Is Gold a Good Investment? Why Central Banks' Reason Isn't Yours

In the second quarter of 2026, central banks added an estimated 289 tonnes of gold, net of what they sold. The World Gold Council called it a record for a second quarter. If you have read anything about gold this year, you have met the argument that travels with that number: the most cautious institutions on earth are buying, so it must be a good investment.

The buying is real. The conclusion does not follow from it, and the reason fits in two short sentences. Central banks are solving a reserve problem. You are solving a portfolio problem. Same metal, two jobs — and the answer for the second one is much less exciting than the headline.

Is gold a good investment? First decide what job it has

The gold argument has two camps, and both have been right for a very long time. Camp one says gold has outlasted every currency, every government and every crisis it was held through. Camp two is the Buffett position: gold produces nothing. No dividend, no interest, nothing built. You dig it out of one hole in the ground and pay somebody to guard it in another.

Both are true, which is why the argument never ends. Notice that both camps argue about what gold is. Neither asks what it would do in your portfolio, and that is the only question a bank asks about any asset.

The version I learned sorts everything with one question: is it an engine, or is it insurance? An engine pays you out of what it produces. Insurance pays you nothing until the day you need it, and costs you something every year you don’t.

Asset	Pays you out of	Compounds?	Its job in a portfolio
Shares	Company earnings, paid out or reinvested	Yes	Engine: long-run growth
Bonds	The borrower’s coupon	Yes, at the agreed rate	Engine with a brake: income and ballast
Rental property	Rent, after the landlord’s costs	Yes	Engine, with a second job attached
Cash	Short-term interest	Barely, after inflation	Liquidity: bills and the emergency fund
Gold	Nothing; only a later buyer’s price	No	Insurance

Every gain you will ever make on gold comes out of the next buyer’s pocket. That is not a criticism of gold. It is a description of what you own.

Over a long enough period the description turns into a number. The UBS Global Investment Returns Yearbook 2026, by Elroy Dimson, Paul Marsh and Mike Staunton, puts the real return of gold in US dollars at 1.3% a year from 1900 to 2025. Over the same 126 years, US equities returned 6.6% a year after inflation, US bonds 1.6% and Treasury bills 0.5%.

Real returns, USA, 1900–2025 · UBS Global Investment Returns Yearbook 2026

1.3% against 6.6% a year looks close. Over 126 years it is $5 against $3,296

Left, the return per year after inflation. Right, what one dollar became over the whole period, on a scale where each step is ten times the last.

This is what insurance looks like, not a flaw. Gold kept its purchasing power and then some. It did not compound, because nothing inside it produces anything that could be reinvested.

A fifth of the yearly return sounds like a gap you could live with. Compounded over a lifetime and a half, it is the difference between a dollar that became $5.20 and a dollar that became $3,296. Insurance is supposed to look like this. Nobody expects a fire policy to pay a return either.

One footnote most gold pages miss — and it matters more than it looks. After inflation, gold’s price doubled between December 2023 and December 2025 (my calculation from World Bank monthly prices and the US consumer price index). Take that doubling out of the Yearbook’s 5.2-fold and the 124 years to the end of 2023 come to roughly 0.8% a year. Nothing about gold’s history changed — two years were added, and they lifted the long-run average by more than half. An average that a single rally can move that far is telling you how lumpy the asset is. It is also a warning about quoting any gold return without its edition and its window.

Does gold protect you in a stock market crash?

Sometimes, briefly, and less reliably than its reputation. Gold’s correlation with shares is low on average, but “low on average” and “rises when shares fall” are different claims.

The two crashes most people remember make a fair test. Both are measured the same way below, gold and the S&P 500 on the same closing dates.

Window	Gold	S&P 500
2008: gold’s high on 18 March to its low on 13 November	$1,004.30 → $705.00, −30%	−32%
12 to 22 September 2008 (Lehman files on the 15th)	$764.50 → $909.00, +19%	−4%
22 September to 23 October 2008	$909.00 → $714.70, −21%	−25%
2020: stocks’ peak on 19 February to their low on 23 March	$1,611.80 → $1,567.60, −3%	−34%
9 to 18 March 2020, gold’s sharpest fall	$1,675.70 → $1,477.90, −12%	−13%

Gold: COMEX gold futures, front month, daily close; S&P 500: daily close. Both via Financial Modeling Prep, my calculation.

