Understanding gold

1. A century of gold returns, analysed

One hundred dollars placed in gold at the end of 1928 grew to about $21,000 by the end of 2025, a nominal return of roughly 5.7% a year, or about 2.5% a year once inflation is stripped out, meaning over and above the rising cost of living. The same $100 in US stocks, with dividends reinvested, grew to roughly $1.16 million, about 10% a year nominally and 6.8% in real terms [1]. Over a full lifetime, equities compounded at several times gold's rate, for a structural reason: a share is a claim on a business that reinvests its earnings, while an ounce of gold pays no dividend, coupon or rent and so cannot compound. What gold did instead was hold its purchasing power across a century that broke the gold standard, abandoned fixed exchange rates, and lived through repeated bouts of inflation.

But an average return hides as much as it reveals. Look beyond the century-long figure and what you find is a sequence of very different decades. To understand gold you have to understand those decades, and the story begins not in a market but in a law.

For the first four decades on the chart, gold's price barely moved, because it was not set by a market at all. It was fixed by the US government: at $20.67 an ounce under the classical gold standard, and then, after the Gold Reserve Act of January 1934, at $35 an ounce. That act was a deliberate devaluation. It cut the gold value of the dollar by roughly 40% to fight Depression-era deflation [12], and the $35 price then held for thirty-seven years. A flat line on the chart from 1934 to 1971 does not mean gold was stable in any economic sense; it means its price was administered.

The peg ended on 15 August 1971, when President Nixon suspended the dollar's convertibility into gold and effectively closed the gold window. The reason was arithmetic. The United States had issued far more dollars than its gold reserves could back, partly to fund the Vietnam War and domestic spending, and foreign governments, France most aggressively, had begun redeeming their dollars for American gold and draining the reserves. Rather than keep honouring $35 an ounce, Washington let the link break. For the first time in the modern era, gold was free to find its own price.

Its initial move was wild. From $35 in 1971 gold ran to a peak around $850 an ounce in January 1980, a more than twentyfold rise in nine years. The decade gave it every reason to climb: stagflation, the combination of high inflation and stagnant growth the post-war framework had no answer for; two oil shocks, the 1973 OPEC embargo and the 1979 disruption after the Iranian Revolution, which pushed energy prices and inflation higher; a falling dollar; and a run of geopolitical shocks, the Iran hostage crisis and the Soviet invasion of Afghanistan, both in late 1979. Gold was what people reached for when paper money and the political order both looked unreliable.

Then it had two horrible decades. From the 1980 peak gold fell and drifted lower until it bottomed near $250 an ounce in 1999. The main cause was monetary policy: Federal Reserve Chairman Paul Volcker had pushed interest rates toward 20% to break inflation, and a high real return on cash and bonds is the natural enemy of an asset that yields nothing. Inflation fell, the dollar strengthened, and equities began a historic bull market, so investors had little reason to hold gold. Central banks felt the same way: many treated their bullion as a relic and became net sellers, and the selling itself drove the price down. It fell so far that in September 1999 European central banks signed the Washington Agreement to cap their own sales and halt the rout [13], while the United Kingdom sold roughly half its reserves near the very bottom, a sale still remembered as Brown's Bottom.

The low of 1999 to 2001 began the next climb. From around $270 an ounce gold rose, with only short interruptions, to a record of $1,921 on 6 September 2011 [14]. The drivers were the opposite of the forces that had held gold down in the 1990s: a weakening dollar, the bursting of the dot-com bubble in 2000 and 2001, the shock of 9/11, the 2008 failure of Lehman Brothers and the financial crisis that followed, successive rounds of quantitative easing, and real interest rates that turned negative. The 2011 peak itself sat on top of three fears at once, the euro-zone sovereign-debt crisis in Greece, Ireland and Portugal, the loss of the United States' AAA credit rating in August 2011, and money-printing on an unprecedented scale, all of which made a finite, unprintable asset look attractive. By then central banks had switched sides again, from net sellers to net buyers, where they have stayed.

After 2011 gold fell once more, to around $1,050 by the end of 2015, as the Federal Reserve signalled the end of easy money (the 2013 taper tantrum), the dollar strengthened, and the inflation many feared failed to arrive. The most recent climb began there and has been the largest yet: gold passed $2,000 during the 2020 pandemic, $3,000 in 2025, and spiked above $5,000 an ounce in February 2026 during the war between the United States, Israel and Iran, before falling back below $4,000 by the middle of that year [5]. Russia's full-scale invasion of Ukraine in February 2022 gave the climb a further, structural push: the sanctions that followed froze a large part of Russia's central-bank reserves held abroad, a reminder to every other central bank that dollar and euro reserves can be frozen, which accelerated record official-sector gold buying and a broad move to diversify reserves away from the dollar.

Seen this way, gold's century is not a smooth upward line but a sequence of negative periods broken by a few near-vertical climbs, and almost the entire return is delivered in those bursts. The decade an investor happens to hold gold in matters more than the century-long average. The chart below shows the free-float era from 1971 onward, with the events that moved the price marked. A note on the levels: they are annual figures, so they sit below the intraday extremes. Gold touched about $850 in January 1980 and $1,921 in September 2011, both higher than the yearly numbers shown.

2. Does gold keep pace with inflation?

Gold is a hedge against inflation: it is one of the most often repeated claims in all of investing. Part of why the belief is held so firmly is that gold genuinely rescued the portfolios that owned it in the early 1970s, when inflation ran hot, equities, bonds and cash all lost ground in real terms, and gold's price surged at the same time. That episode was real, and it left a deep mark. But examine the other inflationary stretches in the record and the relationship turns out to be far more nuanced than the slogan suggests.

