Goldman's 79-Tonne Gold-Buying Estimate Exceeds Reported August Purchases

Goldman Sachs estimates central banks bought 79 tonnes of gold in August, almost double the 39 tonnes of net additions identified in official disclosures by the World Gold Council. The difference matters to investors assessing the strength of the official-sector bid beneath bullion prices, but it is not evidence that 40 tonnes of secret purchases have been found. The figures measure the market differently. 45

Goldman's October research note describes the 79-tonne figure as a nowcast, using trade and flow information to estimate buying beyond transactions disclosed by reserve managers. It compares with the bank's estimated 44 tonnes in July and a pre-2022 monthly average of 17 tonnes. Its three-month seasonally adjusted pace was 85 tonnes a month. These are estimates, not a register of identified central-bank transactions, and can change as more information arrives. 25

The World Gold Council's October 6 update instead compiled 39 tonnes of August net buying from reported reserve changes, using IMF and central-bank data available at the time. Its reported year-to-date total was 170 tonnes. That series is narrower than an attempt to model unreported activity. Subtracting it from Goldman's nowcast therefore produces a numerical difference, not a verified quantity of undisclosed bullion. Reporting lags, different coverage and estimation error could all affect the comparison. 4

China illustrates the distinction. Goldman estimated roughly 25 tonnes of Chinese buying in August; the People's Bank of China reported an addition of about 20 tonnes, or 20.2 tonnes on the World Gold Council's more precise calculation. The difference does not establish that China acquired another five tonnes off the books. It indicates how a model's inferred flows can diverge from the change in reserves disclosed by the institution itself. 145

The broader pattern is not uniform buying. In the World Gold Council's reported figures, Poland added eight tonnes in August and led year-to-date purchases at 98 tonnes, while Russia sold six tonnes that month and Turkey remained a net seller for the year. China's year-to-date reported buying stood at about 80 tonnes. Those divergences matter: a bullish aggregate estimate should not be read as evidence that all reserve managers are accumulating gold at the same rate. 4

Goldman argues that emerging-market central banks still have reasons to diversify away from major reserve currencies, including geopolitical and financial risk. That is an investment thesis, not a guaranteed price floor. Official demand can support gold over time, but the metal also responds to real yields, the dollar, exchange-traded fund flows and forced selling. A modelled buying estimate cannot by itself explain a particular day's price move. 15

For investors, the next test is whether subsequent official disclosures and revisions corroborate Goldman's elevated pace. The 79-tonne nowcast offers a potentially important signal about demand that is difficult to observe. The 39-tonne reported figure is a separate, more directly documented measure. Keeping them apart is more informative than treating either as a definitive count of every bar central banks bought in August. 45