So the story is right that gold jumped when Lehman failed. Then it handed all of that back — and more — in the month when shares fell hardest, and by 23 October it stood 6.5% below where it had been on the Friday before Lehman. From its March high to its November low, gold lost almost as much as the stock market did over the same dates.

2020 is messier, which makes it more useful. Measured over the whole crash, gold did its job: shares lost a third, gold barely moved. Inside that crash, for the seven trading days from 9 to 18 March, gold fell as fast as shares did. It was one more thing people sold when they needed cash. Both statements are true, and which one you remember depends on which week you were living through.

The research says the same thing more formally. Dirk Baur and Brian Lucey (The Financial Review, 2010) found gold to be a hedge against stocks on average and a safe haven in extreme market conditions, with one qualification in their own words: “the safe haven property is short-lived”. The working-paper version of the same study puts a number on it: gold was a safe haven “only for 15 trading days after an extreme shock”. Baur and Thomas McDermott (Journal of Banking & Finance, 2010) found the safe-haven effect in the US and the major European markets, but not in Australia, Canada, Japan or large emerging markets such as the BRIC countries.

Fifteen trading days is useful to someone rebalancing a fund. It is not what most people mean when they say gold will protect them. And if your portfolio is priced in Australian or Canadian dollars, the research suggests the protection may not have shown up at all.

On inflation, the channel has a separate article on what gold actually tracks. The short version from the same Yearbook: of the 28 years since 1900 in which US inflation ran above 3%, gold lost money in 13.

Why are central banks buying gold?

Start with the size, because it is genuinely large. In 2025 central banks bought an estimated 863 tonnes of gold, net. That is about 80% more than their average year from 2010 to 2021 (473 tonnes), and it is also 21% less than in 2024. The first half of 2026, record second quarter included, was the weakest first half since 2022.

Central-bank net gold purchases, tonnes · Metals Focus / World Gold Council

More than half of 2025's central-bank buying was never reported by a central bank

Three years above 1,000 tonnes, then a step down in 2025, still far above the 2010–2021 average. The amber part of 2025 is the estimate of purchases no central bank disclosed.

The 2026 numbers move after publication. The first quarter of 2026 was published at 244 tonnes and revised three months later to 57. The second quarter came in at 289 tonnes, a record for a second quarter, and the half-year total of 345 tonnes was still the lowest for a first half since 2022. Half-years and full years are different windows, which is why 2026 is not a column here.

Two things are worth knowing about that number before anyone uses it to sell you something.

Most of it is an estimate. The World Gold Council’s own full-year report describes “substantial opaque activity”, 57% of the annual total in 2025. More than half of the 863 tonnes was not reported by any central bank. The figure is modelled by Metals Focus, an independent precious-metals consultancy, for the World Gold Council, which is the market-development body of the gold-mining industry. I am not suggesting anyone invents purchases. Gold Demand Trends is the standard reference in the market. But the series moves when it is revised: the first quarter of 2026 was first published at 244 tonnes and revised three months later to 57, a cut of 77%. The honest description of the headline is “estimated net purchases”, and most of the estimate concerns buyers who chose not to say.

And the buyers are not talking about returns. I have never heard a reserve manager talk about return. They talk about access, about counterparties, about what happens if a settlement system stops taking your call. That is a different vocabulary from the one you and I use for our own money, and it points to the real reason.

It starts on a date: February 2022. After Russia invaded Ukraine, the European Union, the other G7 countries and their partners immobilised Russia’s central-bank reserves held in their jurisdictions — close to €300 billion worldwide by the European Parliament’s research service’s count, €210 billion of it in the EU. Immobilised, not confiscated — the capital is frozen, and its legal future is still being argued.

Here is why that one decision reached every reserve manager on earth. A currency reserve is not cash in a vault. It is mostly other governments’ bonds and deposits, held inside other countries’ financial systems. A bond is somebody else’s promise, and a promise is only as safe as the one who makes it, including their willingness to keep making it to you. Gold held in your own vault, under your own law, has no issuer. It is the one reserve asset that is nobody’s promise.

So yes, central banks are buying insurance. They are insuring a currency reserve against the day access itself becomes political.

Be honest for a moment: is that your risk? For most people reading this, the risks that can actually damage a financial life are a bout of inflation, a market crash, the loss of an income, and their own behaviour in a panic. Gold is an imperfect answer to the first two and no answer at all to the last two. Same metal, a completely different policy, against a completely different problem.

If you want proof that these really are two different problems, one country holds the whole argument.