The answer depends entirely on the horizon. Over very long spans, gold has roughly held its purchasing power: the same ounce buys a broadly similar basket of goods across generations. Over the horizons most investors use, five to twenty years, the link is weak. In the most cited study of the question, The Golden Dilemma, issued in 2013 by  Claude Erb and Campbell Harvey found that the correlation between gold and realised inflation over typical investment horizons is close to zero, and that gold's inflation-adjusted price wanders far from any stable level for decades at a time [6]. Two episodes make very good examples. Through the long bear market that followed the 1980 peak, gold lost roughly 70% of its value from peak to trough by 1999, even as US consumer prices themselves roughly doubled over those two decades. More recently, and on a far shorter horizon, gold fell from a February 2026 high above $5,000 an ounce to below $4,000 within months, even though the war with Iran had just pushed inflation sharply higher.

So the relationship is real but slow. An investor using gold as a hedge against next year's consumer price index is relying on a short-run relationship the data does not show; the long-run store-of-value attribute and the short-run inflation-hedge attribute are not the same thing.

3. Gold and real interest rates

The clearest ongoing relationship gold has is with real interest rates, the yield on inflation-protected government bonds. Because gold pays no income, its main competitor is the real return available on safe bonds. When real yields fall, the opportunity cost of holding a zero-yield asset falls with them, and gold has tended to rise; when real yields rise, that cost increases, and gold has tended to fall. On the chart, the gold price and the 10-year US TIPS real yield, plotted directly on a second axis, have moved broadly opposite to each other for most of the past two decades [4].

The relationship is strong but not mechanical, and it has broken down at times. The clearest break came from 2022 onward, when real yields rose sharply yet gold rose as well, the opposite of the usual pattern. The widely cited explanations are heavy central-bank buying and geopolitical demand, forces the real-yield model does not capture (these are taken up in sections 6 and 9). The point for a reader is that real yields explain much of gold's movement in normal times, and that when the relationship breaks, it is usually a sign that a different driver has taken over.

4. Gold and the dollar

Gold is quoted in US dollars, so the dollar's own value is part of every gold price. A stronger dollar tends to coincide with a lower gold price in dollar terms, and a weaker dollar with a higher one, an inverse relationship that holds loosely over time. The mechanism is partly arithmetic: when the dollar strengthens against other currencies, gold becomes more expensive for buyers outside the dollar bloc, which tends to weigh on demand.

Priced in different currencies, gold tells the same story at different volumes. Over the past two decades it has risen in dollars, euros, Swiss francs and pounds alike, but by different amounts: it has gained most in the currencies that weakened most against the dollar, and least in the strongest currencies, such as the Swiss franc [7]. For a non-dollar investor this is the practical lesson: the gold return you actually receive combines the dollar gold price with the move in your own currency, and the two can pull in opposite directions.

5. Safe haven: truth or myth

"Safe haven" implies gold rises when equities fall. The record across eighteen equity-market shocks since 1973 is mixed. Measured from the day before each shock, gold rose strongly through the inflationary crises of the 1970s and early 1980s, through the 2008 Lehman crisis (up about 16% within a week and still higher a year later) and during the 2011 euro-zone and US-downgrade episode. In other shocks it did much less, and in some it fell: it barely moved through Black Monday in 1987 and did essentially nothing in the 1994 bond rout. During the COVID crash of early 2020 gold dropped around 7% at the thirty-day mark as investors sold liquid assets to raise cash, before recovering over the following two months [8].

Two patterns are worth drawing out. First, in a genuine liquidity panic gold is often sold alongside everything else in the first days, because it is easy to sell, and only acts as a haven later. Second, timing matters: in the 2026 Iran war, gold had already risen sharply in anticipation of the conflict, so an investor buying after the headlines bought near the peak and then saw the price fall by roughly 14% over the following months as a ceasefire took hold [5]. Gold's correlation with equities is low on average, which is the real diversification benefit, but low average correlation is not the same as a reliable negative move on any given day of stress.

6. Where gold comes from, and where it goes

Annual gold flows are smaller than the metal's prominence suggests: total demand in 2024 was about 4,970 tonnes [3]. Supply comes from two sources. Mine production provides roughly three quarters of it and grows only slowly, because new deposits are scarce and mines take years to develop. Recycling, mostly of old jewellery, provides most of the rest and rises when the price is high. Almost all the gold ever mined still exists, so the above-ground stock is large and the annual flow adds only a small percentage to it each year.

On the demand side the chart shows four uses. Jewellery is the largest, concentrated in India and China. Investment, in bars, coins and exchange-traded funds, is the most price-sensitive and swings the most from year to year. Central banks are the third pillar and have become a major buyer (section 9). Technology, where gold is used in electronics, is the smallest and most stable. Because supply is nearly fixed in the short run, the gold price is set mostly by shifts in demand, and within demand by the investment and official-sector components that change the fastest.

7. Who holds the gold

The world's central banks and the IMF together hold about 36,250 tonnes of gold in official reserves [7]. The holdings are concentrated. The United States reports by far the largest single stock at around 8,100 tonnes, followed by Germany, the IMF, Italy and France; Russia and China are the largest holders outside that group, and China has been adding steadily. For the long-standing holders, gold makes up a large share of total reserves, often more than half for the United States and the major European holders, while for many emerging-market central banks it is still a small share they have been raising.

The reason this matters for the price is sheer scale: official-sector holdings represent roughly a fifth of all the gold ever mined, and these are holders who buy and sell for policy reasons rather than for return. A shift in their collective behaviour, from selling to buying, is one of the larger forces that can act on the market over a span of years, which is the subject of the next section.

