Published 16 September 2026
In 2017 half of new three-bedroom buyers came from an HDB address, and by 2025 it was under a third
By Realila
Of everyone who bought a new-launch three-bedroom in 2017 and whose address type the caveat records, 52.1% were buying from an HDB flat. In 2025 that was 28.9%.
Resale fell far less: 36.9% to 27.2%. The two legs crossed in 2022, and for three years the HDB buyer was under-represented at a new launch rather than over-represented. The rest of the three-bedroom picture is on the three-bedroom condo guide: what it costs, what it rents for, what it yields and how big it is.
Key takeaways
- Who was buying. New-launch HDB share fell 23.2 points against resale's 9.7, and the two legs crossed in 2022. Who was buying
- Smaller homes, more upgraders. The HDB buyer is more present the smaller the home, in every year and both legs, without exception. Smaller homes, more upgraders
- Where they went. The retreat is a new-launch phenomenon. Four-bedroom resale barely moved. Where they went
- What the two legs returned. Buying at launch returned 3.24% a year against resale's 3.95%. What the two legs returned
Who was buying
Share of buyers giving an HDB address, of those whose address type is recorded.
- New launch
- Resale
| Year | New launch | Resale |
|---|---|---|
| 2017 | 52.1% | 36.9% |
| 2018 | 49.7% | 36.8% |
| 2019 | 51.5% | 33.6% |
| 2020 | 46.9% | 37.3% |
| 2021 | 39.5% | 34.8% |
| 2022 | 32.3% | 34.3% |
| 2023 | 26.3% | 29.6% |
| 2024 | 28.2% | 31.6% |
| 2025 | 28.9% | 27.2% |
For five years the new-launch share ran well above resale. In 2022 it went below, stayed below through 2024, and in 2025 it is above again. A buyer moving out of an HDB flat used to be a good deal more likely to choose a new launch than an existing one. For three years they were slightly less likely, and they are now slightly more likely again.
The fall is not uninterrupted either. New-launch share bottomed at 26.3% in 2023 and has risen in both years since, by 2.6 points in all.
Smaller homes, more upgraders
The same measure across bedroom counts, new launch.
- 1BR new launch
- 2BR new launch
- 3BR new launch
- 4BR new launch
| Year | 1BR | 2BR | 3BR | 4BR |
|---|---|---|---|---|
| 2017 | 56.1% | 53.6% | 52.1% | 50.2% |
| 2018 | 61.6% | 50.8% | 49.7% | 41.9% |
| 2019 | 58.2% | 51.6% | 51.5% | 34.1% |
| 2020 | 55.1% | 49.0% | 46.9% | 34.2% |
| 2021 | 43.3% | 41.5% | 39.5% | 31.3% |
| 2022 | 41.1% | 36.6% | 32.3% | 25.9% |
| 2023 | 45.4% | 35.8% | 26.3% | 17.9% |
| 2024 | 47.3% | 39.7% | 28.2% | 22.5% |
| 2025 | 42.6% | 37.7% | 28.9% | 23.3% |
Read any row and the order is the same: one-bedroom highest, four-bedroom lowest. That holds in every year of both legs, eighteen year-and-leg combinations, with no exception.
2017 is the one year the four sizes moved nearly together, 5.9 points between the smallest and the largest. Every year since has had them between twelve and twenty-eight points apart, and 2018 was already 19.7 apart, so this is not a gap that opened gradually.
The falls differ by size. One-bedroom is down 13.5 points over the nine years, two-bedroom 15.9, three-bedroom 23.2 and four-bedroom 26.9. The larger the home, the further the HDB share fell.
Where they went
The same comparison on resale tells a different story.
| Bedrooms | Resale 2017 | Resale 2025 | Change |
|---|---|---|---|
| 1BR | 46.2% | 40.1% | -6.1 |
| 2BR | 40.3% | 33.1% | -7.2 |
| 3BR | 36.9% | 27.2% | -9.7 |
| 4BR | 23.7% | 20.3% | -3.4 |
Four-bedroom resale has the smallest fall of the four, at 3.4 points, against 26.9 for four-bedroom new launch. So this is not a claim that HDB buyers stopped buying large private homes. They stopped buying them new.
By region, the new-launch fall is steepest outside central and shallowest in the core.
| Year | CCR | RCR | OCR |
|---|---|---|---|
| 2017 | 22.6% | 46.8% | 65.2% |
| 2019 | 16.4% | 46.1% | 61.5% |
| 2021 | 14.2% | 36.3% | 51.5% |
| 2023 | 13.5% | 24.9% | 32.3% |
| 2025 | 17.6% | 24.6% | 36.6% |
Outside central went from two in three to roughly one in three, a fall of 28.6 points, and the city fringe fell 22.2. The core central region fell too, by 5.0 points, but it starts low and ends low, and the shallower fall is a floor rather than a different story. Its thinnest year is 2018, which prints 7.5% on 93 caveats recorded HDB or private out of 107 in all, and should not be read as a movement.
