In the year to the March quarter of 2026, 199,234 dwellings were approved in Australia and 197,456 were begun: 99.1 starts for every 100 approvals. Over the 42 years to 2025 the same comparison gives 97.0 starts per 100 approvals, and the rate has drifted up rather than down: 94.9 in the ten years to 1993 against 99.3 in the ten years to 2025. In 41 of the 169 quarters since the start of 1984, more dwellings were begun than approved. That is the clearest sign that this ratio measures timing rather than projects abandoned.
Approvals and commencements are separate collections with different timing: a dwelling approved in one quarter is usually begun in a later one, and the two are not matched project by project. The ratio between them is a timing measure, not a count of projects that did or did not proceed.
Two collections, not one pipeline
Dwelling approvals come from councils and private certifiers, and the ABS counts them in the month the approval is given. Dwelling commencements come from a survey of building activity, and count a dwelling when site work begins. A dwelling approved in March may be begun in June, or in the following year, or not at all; and a dwelling begun this quarter was approved in some earlier one.
The two collections are therefore not matched to each other. Comparing them gives a useful ratio at the national level over a year or more, but it does not say which approvals proceeded. Nothing in either collection tracks an individual approval through to a start, so the share of approvals that lapse is NIL RETURN on this data.
Over a year, nearly every approval becomes a start
Across the 42 calendar years from 1984 to 2025, 7,184,083 dwellings were approved and 6,965,210 were begun, which is 96.95 starts for every 100 approvals. The annual ratio has ranged from 91.5 in 1988 to 105.8 in 2018, and in six years it exceeded 100: 2000, 2011, 2018, 2021, 2023, 2025.
The drift in it runs the opposite way to the one usually assumed. The ten years to 1993 averaged 94.9 starts per 100 approvals; the ten years to 2025 averaged 99.3, 4.4 points higher. Part of that may be the two collections themselves: both have been revised over four decades, and commencements are estimated from a survey rather than counted. What the figures do not show, on any reading, is a widening shortfall of approvals that never become starts.
Dwellings approved and dwellings begun, thousands, calendar years 1984 to 2025
Sources: ABS Building Approvals, Australia (approvals, monthly aggregated to quarters) and ABS Building Activity, Australia (commencements). Both series are total dwelling units, original.
Dwelling commencements per 100 dwelling approvals, calendar years 1984 to 2025
Sources: as above. The dotted line marks one start for each approval; readings above it are years in which starts drawn from earlier approvals outnumbered the approvals given that year.
Quarter by quarter, the ratio is noise
The quarterly picture is far less stable than the annual one. Across the last twelve quarters the ratio has run from 89.9 to 106.3, a spread of 16.4 points, with no pattern to it: the December quarter of 2025 recorded 106.3 starts per 100 approvals, and the December quarter of 2024 recorded 89.9.
Some of that is seasonal, some is the gap between approval and start moving by a few weeks either way, and some is survey sampling. Nothing in it supports a quarterly reading of how many projects are going ahead, which is why this piece reports the annual figures and treats the quarterly ones as noise.
Dwelling commencements per 100 dwelling approvals, the last twelve quarters
Sources: as above. The dotted line marks parity. Quarterly readings mix the timing of two different collections and should not be read as a project-level conversion rate.
| Year | Dwellings approved | Dwellings begun | Difference | Starts per 100 approvals |
|---|---|---|---|---|
| 2016 | 234,869 | 234,113 | 756 | 99.7 |
| 2017 | 224,609 | 216,640 | 7,969 | 96.5 |
| 2018 | 211,539 | 223,795 | −12,256 | 105.8 |
| 2019 | 176,611 | 175,263 | 1,348 | 99.2 |
| 2020 | 187,365 | 182,998 | 4,367 | 97.7 |
| 2021 | 228,995 | 229,378 | −383 | 100.2 |
| 2022 | 191,664 | 182,441 | 9,223 | 95.2 |
| 2023 | 164,332 | 165,856 | −1,524 | 100.9 |
| 2024 | 174,362 | 169,607 | 4,755 | 97.3 |
| 2025 | 196,832 | 197,186 | −354 | 100.2 |
Sources: as above, calendar years. The difference column is approvals less commencements in the same year; a negative difference means more dwellings were begun than approved that year.
What the gap does measure
APN reads approvals, commencements and the difference between them as Building Approvals & Commencements (21510). At the March quarter of 2026 that composite stands at +0.0034σ, a position of 80.0 against the 171 quarters of the series, and its three parts pull in different directions: approvals at −0.1521σ and commencements at −0.3613σ are both slightly below their 2011–2025 baselines, while the difference between them, at +0.5237σ, is above its own.
In dwellings, that difference was 2,385 in the March quarter of 2026, against a 2011–2025 average of about 770. A larger difference in a single quarter is consistent with approvals arriving faster than builders can start, and equally consistent with the timing of the two collections drifting apart. The quarterly series cannot separate those, and the annual ratio above shows no accumulation of unbuilt approvals over time.
What this measure does and does not show
Three limits. The two collections differ in coverage as well as timing: approvals include dwellings that are later amended or abandoned, and commencements are estimated from a sample survey with its own revisions. The comparison is national and total; the ABS tables behind these series do not support a house-against-apartment split at this level, so that question returns NIL RETURN here. And both series are original, not seasonally adjusted.
What the comparison does show is a national conversion rate that is close to complete, that has drifted up rather than down across four decades, and that is unhelpful read quarter by quarter.
Findings are presented on the basis of data and evidence alone.