8. From sellers to buyers

The official sector has reversed its stance on gold within a single generation. Through the 1990s and 2000s central banks, mainly European, were net sellers, to the point of coordinating their sales to avoid disrupting the market. Since the 2008 financial crisis they have been net buyers, and the buying accelerated after 2022, with annual net purchases of roughly 1,080, 1,050 and 1,090 tonnes in 2022, 2023 and 2024 [9]. In 2025 net buying eased to about 863 tonnes, down 21% from 2024 but still the fourth-largest annual total on record and well above the 2010 to 2021 average of around 470 tonnes [9].

The buyers have also shifted, and the buying has been broad. Over 2023 to 2025 the largest additions to reported reserves came from Poland, China, Azerbaijan, India and Turkey, with more than twenty central banks adding in all, a list dominated by emerging markets diversifying away from the dollar [9]. Early 2026 brought the first crack in the trend: Turkey, Russia and Azerbaijan turned net sellers in the first quarter. That reversal appears linked to the Iran crisis and the jump in energy prices it triggered: for a large energy importer like Turkey, costlier oil and gas widened the budget and external deficits, and the central bank drew on its gold reserves to help cover them. Even so, the official sector as a whole still bought a net 244 tonnes that quarter [10]. The pattern to take from this is directional rather than precise: central-bank demand has been a structural support under the gold price for over a decade, it is large enough to matter, and like any demand it can slow or partly reverse.

A new kind of buyer has also appeared, and it is not a state. Tether is the issuer of USDT, the largest dollar stablecoin, used by hundreds of millions of people worldwide; to keep each token redeemable for one dollar it holds one of the largest reserves of US Treasuries on earth, with total exposure above $130 billion by late 2025, enough to rank it around the seventeenth-largest holder of US government debt, ahead of countries such as Germany and South Korea. In January 2020 it launched a sister token, Tether Gold (XAU₮), backed one-for-one by physical bars held in Swiss vaults [16].

Through 2025 Tether went much further, buying gold for its own reserves rather than only to back the token. By several estimates the single largest buyer of gold that year was not a central bank but Tether, which added more than the roughly 100 tonnes bought by Poland, the largest sovereign buyer; by the end of 2025 its gold was worth around $17 billion, and by early 2026 it held close to 148 tonnes and was buying at a pace of about two tonnes a week, more than a billion dollars a month, stored in a high-security vault in Switzerland [29]. It is early to know how gold-backed stablecoins will reshape the overall demand picture, but they plainly make gold exposure easier for a large share of the world's population, and these purchases appear to have added to the surge in the gold price through 2025.

9. Ways to hold gold, and how to keep it

There are different ways to hold gold, and they do not behave alike. Physical bars and coins give direct ownership but carry storage, insurance and dealing costs. A gold exchange-traded fund such as GLD tracks the metal price closely for a small annual fee and trades like a share. Gold-mining shares, and the funds that hold them, are a different asset again: they are equities whose profits rise and fall faster than the gold price. And a gold account at a bank or dealer lets you own metal without taking delivery, which sounds simple but hides the most important distinction of all.

That distinction is between allocated and unallocated gold, and it decides not just your costs but who actually owns the metal and what happens if your custodian fails. In an allocated account, specific numbered bars are set aside in your name and held in safekeeping. In law this is a bailment: the bars remain your property, sit off the custodian's balance sheet, and in most jurisdictions do not form part of its estate if it goes bankrupt. In an unallocated account, often exactly what a bank means by a "metal account" or an "XAU" balance, you own no specific bars at all. You hold a credit balance: the institution owes you a quantity of gold, and you rank as an ordinary unsecured creditor if it fails, no different from a depositor whose cash is treated as a claim rather than as property. Unallocated is cheaper and more liquid, which is why the great majority of London bullion trading clears that way [15]. The trade is convenience and cost against ownership and safety.

Where the vault sits matters too. Gold can be held in a bank's own safekeeping, with a specialist vault operator such as Brink's, Loomis or Malca-Amit, or through a vaulting platform such as BullionVault or GoldMoney. The strong form in every case is the same: allocated, segregated metal, with a bar list you can inspect and an independent audit. The moment storage becomes pooled or unallocated, whoever the provider is, counterparty risk comes back in through the side door.

As for indirect exposure to gold through gold-mining shares or the funds that hold them, a few things are worth understanding. Mining shares carry operating leverage: a given move in the gold price produces a larger move in a miner's profits, because much of a mine's cost is fixed. That makes miners more volatile than the metal in both directions, and it adds risks the metal does not have, including management decisions, debt, energy costs and country risk. The chart below shows the effect. Over the period shown, the miners (GDX) and especially the junior miners (GDXJ) were considerably more volatile than the metal yet did not reward that extra risk with higher long-run returns; gold itself, through GLD, produced the steadier path [11]. Leverage to gold cuts both ways, and historically the downside has shown up more reliably than the upside.

A newer wrapper is the gold stablecoin, a token on a blockchain that claims to represent metal one for one. The two that dominate are Tether Gold (XAU₮), issued by a Tether subsidiary and backed by bars in Swiss vaults, and Pax Gold (PAXG), issued by the New York-regulated Paxos Trust and stored with custodians such as Brink's; together they are roughly nine-tenths of a tokenised-gold market worth over four billion dollars in late 2025 [16]. Each token claims one fine troy ounce of LBMA gold and can move around the clock in seconds, which is the appeal. But a token does not remove the custodian, it adds one: you are trusting the issuer's vault, its attestations and its redemption mechanics rather than holding metal yourself. Regulatory standing varies by issuer and by jurisdiction. Paxos is supervised by the New York State Department of Financial Services. Tether's main dollar token was fined by the Commodity Futures Trading Commission in 2021 over reserve disclosures and long sat outside United States regulation, though in early 2026 Tether launched a separate, GENIUS Act-compliant dollar stablecoin for the American market, USAT, issued through a federally chartered bank [23]. The European Union has been stricter: under the MiCA regime phased in across 2024 and 2025, Tether chose not to seek authorisation, and regulated EU exchanges removed its main dollar token USDT for customers in the bloc. A gold token counts under MiCA as an asset-referenced token and falls under the same authorisation requirement, so Tether Gold cannot be offered to EU customers through a licensed venue either, even though simply holding or transferring it is not banned [24].

