Two thirds of Sentosa Cove resales lost money. The losses were set in 2007, not 2023.

Two thirds of Sentosa Cove homes resold at a loss. That figure was reported by The Business Times on 13 August 2026 and carried widely since, usually alongside the 2023 stamp duty rise as the explanation.

We can reproduce the number. We get 65.1% of non-landed resales loss-making since May 2023, against a reported 64.5% including landed homes. The arithmetic holds.

The explanation does not. The people selling at a loss today bought in 2007 and 2008. Their loss was set at purchase, sixteen years before the stamp duty changed.

Key takeaways

• We reproduce the loss rate independently: 65.1% of non-landed resales since May 2023. Reproducing the claim
• Mogul.sg and Newmark, both quoted in the same report, disagree on direction. One found the loss rate rising, the other found it falling. Ours agrees with Newmark. Two answers, one question
• Loss depends almost entirely on the year of purchase. Buyers from 2007 lost money 90% of the time. Buyers from 2017 made money two thirds of the time. When you bought decides it
• Every published figure excludes stamp duty, which for a foreign buyer since April 2023 is the largest single cost in the transaction. The cost nobody counted

Reproducing the claim

Before disagreeing with a number, we check we can produce it.

Matching each Sentosa Cove non-landed unit to its own previous sale, and splitting on the same dates the published analysis used:

	Mar 2020 to Apr 2023	May 2023 to Jun 2026
Resale pairs	162	106
Loss-making	67.3%	65.1%
Median outcome	-$269,520	-$369,640
Median loss, when losing	-$642,700	-$789,290
Median gain, when gaining	+$359,745	+$374,500

The median outcome across all sales is a loss of $369,640 in the recent window. Volume nearly halved, from 162 pairs to 106. The typical loss got larger. Those parts of the story are real and our figures agree with the published ones closely enough that there is nothing to argue about.

Two answers, one question

The report quotes three research houses. Two of them answer the same question and reach opposite conclusions.

Mogul.sg found the loss-making share rising, from 62.8% to 64.5%, counting landed and non-landed homes together. Newmark, looking at non-landed homes only, found it falling, from 67.3% to 65.5%. Cushman & Wakefield used a different window again, and reported 83 of 244 non-landed resales profitable between 2021 and the first half of 2026.

Ours, also non-landed only, finds the share falling from 67.3% to 65.1%. Our earlier-window figure matches Newmark's to one decimal place, and our median outcome of -$269,520 for that window matches theirs to the dollar. Two independent datasets and two independent methods, landing on the same numbers.

So on the question the headline answers, the sources contradict each other, and the coverage carried the one that said conditions worsened. Whether the loss rate rose or fell depends on whether landed homes are counted, and Sentosa has roughly 392 landed homes against 1,766 non-landed ones.

This is not a criticism of any of the three. Each published its scope, and the divergence comes from methodology rather than from anyone being careless. It is a caution about the number that travelled without its scope attached.

When you bought decides almost everything

Here is what the loss rate actually tracks.

Bought in	Resale pairs	Loss-making	Median outcome
2006	53	32.1%	+$161,040
2007	51	90.2%	-$1,017,040
2008	17	100%	-$1,120,000
2009	21	71.4%	-$157,200
2010	38	94.7%	-$816,145
2011	11	90.9%	-$970,000
2012	13	100%	-$521,112
2013	9	66.7%	-$178,000
2015	8	50.0%	-$75,000
2017	11	36.4%	+$155,000

Every sale in this table happened in the same market, since 2020. What separates a 32% loss rate from a 100% one is not when the owner sold. It is when the owner bought.

Sentosa Cove sold hard between 2007 and 2012. Buyers in those six years lost money between 67% and 100% of the time. Buyers in 2006, before the run, and in 2017, after it, mostly did not.

The median holding period behind these figures is 13.3 years in the earlier window and 15.1 years in the later one. So the seller taking a loss in 2025 bought around 2010. Nothing that happened in 2023 caused that outcome. It was fixed when they signed.

That reframes the loss rate. It is not a live signal about Sentosa today. It is the group who bought between 2007 and 2012 still working its way through the market, and it will keep producing loss-making sales for as long as those owners keep selling, regardless of what happens to demand.

The cost nobody counted

Every figure quoted in the reporting, ours included so far, is a gross figure. The published analyses say so plainly: they exclude stamp duty, property tax, legal fees and agent commissions.

For most properties that omission shifts the answer. For Sentosa Cove it can invert it.

Additional buyer's stamp duty for most foreign buyers was raised to 60% in April 2023. On a $3.68 million purchase, the median buy price behind the recent window, that is roughly $2.2 million paid at purchase, on top of the price, recoverable only if the property appreciates by more than that. A gross gain of $374,500, the median for a profitable sale in this window, does not begin to cover it.

That does not change the numbers above, because those sellers bought long before 2023 at far lower rates. It changes what the numbers mean for anyone buying now. A foreign buyer entering Sentosa Cove today needs the price to rise by more than half before the sale breaks even, and the table above says the last people to need that kind of appreciation were waiting sixteen years and had not got it.

We have measured this effect before. In our work on shoebox units, the stamp duty a buyer paid changed their return more than the choice of property did.

What this does not tell you

• What it means for anyone at Sentosa Cove now. The loss rate describes people who bought fifteen years ago. It says little about a purchase today, which faces a different price and a different tax. For an owner weighing whether to sell, the number that matters is not the enclave's loss rate but the year they bought. For an agent pricing a listing, the last comparable sale may have been made by someone whose purchase year makes their outcome a poor guide to anything.
• Who is buying or selling. Caveat records do not identify nationality, residency or motive. Every claim about foreign demand in this story, including the ones in the original reporting, is an inference from price and timing.
• Landed homes. Our data covers non-landed only. Sentosa's landed segment is small, roughly 392 homes, and the published analyses report it behaving differently.
• Net outcomes. We have not computed the full cost stack for these sales, only shown the size of the largest cost that everyone excluded.

Method

Figures are transacted prices from URA caveat data for private condominiums and apartments in Sentosa Cove, excluding collective sales.

A resale pair is a unit matched to its own previous recorded sale. Gain is the second price minus the first, gross of all costs. This is the same method used in our repeat-sales work on shoebox units.

Windows match the published analysis: March 2020 to April 2023, and May 2023 to June 2026, split on the April 2023 stamp duty change.

Purchase-year figures cover all pairs sold from 2020 onward, grouped by the year of the earlier sale, and show only years with at least eight pairs.

Median holding period is the median across all pairs in each window.

Sale data to 4 August 2026.