Key takeaways
- The World Gold Council's 2026 study back-tests gold sleeves over 2005 to 2025. A 5% weight lifted annualised return from 6.7% to 7.0% and cut the worst drawdown from -34.9% to -32.7%.
- That 20-year window starts in 2005 and excludes what matters most: an investor who bought at gold's January 1980 monthly average waited until February 2025 — 45 years — to break even after inflation.
- The real drawdown in between was -82.9%, from $2,879/oz in January 1980 to $491/oz in April 2001, both in June-2026 dollars.
- Gold's diversification is unreliable at the moment of stress. Between 17 March and 24 October 2008 gold fell 29.5%, from $1,011.25 to $712.50, in the middle of the crisis it is supposed to hedge.
- Gold is 25.3% below its record. The LBMA afternoon fix was $4,039.75/oz on 13 July 2026, against a record fix of $5,405.00 on 29 January 2026.
Two people buy gold. The first buys in January 1980, at the top of an inflation panic, at the month's average price of $675.31 an ounce. The second never buys a lump at all: she holds a 5% sleeve and rebalances it, year after year, decade after decade. Which one would you rather have been?
You can probably guess. What's harder to guess is how long the first one waited to find out. In June-2026 money, that $675.31 ounce cost $2,879. By April 2001 it was worth $491. And it didn't climb back to $2,879 until February 2025 — 45 years and one month later.
Hold on to that pair, because almost every argument about how much gold to own is really an argument about which of the two you'd turn out to be.
How much gold in a portfolio: the evidence is thinner than the number suggests
If someone has told you to hold 5% to 10% of your portfolio in gold, the number traces back to a specific body of research, and that research is real. It's also narrower than the confidence with which the number gets repeated.
The current version is the World Gold Council's Gold as a Strategic Asset: 2026 Edition. It builds a hypothetical portfolio — 50% stocks, 40% bonds, 10% alternatives — and back-tests what would have happened if 2.5%, 5%, 7.5% or 10% of it had been gold, using monthly US-dollar returns from 31 December 2005 to 31 December 2025.
Here's the whole result, in the Council's own table. Over those 20 years the portfolio without gold returned 6.7% a year with 9.9% volatility and a worst drawdown of -34.9%. The same portfolio with a 5% gold sleeve returned 7.0% a year with 9.6% volatility and a worst drawdown of -32.7%.
That's the case. Three-tenths of a percentage point of extra annual return, and about two percentage points shaved off the worst fall. Is that enough to build a rule around? It's your call, and it's an easier call once you've seen the improvement at its actual size rather than at the size the headline implies. It is a genuine improvement, and it is a small one.
Notice what the Council does not say. It tests four weights and reports what each of them did. It does not recommend 5% to 10%; its own wording is that the "'optimal' amount of gold varies according to individual asset allocation decisions." The tidy range is something the retelling added.
The World Gold Council's 29 members are gold miners and gold royalty companies
Does that make the arithmetic wrong? No. The numbers come from Bloomberg and ICE Benchmark Administration data and the method — a resampled optimisation run in Portfolio Visualizer — is disclosed on the page. Anyone can check it.
It does mean you're owed the provenance. The World Gold Council is a trade body. It has 29 members — gold miners including Barrick, Newmont and AngloGold Ashanti, alongside gold royalty and streaming companies such as Franco-Nevada, Wheaton Precious Metals and Royal Gold, which finance mines rather than dig them. The Council's own line is that its members "are some of the world's most forward-thinking gold mining companies." It exists, in its own words, to champion the role gold plays as a strategic asset. Research that concluded gold had no place in a portfolio would be a strange product for it to publish.
The honest way to handle a source like that isn't to dismiss it. It's to read its evidence carefully and then go looking for what it left out. In this case, what it left out is easy to find, because it's the choice of start date.
The gold drawdown that took 45 years to break even after January 1980
The study window opens on 31 December 2005. Gold's great disaster ended in 2001.
So here is your first buyer in full, with the method attached, in case you want to redo it yourself. Take the LBMA gold price, the industry's own benchmark, and deflate it by US consumer prices. In January 1980, at the top of the inflation panic, gold averaged $675.31 an ounce. In June-2026 money, that's $2,879. By April 2001 the real value of that same ounce had fallen to $491. A drawdown of 82.9%, after inflation, spread over 21 years.
Then the wait. It wasn't until February 2025 — 45 years and one month after the peak — that the real gold price climbed back to where it had been in January 1980. Someone who bought at the peak London PM fix of $850 on 21 January 1980 waited longer still: that price is worth about $3,624 in today's money, and gold's monthly average didn't clear it until September 2025.
Now imagine living inside that. Not the arithmetic — the experience. Twenty-one years of the ounce in your safe quietly losing ground, then more than twenty of it clawing the loss back, and nothing arriving in the post the whole time to soften any of it. Compound the whole 46 years and gold's real return comes to roughly 0.8% a year. Positive. Barely.
None of this proves gold is a bad holding. It proves that the answer you get depends enormously on where you start counting, and that a 20-year study starting in 2005 cannot tell you what a 45-year drought feels like from the inside.
Gold pays nothing, and that is why the waiting is so expensive
An ounce of gold in 2026 is exactly the same ounce it was in 1980. It doesn't pay a dividend, it doesn't pay a coupon, it doesn't retain earnings and reinvest them. Its entire return is the price someone else will pay you for it.