All of which leads to the question under every paper and digital claim on gold: is the metal actually there? Examine it rather than assume it. Unallocated balances are fractionally backed by design, the bank holding a fraction of the metal it owes. Even sovereign hoards invite the question. The United States gold at Fort Knox has had no full independent audit since 1953, when a panel spot-checked only a small share of the bars, and a 2025 bill, the Gold Reserve Transparency Act, seeks the first comprehensive assay in over seventy years; the Treasury maintains the gold is "present and accounted for" [17]. The point is not that the gold is missing. It is that trust is doing work that examination usually does not test.

Holding gold and keeping gold are different problems. Metal that sits in someone else's jurisdiction can be frozen or seized when politics turn, and recent history is full of examples. After Russia's full-scale invasion of Ukraine in 2022, the G7 immobilised around 300 billion dollars of Russian central-bank reserves; the part that escaped was largely the gold Russia held inside its own vaults and balances kept in China [18]. About two billion dollars of Venezuelan gold, worth far more today, has sat frozen at the Bank of England since 2020, after Britain recognised Juan Guaidó rather than Nicolás Maduro and the UK Supreme Court upheld that position in 2021 [19]. Iranian and Afghan reserves have been blocked in similar ways. And the longest-running case is the oldest: the dormant accounts of Holocaust victims and the gold looted by Nazi Germany and laundered through Swiss banks, which heirs and survivors spent half a century trying to recover, ending in a 1.25 billion dollar settlement by UBS and Credit Suisse in 1998 after a commission led by Paul Volcker found tens of thousands of suspect accounts [20]. Even neutral, "safe" Swiss custody was not immune to politics and obstruction.

The response from states that can afford it has been to bring gold home. Germany repatriated about 674 tonnes from New York and Paris between 2013 and 2017, and the Netherlands moved a large block of its reserves from New York to Amsterdam. The Banque de France has reportedly completed the same move, bringing the last of its monetary gold back from the New York Fed across 2025 and into 2026, so that essentially all of France's roughly 2,437 tonnes now sits in Paris [21]. Italy's case is different and often misdescribed. Its roughly 2,452 tonnes are partly held abroad, but the recurring Italian fight is not about location, it is about ownership: by an unusual arrangement the gold belongs to the Banca d'Italia, the central bank, rather than to the State, and proposals to assert State control, raised in 2019 and again in late 2025, have drawn formal objections from the European Central Bank as incompatible with central-bank independence under the EU treaties [22].

What survives examination is a single trade-off. The cheapest and most liquid ways to hold gold, unallocated accounts, tokens, and pooled vault products, are all claims on someone else's metal, and a claim is only as good as the custodian who holds it, the jurisdiction it sits in, and that jurisdiction's willingness to honour it when it would rather not. Direct, allocated, audited ownership in a place unlikely to freeze you costs more and yields nothing, which is precisely the price of removing the counterparty. The reader's question is therefore not "which form is best" but "how much counterparty and political risk am I actually carrying, and is the convenience worth it?"

10. Gold versus bitcoin versus silver

Bitcoin is often described as "digital gold," a scarce asset held outside the banking system. The analogy is deliberate: bitcoin's supply is capped at 21 million coins and grows only slowly, and the proof-of-work system by which new coins are "mined" is designed to echo the real effort of digging, melting and refining gold, costly work that cannot be faked and that limits how fast new supply can appear. The label captures that similarity of intent, but the two assets have very different profiles. Over the period both can be measured, bitcoin has delivered far higher returns and far deeper losses: drawdowns of 70% or more have happened several times, against gold's more contained declines [11]. On any risk measure, volatility, maximum drawdown, worst year, bitcoin is in a different category.

Their relationship to each other is also unstable. Bitcoin and gold are sometimes described as moving together because both can respond to concerns about currencies and policy, but the correlation between them has been low and inconsistent, and bitcoin has often traded more like a high-risk technology asset than like a haven. The two can play overlapping roles in an investor's thinking, but they are not substitutes in behaviour: gold has been recognised as a store of value and a medium of exchange for thousands of years and has a deep official-sector market behind it, while bitcoin has roughly fifteen years of history and a far wider range of outcomes. That official-sector gap is the starkest difference. Central banks hold about 36,000 tonnes of gold and have been buying more than a thousand tonnes a year, yet their adoption of bitcoin as a reserve asset is close to non-existent: the exceptions sit at the level of national treasuries rather than central banks, El Salvador, which made bitcoin legal tender in 2021 before scaling the policy back under an IMF programme, a United States strategic reserve built largely from seized coins, and Bhutan's state-linked mining [25].

The gap in size is just as wide. With gold near $4,000 an ounce in mid-2026, the whole above-ground stock, around 216,000 tonnes, is worth on the order of twenty-eight trillion dollars, against a bitcoin market value of a little over one trillion: gold's market is still roughly twenty times larger [26]. Whether bitcoin grows into a rival or stays a separate, higher-risk asset is a question about the future, not a settled fact about the present.

