Ask a question

What can I afford, and where?

Is it a good time to sell my flat?

What does the condo jump cost?

Where is the most life per dollar?

Should I wait for a BTO instead?

Metric Flat type Walking distance to MRT Period Remaining lease

scroll to zoom · drag to pan · click a station for details

What this map shows

MRT access and resale pricing in Singapore, built from 234,628 HDB resale transactions (2017 to 2026), 65,567 condo and apartment resales (2021 to 2026, URA), and 10,600+ real walking routes computed via OneMap pedestrian routing.

The layers

Median prices. Every colored patch is the walking catchment of an MRT station. Color shows the median resale price (or price per sqft) of homes within your selected walking distance, under your filters. Distances are actual on-foot routes, not straight lines. That distinction matters: the median walk is 1.37x the straight-line distance, and in extreme cases far more. Blocks 234 and 235 on Ang Mo Kio Ave 3 sit 450 m from Bright Hill station as the crow flies but 2.1 km away on foot.

Mispricing overlay. A price model (fixed effects hedonic, R-squared 0.94 for HDB, 0.92 for condos) predicts each transaction from size, storey, remaining lease, walking distance to MRT, location group, and month. The overlay shows the residual gap: how much homes near announced-but-unopened stations trade above or below prediction, versus matched controls in the same town far from the future station. Teal means trading below model price, a candidate for incomplete capitalization of the coming station. Orange means the premium is already in.

Four more catchment layers. Yield (condo) colors each catchment by gross rental yield, URA median rent per sqft over median resale per sqft (gross, before costs). Growth shows the change in median price per sqft from the first 12 months of data to the last. Lease risk (HDB) shows the share of recent sales with under 60 years of lease left, where price decay is steepest. Liquidity shows annualized transactions per year over the last 24 months; thin markets are where mispricing persists longest. In the mispricing overlay, HDB stations that opened during the data window (Thomson East Coast Line stages 1 to 4) are ringed in teal; click one for a half-year event study of its new-access blocks with the opening date marked.

The headline finding

HDB buyers price a future MRT station late; condo buyers price it early. For HDB flats, the discount near future stations persists until roughly 1 to 2 years before opening, then closes: flats near Bedok South (opened 2026) went from discounted to +12% versus controls, and Jurong Region Line areas (opening 2028) still trade about 3% below model today. Condos near Thomson East Coast Line stations showed the same arc but completed it 2 to 4 years before opening, and Cross Island Line condo areas moved from -1.8% (2022) to +2.2% (now), four years ahead of the 2030 opening. Whatever window exists in the private market closes early; in the HDB market it stays open into the final years.

Other things the data showed

1. Walking distance is worth about 1.3% per 100 m in both markets (HDB -1.27, condo -1.30 percent per 100 m walked). On a $600k flat, moving 400 m further from the station costs roughly $30k.

2. Lease decay is not a straight line. HDB prices fall about 1.6 to 1.8% per year of remaining lease below 60 years, but only 0.7% per year in the 60 to 80 band. The steep zone starts near the 60-year mark.

3. The center-periphery gap compressed. Since 2017, HDB price per sqft near Pasir Ris rose 88% while Holland Village and Buona Vista catchments rose under 20%. Median catchment: +44%.

4. Condos cost a median 2.57x HDB per sqft in the same catchment, ranging from 1.47x (Canberra) to 3.7x (Braddell, Mattar and Jurong East); full ranking in Deep dive 12.

5. Old leasehold condos carry an en-bloc option premium of 2 to 7 points that tracks the collective-sale cycle: high in 2017 to 2019, zero during COVID, +2 to +2.5 now.

6. New launches do not lift nearby resale prices within a year; across 52 major launches the effect was slightly negative (-1.4%).

7. Some gaps are geography, not mispricing. Condo gaps near CRL stations were identical before and after the line was announced, so persistent location effects can masquerade as signal. Trends and sample sizes matter more than levels.

8. Rental yield runs opposite to price. Gross condo yields average 3.4% and move inversely with price per sqft (correlation -0.5): prime catchments like Napier and Orchard Boulevard yield 1.6 to 2.3%, while city-fringe and mass-market pockets like Khatib, Yew Tee and Canberra reach 4.0 to 4.2%. Cheaper locations buy income; expensive ones buy capital.

