Skip to content
dailystudysgDaily Study SG

Articles · Buying, Rates, Refinancing

Your lock-in date decides your refinance. The rate forecast barely matters

Three-month SORA peaked at 3.75 per cent and is 1.23 — 252 basis points nobody on a 2023 package has captured. But a 1.5 per cent lock-in penalty moves the breakeven from six months to twenty-two, which is more than the entire plausible range of rates does.

Oct 22, 2026 · Daily Study SG

Three-month compounded SORA — the benchmark almost every floating home loan in Singapore is priced off — peaked at 3.75 per cent in Oct 2023. It is 1.23 per cent as of Sept 2026. That is 252 basis points of compression, and anyone still paying a rate set in 2023 has not captured any of it.

Which raises the question everybody actually asks, and which nobody can answer: is this the moment to refinance, or will rates fall further? The useful discovery is that for most borrowers the question does not need answering. The arithmetic of switching is dominated by something that has nothing to do with the rate cycle.

What rates have actually done

The series runs from Oct 2005. It contains one fact that explains most of the confusion about Singapore mortgages: 10 of the 21 years since 2006 had an average three-month SORA below half a per cent. The all-time low is 0.036 per cent, in April 2012. Anybody who bought between the financial crisis and 2021 borrowed in conditions with no precedent anywhere else in the record, and advice formed in that era does not transfer.

Show the figures
YearAverage 3-month SORA
20063.19%
20072.43%
20080.75%
20090.21%
20100.12%
20110.06%
20120.06%
20130.05%
20140.08%
20150.24%
20160.21%
20170.52%
20181.01%
20191.54%
20200.40%
20210.16%
20221.31%
20233.59%
20243.52%
20251.96%
20261.12%

The shape is three eras and a turn: two-and-a-half to three per cent before the crisis, then the long stretch near zero — 10 years averaging under half a per cent, from 2009 to 2021 — then the fastest tightening in the series, 1.31 per cent average in 2022 against 0.16 in 2021, peaking in Oct 2023 and falling since.

And where they are now

The fall has stopped. Three-month SORA bottomed at 1.03 per cent in April 2026 and is 1.23 per cent now — up 20 basis points, and every one of the five months since the trough moved upward.

3-month SORA, Sept 2026
1.23%
Up 20bp from 1.03% in April 2026.
2-year government bond
1.99%
Bottomed in Nov 2025, up 61bp since.
Still uncaptured
252bp
From the Oct 2023 peak to now. The refinancing pool.

All four instruments turned, which is why this is worth stating. A single series bottoming is noise; the one-year bill, the two-year bond, the ten-year bond and the loan benchmark all bottoming and all rising is a change in conditions. And the order matters: the two-year bond bottomed five months before the loan benchmark did. The government curve is the cheaper early warning, and it is free to watch.

Show the figures
InstrumentTroughNowMove
3-month SORA1.03% (April 2026)1.23%+20bp
1-year T-bill1.35% (Sept 2025)1.73%+38bp
2-year bond1.38% (Nov 2025)1.99%+61bp
10-year bond1.83% (Aug 2025)2.47%+64bp

Why none of that decides your refinance

Here is the arithmetic, on $500,000 outstanding with 25 years left, about $3,000 of conveyancing and valuation, switching to an illustrative 2.20 per cent. The left column is where you are now. The two right-hand columns are the same switch, out of a lock-in and inside a 1.5 per cent one.

Show the figures
Rate you pay nowSaved monthlyBreakeven, no lock-inBreakeven, in lock-in
4.0%$4716.4 mo22.3 mo
3.5%$3359.0 mo31.4 mo
3.0%$20314.8 mo51.8 mo
2.5%$7540.1 mo140.4 mo

Read the two right-hand columns against each other, because that is the finding. At 4.0 per cent the switch repays itself in 6.4 months out of lock-in and 22.3 months inside one. At 2.5 per cent it is 40.1 months against 140.4.

The penalty is 1.50 per cent of the balance — $7,500 on this loan, several times the conveyancing. It changes the answer by more than the difference between the best and worst rate anybody is likely to be offered. So the decisive fact is a date on your existing loan agreement, not a view on the rate cycle.

The honest substitute for a forecast

Instead of predicting where rates land, price the switch across the range they might land in. From 4.0 per cent, out of lock-in, on the same loan:

Show the figures
If you can getSaved monthlyBreakeven
1.50%$6404.7 mo
2.00%$5205.8 mo
2.50%$3967.6 mo
3.00%$26811.2 mo
3.50%$13622.0 mo
4.00%—never

Every rate in that band repays the cost of switching inside the remaining term. For a borrower on four per cent, the case does not depend on the forecast at all — which is the most useful thing a rate series can tell you, and it is not a prediction.

Work your own numbers on the refinancing breakeven calculator. It takes your balance, your rate, the rate you have been offered and your lock-in status, and returns the breakeven and the band. It will not tell you what to do.

What this cannot tell you

What any bank will actually offer. Singapore publishes benchmarks — SORA and the government curve, which is where every figure above comes from — not the packages banks quote. Those move weekly and are often given only on enquiry. The 2.20 per cent used in the tables is an illustrative rate, not a quote: it is roughly the benchmark plus a spread, and the spread is not something this site has measured.

Where rates go next. The curve prices expectations — the one-year bill at 1.73 per cent against the ten-year at 2.47 per cent says the market expects higher short rates ahead — but a price is not a forecast, and the market is frequently wrong. This site computes; it does not advise, and it does not predict.

Anything that is not the rate. Moving between an HDB concessionary loan and a bank loan changes eligibility, deposit rules and what CPF may be used for. Those can matter more than the rate, and none of them are in the arithmetic above.

Rates are Singapore Department of Statistics, table M700071, last updated 14/10/2026, covering Oct 2005 to Sept 2026. Figures recomputed whenever the table refreshes, so quote the date you read them. See the method.