How to read Singapore's official home price indices and supply data

In this article

Written by

Kooky

Builder of Shaka, the payment router that pays every agent their commission on closing date.

About Kooky and Shaka →

Four times a year, two public bodies in Singapore put a number on home prices. The Urban Redevelopment Authority (URA) publishes a price index for private homes, and the Housing and Development Board (HDB) publishes a resale price index for public flats. Each gives an early reading first, called a flash estimate, and a fuller set of figures some weeks later. Around the two indices sit the tables that explain them: rents, sales counts, vacancy and the homes still to be built.

A client who asks "are prices going up?" is asking a question these figures answer only in part. An index is a measure of a whole market over a whole quarter, built by a method that has its own rules about what is counted and what is left out. A salesperson who knows those rules can say what a figure means, and can also say where it stops.

This guide goes through each indicator in turn: how the URA builds its price index, what landed, non-landed and the three regions cover, how far a flash estimate has stood from the full figure in 2026, the release calendar, the rental index, the sales counts, the vacancy and pipeline tables, and the HDB index. The latest figures each official page shows are given with their dates.

URA flash estimates of 1 April and 1 October 2026 and full statistics of 24 July 2026; HDB resale price index table, base first quarter of 2009 = 100.

Two indices for two housing markets

Singapore's official price figures are split by who built the home. The URA's methodology note for its REALIS database states that its private residential statistics leave out HDB flats and other public sector buildings. The HDB's resale price index covers the other side: public flats sold on the resale market.

Neither index can stand in for the other. In the second quarter of 2026 the URA's full statistics, released on 24 July 2026, showed private home prices up 0.5%. The HDB table shows its resale price index down 0.3% for the same quarter. Both figures are correct, and they describe different homes bought under different rules.

Executive condominiums (ECs) sit between the two in the URA's tables, and the treatment changes from one table to the next. The methodology note places ECs more than 10 years old inside the non-landed category of the price index. The sales and supply tables of the quarterly release state whether a count includes or excludes ECs, and the answer differs by table. Reading the label on each figure is the first habit worth having.

How the URA price index is built

The methodology note names the raw material. The index rests on caveats lodged with the Singapore Land Authority, supplemented by stamp duty data from the Inland Revenue Authority of Singapore (IRAS). Developers supply the data for new sales. The description the URA attaches to the index on the government's open data portal adds that a caveat is normally lodged by a purchaser shortly after the option to purchase is exercised.

The price that enters the index is the agreed purchase price in the contract. The note states that stamp duties, legal fees, agency fees and other professional fees are excluded. For new sales the URA uses nett prices, which it defines as prices after deducting indirect discounts or benefits. A transaction with no caveat is still counted, the note explains, because the contract goes to IRAS for stamp duty.

Some transactions are removed. The note lists en bloc sales for redevelopment, units with short tenures, and transactions that are not representative of general price movements. It also says that outlier records with very high or very low prices may be excluded.

Since the first quarter of 2015, the note says, the index has been computed with a "stratified hedonic regression methodology". In plain terms, the homes sold in one quarter are never the same homes as those sold in the quarter before. A quarter with more large or newer units sold would show a higher average price even if nothing had become dearer. The regression controls for attributes of the unit, and the note gives age and unit size as examples, so that the change measured is closer to the change in price for a like home. The market is divided into strata, a change is estimated for each, and the results are combined.

The combining step uses weights. The methodology note says price movements are aggregated using "5-quarter fixed weights". The open data description puts it this way: the value of properties transacted in the past five quarters is used as weights, the weights are revised every three years, and the latest revision was in the first quarter of 2015. The two pages agree on the five quarters. The second adds a revision cycle and a date that the first does not mention, and neither page read for this guide shows a more recent revision.

The index is expressed against a base. The open data portal publishes it with prices in the first quarter of 2009 set at 100, in a series that runs from the first quarter of 1975 to the second quarter of 2026 and was last updated on 27 July 2026. The URA's media releases, by contrast, give the percentage change from one quarter to the next and no index level.