Silver sits between the two. Like gold it has been money for thousands of years and appeals to the same instinct for a hard, tangible asset, which is why it is often called the poor man's gold. But it behaves differently. It is a far smaller and more volatile market, so it tends to amplify gold's moves in both directions. More than half of its demand, around 60% in 2025, is industrial, in solar panels, electronics and electric vehicles, which ties silver to the economic cycle in a way that gold's mostly monetary and jewellery demand is not [27]. And central banks, the buyers that anchor gold, hold no silver in their reserves. The link between the two metals is usually summarised by the gold-to-silver ratio, the number of ounces of silver that one ounce of gold will buy. Under the bimetallic standards of the eighteenth and nineteenth centuries it was fixed near 15 to 1; in the floating era it has swung widely, from about 17 to 1 when silver spiked in 1980 to roughly 100 to 1 in 1991 and a record near 125 to 1 in the 2020 panic, before compressing again as silver rallied into 2026 [28].

11. Gold in a portfolio

Ray Dalio is about as close to an oracle as Wall Street has. The firm he founded in 1975, Bridgewater Associates, grew into the largest hedge fund in the world, and when he speaks about markets, allocators listen. In 2025, with public debt mounting and equities stretched, he gave an unusually specific recommendation:

He went further, suggesting that the present moment rhymes with the early 1970s, the very decade where this study begins. It is a striking claim from a serious source. But this is Socrates on Investing, and a claim is not made truer by the authority of the person making it. So rather than take Dalio's number on trust, we are going to examine it against more than half a century of data. The results may surprise you. They surprised me. So how much gold is optimal in a portfolio?

There is a further reason the question deserves examination: gold frightens a good many investors, and not without cause. It is a genuinely volatile asset, considerably more so than the S&P 500. Across 1972 to 2025 its yearly outcomes ranged from a fall of 32.6% to a gain of 126.6%, a spread no other asset class in this study comes close to, and after its 1980 peak it spent two full decades losing value while equities soared. Among the everyday building blocks of a portfolio, stocks, bonds and cash, gold is the jumpy one, even if it stays well short of the swings of a single stock or a cryptocurrency like bitcoin. So how can an asset this volatile, held in the right proportion, reduce the overall risk of a portfolio rather than add to it? That is the puzzle worth resolving, and the only honest way to resolve it is with evidence. So I tested it.

I built twenty portfolios and ran them through more than half a century of data, and they fall into three families by the role gold plays. In the first, gold is a minority holding: a sweep of sixteen blends starting from the classic 60% equities and 40% bonds, where equities means the S&P 500 and bonds means ten-year US Treasuries, both as total returns, and adding gold two points at a time, funded equally from each, up to 30% gold (45% equities, 25% bonds, 30% gold). In the second, gold stands as an equal: the Permanent Portfolio, an equal split of 25% equities, 25% bonds, 25% gold and 25% cash (cash being three-month US Treasury bills), and a portfolio holding a third each in equities, bonds and gold with no cash at all. In the third, gold is the dominant asset, the single largest holding at 40%, in two portfolios we will come to last. I then measured each one from 1972, the first full year gold traded freely, to 2025, not only over the whole span but over the shorter horizons that real investors live in: one, three, five and ten years. Here is the result.

Gold as a minority allocation

The clearest way to see what gold does is to take the smallest meaningful step and look at the whole distribution of outcomes it produced, not just the average. The first chart is plain 60/40: every year from 1972 to 2025 dropped into its return bracket, the good years stacking to the right, the bad to the left. It averaged 9.4% a year, with a worst year of minus 18.0% in 2022 and a best of plus 31.7% in 1995. One clarification holds throughout this section: a "worst year" is the worst a portfolio finished a calendar year down, not its worst drawdown. Markets routinely fall much further within a year before recovering, and annual data cannot see that deeper intra-year low. Every worst-year figure here therefore understates the most painful moment an investor would actually have lived through. The second chart takes a tenth of the portfolio and moves it into gold, funded equally from stocks and bonds, giving 55% equities, 35% bonds and 10% gold. The distribution does not transform; it tightens. The average return edges up to 9.8%, volatility falls from 11.1% to 9.9%, and the worst year shrinks from minus 18.0% to minus 16.1%. A modest change, but every one of those moves points the same way: slightly more return, slightly less risk, a shallower hole in the bad year. That is the question worth pursuing. If a tenth helps a little, what does the full range do?

So I slid the gold weight from zero to 30% and tracked the four numbers that matter as it moved. The picture is consistent and, at first glance, almost too flattering. The average return rises the whole way, from 9.4% at 60/40 to 10.2% at 30% gold. Volatility falls to a minimum at roughly 18% gold, around 9.7% a year, before edging up again as the gold weight grows large. And the worst year keeps shrinking, from minus 18.0% to minus 12.4%, because the bad year was 2022 in every case and gold finished 2022 roughly flat while both stocks and bonds fell about 18%. The portfolios that held gold were not rescued by a gold rally that year; they were rescued by gold's refusal to fall when everything else did. That is the property worth holding in mind. Gold's average return over this period was not modest at all: it compounded at almost 9% a year in nominal terms, not far below the S&P 500 and well above bonds or cash, though it delivered that return through an unusually wide range of yearly outcomes. What set it apart was less the size of the return than its timing. Gold's gains tended to arrive at different moments from those of equities and bonds, and above all at different moments from inflation, which is what makes the difference when you measure real returns rather than nominal ones. Notice too the best-year panel, where the peak migrates: at low gold weights the best year is an equity year, 1995, but as gold rises the best year becomes 1979, when gold alone rose 126.6%. That migration is a clue I will return to.

None of this means gold is a safe asset, and one figure settles the point. Held on its own, gold was the most volatile of the four building blocks in this study, stocks, bonds, gold and cash, swinging 27.1% in an average year against the S&P 500's 16.9%. Measured this way it is more volatile than the broad stock market, though still far calmer than a single company's shares, with long stretches of years going nowhere in between its lurches. So return to the puzzle posed at the start of this section: how can an asset that volatile lower the volatility of a portfolio rather than raise it? That is the heart of diversification, and it is worth examining rather than taking on faith. The answer is that gold's swings were largely unrelated to those of equities and bonds, often moving when they did not and rising when they fell. Risk in a portfolio is not the sum of the parts' risks; it depends on how the parts move together. A holding can be turbulent on its own and still steady the whole, provided it marches to a different drum. The lesson is not that gold is calm. It is precisely that gold is not calm, and was useful anyway.

