribution is not.
• Why three forecasters disagree. The final shortfall is put at 220,000 by the council, 190,000 by one national industry body and 166,000 by another. No source opened here reconciles the three, and no procedure exists in anything opened that would require them to be reconciled.
• What follows from missing the target. This is a silence read directly rather than a gap in research. The agreement page carries no penalty, no correction mechanism and no review triggered by a shortfall, and the one payment attached to the target is available only to jurisdictions that exceed their allocation.
• How the greatest-need figure is derived. The commission page states that it does not publish greatest need as a composition of the waiting list. The figure of 122,457 used above comes from a secondary report citing the commission, and the method behind it could not be reconfirmed in this round.
• Whether the 2026 forecast holds. The release that raised the forecast by 42,000 homes named an oil price rise arising from events in the Middle East as a downside risk in the same document, and did not quantify it.
• What the 1,432 fund completions consist of. One source describes many of the early completions as purchases of homes already under construction rather than new builds, and gives no split. The share that represents housing that would not otherwise exist is therefore unknown.
• What the audit found. An audit of the design and delivery of the fund was reported as due around June 2026. Its result was not confirmed at 2026-08-15.
• How many people stand behind the household counts. Every count opened here is of homes, dwellings or households. Nothing states an average household size for the waiting list or for the shortfall, which is why the population file for this document records the estimate as not derivable.

See the evidence

No primary statistical microdata was read here, and one primary policy document was. The Treasury accord page was opened directly, and the finding that the agreement carries no consequence for missing the target rests on that reading rather than on any report about it. The Australian Bureau of Statistics release for the December quarter of 2025 and the council release of 2026-04-30 were also opened directly and carry the quarterly figures and the forecast respectively. Everything else is professional, industry or trade reporting of those series. Where sources overlap they agree: the 220,000 forecast shortfall appears in the council release and in secondary reporting of it, the 1,432 fund completions appear alongside the contracted totals on the agency funding page, and the finding that no jurisdiction is rated on track appears in industry reporting of the council quarterly report. Where they disagree the disagreement is left visible rather than resolved — the final shortfall is put at 220,000, 190,000 and 166,000 by three different bodies with no reconciliation between them, and the greatest-need figure of 122,457 is reported against a commission page that states it does not publish that composition. Five rows above carry no URL, and each names why. None of the five was deleted and none was replaced with a substitute link. This is a Path A output, 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