Landed, non-landed and the three regions

The overall index hides two markets that can move in opposite directions. The methodology note defines landed properties as detached houses, semi-detached houses, terrace houses and strata landed houses. Non-landed properties are apartments, condominiums and ECs more than 10 years old. A townhouse with condominium status is classed as a condominium, and one without it as a terrace house. Strata and cluster units follow the form of the building.

Non-landed homes are then split into three regions, which the note defines as follows:

• Core Central Region (CCR): postal districts 9, 10 and 11, the Downtown Core Planning Area and Sentosa.
• Rest of Central Region (RCR): the remainder of the Central Region.
• Outside Central Region (OCR): everything outside the Central Region.

The second quarter of 2026 shows why the split matters. According to the URA's full statistics, the overall index rose 0.5%, yet non-landed prices fell 0.1%. The rise came from landed homes, up 2.5% after a fall of 0.4% in the first quarter. Inside the non-landed market the three regions went three ways: the CCR rose 1.8%, the RCR fell 1.2% and the OCR fell 0.1%.

An owner of a suburban condominium who heard only "prices rose 0.5%" would have been told something that the figure for the OCR did not show. The regional sub-index is the nearest official measure to that owner's home, and it remains a measure of a whole region.

Flash estimate and full figure: the 2026 record

The URA publishes its flash estimate on the first day after the quarter ends: 1 April, 1 July and 1 October in 2026. Each release explains what the estimate is made of. The October release says the flash estimates are compiled from transaction prices submitted for stamp duty payment and from data on units sold by developers up till mid-September. The April and July releases give mid-March and mid-June. A flash estimate therefore leaves out the second half of the quarter's last month.

Each release carries the same warning: past data have shown that the flash estimates may differ from actual changes, and the public is advised to interpret them with caution. The URA does not say by how much. The two quarters of 2026 for which both readings exist give an idea.

The gaps that follow are this guide's own subtractions from the two sets of releases; the URA publishes no such comparison. The two quarters behaved differently. In the second quarter no line moved by more than 0.2 of a percentage point, and the overall figure did not move at all. In the first quarter the overall rise tripled, from 0.3% to 0.9%, a gap of 0.6 of a point. The landed figure moved by 1.4 points and the OCR figure by 0.9 of a point. The direction of each line held in both quarters, including the fall in landed prices in the first quarter.

Two quarters are not a rule. What they show is that the gap can be nil or can be larger than the estimate itself, and that the pages give no way of knowing in advance which it will be. The flash estimate for the third quarter of 2026, released on 1 October, should be read in that light: overall up 1.4%, landed up 2.8%, non-landed up 0.9%, with the CCR down 0.1%, the RCR up 0.2% and the OCR up 2.2%. The URA states that the statistics will be updated on 23 October 2026.

A worked example shows how quarterly changes add up. Assume the full figures for the first two quarters, 0.9% and 0.5%, and assume, for illustration only, that the third-quarter flash of 1.4% is confirmed. Multiplying 1.009 by 1.005 by 1.014 gives 1.0282, a rise of about 2.8% over the three quarters. The first two factors alone give 1.014, which matches the 1.4% the URA reports for the first half of 2026. Quarterly changes are compounded, never simply read one at a time.

When each figure comes out

The URA's Data Release Calendar lists what is published and how frequently. For the private residential market, eight items were read on it for this guide. Two are quarterly headline releases: the flash estimate of the price index and the Quarterly Property Market Updates. Two more quarterly items arrive with the updates: median rentals of private non-landed homes by street name, and the list of private residential projects in the pipeline. Two items are monthly: rental contracts, and the units launched and sold by developers. New sale transactions are updated weekly, and resale and sub-sale transactions twice a week.

The three full quarterly releases of 2026 are dated 24 April, 24 July and 23 October. Each is the fourth Friday of its month, by this guide's own check of the dates. The calendar states that rule in words for the quarterly updates on commercial property: the fourth Friday of January, April, July and October, moved to the next working day when that Friday is a public holiday. The monthly figures follow a mid-month pattern: 15 September for August data, 15 October for September data and 16 November for October data.