9. Opened stations confirm the late-HDB pattern directly. Tracking blocks that gained new rail access, Mayflower moved from -2% before its 2021 opening to +5.3% after, and Upper Thomson from -5.4% to +1.5%. Stations sited at existing interchanges (Woodlands, Caldecott, Outram Park) gave no new access and show no effect. The countdown is about new access, not the station itself.

10. The lease cliff is geographic. In the oldest central estates (Jalan Besar, Chinatown, Tanjong Pagar, Marine Parade) every recent HDB resale already has under 60 years of lease left, with medians of 50 to 58 years, whereas 17 catchments have no sub-60-year sales at all. Lease-decay risk is concentrated, not spread evenly.

11. Turnover concentrates in the young towns. HDB resale volume peaks in Admiralty, Punggol and Sengkang (roughly 270 to 380 sales a year per catchment) and thins to a handful in central pockets like Dover, Bras Basah and Bencoolen. Condo turnover is highest around Aljunied, Queenstown and Great World. Thin catchments are where mispricing lingers longest.

Caveats

Residual gaps control for observed characteristics only; renovation, view, and micro-location can account for 5 to 10% of price. Individual homes deviating 5 to 8% from model is normal noise. Future station effects use straight-line distance (walking routes do not exist for unbuilt stations). The three Circle Line stations that closed the loop in 2026 (Keppel, Cantonment, Prince Edward Road) now use real OneMap pedestrian routes like every other station (153 routes computed Jul 2026). Condo building age is unobserved. Expected opening years are assumptions and can slip. Rental yields are gross (before vacancy, tax, maintenance and fees) and joined to catchments by rental-project location, with district medians where no project matches. This is a research exhibit, not investment advice.

Sources: HDB resale transactions (data.gov.sg), URA private residential transactions (URA Data Service), LTA station coordinates, OneMap routing and geocoding.

Deep dive: what the price model says beyond the map

Thirteen sections and a transaction-volume panel from the same verified pipeline, ordered from market to decision (89,352 model transactions near MRT; 234,628 all geocoded HDB resales, 2017 to 2026). Companion note: hdb_deepdive_report.md.

1. Ten years of quality-adjusted prices and five cooling measures

The model's month effects form a constant-quality HDB price index (2017-01 = 100). Diamonds mark cooling measures; hover for the 12-month growth rate before and after each. The partial month 2026-07 is excluded.

The long view, from the official index. The chart above starts in 2017 and has never seen a falling market. The official HDB Resale Price Index (below, 1990-2026, quarterly) has: -27% in 1997-98, and a six-year slide of -12% from 2013 to 2019. The grey line is the 3-month SORA interest rate, the anchor for floating mortgages. Our quality-adjusted index tracks the official one within 3.5 points across the 37 overlapping quarters, an independent check that the model is measuring the market.

Descriptive, not causal. There is no control group for economy-wide policy, and the 2022-2023 deceleration coincides with the mortgage-rate surge, a confound that can carry the whole effect. What the data does show: prices were flat 2017-2020, rose fastest into late 2021 (+12.4%/yr), and each measure from Dec 2021 onward preceded a further deceleration, largest after the Sep 2022 package (-4.4 pp), the only one aimed squarely at HDB demand. The notable current fact: the index has been flat for the last 12 full months (154.3 in Jun 2025 and Jun 2026) while raw medians still print gains on mix. Jul 2018, which targeted private property, is a useful placebo: HDB was unaffected. And the rate lens: the 2022-2023 deceleration coincided with SORA rising from 0.2% to 3.7% in eighteen months, the fastest tightening in the data; rates have since fallen back to about 1.1% (May 2026) while the official index sits flat at 203.4, three quarters off its peak. Cheap money returning without prices moving is itself information.

Transaction volumes: is the market busy?

A companion to the price index above and the rent index below: the count of resale transactions each quarter, within the same MRT walking catchments used everywhere on this map. HDB runs from 2017; condo begins 2021Q3, where the current URA capture starts. The three buttons normalise the view.

