Every few weeks a headline announces that central banks are buying gold at a record pace. The claim is usually true and almost always useless, because the number is quoted without the two things that make it meaningful: what is being measured, and who is doing the buying. This edition is an attempt to fix that.
What the data is, and what it is not
The figures that circulate come overwhelmingly from one place: the World Gold Council's quarterly Gold Demand Trends, which in turn builds on IMF International Financial Statistics reporting and on Metals Focus estimates. Two features of that dataset matter before you quote any number from it.
First, it is a net figure. A headline of several hundred tonnes in a quarter is purchases minus sales across all reporting institutions. A single large seller can mask several determined buyers, and frequently has.
Second, reporting is voluntary in practice. Some institutions report monthly to the IMF, some report with a lag, and some accumulate for extended periods before disclosing a revised reserve figure in one step. That is why estimates for unreported buying exist at all, and why they are estimates rather than observations. When a quarterly number is later revised, this is usually why.
The observable, uncontroversial picture from the published record: after two decades in which official institutions were net sellers — the era of the Central Bank Gold Agreements, when European sellers coordinated disposals to avoid disorderly markets — the official sector turned net buyer around 2010 and has stayed there. In 2022 and 2023 annual net purchases exceeded a thousand tonnes, the strongest two-year stretch in the modern series, and 2024 stayed in the same region. That is the fact. Anything about the current quarter that has not yet been published is a forecast, and we will label it as one.
Why reserve managers are not like other buyers
A hedge fund buys gold because it expects the price to rise. A reserve manager buys gold for reasons that have almost nothing to do with the next twelve months of price action, and understanding that difference is the whole point of tracking this demand bucket.
Reserve management is conventionally described as a hierarchy of three objectives: safety, liquidity, and return — in that order. Gold scores unusually on all three. It has no issuer and therefore no credit risk. It is deeply liquid in size, in London, at almost any hour. And its return profile is uncorrelated enough with the rest of a reserve portfolio to reduce the variance of the whole.
Then there is the fourth reason, which reserve managers rarely state in public and everybody understands anyway: gold held in your own vault cannot be frozen by the issuer of another country's currency. The events of 2022 turned that from a theoretical property into an operational one, and reserve committees that had never seriously debated their gold weighting suddenly had a mandate to.
The practical consequence is that this demand is price-insensitive over the horizons traders care about. A reserve committee that has approved a multi-year programme to move gold from four per cent of reserves to eight does not suspend it because spot rallied fifteen per cent. It may slow the execution. It does not reverse the decision. That behaviour puts a floor of persistent bid under drawdowns that would otherwise run further.
The attribution trap
It is tempting to convert all of this into a dollar figure — sovereign demand accounts for X of the gold price. We won't, and you should be sceptical of anyone who does. The drivers are not separable. Real yields, dollar direction, ETF flows, jewellery demand in India and China, recycling supply, and official-sector buying all move at once and interact. Any single-driver attribution is a modelling assumption wearing the costume of a fact.
What can be said honestly is directional and still useful: a demand bucket that used to be a net source of supply is now a persistent net sink, it absorbs a material share of annual mine output, and the institutions doing it have longer time horizons than any other participant in the market.
How to follow this yourself
You do not need a subscription to track this. The IMF publishes International Financial Statistics monthly. Individual central banks — the People's Bank of China, the Reserve Bank of India, the National Bank of Poland, the Central Bank of Turkey — publish their own reserve statements. The World Gold Council publishes Gold Demand Trends quarterly and its central bank statistics free of charge. Reading two of those primary sources will put you ahead of most of the commentary written about them.
One habit worth building: when you see a tonnage figure, ask whether it is gross or net, whether it covers the official sector only or includes sovereign wealth funds, and what period it covers. Three questions, and most misleading headlines fail at least one.
On the programme
The reserves and geopolitics panel on 10 October is built on the primary data rather than on commentary about it, and we have asked the participants to bring the series they actually watch rather than the ones that make the neatest slide. Confirmed speakers are announced individually as contracts complete.
Housekeeping
The Academy course on macro drivers has been extended with a lesson on reading reserve statistics, including the revision behaviour described above. General Admission to the Congress remains free; Pro and VIP places are arranged by manual invoice after you submit the registration form.