Stretched over fifty-four years, small annual differences compound into large ones, and the growth-of-$100 chart shows the gap open up. It also shows the single most important caveat in this section, if you know where to look. The gold-tilted portfolios pull ahead, but trace when they do it: almost the entire lead is established in the 1970s and again after 2000, and in the long middle stretch, from gold's 1980 peak to roughly 2000, the gold-heavy lines flatten and the lean 60/40 line quietly catches up. The flattering full-period averages rest heavily on a single inflationary decade at the start of the record. An investor who loaded up on gold in 1980 on the strength of the previous ten years then spent twenty years watching that decision cost them. This is the central caution of the whole section: what worked here worked in this window, and the window contains one extraordinary gold decade that may not repeat.

Equal-weight portfolios

Assessing what gold contributes to a portfolio is genuinely hard, because the answer depends so heavily on what else the portfolio holds. The number of possible combinations is effectively endless, and chasing them all leads to confusion rather than insight. That is why I chose a deliberately simple path: start from 60/40 and, for every two points of gold added, take one point from equities and one from bonds. I could have tested millions of other portfolios, but more variations would only have clouded the conclusion rather than sharpened it. That said, there is one portfolio outside this neat family, built around a large 25% allocation to gold, that is interesting enough to bring into the analysis: the Permanent Portfolio.

With a permanent quarter each in stocks, bonds, gold and cash, it was the steadiest of all twenty by a wide margin: volatility of 7.8% a year against 11.1% for 60/40, and a worst year of only minus 8.3% in 2022 when 60/40 lost 18.0%. Look at its histogram and the difference is visible at a glance, the whole distribution pulled in tight around the middle, with no extreme years at either end. There is something deeply counter-intuitive, even a little ironic, in this: a portfolio that hands a full quarter of its money to gold, an asset far more volatile than stocks or bonds, turns out to be the most defensive of all twenty. The diversification effect is strong enough that the turbulence of the ingredient is swallowed by the steadiness of the blend. If the goal is the smoothest possible ride, the equal-weighted four-asset split delivered it. But smoothness has a price, and the price is return. The Permanent Portfolio compounded at 8.5% a year nominally and 4.4% in real terms, the lowest of the group, because a permanent quarter in cash is a permanent drag in any decade that rewards risk. There is no portfolio here that is best on every measure. There is only the question of which measure matters most.

The Permanent Portfolio's steadiness deserves a caveat, because a good deal of it has nothing to do with gold. A full quarter of the portfolio sits in cash, and cash lowers volatility for the dullest of reasons: it barely moves, it simply sits out the market. That makes the Permanent Portfolio a poor lens for the question this section is actually asking, which is what gold itself contributes. To isolate that, take the cash away entirely and split the money a third each into equities, bonds and gold. That is a great deal of gold, 33.3%, more than any blend in the sweep, with no cash cushion to hide behind. Here is how it behaved.

The result is striking for how moderate its risk turns out to be. With a third in gold and no cash at all, the portfolio averaged 10.2% a year, ahead of both plain 60/40 and the Permanent Portfolio. Its volatility was 10.3%, below 60/40's 11.2% and only a little above the cash-cushioned Permanent Portfolio's. Its worst year was minus 11.8% in 2022, shallower than 60/40's minus 18.0%, and its best was an extraordinary plus 48.6% in 1979. This is the cleanest demonstration in the section of what gold actually does. Take away the cash crutch, load a full third of the money into the most volatile of the assets in the mix, and the portfolio was still calmer than a conventional 60/40 while out-returning it. The steadying was gold's doing, not cash's. One caveat keeps it honest, though. Over the whole half-century the one-third-each portfolio looks remarkably solid, but the heaviest gold weight here is also the most exposed to gold's own bad decades, and it shows up in a single statistic worth dwelling on: there was one five-year stretch, the years to 1984, when this portfolio lost purchasing power in real terms as gold gave back its 1980 spike. A single ugly year is uncomfortable but quickly behind you; five years of quietly going backwards in real terms is a slower, harder thing to sit through, and it tests an investor's conviction far more than a sharp one-year fall. The one-third portfolio is strong, but it is not invulnerable, and the form its vulnerability takes is precisely the form investors find hardest to hold through.

Gold-led portfolios

So far gold has played a supporting role. In the sweep it was the minority holding, a slice carved out of a portfolio still dominated by stocks and bonds; in the Permanent Portfolio and the one-third-each it was an equal partner, no larger than any other asset. The natural last question is what happens when gold stops being a partner and becomes the principal. To answer it, here are two portfolios where gold is the single largest holding, at 40%. The first keeps stocks and bonds balanced behind it, 30% each (30/30/40). The second tilts the rest towards bonds, 40% bonds and 20% stocks (20/40/40), a 60/40 of sorts in which the 60% held in risk assets is mostly gold rather than mostly equities.

