August 2026 recorded 2,521 HDB resale transactions in the official dataset, 4.9 per cent fewer than July’s 2,651. The islandwide median moved from S$630,000 to S$635,000, an increase of about 0.8 per cent. Those two movements do not prove that every flat became dearer while activity fell.
Little Big Red Dot downloaded the complete HDB resale-price dataset on data.gov.sg on 3 September 2026, isolated registrations dated 2026-07 and 2026-08, and calculated counts and unweighted medians from the resale_price field. The file contained 239,688 rows from January 2017 onward. The August result is a registered-transaction snapshot, not a valuation index.
August versus July
| Month | Registered rows | Islandwide median |
|---|---|---|
| July 2026 | 2,651 | S$630,000 |
| August 2026 | 2,521 | S$635,000 |
The count fell by 130. The median rose by S$5,000. Because the mix of towns, flat types, floor areas, storeys and remaining leases changes each month, the median can move even when comparable individual flats do not. This is why the dataset should frame a comparison set, not produce an automatic asking price.
The busiest towns were not the most expensive
| Town | August rows | Median resale price |
|---|---|---|
| Tampines | 229 | S$688,000 |
| Punggol | 192 | S$680,000 |
| Woodlands | 184 | S$584,000 |
| Sengkang | 174 | S$650,000 |
| Yishun | 166 | S$556,500 |
| Bedok | 152 | S$585,000 |
| Jurong West | 139 | S$555,000 |
| Bukit Batok | 135 | S$575,000 |
These are town-wide medians across different flat types and lease profiles. Tampines’ S$688,000 median should not be compared directly with a Woodlands figure without first matching flat type, floor area, storey range, model, street and lease commencement year.
Flat type changes the comparison immediately
| Flat type | August rows | Median |
|---|---|---|
| 2-room | 58 | S$368,000 |
| 3-room | 559 | S$440,000 |
| 4-room | 1,155 | S$628,000 |
| 5-room | 578 | S$740,000 |
| Executive | 170 | S$925,444 |
The dataset contained one 1-room August row at S$279,000. Reporting that single transaction as a stable 1-room median would invite overinterpretation, so it is excluded from the comparison table and disclosed here. Sample size belongs beside every median.
A five-filter comparable set
- Start with the same town, then narrow to the same street or nearby blocks.
- Keep only the same flat type and a sensible floor-area band.
- Compare similar storey ranges.
- Account for lease commencement year and remaining lease.
- Read individual transactions, not only the group median.
If the group becomes very small, widen one filter at a time and label the compromise. Do not silently combine a low-floor 1970s flat with a newer high-floor unit just to produce a larger sample.
Why this is not the HDB Resale Price Index
HDB’s resale statistics page separates median prices, registered application counts and the Resale Price Index. The RPI is designed to track overall market movement. Our monthly median is a simple descriptive calculation from that month’s registered transactions. It does not control for changing composition and should not be labelled a price index.
Buyers should also run the lease and age-95 financing check before converting a comparable price into a budget. Sellers approaching completion can use the guide to electronic HDB resale cash proceeds to prepare the receiving account.
What the August data can support
It can show registered prices for a defined comparison set and reveal how much evidence exists. It cannot observe renovation quality, unrecorded condition, view, noise, urgency or negotiation terms. Use it to ask sharper questions and set a range. A valuation or market advice for a named flat still needs property-specific work.
Reproduce the calculation
The monthly comparison used the month field exactly as published, counted every August and July row, converted resale_price to a number and took the middle value after sorting. For an even number of rows, the median is the average of the two middle observations. No weighting or quality adjustment was added.
That method is transparent but limited. A repeat analysis should save the dataset download time and file hash, because later corrections or added records could change the result. The source file used here was downloaded on 3 September 2026 and contained data through September, which was still incomplete and excluded from the comparison.
Build a property-specific table
| Filter | Why it matters | How to widen safely |
|---|---|---|
| Town and street | Location and amenities | Add nearby streets with similar access |
| Flat type and area | Usable space and buyer segment | Use a stated square-metre band |
| Storey range | View, noise and lift dependence | Add one adjacent storey band |
| Lease start | Remaining tenure and financing | Use a narrow year range |
Show at least the count, median, low and high for the final set. Open the individual rows. An outlier may be a genuine premium flat, a data issue or simply a property unlike the subject.
Do not label registration month as contract month
The dataset is based on registration date. Negotiation, option and completion may occur on different dates. When relating a transaction to interest rates or policy announcements, avoid implying the buyer agreed the price on the month shown unless other evidence establishes it.
Use the median as a question generator
If a specific asking price sits far above the narrow-set median, ask what evidence supports the premium: floor, view, condition, layout or scarcity. If it sits below, ask about lease, defects, restrictions or urgency. The number does not answer those questions, but it helps identify them.
Record the comparison boundary
On any chart or briefing, state the two months, the number of transactions, the median calculation and the filters applied. If the audience later asks about a town or flat type, rerun the narrow set instead of reading an answer from the all-Singapore median. This protects a descriptive calculation from becoming an unsupported valuation claim.