HDB Condo / apartment

Normalising. Both lines are near-MRT resales from the same dataset, a consistent slice of the market rather than the total. Indexing each series to its own 2022 average removes the roughly 1.5x level gap so the trends compare directly. The 4-quarter average strips the seasonal pattern (Q1 is always the quietest quarter). A true turnover rate (sales divided by housing stock) would be the ideal normalisation but needs per-segment stock counts not in this dataset.

2. The rental market: three years of frozen rents

A chained index built from roughly 400,000 individual URA rental contracts (condo and apartment, 2021Q3 = 100). Each step compares only projects that rented in both adjacent quarters, so the mix cannot distort it.

The shape is the story. Rents exploded +43% in seven quarters to a 2023Q2 peak, then went flat, and have moved less than 2% in three years. Combined with the flat price index (section 1), the whole housing economy is in stasis: prices flat, rents flat, so gross yields are frozen at a median 3.4% (737 projects with both rent and price data; hover the condo dots on the Neighbourhoods layer for each project's yield). What it means practically: landlords can no longer count on rent growth to rescue a thin yield, and the 2021-2023 window where tenants were forced buyers is closed. Every yield on this site uses these contract-level figures, gross of tax, fees and vacancy. The dashed line is the official URA rental index (rebased to the same start): the two agree on the whole shape, an independent check on our contract-level series, and the official history adds a warning ours cannot show: rents fell 13% between 2013 and 2020. Rents can fall for seven straight years.

3. The price of height: a block of flats, floor by floor

Premium over an identical flat on floors 01-03, from the hedonic model with storey bands (same controls as the mispricing overlay: size, lease, walk distance, town x flat type, month). Pick a base price to see what each band implies.

Price of this flat on floors 01-03:

The curve is concave: the climb from floor 1 to 12 buys about half the total premium; above floor 22 each band adds under 2 points. The dashed line re-estimates the curve comparing units within the same block only (6,689 block-type groups); it tracks the main curve within 2 points at every band, so this is the price of height itself, not of better blocks being taller. The thin straight line is the production model's linear term (+0.72% per storey), which overprices the top and underprices the low-floor discount. The 37+ tier (+26.1%) is a scarcity premium concentrated in a handful of very tall central projects. Standard errors are 0.1 to 0.5 points per band.

4. What living near the MRT is worth, town by town

The model's headline says HDB prices fall about 1.3% per 100 m of walking distance from the station. That single number hides a 7-point spread. Each bar re-estimates the walk premium inside one town (same controls, whiskers are 95% intervals).

Read each bar as the price drop for every extra 100 m you walk from the station. On average a flat loses about 1.3% per 100 m, but that average hides a 7-point spread. In younger, MRT-sparse towns the price falls much faster than average (Sengkang -3.7%, Bukit Panjang and Geylang -3.0% per 100 m), which is the market saying MRT proximity is worth more there. In mature, MRT-dense towns it falls close to nothing (Bedok 0.0, Ang Mo Kio -0.6), so the walk barely changes the price. Central Area's positive bar is a confound, not a finding: near the CBD, blocks further from stations are often newer or better located in other ways, and the within-town model cannot separate that. For the map, the pooled -1.3%/100 m over-adjusts mature towns and under-adjusts the young ones, so residual gaps in Sengkang-type towns are conservative.

5. Which flats appreciated, and the MOP mix trap

Each cell answers a simple question: if you compare typical resale prices in 2017 with the last 12 months, how much did they rise in this town, for this flat size? Darker purple means less growth, bright yellow means more. But there are two ways to measure it, and they tell different stories. Use the toggle.

Worked example: where a headline number comes from

Why this matters. Once the new-flat effect is stripped out, the market-wide result flips the folk wisdom: bigger flats appreciated more, not less (Executive +48.8%, 5-room +46.5%, 4-room +45.2%, 3-room +39.3%, 2-room +33.0%). That national ordering is partly geography, though: Executive stock sits disproportionately in fast-growing outer towns, and compared within the same town Executive and 4-room are indistinguishable (4-room ahead in 11 of 20 towns, mean difference 0.6 pp of total growth). Which size leads is a town-level fact, not a national one, as the pair test below shows. And any town ranking based on raw medians, including this map's own Growth layer, partly ranks BTO completion geography rather than value gains. Cells need 30 or more sales in both windows; grey cells have too few older flats to measure.