This is where the free lunch runs out. Up to about a third in gold, every step had added return and subtracted risk. Push gold to 40% and the pattern reverses. Volatility, which had been falling, climbs back above plain 60/40: 11.5% for the balanced 30/30/40 and 11.3% for the bond-tilted 20/40/40, against 60/40's 11.2%. Returns stall, then slip. The 30/30/40 still edges 60/40 on return (9.8% against 9.4%), but the 20/40/40 actually earns less, 9.2%, while carrying more risk: it is beaten by the very 60/40 it was meant to improve on, and comprehensively beaten by the 30%-gold blend. Notice too where the worst year moves. For every portfolio up to a third in gold, the worst year was 2022, a stock-and-bond shock. For both 40%-gold portfolios it becomes 1981, the year gold's bubble burst. The portfolio's chief danger is no longer a crash in markets; it is gold itself unwinding. The rolling record confirms it: the bond-tilted 20/40/40 suffered two negative real five-year periods, one of them running to 2016, gold's own modern bear market, a stretch none of the lower-gold portfolios endured. Past a certain point you are no longer diversifying with gold, you are concentrating in it, and you inherit its long, dispiriting droughts. The evidence puts that turning point at roughly a third. Gold earns its keep as a diversifier; it does not reward being made the main event.

Returns across different holding periods

A full-period average is a comfortable number, and comfortable numbers hide things, because almost nobody holds a portfolio for fifty-four uninterrupted years. They hold it for the three or five or ten years in front of them, and judge it on those. The final chart reorganises the whole history into rolling windows so you can see what each portfolio actually delivered over a human horizon, with a toggle between one, three, five and ten years, and between nominal and real returns. Two things emerge that the averages conceal. First, the long penalty: across the heatmap there is an almost unbroken stretch from the early 1980s to around 2000 where the gold-heavy portfolios trailed plain 60/40 in window after window. Gold went nowhere for two decades while equities compounded, and more gold simply meant less return. Second, and pulling the other way, switch the toggle to Real and look at the 1970s. Plain 60/40 did not merely wobble; it lost real wealth for years at a stretch, shrinking a reader's purchasing power by 4.9% a year over the five years to 1977, with five separate negative real five-year windows in all. Gold steadily removed that vulnerability: among the portfolios here, those carrying a gold weight between 20% and 30% never had a single negative real five-year period, despite the punishing decades gold endured through the 1980s and 1990s, and the cash-cushioned Permanent Portfolio had none either. Even the one-third-each portfolio, with a full 33% in gold, had just one, the five years to 1984, while gold was still unwinding its 1980 spike.

Conclusion

Many investors start from an intuition that gold, being a volatile asset, has no business making up a meaningful share of a serious portfolio. The hardest-won conclusion of this analysis is the opposite. For a balanced risk budget, a volatility of around 10% a year, a gold allocation as high as 30% produced some of the most efficient portfolios in the whole study, provided it was paired sensibly with stocks and bonds. The asset that looks reckless in isolation became, in the right proportion, one of the most useful holdings in the mix.

That is the finding that survives examination. Gold did not always help, and for one long stretch it plainly hurt. But it helped in precisely the conditions, inflationary and equity-hostile, that a conventional portfolio is least equipped to handle, and those are exactly the conditions investors find easiest to assume away after a long calm. So is there an optimal gold allocation? Over this particular history, if the single goal is to minimise volatility, the answer is unusually precise: about 18% gold, but only among the sixteen cash-free blends I swept, where stocks and bonds still carry most of the portfolio. Allow cash into the mix and you can be quieter still: the Permanent Portfolio, a quarter of it in cash, was the least volatile of all twenty, though it bought that smoothness with noticeably lower returns. If the goal is never to have lost purchasing power over any five years, a gold weight of roughly 20% to 30%, or the Permanent Portfolio's structure, did that instead. And if the goal is the highest compound return, the data points to more gold than most investors hold, but only on the strength of one inflationary decade that may not return. The honest answer is therefore not a number but a reframing: across this window gold behaved less like a return engine and more like insurance, a holding that gave up little and occasionally paid out exactly when the rest of the portfolio could not. This is one historical record, not a forecast. The reader's task is not to copy an allocation but to decide which of these objectives is theirs, and how much they are willing to pay, in forgone return or in forgone safety, to pursue it.

Glossary of key terms

Spot price: the current market price to buy or sell gold for immediate delivery, quoted per troy ounce in US dollars.

Troy ounce: the standard unit for precious metals, about 31.1 grams, slightly heavier than a normal ounce.

Real return: a return after subtracting inflation. It measures change in purchasing power rather than in nominal dollars.

Real interest rate: the interest rate after expected inflation, approximated by the yield on inflation-protected government bonds (TIPS).

Safe haven: an asset expected to hold or gain value when riskier assets fall. Gold's record in this role is mixed and crisis-dependent.

Correlation: a measure from -1 to +1 of how two assets move together. Gold's low correlation with stocks is the basis of its diversification value.

Drawdown: the peak-to-trough fall in an asset's value, a common measure of downside risk.

Central-bank reserves: gold and foreign currency held by a central bank. Official-sector gold is about a fifth of all gold ever mined.

Bullion: gold in bar or coin form, valued by weight and purity rather than as jewellery.

Gold miner: a company that extracts gold. Mining shares have operating leverage to the gold price and are more volatile than the metal.

ETF: exchange-traded fund. A gold ETF such as GLD holds bullion and trades like a share, tracking the metal for a small fee.

Recycling: gold recovered from old jewellery and electronics, the second source of supply after mining.

Sources

1. Aswath Damodaran, NYU Stern, Historical Returns on Stocks, Bonds and Bills: 1928 to 2025 (histretSP dataset). https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/histretSP.html

2. World Gold Council, gold price data and history. https://www.gold.org/goldhub/data/gold-prices

3. World Gold Council, Gold Demand Trends Full Year 2024. https://www.gold.org/goldhub/research/gold-demand-trends/gold-demand-trends-full-year-2024

4. Federal Reserve Bank of St. Louis (FRED), 10-Year Treasury Inflation-Indexed Security real yield (DFII10). https://fred.stlouisfed.org/series/DFII10

5. Gold price during the 2026 US-Israel-Iran war: BullionVault, Gold Falls Through $5000 Even as War Spikes Stagflation Fears (March 2026), https://www.bullionvault.com/gold-news/gold-price-news/gold-5000-iran-stagflation-031620261 ; Newsweek, Why the Gold Price Is Falling During the Iran War.