The calendar read for this guide covered September to November 2026 only. It gave no time of day for a release and no dates for 2027.

The rental index

The URA's rental index for private homes is built in a similar way from a different source. The methodology note says it is based on tenancy agreement data from IRAS and that, since the first quarter of 2015, it uses the same stratified hedonic regression, with controls such as age and unit size and with five-quarter fixed weights. It has the same landed and non-landed split and the same three regions.

There is one difference of scope. In the rental index, the non-landed category includes ECs without the ten-year limit that applies in the price index.

The latest full figures are for the second quarter of 2026. According to the URA release of 24 July 2026, rentals of private homes rose 0.7%, after 0.3% in the first quarter. Landed rentals rose 2.7%. Non-landed rentals rose 0.4%, the same pace as the quarter before, with the CCR up 1.2%, the RCR unchanged and the OCR down 0.3%. There is no flash estimate of the rental index on the URA's calendar: the monthly rental contracts are the only rental figures it lists between two quarterly releases.

Read side by side, the two indices describe the same quarter from two angles. In the OCR, prices fell 0.1% and rentals fell 0.3%. In the CCR, prices rose 1.8% and rentals 1.2%. The indices measure change. They give no rent in dollars and no yield, and nothing in the pages read here links the two into one.

Sales counts: new sale, resale and sub-sale

The quarterly release counts transactions in three groups, and the methodology note explains where each count comes from. Since 25 May 2015, new sale transactions for private homes and ECs come from information that licensed developers provide on options issued. Resale and sub-sale transactions come from caveats lodged with the Singapore Land Authority.

In the second quarter of 2026, the URA counted 2,141 units sold by developers, excluding ECs, against 1,783 uncompleted units launched. There were 3,813 resale transactions, or 62.0% of all sale transactions, and 194 sub-sales, or 3.2%. By this guide's addition, the three counts come to 6,148, the figure the URA's October flash release gives for the second quarter.

That total matters for reading a flash release. The flash estimate of 1 July 2026 gave 5,420 sales for the second quarter, counted up to mid-June. The full quarter ended at 6,148, so the early count held about 88% of the final one; that percentage is this guide's arithmetic, not a URA figure. The flash release of 1 October then set 4,296 units, counted up to mid-September, against 6,148 for the second quarter, and described a fall of about 30%. The 6,148 is the full-quarter total. The comparison, on this guide's reading, is thus between a part of one quarter and the whole of another, and the footnote of the release says so by giving the cut-off.

The methodology note adds a limit of its own: lodging a caveat is voluntary, so some transactions may be missing from the caveat records.

Vacancy and stock

The vacancy table answers a narrow question: of the private homes that are completed, what share stands empty? The methodology note explains both halves. The stock of available homes is updated every quarter from Temporary Occupation Permits (TOP), Certificates of Statutory Completion and permits for change of use or demolition. Vacancy is inferred from the water and electricity consumption of completed units. It is an estimate drawn from utility use.

At the end of the second quarter of 2026, the URA put the vacancy rate of completed private homes, excluding ECs, at 6.4%, against 6.2% a quarter earlier. The CCR stood at 8.3%, the RCR at 6.1% and the OCR at 5.6%.

The release also gives the two movements behind the rate. The completed stock grew by 416 units in the quarter and the occupied stock fell by 387 units. Together, by this guide's arithmetic, they mean 803 more vacant units than three months before. A rise in vacancy can come from new completions, from homes being vacated, or from both, and the table shows which.

The rate says nothing about why a home is empty. A unit between two tenants, a unit held unused and a unit just completed all count the same way.

The pipeline supply tables

The pipeline tables count homes that do not exist yet. The methodology note sorts them into two groups. A project is under construction once the Commissioner of Building Control has issued a permit to commence building works. Planned developments are those on which construction has not commenced. They include projects with Provisional Permission, Outline Provisional Permission or Written Permission from the URA, projects whose plans are under consideration, Government Land Sales (GLS) sites that have been awarded with no plans submitted yet, and GLS sites announced or put to tender.

