Most confusing gold price behaviour is not economic. It is structural — an artefact of three markets with different conventions being described as if they were one. This edition is a plain-language tour of the plumbing, because the alternative is mistaking a settlement quirk for a monetary event.
London: where the metal actually is
The London over-the-counter market is the centre of gravity for physical bullion. It trades unallocated positions between members, settles by book entry against metal held in a small number of vaults, and prices in a twice-daily auction that produces the benchmark most contracts reference.
Two features matter for anyone reading the market from outside. First, most London activity is not delivery: it is transfers of claims, which is efficient and also means volume figures do not describe metal moving. Second, the eligible bar standard is the large good-delivery bar, which is why London liquidity and retail coin availability can diverge sharply during a demand spike — the two are not the same product.
COMEX: where the headlines are set
Futures on COMEX dominate price discovery for short horizons because that is where leverage and speed live. This is also the source of the standing accusation that paper gold suppresses the physical price. The honest version of the argument is narrower and more useful: futures allow exposure without metal, so the marginal price-setter in a quiet week is a position rather than an ounce.
What to watch instead of the conspiracy framing: registered against eligible inventories, and how open interest behaves during a price move. Rising open interest into a rally means new positions are financing it; falling open interest into a rally means positions are being closed, which is a different market with the same chart.
Shanghai and Mumbai: the physical read
Chinese and Indian premiums or discounts to the London benchmark are the cleanest available signal of whether physical demand is absorbing a Western-led move or resisting it. A sustained Shanghai premium during a rally says domestic buyers are chasing metal. A discount during the same rally says the move is financial and the physical market is stepping back.
The same logic applies to Mumbai, with the additional complication of import duty changes and a strongly seasonal wedding and festival pattern. A discount in the weeks after a duty change is a policy artefact, not a demand collapse.
The three dislocations worth noticing
First, a persistent gap between futures and spot beyond financing cost, which historically signals stress in the ability to move or borrow physical metal rather than a view on price. Second, lease rates spiking, which says borrowed metal has become scarce. Third, retail premiums on small formats blowing out while large-bar premiums stay flat, which is a fabrication and logistics bottleneck rather than a bullion shortage — and the fastest way to overpay if you mistake one for the other.
The practical takeaway
You do not need to trade any of these venues to benefit from watching them. The plumbing tells you who is moving the price on any given week, and that is usually more actionable than the price itself. When the venues agree, the move is broad. When they disagree, one of them is telling you something the headline has not caught up with.
A week in the life of one price
It helps to trace a single trading day. Asian hours open with Shanghai quoting a premium or discount to the London benchmark, which is the first read on whether physical buyers are engaged. London then sets the reference in its morning auction, where refiners, banks and central-bank agents transact in unallocated positions against vaulted metal. When New York opens, futures take over the tape: volume concentrates in the front month, leverage arrives, and the intraday range widens even though no additional metal has changed hands. The close prints in New York, gets reported worldwide as "the gold price", and is then quoted back to a Shanghai buyer the next morning as if it were a physical clearing price. It is not — it is the settlement of a leveraged contract in a different time zone.
Once you can see that sequence, several familiar puzzles dissolve. Prices that move violently while vault inventories sit unchanged. Rallies that die at the New York open. Divergences between what a coin dealer quotes you and what your screen says. None of these require a hidden hand; they follow from three venues with different participants, different settlement conventions and different minimum tradeable units.
What each venue can and cannot tell you
London tells you about the metal: vault holdings, good-delivery standards, and whether unallocated claims can be converted without friction. It tells you almost nothing about short-term direction, because most of its activity is bilateral and reported with a lag.
COMEX tells you about positioning and financing: who is levered, in which direction, and at what cost. It tells you very little about physical availability, because the overwhelming majority of contracts are closed or rolled rather than delivered.
Shanghai and Mumbai tell you about the marginal physical buyer. They tell you nothing reliable about Western institutional flows, and they are distorted by domestic policy — import quotas, duty changes and currency controls — often enough that a single week's premium should never be read alone.
The mistake worth avoiding is asking one venue a question it cannot answer: reading direction from vault data, or availability from open interest.
How dislocations resolve
The useful question about a dislocation is not whether it is unusual, but what the arbitrage is that should close it and what is preventing that arbitrage from working. Futures trading persistently above spot beyond financing cost implies someone should sell futures and buy metal; if nobody does, the constraint is usually logistical — freight, refining capacity, bar-size mismatch or balance-sheet limits at the institutions that would normally warehouse the trade. Those constraints are physical and they take weeks, not minutes, to clear.
That is why the honest interpretation of a 2020-style dislocation was mundane rather than dramatic: aircraft were grounded, refineries were closed, and the deliverable bar in New York was a different size from the deliverable bar in London. The gap closed once bars were recast and shipped. A reader who understood the plumbing was not frightened by it, and a reader who did not was told for months that the market had broken.
The series to follow, all public
Exchange-reported registered and eligible inventories. Weekly commitment-of-traders positioning. Published vault holdings from the London market administrator. Daily benchmark auction prices. Shanghai Gold Exchange benchmark quotes and withdrawal figures. Indian import statistics and current duty rates. Gold forward offered rates or published lease-rate curves where available.
Two habits make that list useful rather than decorative. Read each series against its own history rather than against a headline, and write down in advance what level would change your mind. A number without a prior expectation attached to it is entertainment.
What would prove this framework wrong
If venue-level divergence stopped preceding or accompanying stress episodes — if premiums, lease rates and inventories moved randomly with respect to disruptions in physical settlement — then the plumbing would be noise and the framework should be discarded. We watch for that. So far the relationships have held often enough to be worth the effort, and we will say so plainly here if they stop.
On the programme
Market structure has a dedicated block on 10 October, built on venue data rather than commentary, including the settlement mechanics that surface only under stress. This edition is educational only and not investment advice.