Gold’s reserve share rose. How much was an allocation decision?

A reserve percentage is a relationship

The ECB’s June 2026 review offers a useful historical comparison: gold represented 27% of the official foreign-reserve aggregate it examined at the end of 2025. Revaluing gold using its end-2023 price reduced that share to 16%. The report also warns that the gold and foreign-exchange datasets have different country coverage. These are dated research observations, freshly checked for this article, not November reserve readings.

The analytical question is different from asking whether central banks bought gold. Purchases can occur while valuation effects also lift the portfolio weight. A share tells us how one measured value relates to an aggregate; it cannot, by itself, reveal how much of the change reflects a deliberate allocation decision. That distinction matters when a headline is used to support a claim about future demand.

Source notes: [1] European Central Bank — The international role of the euro, June 2026

Hold the ounces still and change the price

Consider a purely hypothetical reserve portfolio with gold worth 20 units and other reserves worth 80. Gold begins at 20% of the total. If the gold price rises by 25% while its quantity and the other assets’ values remain fixed, gold becomes worth 25 and total reserves become 105. Its share rises to 23.81% with no additional ounce purchased.

Now hold gold at 20 and reduce other reserves from 80 to 60. Its share becomes 25%, again without a gold purchase. The second example could have many explanations in a real institution, including transactions or valuation changes elsewhere; the arithmetic alone does not select one. Conversely, buying more gold while accumulating other reserves even faster can leave the gold share lower. Direction of share and direction of purchase are different observations.

Do not borrow a denominator from a different dataset

The IMF’s COFER dataset describes the currency composition of official foreign-exchange reserves. Its metadata explicitly exclude monetary gold, SDR holdings and the reserve position in the IMF from that foreign-exchange category. Therefore a currency’s COFER share and gold’s share of a broader reserve aggregate do not automatically have the same denominator. Putting the percentages beside each other without that distinction can manufacture an apparent substitution.

My working table would retain the coverage universe for every column: reporting entities, asset categories, valuation currency and date. If I cannot reconcile those universes, I would show the figures separately rather than produce a spurious combined total. This is particularly important when comparing a currency category with a particular security type: US Treasuries and all dollar-denominated reserves are not equivalent labels.

Source notes: [2] International Monetary Fund — COFER dataset metadata

Build three versions of the same portfolio

I would first preserve the reported current-value share. Second, I would calculate a constant-gold-price version using a clearly stated reference price, while leaving the remaining assumptions visible. Third, where quantities and currency composition are available, I would construct a broader constant-price and constant-exchange-rate comparison. Each version answers a different counterfactual question; none should replace the original observation silently.

An exact decomposition needs care because changing quantities and prices together creates interaction terms. Applying one change before another can assign those interactions differently. I would state the chosen order, or use a symmetric attribution method, and keep the residual visible. A simpler report can avoid false precision by presenting the separate scenarios without pretending their contributions are uniquely additive. The purpose is to reveal sensitivity, not disguise it beneath decimals.

An allocation intention needs evidence beyond the ratio

My interpretation is that a sustained increase in gold weight becomes more informative about policy when supported by quantity records, transaction disclosures and a stated strategic framework. Even then, one institution’s decision should not be generalized to all reserve managers. A global weight is an aggregation, and the same change can arise from many different country-level paths.

A useful follow-up is to ask whether the institution allows weights to drift or rebalances them around a target. Those policies would imply different responses to an identical price rise. I have not verified a new target or purchase programme for any central bank in this edition. The discussion therefore defines the evidence needed to assess a claim; it does not announce an undisclosed policy or forecast the next official-sector purchase.

Use the result without turning it into a trading promise

For gold analysis, the practical improvement is a better demand question. Before saying that reserve diversification accelerated, identify the part of the evidence that measures transactions rather than repricing. Before saying it reversed, test whether a falling share merely reflects other assets growing faster. Preserve older observations under their own dates and request missing quantities rather than infer them from a valuation total.

This edition complements the weekly report’s opportunity-cost analysis and the Daily Gold Brief’s session preparation. It does not revise their market observations or assert a current official-sector flow. The Dubai custody guide addresses the separate issue of what an owner actually holds. Portfolio weight, legal ownership and executable trading results require different evidence; no CY66 win rate, broker fill or P&L follows from the reserve arithmetic here.