What the two legs returned
The leg the HDB buyer left is the one that paid less.
| New launch | Resale | |
|---|---|---|
| Pairs measured | 13,266 | 12,850 |
| Sold above purchase | 95.15% | 95.89% |
| Median gain | $403,962 | $482,000 |
| Median gain, per cent | 30.05% | 36.61% |
| Median hold | 9.7 years | 9.0 years |
| Median return per year | 3.24% | 3.95% |
| Median purchase price | $1,237,225 | $1,245,000 |
| Median floor area bought | 1,130 sqft | 1,281 sqft |
For nearly the same money, a launch buyer got 151 square feet less, held it longer and made less per year. That is the trade the upgrader was taking in 2017, when half of them took it, and it is the trade fewer take now.
This is not a claim about what any of them should have done. A new launch is bought years before it exists, with a payment schedule and no rental income in between, and those are real reasons to choose one that a return figure does not capture.
What this does not tell you
- What "N.A" means. The caveat records an address indicator with three values, HDB, private and N.A. Nothing in our sources documents what N.A denotes, so this note excludes it from the denominator rather than guessing. The Method sets out what it looks like.
- Whether the buyer was an upgrader. An HDB address at the time of purchase is not the same as selling a flat to fund the purchase. It is the closest the data comes and it is not the same thing.
- Why they left. Four cooling-measure regime breaks fall inside the nine years this note covers: July 2018, December 2021, September 2022 and April 2023. This note measures the change, not its cause, and nothing in our sources attributes any part of it to any one of them.
- Anything about a specific project. These are shares across hundreds of developments a year.
Method and data notes. Sale data to 23 August 2026.
Shares from URA caveat data for private condominiums and apartments, by the purchaser address indicator the caveat records, split by whether the sale was a new launch or a resale.
The indicator takes three values: HDB, private, and N.A. This note reports the HDB share of those recorded as HDB or private, excluding N.A from the denominator, because N.A appears almost exclusively on new-launch caveats, 25.7% of them in 2025 against a resale leg that has never carried more than 2.0%, in 2019, and carries none at all in 2025. Including it would divide one leg by a denominator carrying a block the other does not, which depresses the new-launch share against resale in every year for a reason unrelated to who was buying.
On the included basis the 2025 new-launch share reads 21.4% rather than 28.9%, the fall reads 21.0 points rather than 23.2, and the two legs appear to cross in 2021 rather than 2022. The third option, showing N.A as its own row, gives 2017 new launch as 42.4% HDB, 39.0% private and 18.6% N.A.
Nothing in our sources states what N.A denotes. What can be measured: of the 72 projects with a new-launch three-bedroom sale in 2025, N.A appears on 57, and on 56 of those it is mixed with HDB and private inside the same project rather than accounting for the whole of it. So it is not one developer reporting differently, and it is all but absent from resale.
Bedroom counts are those resolved from the unit's own record. Nine of the cells plotted above rest on fewer than 600 caveats recorded HDB or private, and they are named here rather than left to be found: the core central region in 2019 on 177, the core central region in 2017 on 221, the core central region in 2023 on 244, the core central region in 2025 on 375, 1BR new launch in 2024 on 406, the core central region in 2021 on 479, 4BR new launch in 2019 on 481, 4BR new launch in 2018 on 515 and 4BR new launch in 2023 on 521. The whole core central column is in that list, so read it for shape and not for level.
The return figures are matched pairs of the same unit bought and resold, with the sale completing between 1 January 2020 and 19 June 2026, computed on data as at that date, and reported in full in what three-bedroom owners made.
This page therefore carries two vintages. The 882 purchaser-address cells it measures are at 23 August 2026, the latest caveat in the corpus. The sixteen return figures it borrows are at 19 June 2026, the latest three-bedroom resale the repeat-sales run matched.
Sale data to 23 August 2026.
More from Realila
- Four ways to measure whether Singapore property prices went up, and why they disagree by 18 points
Median price per square foot says private condominium prices rose 45.9% between 2017 and 2025. Hold the mix of what sold constant and it says 47.2% or 51.5%, depending on whose basket you hold constant. Repeat sales, tracking the same units sold twice, says 33.5%. Four measurements of one market over nine years, 18 percentage points apart, and composition explains barely a tenth of it.
- Buying a 2 bedroom at launch beat buying the same condo later in 128 of 228 developments, and lost in most of the OCR
For two-bedroom condominiums, the buyer who bought in the launch year beat the buyer who waited one to four years for the same building in 128 of 228 developments on gain, and in 126 of 218 on return on the money put in. The advantage is a central-region one. In the CCR and RCR the launch buyer was ahead on both measures and by 6 to 23 percentage points a year on return; in the OCR the later buyer won more developments than the launch buyer, 49 of 90.
- A resale 2 bedroom condo returned more on the money put in than a new launch in every year from 2017 to 2021
For two-bedroom condominiums bought between 2017 and 2021 and resold from 2022, a resale unit returned more on the capital committed than a new launch bought in the same year, in every one of the five years and at every interest rate tested. At 2.5%, resale returned 11% to 15% a year and new launch 5% to 9%. In the CCR, new launches returned nothing or less on the money put in. Rent from the first month, paid at once against paid progressively, is most of the difference.
Every number in this post comes from Realila Research, dated to when it was true. For now we publish research notes from the platform to answer the community's questions; the platform itself opens to the public later.
Join the waitlist
Get each new note by email, and be first in when the research platform opens.
Ask for a research note
A question about the market, or data you would like to see.
We read every request. We cannot write a note on every one, but the ones we do write will appear on the Feed, and we will email you if yours does.