Diogenes of Sinope was a Greecian philosopher from the Cynic school of thought. He did crazy iconic acts as a way of attacking the materialism and pretentiousness of Athenian society. One of his famous outrageous acts was walking about the streets in the day holding a brightly lit lamp. When asked, he would say he was searching for an honest man, or searching for truths.
If Diogenes were to appear in Singapore, would he be able to find someone who can tell him how much is Singapore's national reserves? Some have proposed estimates based on their own assumptions (Kenneth Jeyaratnam, Leong Munwai etc), most simply take figures they find quoted by various foreign publications without any idea how reliable the data is.
It is strange no one in Singapore academia in the fields of finance, economics and statistics, have shown any curiosity, absolutely none. Anyone with a math or statistics degree who hasn't tried figuring this out ought to be ashamed. I mean, if I were an engineer, I would be fascinated with China's mega projects, or as a technologist I would want to understand AI at the tech level. As a Singaporean I would want to know the size of the national reserves. Why. Because the government created the Brabara Streisand effect.
From army training of decades ago I learnt how triangulation of two data points allow a third to be discovered. Having taken statistics as a subject (which I had to take 3 sittings because back then it was the most boring subject in the world), I understood there are various tools available that can assist in determination of our reserves. I'm mouthing stuff like inferential statistics, regression analysis, Bayesian inference, interpolation and extrapolation, time series analysis. Our government's determination to remain opaque forces on any attempt to estimate the size of the reserves to make certain assumptions which of course impacts accuracy.
There are 3 data points required - the portfolio size of the sovereign wealth funds + MAS official foreign reserves (OFR), the NIRC (net investment returns contribution), and the interest rate. Knowing 2 of these, one can figure out the third. We all know the size of GIC is a tightly held state secret. And the MOF does not tell us what is the interest rate they use in computing the NIRC.
The math should be quite simple:
NIRC(Total) = NIRC(Temasek) + NIRC(MAS) + NIRC(GIC)
NIRC(Temasek) = 0.5 x Rate(Temasek) x Portfolio(Temasek)
NIRC(MAS) = 0.5 x Rate(MAS) x Net OFR(MAS)
NIRC(GIC) = 0.5 x Rate(GIC) x Net assets(GIC)
The knowns:
The variables in red are the knowns. The total NIRC is in the fiscal reports of MOF. Temasek's portfolio size is in their financial reports. The size of MAS' OFR is the mosr transparent, in their financial reports and updated in their website. monthly.
The rest are state secrets.
Estimating the total reserves and the GIC portfolio boils down to guessing the interest rates used. The closer one gets to the rate used by the government, the more reliable the estimates. Let's try to understand the NIRC.
It comprises of :
* Up to 50% of the Net Investment Returns (NIR) on the net assets invested by GIC, MAS, and Temasek; and
* Up to 50% of the Net Investment Income (NII) derived from past reserves from the remaining assets.
The NIR is based on 'net assets' which means net of borrowings. In the case of Temasek, we can assume the portfolio as the base. Temasek has some borrowings but it is negligible. In the case of MAS, it's earning assets is basically its official foreign reserves (OFR), but its open market operations are always sanitised by MAS Bills, which are debt that should be deducted to arrive at net OFR. In the case of GIC, its massive gross portfolio is bloated by funds from government securities (debt) which should be deducted to arrive at the net assets.
The NIR return rates are long term 20-year expected real rate. "Expected" means it is a projected rate, not historical. "Real" means it is adjusted for projected inflation. The 'return' includes capital gains, ie unrealised asset appreciation.
The NII is based on "remaining assets", ie, other than GIC, MAS and Temasek. This refers to other assets managed by other entities. The return is based on historical rates and does not include capital gains, in other words, it is based on actual cashflows. This would comprise all the other legacy investments of other government entities, such as rentals of EDB etc.
The fiscal reports do not break down the NIR and NII. However, the NII is expected to contribute only a small number to the NIRC. In our estimates here, NII is ignored.
Deciding on the long term rate of return:
The government has always kept us guessing is it 5, 10, 15 or 20 year long term expected real return rate they use? What is the difference? The lower the years, the more the returns are impacted by economic volatility. Therefore, it is a higher risk to distribute too much of the computed NIRC. On the upper end of the spectrum, the 20-year term returns are more smoothened out so higher amount can be paid out of the computed NIRC. Then again, on the other hand, the further into the future one projects, the more wobbly the figures get. A 20 year long term rate is more aligned with what the government has consistently mentioned, that the NIRC is a methodology for transferring wealth over generational lines.
The exact methodology they use is not publicly disclosed but one can reasonably infer the kinds of models, simulations, and economic assumptions they are likely to use, based on global best practices and what has been publicly shared.
The kinds of models they might use would probably include :
* Strategic Asset Allocation Models - this stimulates expected return distributions for different asset classes (equities, bonds, real estates, etc)
* Monte Carlo Simulations - this method runs thousands of simulations of economic outcomes from factors like interest rates, inflation, GDP growth, equity/bond returns, etc. Use of historical data and economic forecast and probability distribution allows for stress testing worst case scenarios.