6. Claude B. Erb and Campbell R. Harvey, The Golden Dilemma, Financial Analysts Journal / NBER Working Paper 18706, 2013. https://www.nber.org/papers/w18706

7. World Gold Council, Gold reserves by country and central-bank statistics (via gold.org), end-2025 reported holdings. https://www.gold.org/goldhub/data/gold-reserves-by-country

8. Author's calculation from COMEX gold (GC=F) daily prices via Yahoo Finance (safe-haven crisis windows, 2001 to 2026).

9. World Gold Council, annual central-bank net gold purchases; 2025 full-year figure via World Gold Council and BullionVault. https://www.gold.org/goldhub/research/gold-demand-trends

10. World Gold Council, Gold Demand Trends Q1 2026, Central banks. https://www.gold.org/goldhub/research/gold-demand-trends/gold-demand-trends-q1-2026/central-banks

11. Price data for GLD, GDX, GDXJ and Bitcoin (BTC-USD) via Yahoo Finance, month-end.

12. Gold Reserve Act of 1934, Federal Reserve History. https://www.federalreservehistory.org/essays/gold-reserve-act

13. World Gold Council, Central Bank Gold Agreement (the 1999 Washington Agreement on Gold); and the 1999 to 2002 sale of British gold reserves (Brown's Bottom). https://www.gold.org/what-we-do/official-institutions/central-bank-gold-agreements/first-central-bank-gold-agreement

14. Gold's intraday record of $1,921 on 6 September 2011 and its drivers (the euro-zone debt crisis, the US AAA downgrade, and quantitative easing).

15. London Bullion Market Association, The OTC Guide, Precious Metal Accounts (allocated versus unallocated accounts and their treatment on insolvency). https://www.lbma.org.uk/publications/the-otc-guide/precious-metal-accounts

16. Tokenised-gold market data and issuer disclosures for Tether Gold (XAU₮, TG Commodities) and Pax Gold (PAXG, Paxos Trust); CoinDesk, Tokenized Gold Market Tops $2.5B (September 2025). https://www.coindesk.com/markets/2025/09/01/tokenized-gold-market-tops-usd2-5b-as-the-precious-metal-nears-record-highs

17. Gold Reserve Transparency Act of 2025 (Rep. Thomas Massie), and reporting on the absence of a full Fort Knox audit since 1953, with the US Treasury's statement that the gold is present and accounted for. Mining.com, US lawmakers push for comprehensive audit of Fort Knox gold.

18. Brookings Institution and European Parliament briefings on the roughly 260 to 300 billion euro of Russian sovereign assets immobilised after February 2022, and on the gold and renminbi reserves that were not reachable.

19. UK Supreme Court, "Maduro Board" of the Central Bank of Venezuela v "Guaidó Board" (2021), on the Venezuelan gold held at the Bank of England; contemporary reporting on the continued freeze.

20. Swiss Banks Settlement of 1998 (UBS and Credit Suisse, US$1.25 billion) over Holocaust-era dormant accounts and Nazi-looted assets, and the Volcker Committee audit. swissinfo.ch and the Claims Conference, Swiss Banks Settlement.

21. Reporting by gold-market analysts and press on the Banque de France's repatriation of its remaining monetary gold from the New York Fed (2025 to 2026); Mining.com and The Gold Observer.

22. European Central Bank, Opinion on the ownership and management of Italy's official reserves (CON/2019/23, with further 2025 opinions), objecting to State-control proposals on central-bank-independence grounds; Euronews, Who does Italy's gold belong to? (November 2025).

23. United States stablecoin regulation under the GENIUS Act (2025), and Tether's launch in early 2026 of USAT, a GENIUS Act-compliant US dollar stablecoin issued through Anchorage Digital Bank, with Cantor Fitzgerald as reserve custodian. Cointelegraph and PYMNTS.

24. Markets in Crypto-Assets Regulation (MiCA): stablecoin provisions phased in across 2024 and 2025, the resulting removal of Tether's USDT for EU customers by MiCA-regulated exchanges (Coinbase, Crypto.com, Binance), and the classification of gold-backed tokens as asset-referenced tokens. European Securities and Markets Authority guidance and exchange delisting notices.

25. Sovereign bitcoin holdings: El Salvador's adoption of bitcoin as legal tender in 2021 and its scaling-back under the 2024 IMF programme; the United States Strategic Bitcoin Reserve (2025), built largely from seized assets; and Bhutan's state-linked mining. Central banks otherwise hold essentially no bitcoin in their reserves. Chainalysis and IMF reporting.

26. Market capitalisations of gold and bitcoin: gold's above-ground stock (about 216,000 tonnes, World Gold Council) valued at the prevailing LBMA gold price, and bitcoin's circulating supply at its market price (CompaniesMarketCap and MacroMicro). Both figures move with price.

27. Silver demand and its industrial share: The Silver Institute, World Silver Survey 2025 (industrial and technology uses around 60% of total demand, led by solar photovoltaics, electronics and vehicles).

28. Gold-to-silver ratio, long-run history (about 15 to 1 under the bimetallic standards, peaks near 100 to 1 in 1991 and roughly 125 to 1 in 2020): Macrotrends and Britannica Money.

29. Tether's gold reserves and 2025 to 2026 buying: Tether attestation reports (tether.io) with reporting and estimates by CoinDesk, Bloomberg and Jefferies, including year-end 2025 gold holdings of around $17 billion and an early-2026 buying pace of roughly two tonnes a week. https://www.coindesk.com/business/2026/01/28/tether-is-buying-up-to-usd1-billion-of-gold-per-month-and-storing-it-in-a-james-bond-bunker