Which of those differences are actually distinguishable from noise?

The grid above gives each flat size its own growth number. It does not say whether two sizes in the same town differ by more than sampling noise. This does. Pick a town: every pair is tested, and only pairs that keep their sign under six different ways of computing them and have a 95% interval excluding zero are reported as findings.

Town

Different measurement from the grid above. These are annualised, in percentage points per year, over 2017Q1 to 2026Q2 using four-quarter averaged endpoints, so they will not reconcile cell-by-cell with the total-growth percentages above. Same-stock basis (pre-2013 lease) in both cases. A gap of 1.00 pp/yr compounds to roughly $60,000 on a $500,000 flat over 8.5 years. Pairs not listed as findings showed no detectable difference, which is not the same as a small one: 114 of the 150 town-and-pair combinations fall in that bucket. These are averages across the window and do not establish that any gap is still widening today.

6. Affordability: years of local income for a 4-room, measured in the same year

For each town: the median 4-room resale price in 2020 divided by the annual median household income of that planning area in 2020 (the census year). Both sides of the ratio come from the same moment, so this is a clean measure of how far a flat sat out of reach of the community around it.

Reading it. Marine Parade, Choa Chu Kang and Pasir Ris sat near 3.1 to 3.2 years of local income in 2020; Bukit Merah (6.6) and Queenstown (6.8) more than double that. The spread, not the level, is the finding: central flats cost twice the local earning power of outer ones, and that ranking is stable over time. Two caveats: planning-area income includes condo households, so mixed areas like Marine Parade look more affordable than they feel to an HDB buyer specifically; and prices have risen about 45% since 2020 while incomes have grown too, so do not read the 2020 levels as today's burden. The affordability view on the Prices layer shows the same aligned measure per catchment. Bridged to today: national median household income grew +30.6% from 2020 to 2025 (DOS); applying that growth to every area, the estimated current range runs from about 3.1 years (Pasir Ris area catchments) to 10 years (central), with the caveat that income growth was assumed uniform across areas. Catchment tooltips on the map carry both numbers.

7. Where the money lives: household income by neighbourhood

Median monthly household income from work, census 2020, for every planning area with 3,000+ resident households. Hover for the share of households earning above $20,000 a month. Also on the map: the Prices layer's "Local income" metric colors every HDB catchment by its town's income.

The surprises, in order of how much they should update you. First, the richest HDB-town incomes are Marine Parade and Bishan ($13,007 and $12,330); Pasir Ris ($11,276) ranks 8th of 29 areas but still pairs a high income with low flat prices, so it is the affordability champion of section 6. These are planning-area incomes, so they count each area's private condos and landed homes, not HDB households alone. Second, Outram is the lowest-income area in the country ($7,031) while containing some of its most expensive flats: one-room rental blocks and million-dollar Pinnacle resales share the same postal district, a reminder that the income of residents is not the price of property. Third, income follows estate age, not prestige: young Sengkang and Punggol ($9,600) out-earn mature Toa Payoh ($8,357) and Ang Mo Kio ($7,577), because working-age dual-income families cluster where the new flats are; this age gradient drives much of the affordability spread. Fourth, the expected top is very top: Tanglin, River Valley and Bukit Timah medians exceed $20,000, with a majority of their households above $20k. Fifth, the inequality tell: Bedok and Queenstown show high top-bracket shares against modest medians, meaning private enclaves (Bayshore, one-north fringe) sit inside otherwise ordinary-income towns. Caveat: income from work per resident household, 2020, includes condo households.

8. Do cheap blocks stay cheap? The mispricing signal, tested

The map's mispricing layer rests on one bet: when a block trades below its model price, the discount is temporary. This chart tests that bet on 1,020 blocks with 8+ sales in both periods. Left-right: how cheap or dear the block was in 2017-2019 versus what the model predicts from its size, floor, lease, walk to MRT, town and date. Up-down: the same check on the same block, six years later.

