Insurance you can hold: why physical gold carries less risk than paper gold

You buy gold to protect yourself from political risk, inflation and financial-system stress. Paper gold quietly reintroduces the very risks the insurance was supposed to remove.

01Start with why you own gold at all

Nobody buys gold because it pays a coupon. It pays nothing, costs money to store, and can go sideways for a decade. People buy it for exactly one reason: it is insurance — against currency debasement, against banking-system failure, against confiscatory politics, against the moments when the normal financial machinery stops working.

That framing matters, because it defines what “risk” means for a gold position. For a growth asset, risk is volatility. For an insurance asset, risk is something else entirely: the chance that your insurance doesn’t pay out on the day you actually need it. An insurance policy that works 99% of the time — except during the fire — is not cheap insurance. It is worthless insurance.

Judged by that standard, the various forms of “paper gold” — ETFs, futures, unallocated bank accounts, CFDs, gold certificates — are structurally weaker products than metal in your possession. Not because their prices track gold poorly in normal times (they track it very well), but because every one of them routes your claim through the same financial and legal system that gold insurance is meant to protect you from.

02What “paper gold” actually is

The phrase covers several instruments with very different mechanics, and it is worth being precise, because their risks are not identical:

Gold ETFs (GLD, IAU, and European equivalents) are trust shares. The trust holds allocated bars with a custodian — often with sub-custodians beneath it — and you hold a security that sits in a brokerage account. You cannot redeem shares for metal unless you are an authorised participant dealing in baskets worth millions. The GLD prospectus itself warns, in plain language, that shareholders have limited recourse if a sub-custodian loses gold, and that the trust may not be fully insured against loss.¹

Futures are standardised contracts where the overwhelming majority of positions never touch metal. In most delivery months, typically under 5% of outstanding COMEX contracts actually go to physical delivery; the rest are rolled, closed, or cash-settled.² The exchange also retains emergency powers — position limits, margin changes, forced cash settlement — that can be invoked precisely during the dislocations gold owners are hedging against.

Unallocated accounts — the workhorse of the London bullion market — are the weakest form of all. You are an unsecured creditor of a bank, holding a gold-denominated IOU. If the bank fails, you queue with the other creditors. The metal “backing” your account is the bank’s general pool, lent and re-lent.

CFDs and synthetic trackers add a further layer: your counterparty is a broker, often offshore, and your claim never touches gold at any point in the chain.

03More claims than metal

The paper gold market is far larger than the metal that stands behind it. On COMEX, open interest — the total outstanding paper claims — routinely exceeds the registered gold actually available for delivery several times over, and the ratio blows out spectacularly under stress: in early 2016 it briefly reached over 500 paper ounces per registered physical ounce.³ In calmer periods it compresses toward 2–4 claims per deliverable ounce — still a fractional-reserve structure by construction.⁴

In London, average daily clearing alone runs to more than 20 million ounces of gold⁵ — meaning the paper market turns over the equivalent of the entire annual global mine supply roughly every five to six trading days. The system works because almost nobody asks for the metal. It is, functionally, a fractional-reserve system for gold: perfectly smooth in normal times, and untested at scale in the one scenario that matters to an insurance buyer — a simultaneous rush for physical delivery.

Defenders of the system correctly point out that most futures holders never want delivery, so high ratios are normal and not fraudulent. True — and irrelevant to the insurance buyer. The question is not whether the system functions on an average Tuesday. The question is what happens on the day everyone wants the same exit. In 2025, as delivery demand surged and metal flowed from London to New York, COMEX futures at times traded at premiums of $20–100+ per ounce over London spot, refiners rationed large orders, and settlement times in the London market reportedly stretched from days to weeks.⁶ That is what the early stage of a paper–physical divergence looks like.

05The insurance-payout problem

Now put the pieces together. The scenarios in which you actually need your gold insurance — a banking crisis, capital controls, a sovereign debt event, a currency crisis, wartime financial measures — are precisely the scenarios in which:

— your broker may restrict withdrawals or halt trading;
— the exchange may impose forced cash settlement at an administered price;
— the ETF may suspend redemptions or trade at a discount to metal;
— the bank behind your unallocated account may be the thing that’s failing;
— and the state may freeze, tax or convert in-system claims by decree.

Paper gold’s risks are not merely additional — they are correlated with the trigger event. That is the worst property an insurance contract can have. It is the difference between owning a lifeboat and owning a voucher for a lifeboat, redeemable at the purser’s office, subject to availability.

06What the insurance is for: the long debasement

None of this matters if the underlying insurance case is weak. It is not. Since the 1971 default on gold convertibility, the US dollar has lost roughly 87% of its purchasing power, while gold has gone from $35 to above $5,000 per ounce in early 2026 — an appreciation of more than a hundredfold in nominal terms.⁷ Gold did not “go up”; the measuring stick shrank. The chart below indexes both to 1971.

07The honest costs of physical — and how to think about them

Physical gold is not a free lunch, and pretending otherwise would undermine the argument. You pay dealer premiums of roughly 2–8% on coins and small bars, and you face a bid/ask spread on the way out. You take on custody: theft risk if you store at home, and fees of roughly 0.1–1% per year for insured private vaulting. It is slower to sell than clicking a button, and large positions are cumbersome. And a bank safe-deposit box, note well, is not physical possession in the sense this article means — it sits inside the banking system, closes when the banks close, and was the primary collection point in 1933.

But price these costs correctly: they are the insurance premium. A few percent upfront and a fraction of a percent per year is what it costs to convert a correlated, conditional claim into an unconditional asset. Compare that to any other insurance you buy. Nobody calls fire insurance “underperforming the index” in the years the house doesn’t burn.

The convenience of paper gold, meanwhile, is real — which is why a pragmatic structure for many investors is a division of labour: paper gold for trading (tactical exposure, tight spreads, leverage if wanted) and physical gold for the insurance core — the part you hope to never sell, held in your own custody or in allocated, segregated, audited storage outside the banking system, ideally with some jurisdictional diversification. The mistake is not owning an ETF. The mistake is believing the ETF is the insurance.

08The bottom line

Gold’s entire value proposition is that it is nobody’s liability. Paper gold takes that one great property and trades it away for convenience — turning nobody’s liability back into somebody’s promise, held at a broker, inside a jurisdiction, subject to a rulebook that can change on a Sunday night. In normal times the difference is invisible. In the specific scenarios gold is bought to insure against — inflation crises, capital controls, banking failures, political rupture — the difference is the whole point.

Buy the insurance in the form that pays out.