Australia Has Nearly Enough Homes Under Construction to Hit Its Housing Target. It Finishes 172,928 a Year.

Australia’s housing debate has one shape. Not enough homes are being built, and the reason is that they are too hard to get approved: councils block density, planning systems take years, objectors win. Nearly every lever governments have reached for points at that stage of the process — planning reform, approval-time targets, density mandates near railway stations, a national scorecard for how fast councils tick applications through.

The Australian Bureau of Statistics publishes the numbers needed to test where the homes are actually being lost. It counts dwellings at three separate stages: approved, commenced, and completed. It also counts the stock sitting part-built at the end of every quarter.

Follow all four and the shortfall shows up in one place, and it is not the approval counter. In the year to March 2026, 198,307 new dwellings were approved and 196,618 were commenced. Only 172,928 were finished.

A record pipeline producing fewer homes#

At the end of the March 2026 quarter, 242,874 dwellings were under construction in Australia. That is the highest figure in the entire series, which begins in 2004. There has never been more housing on site.

The year of peak completions was the year to March 2017, when 220,650 homes were finished. The stock under construction at that point was 217,108 — about 26,000 fewer than today.

So the pipeline is bigger now than it was then, and it is producing 47,722 fewer finished homes a year, a fall of 21.6%.

The measure: how long the pipeline takes to clear#

Divide the dwellings under construction by the completions of the preceding year and the result is a duration: at the completion rate then running, how many years it would take to finish everything currently on site. Call it the build-out ratio.

It is a ratio of a published stock to a published flow, not a stopwatch on any individual building. It rises when the stock grows, when the completion rate falls, or both.

Across the five financial years to June 2019 it averaged 1.03 years. In the March 2026 quarter it was 1.404 — a rise of 36.5%, and within 1% of the series record of 1.413 set in the September quarter of 2022.

The obvious objection is composition. Apartments take longer to build than houses, so if the mix has tilted toward apartments the combined ratio would rise even if nothing had got slower. That is testable, and it does not hold. Houses went from 0.57 years to 0.834, a rise of 45.7%. Other residential went from 1.63 to 2.381, a rise of 46.0%. Both categories slowed, by almost exactly the same proportion. The mix is not doing the work.

The second objection is the baseline. The five years to 2019 were a construction boom, which could make an unusually fast period the comparator. Recomputing across every pre-pandemic quarter on record — 2004–05 through 2018–19 — gives a slower baseline of 0.90 years, against which the current reading is 55.3% higher, not 36.5%. The conservative baseline makes the finding larger. The one used throughout this piece is the cautious choice.

Where the funnel leaks#

Splitting the pipeline into its two transitions locates the loss precisely.

Commencements per approval were 0.991 in the latest year, against 0.986 in 2017–18. That conversion has not deteriorated at all; it is fractionally better. Approved homes are being started at very close to the same rate as before.

Completions per commencement fell from 0.928 to 0.880 over the same comparison. That is the stage giving way.

This does not mean planning systems work well, and nothing here measures how long an application takes to determine. What it does mean is that the volume of approvals is not what is capping the number of finished homes. There were 25,379 more approvals than completions in the latest year. Adding approvals to a pipeline that is already at a record and clearing more slowly than it ever has would lengthen the queue.

What that costs, in homes#

The National Housing Accord commits to 1.2 million new well-located homes over the five years from 1 July 2024 — 240,000 a year. Completions are running at 172,928, or 72.1% of that rate.

Now hold the pipeline where it is and change only the speed at which it clears. At the 1.03-year build-out ratio of 2014–15 to 2018–19, a stock of 242,874 dwellings under construction corresponds to 236,087 completions a year.

That is 63,159 more homes a year than Australia is currently finishing, and it lands within 2% of the Accord target.

This is arithmetic, not a forecast. It is the same division performed with a different denominator, and it assumes the pipeline stays exactly the size it is now. It is not a claim that returning to 2019 build speeds is achievable, or that the Accord would be met if it happened. What it establishes is narrower and harder to argue with: the volume currently under construction is not the thing standing between Australia and its housing target. The rate at which that volume converts into finished homes is.

