HomeResearchLanded › Rebuild economics
New vs old: 2019–2026 land-titled caveats · construction from primary QS sources · URA caveat data

Does tearing down and rebuilding pay?

A brand-new landed house sells for far more per square foot of land than the tired one next door. The rebuild play is capturing that gap for less than it costs to build. Here is the premium the market actually pays for new landed — by district — netted against a construction-cost benchmark.

+46%
median new-vs-old premium
$770
median uplift / land sqft
~$400
construction $/sqft GFA
3/10
districts where it clears costs

The premium the market pays for new

Median land $psf of newly-built landed (Sale Type = New Sale) versus existing stock (Resale), by district, land-titled sales since 2019:

DistrictNew $psfResale $psfPremiumUplift / land sqft
D10$3,134$2,036+54%$1,098
D11$3,227$2,145+50%$1,082
D5$2,538$1,601+59%$937
D20$2,567$1,715+50%$852
D15$2,631$1,849+42%$782
D28$2,242$1,484+51%$758
D14$2,214$1,580+40%$634
D19$2,171$1,557+39%$614
D23$1,844$1,292+43%$552
D13$2,138$1,666+28%$472
The premium is real and large: a newly-built landed home trades at roughly +46% more per land sqft than old stock across districts. That gap — not price growth — is what a rebuild is trying to capture in a single move.

Does the uplift cover the build? — rebuild spread by district

The rebuild only pays if the new-vs-old uplift beats what it costs to build. Netting a central construction benchmark ($400/sqft GFA × 1.4 buildable = $560/land sqft) against the uplift, per land square foot:

DistrictUpliftBuild costGross spreadAfter duties/fees*Read
D10$1,098−$560+$538+$338Positive
D11$1,082−$560+$522+$322Positive
D5$937−$560+$377+$177Positive
D20$852−$560+$292+$92Marginal
D15$782−$560+$222+$22Marginal
D28$758−$560+$198−$2Marginal
D14$634−$560+$74−$126Negative
D19$614−$560+$54−$146Negative
D23$552−$560−$8−$208Negative
D13$472−$560−$88−$288Negative

*Further −$200/land sqft rough allowance for acquisition stamp duty, selling costs and holding — before your own ABSD, which can swing the answer entirely.

Rebuild is a prime-district and own-use play, not a mass-market flip: the spread only clears construction plus costs where the absolute uplift is largest — the top-end districts. In much of the market the new-vs-old gap barely covers the build, so the economics only work if you value living in the finished house or hold for further appreciation.

New-build activity

Land-titled New Sale landed per year — a rough gauge of new-build and redevelopment completions reaching the market:

New-built landed sales per year

Land-titled New Sale caveats, all districts. A proxy for rebuild/new-build supply, not a permit count.
101171161845191820852157226723842461255526

Rebuild pipeline — from planning permissions

The authoritative signal, not a proxy: URA Written Permissions for landed. New erections and reconstructions are teardown/new-build approvals — a leading indicator of redevelopment that shows up years before any sale. Additions & alterations keep the existing house and are excluded from the rebuild count.

27,963
rebuild permissions (2000–2026)
19,912
new erections
8,051
reconstructions
14,763
additions & alterations

Landed rebuild permissions per year

New erections + reconstructions of landed homes (URA Written Permissions). Latest year is year-to-date.
8740079501784027020369404753051014061193071336089600913411016141114291214091311861410051584616875171073181069198552011522112172210882310752410402558426
Rebuild is a live, steady pipeline: roughly 1,076 landed teardown/new-build permissions a year — a rolling supply of freshly-built stock, which is exactly what the new-vs-old premium above prices. A reconstruction between two sales of the same house also inflates that pair’s repeat-sales return (it paid for a new house, not just the land).

Investor verdict

Treat rebuild as a land-and-construction arbitrage that only works at the top. Where the new-vs-old premium is largest — the prime landed districts — the uplift can clear construction and costs with room to spare; elsewhere it is thin to negative once duties and your own ABSD are counted.

Buy the worst house on the best street, price the land off the resale (old-stock) comps, cost the build at the tier your finish demands, and only underwrite the new-build exit at the district’s actual New Sale $psf — not a hoped-for number. The authoritative rebuild-history layer (URA Planning Decisions) is the next build and will let us flag which addresses were already rebuilt.

How to use this before you buy

Price the land off old-stock comps

Your acquisition should reflect resale (old) $psf, not the new-build number — you are buying the land, not the finished house.

Cost the build to your finish tier

Construction runs roughly $330–490/sqft GFA by specification. A luxury finish erases a mass-market spread — match the tier to the district.

Underwrite the exit at real New Sale $psf

Use the district’s actual new-built median above as your exit, and haircut it — a single finished house doesn’t always fetch the top of the range.

Count your own stamp duty

ABSD on the acquisition can dwarf the construction saving. Run your position through the stamp-duty rules before committing.

How is this worked out? — new-vs-old, construction model, sources & limits
Premium
Median land $psf of New Sale vs Resale landed, land-titled, 2019–2026, from URA caveat data. Districts need ≥8 of each. The gap is the market’s premium for a brand-new house on equivalent land.
Construction
Central $400/sqft GFA (mid/good tier; range $330–$490) from primary QS sources (Arcadis, RLB) per the repo’s construction-cost research. Buildable GFA assumed 1.4× land (typical 2-storey landed) → $560/land sqft.
Spread
Gross = uplift − build cost. “After duties/fees” nets a further ~$200/land sqft for acquisition stamp duty, selling and holding — before ABSD, which is buyer-specific.
Pipeline
Rebuild permissions are URA Written Permissions (service Planning_Decision), classified from each proposal into new-erection / reconstruction / A&A by keyword; only landed forms are kept. New erection + reconstruction = a rebuild. Estimated build year = decision date + an ~18-month construction lag.
Limits: New Sale landed includes some developer cluster projects, not only owner rebuilds; buildable ratio and construction tier are assumptions you should set per project; caveats miss some deals. The permission classifier is keyword-based (high confidence for reconstructions, medium for new erections). This is illustrative economics, not a valuation or a build quote.
For educational purposes only — not financial advice. Confirm stamp duties with IRAS and construction costs with a QS.
Keep going: all research › · value a specific unit › · market dashboard ›