Example. Take a block that sold 10% below model price in 2017-19 (left of center). Two extreme stories could explain it. Story one: the discount is hidden quality, a noisy road, a bad layout, something buyers see but the data does not. Then the block should still be 10% cheap today, and its dot lands on the diagonal line. Story two: the discount is pure mispricing. Then buyers correct it, the dot lands on the horizontal zero line, and whoever bought in 2017 pocketed the correction as extra appreciation. The dots actually land about a third of the way up (the orange line, slope 0.37): a typical 10% discount became roughly a 3.5% discount. Two thirds mispricing that closed, one third hidden quality that stayed.

The bars show what the correction was worth. Sort the same blocks by how cheap they started: the cheapest fifth gained a median +49% by 2023-26, the dearest fifth +27%. Within the same town, every 1% below model price predicted roughly +0.5 pp of extra growth (some of that is noise washing out, so treat it as an upper bound). Net reading for the map: a teal patch is a real signal, but discount what you see by about a third for quality the data cannot capture.

The compression objection, tested. A fair challenge: maybe cheap blocks outperformed simply because cheap things get bid up more in a boom, while expensive ones hit a ceiling, which would be price compression, not mispricing. The test: race the two stories in one regression, block growth on (a) the gap versus model and (b) the absolute price level, within the same town and flat type. Result: the model-gap keeps its full +0.5 pp per 1% below model, while the price level carries exactly zero additional signal (coefficient -0.00, se 0.02), despite the two being correlated 0.51. Being cheap in dollars predicted nothing; being cheap relative to what the flat's size, floor, lease and location justify predicted everything. Compression is real between towns (Pasir Ris vs Holland Village) but that is a different phenomenon; within a town, the discount itself is the signal. One refinement this test surfaced: comparing blocks within the same town and type, persistence is 0.53 rather than 0.37, so for neighbour-to-neighbour comparisons assume about half the discount is durable, not a third.

9. Condo catchments: income, growth, or neither

Every dot is a condo catchment with 30+ resales in both windows and a computable rental yield. Right = better rental income today (gross yield). Up = stronger price growth since 2021. The dashed lines are the medians, splitting the market into four quadrants.

Read the top-left with suspicion. The high-growth, low-yield corner is where the condo version of the mix trap lives: Lentor's +101% is mostly brand-new projects entering the resale sample, not appreciation of existing stock. The dependable corner is top-right (above-median on both): Beauty World, Canberra, Farrer Park, Kovan, Sengkang, Lakeside. Yields are gross (URA median rents over resale psf, before tax, fees, vacancy); subtract roughly 1 to 1.5 points for a net figure. Growth baseline is 2021H2 to 2022H1, so this chart cannot see the 2017-2021 cycle.

10. The price of buying new: launch premium and what happened next

The window: all URA transaction records from Jul 2021 to Jun 2026, the five years the API provides. Within it, developers sold new units at a median 49% premium over resales trading in the same district in the same month (94 launches). This chart asks whether paying that premium worked out. Each dot is one project bought new that has 8+ resales in the last 12 months to measure against. Further right = bigger premium paid at purchase. Below the zero line = its buyers have since gained less than ordinary resale buyers in the same district over the identical period.

Two dots tell the story. Avenue South Residence (grey, District 3): people who bought its units new in late 2021 have seen the median price rise just +1.4% in five years, while plain resale buyers in the same district gained +24.4%. No one is underwater in headline cash terms, but after stamp duty and five years of waiting, many of those buyers effectively are. Tembusu Grand (amber, District 15, launched Apr 2023): its buyers are +15.7%, beating district resale buyers by 8.6 points, the best outcome in the chart. The pattern behind the two: units bought new early in the window, at fat premiums into a hot market, underperformed badly; the few winners launched later, into a cooler market, at thinner premiums.

Amber versus grey, defined. Amber (true launch): the project's first developer sales happen inside the Jul 2021 to Jun 2026 window, so the premium is measured at actual launch day. Grey (already selling at window start): the project launched before Jul 2021, so this chart only sees its later developer sales; its real launch price, and anything its earliest buyers gained before mid-2021, are invisible here. Grey dots therefore measure buying new from the developer in 2021 or later, not launch-day buying. The scoreboard: 32 of 35 dots sit below the line (median -12.9 pp); among amber true launches alone it is 5 of 8, a small sample. One more honesty note: this window was a resale boom, and the new-sale premium does buy a full lease and a new building; a longer horizon may treat it more kindly. What the data rejects is the folk belief that buying at launch reliably beats the market.