At the end of the second quarter of 2026, the URA counted 42,472 units, including ECs, in projects with planning approval. Of these, 15,810 were unsold, about 37% by this guide's arithmetic. The other 26,662, a figure obtained here by subtraction, are not counted as unsold: they are supply to come in the sense of homes to be completed, with no bearing on what is left to sell.

A further 18,153 unsold units, including ECs, were in sites without planning approval, among them 4,745 units on the GLS Confirmed List for the second half of 2026. Added together by this guide, the two groups come to 60,625 units, close to the total of about 60,600 that the release gives for expected completions: about 25,900 expected to be completed by 2028 and about 34,700 from 2029 onwards.

Completions are the other end of the pipe. In the second quarter of 2026, 1,212 private units including ECs were completed. Over the first half of the year, 1,611 private units and 872 ECs obtained their TOP.

The methodology note attaches two cautions to these tables. Supply is scheduled by the expected year of TOP, and the number of units may turn out lower than published because of delays or changes. The expected TOP date of a project is withheld when its developer does not consent to publication.

The HDB resale price index

The HDB publishes its resale price index (RPI) as a table that runs from the first quarter of 1990. The base is the first quarter of 2009, set at 100. A note under the table states that the index was computed by a stratification method until the third quarter of 2014, and by the stratified hedonic regression method from the fourth quarter of 2014, one quarter before the URA's own change of method.

Unlike the URA's releases, the HDB table gives the level of the index as well as the change. The flash estimate for the third quarter of 2026 is 202.4 (the HDB page read for this guide shows 30 September 2026 as its date; press reports of the figure are dated 1 October), down 0.2% from 202.8 in the second quarter. The HDB describes it as the third consecutive quarter of decline in resale prices.

The levels put the quarterly changes in proportion. The index stood at 203.7 in the third quarter of 2025 and has not been higher since. From there to 202.4 is a fall of 1.3 index points, or about 0.6%, over four quarters; both figures are this guide's arithmetic on the HDB table. The fourth quarter of 2025 is shown as 0.0%, with the index at 203.6: a change too small to show at one decimal place, yet the level did move by a tenth of a point.

A worked example shows what the base means. At 202.4, the index says that resale flat prices, measured by this method, are 102.4% above their level of the first quarter of 2009. It does not say that a given flat costs twice what it did. Between the fourth quarter of 2024, at 197.9, and the second quarter of 2026, at 202.8, the index rose by 4.9 points; 4.9 divided by 197.9 is about 2.5%. Index points and percentages are different units, and only the second compares across periods.

The HDB flash page could be read for this guide only as far as its headline figures. How the HDB compiles its flash estimate, the period of transactions it rests on, and the date of the full third-quarter figure are open points here.

What the figures cannot tell an agent

Each indicator is exact about what it counts, and the limits follow from the definitions above.

The first limit is scale. The methodology note calls the price index a broad indication of price trends. Its smallest official unit for non-landed homes is a region, and the CCR alone takes in three postal districts, the Downtown Core and Sentosa. A valuation, or a comparison of recent transactions in one building, answers a different question.

The second is timing. A quarterly change compares one three-month period with the one before. A flash estimate leaves out the end of its quarter, and the first quarter of 2026 showed that the missing weeks can move the overall figure by more than the estimate itself.

The third is coverage. Caveats are voluntary, outliers and unrepresentative transactions are removed, and the index for private homes says nothing about public flats. The methodology note points to one more trap: the figures on ownership by nationality are compiled by IRAS from notices of transfer, starting in the second quarter of 1996, and their stock figures differ from the URA's, which are built from TOP records.

None of the indicators looks forward. The pipeline tables come closest, and the URA itself says the units and dates in them may change. A statement about where prices will be next quarter is not something any of these pages supports.

Kooky, from Shaka

Kooky edits Agents Estate and builds Shaka, the payment router he made for real estate professionals. One payment comes in, and every agent, agency and party in the deal receives their signed share on closing date.

About Kooky and Shaka →