So what does it pay you to wait? Nothing, and on a 45-year round trip that turns out to matter more than the price chart does. This is the structural difference that the drawdown numbers above understate. A shareholder in a business that fell 80% still collects dividends while they wait, and the business can buy back stock at the low price. A bondholder collects a coupon. You, holding gold, collect nothing, and the position runs at a small negative carry. How much depends on which form of gold exposure you own. SPDR Gold Shares (GLD), the largest gold ETF, charges 0.40% a year, and the trust pays that fee by selling gold. Its prospectus is explicit about the consequence: the trust "regularly sells gold to pay for its ongoing expenses," and so "the amount of gold represented by each Share has gradually declined over time."
Say you hold it in a taxable US account, too. The IRS taxes net gains on collectibles at a maximum long-term rate of 28%, against 0%, 15% or 20% for ordinary long-term capital gains. Gold bullion counts as a collectible, and so does an interest in a trust that holds it — GLD's prospectus spells out that a gain on the shares is taxed at that same 28% ceiling. On a taxable holding, the same gain hands over more.
A gold allocation diversifies on average and lets go at the worst moment
The strongest version of the gold case isn't the return. It's that gold moves differently from shares, and so it makes a portfolio steadier. On average, over long periods, that is true.
The problem is that "on average" hides the part you actually care about. In 2008 gold did what it was supposed to do over the calendar year: it ended December 2008 about 1.6% higher than December 2007 in monthly-average terms, while US share prices fell 43.7% over the same twelve months. But look inside the year. Between 17 March and 24 October 2008, as the crisis peaked, gold fell from $1,011.25 to $712.50 — a 29.5% fall — before it recovered. In the weeks when a hedge is most needed, gold was being sold too, most likely by people who needed cash and sold what they could.
Picture that from inside your own account. Your shares are collapsing. The thing you bought precisely to offset them is down nearly 30% as well, and nobody can tell you when it stops. Would you still have been holding it in October?
2022 is the better advertisement. The S&P 500 fell 19.4% over the calendar year, from 4,766.18 to 3,839.50, and bonds fell with it. Gold finished essentially flat, up 0.4% between the last LBMA fix of 2021 ($1,805.85 on 30 December) and the last of 2022 ($1,813.75 on 29 December). It did not make money. It simply refused to lose any, which in that year was the whole point.
Two crises, two different behaviours. Anyone who tells you gold reliably rises when equities fall is describing 2022 and quietly skipping March 2008. That inconsistency isn't unique to gold, which is why it's worth checking what actually held up across past crises rather than what is supposed to.
How much gold a portfolio holds and when you buy it are two decisions
This is the distinction that the current mood makes hard to see, and it's the one worth being clear about, because it is the difference between your two buyers.
Gold has had a spectacular few years. The LBMA annual average price was $3,431.54 in 2025, up from $2,386.20 in 2024. Measured year-end to year-end it rose 67% during 2025, from $2,609.10 to $4,367.80. It then set a record fix of $5,405.00 on 29 January 2026. Central bank buying has been a large part of the story — they added a net 244 tonnes in the first quarter of 2026 alone.
And gold has since fallen. The afternoon fix was $4,039.75 on 13 July 2026, 25.3% below that January record. Central banks weren't one-way buyers either: the same World Gold Council report identifies 115 tonnes of official-sector selling in the quarter, with Turkey the largest seller.
A 5% strategic sleeve, rebalanced, is a claim about how gold behaves across decades. Buying gold now, at these levels, is a claim about what happens next. The research answers the first question. It says nothing at all about the second — and a 25% fall inside six months is a reminder that the two aren't the same trade. What gold's own record highs have and haven't predicted is a separate question with its own evidence. If you already run a target weight, the question you have is a rebalancing question, not a gold question, and it has a much duller answer.
What the data cannot tell you about how much gold a portfolio should hold
Every figure above is a backward-looking measurement, and backward-looking measurements are not forecasts. The 1980–2001 collapse followed a genuine mania and the end of a great inflation; it is not obvious that it is the right template for anything. Equally, the 2005–2025 window that makes gold look good contains a financial crisis, a decade of near-zero interest rates and a pandemic — a very unusual 20 years to extrapolate from. Both samples are small, and they disagree. The property version of the same claim has a longer record and a similar shape, which is what our test of the property inflation hedge across five inflation regimes finds: real house prices kept pace over decades and stalled in the high-inflation years themselves.
What would change the conclusion, in either direction:
- If central bank demand reverses. Official-sector buying has underpinned the price. The 244 tonnes reported for the first quarter of 2026 is a net figure — the 115 tonnes of identified selling is already subtracted out of it, which implies gross buying of roughly 359 tonnes. So the selling is not a small footnote to the buying; it is a real and growing offset inside it, and a sustained flip would remove a support that the 20-year back-test never had to live without.
- If gold sells off with equities in the next real crisis. March 2008 says it can. If it happens again, the diversification premise — the entire case, since the return case is 0.8% a year in real terms — weakens badly.
- If real interest rates stay high for years. The opportunity cost of holding an asset that pays nothing rises with the yield on assets that do. That was the mechanism of the 1980s and 1990s.
Critics of the strategic-allocation case argue that the whole thing is a statistical artifact of a favourable start date, and the 45-year real drawdown is their best evidence. The Council's defenders reply that a portfolio is not a single lump-sum purchase at a mania peak, and that a rebalanced 5% sleeve bought steadily through the 1980s and 1990s would have looked nothing like that chart. Both are right.
So which of the two people at the top is you? Nobody sets out to be the first one. They just buy when the price is in the news, which is the only time gold usually is. The difference isn't conviction or timing skill. It's whether you decided the weight before you saw the price — and that is a fact about you, not about gold.