11. The en-bloc clock, updated to 2026

Old leasehold condos trade at a discount to young ones, that is normal lease arithmetic. The interesting part is how the discount moves: when en-bloc fever runs, buyers bid up old projects for their redevelopment potential and the discount compresses. This line tracks the raw discount of 25+ year projects versus under-20 year projects, matched within district, every half-year.

Reading it: stable between -27% and -31% for five years, tightest in 2022H1 and 2023H1 (the small collective-sale revival), widest through 2024-2025H1, and edging tighter now (-28.3% in 2026H1). No froth, mild warming. This is the raw psf discount, a blunter measure than the earlier hedonic analysis (which put the option-value component at +2 to +2.5 residual points in the same era); read the trend, not the level. The most liquid 30+ year leasehold projects right now, the perennial en-bloc names: Melville Park, The Bayshore, Braddell View, Mandarin Gardens, Bayshore Park, Neptune Court, Orchid Park, Pine Grove.

12. The upgrader gap: what the condo jump costs, station by station

Same catchment, same 12 months: median condo price per sqft divided by median HDB price per sqft. A ratio of 2x means the condo jump costs twice the psf. Median across 49 catchments: 2.57x. The narrowest and widest ten:

Canberra (1.47x) and Beauty World (1.73x) are the cheapest HDB-to-condo jumps in the country; Braddell, Mattar and Jurong East (3.7x) the steepest, typically where new-launch condo stock meets old HDB stock. Caveat: this compares whatever traded in each catchment, not identical homes; where the condo stock is much newer than the HDB stock the ratio overstates the like-for-like gap. Both sides use the last 12 months, 20+ sales each.

13. What it all adds up to: a synthesis for buyers and sellers

Everything below is a plain reading of the twelve sections above and the map layers, from 2017-2026 data. It describes what happened, not what will happen. It is a research exhibit, not financial advice.

Buying a resale HDB flat. The quality-adjusted market has been flat for 12 months (section 1), so speed matters less than selection. The mispricing map is a real edge with a known discount rate: blocks trading below model closed about two thirds of their gap historically, and every 1% below model predicted roughly +0.5 pp extra growth (section 8), but assume the last third of any discount is something wrong you cannot see in data. Go view the flat. On floors: the premium is front-loaded, floors 4 to 12 buy most of the height value at the lowest cost; above floor 22 you pay 17%+ mostly for scarcity (section 3). Ignore raw town growth rankings when comparing districts; they partly rank BTO completion geography (section 5).

Selling one. If your town is MOP-heavy (Toa Payoh, Clementi, Kallang), headline medians flatter the market, not your flat: price against same-vintage comparables or the buyer's valuer will do it for you (section 5). If you sit near an announced station, HDB capitalization arrives late, in the final 1 to 2 years before opening; selling before that forfeits the documented closing of the gap (mispricing layer; JRL areas still trade about 3% below model). Liquidity varies 50x by catchment (Liquidity layer): in thin catchments plan for time, not just price.

Upgrading from HDB to condo. The jump costs a median 2.57x per sqft in the same catchment; Canberra (1.47x) and Beauty World (1.73x) are the cheapest bridges in the country, Braddell and Mattar (3.7x) the steepest (section 12). Route matters more than timing: buying the condo new costs a median 49% premium over local resale, and 32 of 35 measurable projects underperformed plain resale purchases over the same years (section 10). The data favors the resale condo route. If MRT access is the thesis, note condos price future stations 2 to 4 years before opening; by the time a line is about to open, the condo move is done.

Living near the MRT. Worth about 1.3% per 100 m on average, but that average misleads (section 4). In Sengkang, Bukit Panjang or Geylang, proximity is worth about 3% per 100 m, so the walk premium is real money both when buying and selling. In Bedok, Ang Mo Kio or Toa Payoh it is worth nearly nothing: there, a flat 700 m out is the value play, the market barely charges for the walk. For rental income, the catchments above the median on both yield and growth: Beauty World, Canberra, Farrer Park, Kovan, Sengkang, Lakeside (section 9, gross yields).