* Macroeconomic models - this is a structural macroeconomic model that simulates the interaction of the economic environment - global GDP, real interest rates, CPI, fiscal and monetary policy impacts, etc.
* Forward-looking Capital Market Assumptions (CMAS) - throw in all the above factors mentioned and stress test for secular stagnation, high inflation, geopolitical shocks.
* Scenario Analysis & Stress Testing - test to see performance under various scenarios such as global financial crisis, prolonged low-growth periods, tech booms or collapses, etc. This is to ensure expected return is resilient.
Each of the 3 entities submit the return estimates to MOF which reviews and adjusts if necessary. An independent President's Council of Presidential Advisers (CPA) oversees and has access to these assumptions to ensure prudent use of reserves.
Temasek:
Temasek's last financial report shows a 20-year historical nominal Total Shareholder Return (TSR) of 7.5%. Using MAS' average historical inflation of 1.9%, Temasek's 20-year real return rate was 5.6%. We are not interested in the historical rate but the projected 20-year real rate. The international benchmarks for real long term returns on equities or blended portfolios mostly have the 20-year expected real return at 3-4%. For our purpose let's take the mid 3.5%.
NIRC(Temasek) = 0.5 x Rate(Temasek) x Portfolio(Temasek)
NIRC(Temasek) = 0.5 x 3.5% / S$389B
NIRC(Temasek) = S$6.8B
MAS:
As at 31 Mar 2024, the foreign currency reserves was S$517.3B. These foreign reserves are invested in very low yield short-term and liquid assets. In the past, built-up of excess foreign reserves is only transferred to GIC for better investing in longer term assets when the capital reserves of MAS had sufficient balances. However, the foreign reserves was building up at a rate much faster than the capital reserves. In 2022, the RMGS (Reserves Management Govt Securities) was used as a way to move excess foreign reserves out of MAS without having to wait for the capital reserves to build up. The RMGS is a non-marketable security issued by the government. and the balance in 2024 was S$260.2B. The OFR (Official Reserves of MAS) comprises of the foreign reserves, RMGS, and minor sums in gold and IMF special drawing rights. For our purpose here we take the gross investable funds of OFR as the foreign reserves + RMGS, ie $517.5B + $250.2B = $767.7B.
In its open market operations to manage the exchange rate, MAS purchase of foreign currencies are always done on a sanitised basis by issuance of MAS Bills. As at March 2024 MAS Bills outstanding was S$324.5B. Thus the net OFR should be foreign reserves $767.7B less MAS Bills S$324.5B = S$443.2B.
Estimating the return rate for MAS is tricky. The government pays interest on the RMGS but the rates are state secrets. Its investable assets is basically the foreign reserves. The bulk of the foreign reserves are parked in non-interest bearing accounts with central banks in other countries. A portion of it would be in low yielding liquid bonds. Most analysts estimate these bond yields which are possible based on market data and they apply such rates to MAS. However, this is fundamentally incorrect because MAS is not an investment entity. It is not a for-profit organisation. Therefore it is incorrect to use investing norms to assess MAS returns. Unlike investment houses, MAS does not hedge its activities. Its various positions are left exposed to market risks because hedging those positions would mean undoing their various policy operations. This means MAS returns are extremely volatile, especially with regards to exchange revaluations. Whilst its bond holdings provide a stable return, it's interest revenue is in fact from the repo transactions from its function as a lender of last resort for daily liquidity needs of banks.
MAS' returns are heavily impacted by exchange valuations. Given the trend of SGD appreciation, and government's policy of maintaining a strong currency, I would assume an expected long term real return in a very low range of 50 basis points. This is purely my estimation.
NIRC(MAS) = 0.5 x Rate(MAS) x Net OFR(MAS)
NIRC(MAS) = 0.5 x 0.005 x S$443.3B
NIRC(MAS) = S$1.1B
GIC:
NIRC(Total) = NIRC(Temasek) + NIRC(MAS) + NIRC(GIC)
Therefore:
NIRC(GIC) = NIRC(Total) - NIRC(Temasek) - NIRC(MAS)
NIRC(GIC) = S$24B - S$ 6.88 - S$1.1B
NIRC(GIC) = S$16.1B
It is now left to estimate the long term expected real return of GIC on its net assets. In 2024 GIC's 20-year long term real return was 3.9%, marking a steep drop from 4.6% the previous year. This 3.9% return is however, on it's total assets. If it is based on net assets, the return rate would have been significantly higher.
The Government Statement of Borrowings show a balance of S$1,168.5B as at Mar 2024 for various securities issued (SGS Market Devt, SSGS (for CPF), RMGS, Spore Savings Bonds). The proceeds of all these borrowings cannot be spent. We assume these funds are managed by GIC, that is, it is part of the GIC's AUM (assets under management), or gross assets.
AUM(GIC) = Debt(GIC) + Net Assets(GIC)
It is more meaningful to estimate the returns of GIC by looking at the Debt and Net Assets separately. Net Assets are funded by land sales, budget surpluses and capital reserves.
