Holding Longer Doesn't Raise Your Return. It Narrows the Spread by Two-Thirds.

Sort five years of private resale caveats by how long the seller held the unit, and the answer looks brutal. A hold of two to three years returned 6.62% a year. A hold of twelve years returned 2.68%. Read literally, every extra year of patience cost roughly a third of a point a year.

Read correctly, it says almost nothing about patience. It says that the people holding twelve years in this window all bought in 2013.

The pooled gradient

Every caveat in TRIBE's resale dataset carries the seller's own purchase price and date, which makes 68,581 of the 117,319 transactions a matched repeat sale of the same unit — a return with no index, no model and no composition drift. Sorted by holding period:

The annualised column falls for five bands and then turns back up. That shape is the tell. Nothing about compounding produces a U.

It is a vintage effect

Holding period and purchase date are the same variable seen from different ends. A twelve-year hold selling today bought in 2014. A four-year hold selling today bought in 2022. Sorting by holding period is sorting by entry vintage, and entry vintage is the single biggest thing in this dataset.

Median annualised return by the year the seller bought:

• 2011: 2.38%. 2012: 2.26%. 2013: 2.19%. 2014: 2.45%. The four worst cohorts in thirty years of entry dates.
• 2005: 5.44%. 2006: 5.57%. Bought before the run-up.
• 2020: 5.51%. 2021: 5.39%. Bought before the last one.

The 2011–2014 buyer paid the pre-cooling-measures peak and then sat through the plateau that followed. Those buyers are now, mechanically, the 10–15 year band. The worst row in the table is not a holding period. It is a date.

Hold the purchase year still

The clean test is to fix the entry vintage and vary the exit. Because the caveat window runs five years, the same cohort can be observed selling at four or five different holding periods.

The sign flips completely. Across fifteen cohorts with enough volume at both ends, the longer hold produced the higher annualised return in eleven. The four exceptions are 2006, whose gap is 0.12 points and therefore nothing, and 2019, 2020 and 2021 — the buyers who entered immediately before the post-pandemic surge and whose earliest years were the best ones they were ever going to have.

So the pooled gradient is not merely confounded. It is pointing the wrong way for most buyers.

And the pile itself never stops growing. The 2011 cohort's median total gain was +17.1% selling in 2021 and +53.0% selling in 2026. The 2013 cohort: +16.1% to +39.3%. No cohort in this window was better off having exited earlier in absolute dollars.

The tightest cut available

Cohort control still leaves the project uncontrolled. So compare only units in the same development bought in the same calendar year, split at that cell's own median holding period — identical district, tenure, age, managing agent and MRT station, and an identical entry market.

Across 1,551 such project-and-vintage cells, the longer hold beat the shorter one by a median of 0.22 points a year, and did so in 994 of them — 64.1%.

That is the honest size of the effect. Patience is worth about a fifth of a point a year, not the four points the pooled table appears to take away.

What time actually does

The level barely moves. The dispersion collapses.

A short hold is a coin flip with a wide rim: a tenth of short sellers made essentially nothing, a tenth made over 13% a year. Hold fifteen years and the gap between a lucky exit and an unlucky one shrinks by 67%. The interquartile range falls from 6.74 points to 2.33.

Holds of twelve years or more cluster at 3.37% a year, with a quarter below 2.15% and a quarter above 4.55%. That is what this market pays someone who simply waits: not a premium, a convergence. Entry timing is most of the variance in a short hold and very little of it in a long one.

The short hold also has a tax

The 6.62% at the top of the first table is before duty, and almost nobody in that row escaped it. Seller's stamp duty applies on a disposal inside three years of acquisition for units bought from 11 March 2017, at 12%, 8% and 4% by year held; for units acquired on or after 4 July 2025 the window is four years and the rates are 16%, 12%, 8% and 4% (IRAS).

529 of the 532 sellers in the 1–3 year band — 99.4% — fell inside a live SSD window. Net of the duty their own holding period triggered, the band's median annualised return falls from 6.62% to 4.41%. The best row in the pooled table loses a third of itself to a tax that exists specifically to make that row unattractive, and after it, the short hold no longer beats the four-year one.

What survives the cuts

• Sub-sales out. Resale only, 63,879 pairs: 6.85%, 5.10%, 4.20%, 3.36%, 2.68%, 3.80%. Unchanged.
• Tenure. 99-year leasehold (49,194 pairs) and freehold (17,007) both trace the same declining-then-recovering shape, freehold about a point lower in every band except the longest, where the two converge.
• Region. The 10–15 year trough is 1.22% in the CCR, 2.33% in the RCR and 3.21% in the OCR. Same shape, three different levels.
• Current market only. On 2026 sales alone the gradient flattens sharply — 4.23% at 3–5 years against 3.91% at 7–10 — which is what a vintage effect does as the bad cohort ages out of the middle bands.

What this does not show

These are nominal returns. At 3.80% a year over nineteen years, a meaningful share of the 113.6% total gain in the bottom row is the currency, not the asset. Nothing here is inflation-adjusted, and the longer the hold the more that matters.

They are also measured only on units that sold. An owner sitting on a loss usually does not transact, so every return here is drawn from a sample that chose to exit, and the loss rates are floors.

The annualised figures are price-to-price. They exclude rent received, and they exclude maintenance, property tax, interest and agent commission — see one-bedders yield more and appreciate half as fast for what happens when income is put on the same table.

And the cohort control cannot fully separate "time passed" from "this particular market recovered." Mean reversion appears in these five years because the 2013 plateau was followed by a recovery. A long hold narrows outcomes in any market; that it narrowed them upwards for the 2011–2014 buyer is a fact about 2021 to 2026, not a law.

How to read it

If you are weighing an exit, the holding-period table is not the input. Your own entry date is. A seller who bought in 2013 and is looking at a 2.6% annualised return is not being punished for holding thirteen years — eleven of those years were spent climbing out of what was paid in the first one, and selling in 2021 would have booked 1.8% instead.

If you are weighing an entry, the useful number is the spread, not the median. Buying with a two-to-three-year horizon puts you in a band where the middle 80% of outcomes span thirteen points a year and the duty alone takes two of them. Buying with a fifteen-year horizon puts you in a band spanning four and a half, centred near 3.8%.

Time is not the lever on how much. It is the lever on how sure.

---

From 68,581 matched repeat sales of the same private residential unit — every caveat in TRIBE's resale dataset (2,325 projects, 117,319 transactions) that records the seller's own purchase price and date, held at least one year, with the sale falling between 11 October 2021 and 2 October 2026. Median holding period 8.63 years; median purchase price $1,033,000; median sale price $1,480,000. All figures are medians; returns are nominal, price-to-price, and gross of rent, costs and financing. Stamp duty rates are IRAS's as published, current as of 4 July 2025. This is general information, not investment advice; confirm any duty position with IRAS or your conveyancing lawyer.