Choosing a town. On same-stock appreciation, bigger flats beat smaller ones everywhere (Executive +49%, 3-room +39% since 2017), and old central 3-rooms were the weakest holding in the dataset (section 5). Towns are not uniform: Toa Payoh contains both $500 and $1,000+ psf pockets (Neighbourhoods layer, 2.1x spread), so town-level talk hides most of the decision. Pasir Ris, Bishan and Marine Parade are the most internally uniform, what you see is what the town costs.

Where the money lives, and what it pays. Income geography is stable and legible: the highest HDB-town incomes are in Marine Parade and Bishan (Pasir Ris ranks high too, and with cheap flats is the affordability standout), Outram is the poorest area amid the priciest flats, and young estates out-earn old ones (section 7). Measured honestly at the 2020 benchmark, a 4-room ran from 3.1 years of local income in the east to 6.8 in Queenstown (section 6). For landlords: rents have been frozen for three years (section 2), so the yield you buy is the yield you keep; the contract-verified project yields are on the Neighbourhoods layer. And renewal is concentrated: new development approvals cluster in Queenstown, Bukit Merah and Kallang (Activity layer), the old south-central belt being rebuilt.

Lease decay: can the value hold? The decay schedule is measured, not hypothetical: about 0.7% per year of price for leases in the 60-80 year band, steepening to 1.6-1.8% per year below 60 (model estimate, section on lease risk). Set that against a market that has gained 0% quality-adjusted in the past year (section 1): a sub-60-lease flat needs the market to rise just to stand still. What the data supports: value held best in flats bought below model price (the discount closes, section 8), with 80+ year leases, in towns where the walk premium is strong. What no dataset can promise: that the 2017-2026 pattern, driven by a once-off boom and specific policy, repeats. The honest answer to "can the value hold" is: above 80 years of lease, history says mostly yes; below 60, arithmetic says only with market growth that is currently absent. And the official record adds what our window cannot: HDB prices fell 27% in 1997-98 and 12% across 2013-2019, and the official index has now been flat for three quarters. Flat is not a floor (section 1).

Official series (HDB RPI, URA private price and rental indices, SORA, household income) from the Singapore Department of Statistics SingStat Table Builder, retrieved Jul 2026. Provenance: all numbers recomputed this session from the state-pack parquets and verified against the pipeline's anchor results (R-squared 0.9445, residual SD 8.67%, walk premium -1.27%/100 m reproduced inside the storey regression). Methods and caveats: hdb_deepdive_report.md.

Sources and credits

Transaction data.
· HDB resale flat prices: Housing & Development Board, published as open data on data.gov.sg. 234,628 geocoded resales, Jan 2017 to Jul 2026.
· Private residential (condo and apartment) resale and rental records: Urban Redevelopment Authority (URA), via the URA Data Service. Jul 2021 to Jun 2026 for the current capture; an earlier URA pull (2017 to 2022) is used where noted.
Geography and transport.
· MRT and LRT station locations and the rail network, including future lines: Land Transport Authority (LTA).
· Pedestrian walking routes, address geocoding, and planning-area boundaries: OneMap, by the Singapore Land Authority. Planning areas are the URA Master Plan 2019 set.
Official statistics.
· HDB Resale Price Index, and URA private property price and rental indices: Singapore Department of Statistics, SingStat Table Builder.
· Median household income from work, by planning area: Census of Population 2020 (Department of Statistics). National income growth 2020 to 2025 from the same source.
· 3-month compounded SORA: Monetary Authority of Singapore.
Method. Prices are modelled with a fixed-effects hedonic regression on size, storey, remaining lease, walking distance to MRT, location group, and month. All figures on this site were recomputed from the source data and cross-checked against the pipeline's anchor results (R² 0.9445, residual standard deviation 8.67% for HDB). Full method notes: hdb_deepdive_report.md.
Terms and disclaimer. Government datasets are used under the Singapore Open Data Licence and the respective agency terms; the data belongs to those agencies, not to this map. Figures reflect the captures listed above and can lag the live market. This is a research exhibit, not investment advice, and not affiliated with or endorsed by any of the agencies named.

What do you want to know?

234,628 HDB resales, 106,832 condo transactions, 400,000 rental contracts, real walking routes. The map behind is already live: pan and zoom any time. Research exhibit, not financial advice.