I am just making a fair estimate GIC is able to make a 50 basis points spread over their cost-of-funds when investing these borrowings.
NIRC(GIC-Debt) = 0.5 x Rate(GIC-Debt) x Debt(GIC)
NIRC(GIC-Debt) = 0.5 x 0.005 x $1,168.5B
NIRC(GIC-Debt) = $2.9B
Therefore:
NIRC(GIC-net assets) = NIRC(GIC) - NIRC(GIC-debt)
NIRC(GIC-net assets) = S$16.1B - S$2.9B
NIRC(GIC-net assets) = S$13.2B
GIC has a lower risk profile than Temasek, thus lower 20-year expected real return on its net assets, We estimated this expected return for Temasek at 3.5%. For GIC we put it at comparatively lower 3%.
Net assets(GIC) = NIRC(GIC-net assets) / 0.5 x Rate(GIC-net assets)
Net assets(GIC) = S$13.2B / 0.5 x 0.03
Net assets(GIC) = S$880B
GIC Net Assets Compared:
Several external analysts have estimated the size of GIC portfolio :
Global SWF - (Jul 2024) US$847B (S$1,101B)
Bloomberg - (Mar 2025) US$880B (S$1,144B)
Reuters - (Jul 2024) US$770B (S$1,001B)
Forbes - US$744B (S$967B)
I assume these are net assets and not the AUM (assets under management) or gross assets. These 4 external estimates of GIC net assets put it at a low of S$967B to S$1,144B with an average of S$1,053B. My estimate is $880B, 16% off the average and only 9% off the lowest estimates of Forbes.. It can only reflect I have been less conservative than the external analysts in rate estimation.
What is GIC Net Assets:
The AUM of GIC comes from 4 sources of funds - debt (proceeds of government securities), land sales, budget surpluses and capital reserves. The net assets are funded by land sales, budget surpluses and capital reserves.
One can try to cull all the data possible from archives and wherever to aggregate the land sales and budget surpluses that have been transferred to GIC to manage. But one will still be left with the opaque compounded gains in capital reserves to derive a more accurate figure for GIC's net assets.
How much then is the national reserves:
It depends on definition.
Reserves in the accounting sense refers to funds not available for general expenditure. These funds are represented by a mass of co-mingled assets. Based on this definition, our reserves in 2024 is :
National reserves (Gross) :
= Temasek portfolio + MAS OFR + GIC gross assets
= Temasek S$389B + MAS S$676.6B + GIC (S$880B + S$1.168.3B)
= Temasek S$389B + MAS S$676.6B + GIC S$2,048.5B
= S$3,114.1B
Reserves in the sense of sovereign wealth, ie national ownership, free of debt, then our reserves in 2024 is :
National reserves (Net):
= Temasek portfolio + MAS net OFR + GIC net assets
= Temasek S$389B + MAS S$443.3B + GIC S$880B
= S$1,712.3B
I think we can take these figures to the bank with a 90% confidence level.
Conclusion:
Key to getting the reserves figure right is the 20-year expected real return rate. The rate used to compute the NIRC is designed to ensure generational equity, a fair distribution of national wealth across both present and future generations. The government employs a wide array of financial models, simulations and risk assessment methodologies across a wide spectrum of investment products and robustly stress-tested against severe predicted economic scenarios into the future.
On paper the whole process looks extremely elegant and disciplined. But in its opaqueness the public cannot verify how the numbers are derived or how conservative the underlying assumptions are, it makes external benchmarking or critique impossible, there is no parliamentary scrutiny, as the government ultimately influences the assumptions the whole process blurs financial and political lines.
There is just a handful of people privy to the information, and fewer still who actually understands them. The President's Council of Presidential Advisers has an overseeing role to help ensure prudent use of our reserves. Do they really understand the whole methodology? They were unable to explain to President Ong Teng Cheong decades ago and the poor chap went away asking for a list of assets that made up the reserves.
Remember the lessons from history from the world of finance. Baring Bros management thought they understood all those risk reports crunched by the system until the day rogue trader Nick Leeson bankrumpted the oldest merchant bank in UK. The 2008 global financial crisis caused by the collapse of the mortgage-backed securities is a clear illustration of the limits of financial modelling, no matter how sophisticated. The crisis was due to a systemic misjudgement of risks, not because the models were mathematically wrong. Despite the presence of highly credentialed risk managers, rating agencies, quants (quantitative analyst specialists) and economists, the system collapsed under risks they thought were under control.
If Wall Street, with all its modeling and talent, failed to see its own crash coming, why should citizens unquestioningly trust opaque assumptions used in sovereign wealth forecasts? This is not a call for panic. It's a call for healthy skepticism, transparency and accountability.
In seeking Truth, Diogenes would have certainly asked: "Can a country claim financial transparency when its most important revenue stream for fiscal spending is built on invisible assumptions?"
This platform has withdrawn it's subscriber widget. If you like blogs like this and wish to know whenever there is a new post, click the button to my FB and follow me there. I usually intro my new blogs there. Thanks.