Norway · two institutions · 2004 and 2025

One country, two institutions, and neither keeps gold as a reserve

One manages the country's long-term savings, the other its currency reserves. Neither decision is about the gold price. Both are about what the money has to do.

The savings

Government Pension Fund Global

Built to compound national wealth across generations

NOK 21,268bn

about $2.1 trillion, 31 December 2025

• Equities 71.3%
• Fixed income 26.5%
• Unlisted real estate 1.7%
• Renewable infrastructure 0.4%
• Gold not an asset class in the mandate

The reserves

Norges Bank

The central bank, which manages the currency reserves

33.5 tonnes sold

gold bars, first quarter of 2004, for $447.4 million

“Gold only accounted for just over one per cent of the Bank's international reserves, and thus contributed little towards diversifying the risk associated with the reserves. The return on gold has historically been low.”

Norway runs one of the largest sovereign wealth funds in the world: NOK 21,268 billion at the end of 2025, a little over $2 trillion at the year-end exchange rate. It exists to compound national wealth across generations, and it holds 71.3% equities, 26.5% fixed income, 1.7% unlisted real estate and 0.4% renewable energy infrastructure. Gold is not an asset class in its mandate. (It will own gold-mining shares as part of its global equity index, which is why “Norway holds no gold” is the wrong sentence.)

Norway’s central bank, which does manage reserves, sold its gold bars, 33.5 tonnes of them, in the first quarter of 2004. It said why, and the sentence is worth reading slowly: gold “only accounted for just over one per cent of the Bank’s international reserves, and thus contributed little towards diversifying the risk associated with the reserves. The return on gold has historically been low.”

One country, two institutions, and neither of them keeps gold as a reserve. That is not Norway rejecting gold twice — the fund never had it, and the bank sold what it had. Today’s buyers have a different reserve problem from Norway’s in 2004. What has not changed is the reason given at the time: when an institution chooses its assets by their job, gold earns its place through what it protects against, never through what it pays.

What does a bad entry price cost?

Insurance sounds harmless. This particular policy has a history of punishing people who bought it at the wrong price.

Gold after inflation · monthly average · December-2025 dollars · 1980–2025

Bought at the 1980 peak, gold took 45 years to get back to even

The dashed line is the January 1980 record, carried forward with US consumer prices. Every month in the shaded area is a month in which that buyer was still behind.

Two rallies that looked like a comeback were not. The 2011 and 2020 highs made headlines in dollars. After inflation, both stopped about 30% short of a price first paid in January 1980. Being right about gold in dollars and being right about it in purchasing power are two different claims.

Gold set its record fix of $850 on 21 January 1980. Five years later its real price was down about two-thirds. The exact figure depends on how you measure, and the two common methods give different answers, so here are both, from the same data:

Peak	Measured as	Five years later, nominal	Five years later, after inflation
January 1980	January 1980 to January 1985, monthly averages	−55%	−67%
	1980 to 1985, calendar-year averages	−48%	−60%
September 2011	September 2011 to September 2016, monthly averages	−25%	−30%
	2011 to 2016, calendar-year averages	−20%	−25%

World Bank monthly gold prices, deflated with the US CPI-U; my calculation. Daily prices give larger falls from the single peak day. Neither basis is wrong, as long as they are not mixed.

Then the number that should stay with you. Adjusted for US inflation, the 1980 record was not passed again until September 2025. Bloomberg News reported the moment at the time, putting the inflation-adjusted 1980 high at about $3,590 against a new nominal record of $3,674.27. That is 45 years. Measured average against average instead of against the single peak day, the crossing comes in February 2025, so the 45 does not depend on the method.

The people who bought at the top in 1980 were buying insurance too. They paid so much for the policy that it cost them a working lifetime. Claude Erb and Campbell Harvey made the general point in their study of gold, The Golden Dilemma (2013): gold may well be a long-run inflation hedge, but “the long-run may be longer than an investor’s investment time horizon or life span.”

That is history, not a forecast. It says nothing about what gold does next. It says the price you pay for insurance is most of the decision.

Should you buy gold right now?

First, the price, with a date on it, because it moves every day. On 14 October 2026 gold traded at about $4,100 an ounce, roughly a quarter below its record of $5,589 on 28 January 2026. Anyone repeating a “record high” headline today is quoting a price from January. You would be buying after a large fall, not at a peak. That is a different decision from the one in the headlines, and in some ways a harder one, because a fall feels like a discount whether or not it is one.

