The Reservoir · a chart of GIC
Survey 2026 · soundings in disclosure
2 m below the surface

The Rain Gauge

The one performance number GIC publishes

GIC publishes exactly one performance number: the rolling 20-year annualised real rate of return, in US dollars, above global inflation. No annual return. No dollar profit. No figure against a benchmark. The choice is defensible — the mandate is generational — and it is also the least falsifiable disclosure a fund can make, since each year one twenty-year window slides forward by one.

The gauge official

3.4%
real, annualised, over the 20 years to 2026-03-31 · nominal 5.6%

19 windows official

4.9% 07/08: 4.5%07/0808/09: 2.6%08/0909/10: 3.8%09/1010/11: 3.9%10/1111/12: 3.9%11/1212/13: 4%12/1313/14: 4.1%13/1414/15: 4.9%14/1515/16: 4%15/1616/17: 3.7%16/1717/18: 3.4%17/1818/19: 3.4%18/1919/20: 2.7%19/2020/21: 4.3%20/2121/22: 4.2%21/2222/23: 4.6%22/2323/24: 3.9%23/2424/25: 3.8%24/2525/26: 3.4%25/26
Report year20y real20y nominalQuoted inSource
FY2007/084.5%SGDreport ↗
FY2008/092.6%SGDreport ↗
FY2009/103.8%USDreport ↗
FY2010/113.9%USDreport ↗
FY2011/123.9%USDreport ↗
FY2012/134.0%USDreport ↗
FY2013/144.1%USDreport ↗
FY2014/154.9%USDreport ↗
FY2015/164.0%USDreport ↗
FY2016/173.7%USDreport ↗
FY2017/183.4%USDreport ↗
FY2018/193.4%USDreport ↗
FY2019/202.7%USDreport ↗
FY2020/214.3%6.8%USDreport ↗
FY2021/224.2%7.0%USDreport ↗
FY2022/234.6%6.9%USDreport ↗
FY2023/243.9%5.8%USDreport ↗
FY2024/253.8%5.7%USDreport ↗
FY2025/263.4%5.6%USDreport ↗
Rolling 20-year annualised real returns as published by GIC each July for the FY ended 31 March. FY2007/08 to FY2018/19 were read from the archived report PDFs; each carries the sentence it came from. Note a basis change: reports before FY2009/10 quote the 20-year return in Singapore dollars, later ones in US dollars — so the early and late halves of this series are not strictly the same measure. GIC publishes no single-year return and no dollar P&L.

One series, two measures official

The chart above spans 19 annual reports, and a join runs through it. Reports up to FY2008/09 quote the 20-year return in Singapore dollars; from FY2009/10 GIC quotes it in US dollars. Currency of measurement changes a long-run return materially, so the early and late halves are not strictly the same number, and a straight line drawn across the join would be measuring two things. The basis is printed against every row for that reason.

What the wiggle actually encodes inferred

Two consecutive 20-year windows share nineteen years. Divide one compounded window by the other and everything shared cancels, leaving a single fact: how the year that entered the window compares with the year that left it, twenty-one years earlier. That ratio is the only single-year information this series ever discloses, and it can be computed exactly:

Report(1+new) / (1+dropped)Reading
FY2025/260.926the leaving year beat the entering one — the year entering lost to the dot-com-recovery year departing
FY2024/250.981the leaving year beat the entering one
FY2023/240.874the leaving year beat the entering one
FY2022/231.080the entering year beat the leaving one
FY2021/220.981the leaving year beat the entering one
FY2020/211.362the entering year beat the leaving one
Derived from the identity (1+r_t)²⁰ / (1+r_t₋₁)²⁰ = (1+g_in)/(1+g_out), valid only across consecutive same-currency windows, so the SGD-era join is excluded. The six most recent of the 16 USD-basis pairs are shown. Set in italic because GIC publishes none of this; it merely follows from what GIC does publish.

How to read a rolling number honestly

The fall from 4.6% in FY2022/23 to 3.4% in FY2025/26 is mostly the calendar, not the portfolio. Strong years at the start of the old window — the recovery after the dot-com bust — have rolled out, and the window now opens in the mid-2000s, just before the global financial crisis. The same arithmetic will push the number up again when 2008 and 2009 roll out in turn. A rolling average describes its whole window; the urge to read its year-to-year wiggle as current skill runs in both directions and should be resisted in both.

The next reading, bounded inferred

Next July's number is not a mystery — it is this year's window with one year swapped. The year that will leave is the year to March 2007; the year that will enter is the year to March 2027, most of which has not happened yet. Neither single-year return is published, so the table spans assumptions for both. Whatever markets do, the arithmetic confines the FY2026/27 print to roughly 2.1–4.1%:

Entering ↓ · Leaving →+0%+8%+15%
-10%2.9%2.5%2.1%
+0%3.4%3.0%2.7%
+5%3.7%3.3%2.9%
+15%4.1%3.7%3.4%

A crash year entering against a strong year leaving (−10% in, +15% out) still only pulls the twenty-year figure to 2.1%; a boom entering against a flat exit lifts it to 4.1%. One year, diluted twenty ways, moves the gauge very little — which is the gauge's design.

Pure arithmetic on the published 3.4%: (1+r)²⁰ scaled by (1+entering)/(1+leaving), re-annualised. Set in italic because every cell rests on assumed endpoints. Graded by GIC's report, expected July 2027 — come back and check it.