Two explanations that only go part of the way#

The natural culprit is the cost shock. Construction output prices rose steeply from 2020: the ABS producer price index for house construction is 48.3% above its December 2019 level.

But the timing does not line up cleanly. The build-out ratio peaked in late 2022, recovered substantially through 2024 — down to 1.195 by the December quarter — and then gave the entire recovery back over 2025 and early 2026, climbing to 1.404. Prices rose fastest during the period when the ratio was improving, and rose only gradually during the period when it deteriorated again. A cost shock that faded three years ago does not explain a pipeline that is jamming harder now than it was in 2024.

The second culprit is builder collapse, and it is real but smaller than it looks. Construction firms entering external administration for the first time rose from 2,568 in 2021–22 to 6,944 in 2025–26, a factor of 2.7. Construction is consistently the largest single industry in the insolvency count.

The complication is that every industry rose at the same time. Across all industries the count rose by a factor of 2.9 — slightly more than construction. Construction’s share of insolvencies did not increase; it went from 26.1% to 24.5%. And the 2021–22 starting point sits inside the pandemic support period, when insolvencies were being suppressed everywhere, so the multiple overstates the underlying change.

Builder failures are a genuine mechanism for stalled sites, and they are running at high absolute levels. But an insolvency wave that is economy-wide, and in which construction’s share slightly fell, is not on its own an explanation for something specific to housing completions.

The prediction that failed#

Seven hypotheses were written down before the numbers were computed, in HYPOTHESES.md in this investigation’s analysis folder. One of them failed.

The expectation was that the slowdown would be uneven enough that at least some states had escaped it — which would have meant a national construction shock was an incomplete story and something state-specific was at work. Every one of the eight states and territories got slower. Not one escaped.

What survives of the prediction is the spread, which is enormous. Western Australia’s build-out ratio rose 87.1%, from 0.717 years to 1.342. Queensland rose 82.3%. Victoria rose 6.1%, from 1.061 to 1.126 — close enough to flat that it barely registers as a change. A single national explanation has to account for a fourteen-fold difference in how hard the same period hit different states.

What could explain this instead#

The stock is inflated by projects that will never finish. If a share of the 242,874 dwellings under construction are on stalled or abandoned sites, the build-out ratio is measuring a pipeline that is partly fictional, and the “years to clear” reading overstates how long live projects take. The ABS counts a dwelling as under construction until it is completed or formally abandoned, and does not publish an abandonment series, so this cannot be tested from the published data. It would not rescue the front-of-pipe story — homes on dead sites are not finished homes either — but it would change the mechanism from “building is slower” to “more building is failing”.

A composition shift within apartments. The ABS publishes unit counts only for houses and total other residential, not for the finer splits between low-rise and high-rise apartments. A tilt within “other residential” toward taller buildings, which take longer, would raise that category’s ratio without anything getting slower at a given building height. This is a genuine limit on the finding. It cannot explain the 45.7% rise for houses, which have no equivalent internal mix.

Labour and materials constraints that prices do not capture. Output price indexes measure what construction sells for, not whether a crew or a component is available. A site can be fully funded, correctly priced and still idle for want of a trade. Job vacancy and skilled-migration data would bear on this; neither is analysed here, and the piece makes no claim about which input, if any, is binding.

Interest rates and finance conditions. Higher rates raise holding costs and can stall projects between approval and completion independently of construction cost. Rate movements are not tested here, and their timing has not been matched against the build-out series.

The comparison year was exceptional. The year to March 2017 sits at the top of the largest apartment boom in Australian history. Comparing today against a record is a hard test by construction. But the finding does not rest on that single comparison: the build-out ratio is measured against a five-year average, and against a fifteen-year average that makes it look worse.

Revisions. ABS building activity data is revised, and recent quarters more than older ones. The March 2026 reading may move. The multi-year deterioration across dozens of quarters does not depend on the latest point.

The homes Australia says it needs are, to a remarkable degree, already being built. A record 242,874 of them are standing part-finished right now — frames up, slabs poured, cranes over them. The number is not the problem. What has changed is how long they sit there, and on that measure the country is close to the worst it has been in twenty-two years of published data. Every additional approval joins the back of that queue.