What I will not give you is a forecast. The most searched gold question is the price prediction, and it is the least useful one: no forecast tells you whether gold has a job in your portfolio, and without a job, a correct forecast is just a trade.

So here is the answer to the question in the title. Gold can have a place, a small one, but not for the central banks’ reason.

• Yes, if you want a piece of insurance, you can name the risk it covers, and you accept that it may go long stretches paying you nothing at all.
• No, if you are buying because central banks are buying. They are insuring a currency reserve. You are building a portfolio.
• No, if you are buying because it has gone up. That is the worst reason to buy anything, and it is the one thing gold has punished reliably.

You do not need gold to build a sound portfolio. You may hold it for a good reason, and two people can hold it for two different good reasons. The strongest argument for gold right now is an argument for somebody else’s balance sheet.

How much gold should you hold?

I won’t give you a number, because the right one depends on what you are insuring against, and that is the point of the exercise. What I can give you is the arithmetic that decides whether a position is sized sensibly.

Gold’s share of your portfolio	If gold falls 30%	If gold rises 30%
2%	−0.6 points	+0.6 points
5%	−1.5 points	+1.5 points
10%	−3.0 points	+3.0 points
25%	−7.5 points	+7.5 points

A 30% fall is not a stress scenario for gold; it is roughly what happened between March and November 2008. Read the table both ways: a position big enough to rescue your portfolio is big enough to wreck it. Size it so it can never save you and never ruin you, and decide the band around it before you buy, so that a headline never has to make the decision. How to set and use that band is in rules over access, and where an insurance position sits in a structured portfolio is in the three-bucket structure. Once the band exists, rebalancing does the rest.

Before any of that, the sentence from the video. Finish it out loud: I’m holding this in case… If you can finish it, you are buying insurance. If you can’t, you are buying a trend.

Before you buy any gold · one card, your answers

If you hold gold as insurance, write the policy down

Five lines. Line 1 is the one that decides. The others only matter once it is filled in.

Policy schedule: gold

held by: me · reviewed: once a year

1 I’m holding this in case…

Name the risk in one line. If it will not finish, you are buying a trend, not insurance.

your answer

2 Share of my portfolio

Small enough that it can never ruin you, which also means it can never rescue you.

your figure %

3 If gold falls 30%, my portfolio loses

Line 2 × 0.3. A round test size: gold fell 30% from its own high in 2008.

your figure points

4 What holding it costs me a year

Fund fee, or storage and insurance for metal. Set it against 1.3% a year, gold’s real return since 1900.

your figure %

5 I add or trim when it is outside

A band around line 2, decided today. Not a headline, not a price target.

your figure % to %

What does it cost to hold gold?

Gold pays you nothing, so every cost comes straight out of a return that is small to begin with. That makes the holding cost far more important for gold than for shares.

How you hold it	What it costs	Against gold’s 1.3% a year real return
The cheapest large US gold funds backed by physical metal	0.10% a year, one a little less	under a tenth of it
The largest such fund	0.40% a year	almost a third of it
Coins and small bars	the gap between the dealer’s selling and buying price, plus your own storage	check it on the day, at your dealer
Metal in a vault	a yearly storage and insurance fee	ask for it as a percentage of the value

Fund fees as published by the issuers in October 2026.

In the video I say the best-known gold fund charges five times what the cheapest ones charge. It is four times: 0.40% against 0.10% a year. The point stands, a little smaller.

The video also gives a range for coins, 5 to 10% to get in and out, plus up to 1% a year to store and insure. I could not tie those figures to a published survey I can link, so they are not in the table. You don’t need my range anyway — you can get a better one in two minutes. Ask a dealer for the price at which they sell a coin and the price at which they would buy the same coin back today. The gap between the two is your round trip, before gold has moved at all, and with a real return of about 1.3% a year it takes years to earn back. Do the same with a vault’s annual fee.

Dealer margins, sales-tax rules and buy-back terms differ a lot between the US, the EU and individual countries. Tax on gold depends entirely on where you live and on the product, and is worth checking with someone qualified before you buy. For the mechanics of choosing a product, the same channel’s gold article goes through them, and the fee calculator shows what a yearly charge does over a long holding period.

The honest summary

Central banks aren’t buying gold because they found a magic investment. They are insuring a reserve against a risk you don’t have. If you own gold, own it on purpose: as insurance for a risk you can name, in a size you chose, at a cost you checked. Never ask it to be the engine. The engine is the rest of your portfolio.