More gold, less value: reading September’s ETF balance sheet

The observation: three figures moving in different directions

The World Gold Council’s report published on 7 October says global physically backed gold ETFs received US$10 billion in September. Holdings increased by 67 tonnes to 4,256 tonnes, while assets under management fell 7% to US$574 billion. These are September observations, not a statement about flows during the latest October trading session. The monthly report supplies no publication hour.

The combination is economically possible: investors can add metal exposure while a lower valuation reduces the dollar value of the much larger existing stock. My starting question is therefore not whether the figures contradict one another. It is which measurement answers the question I am asking.

Source notes: [1] World Gold Council — September 2026 gold ETF report

Build a bridge from opening assets to closing assets

Consider a deliberately simplified example, not observed fund data. A portfolio begins with 100 units of gold worth US$100 each: assets are US$10,000. Investors add two units, and the closing gold price is US$90. Closing assets are US$9,180. Metal holdings have increased by 2%, but assets have declined by 8.2%. Both statements are true.

The exact bookkeeping depends on when the additional units arrive and what is paid for them. That is why multiplying a month-end holdings change by one closing price cannot reproduce every dollar-flow measure. A useful reconciliation separates opening exposure, contributions or redemptions, market revaluation, costs and any other adjustments. Any unexplained remainder is a research question, not a reason to force the numbers to agree.

A flow series is not a performance series

The Council distinguishes changes in gold holdings from dollar fund flows. Its published methodology discusses how flow timing, gold prices and exchange rates can create differences between tonnage changes and dollar measures. The measurement rule matters especially when combining funds across currencies.

For my research worksheet I would retain separate columns for tonnes, net dollar flows and percentage investment return. I would never substitute one for another in a chart simply because all three rise in a convenient example. A holding change helps describe the quantity represented by the covered funds; an investment return answers what happened to an exposure over a specified interval. The two can be related without being interchangeable.

Source notes: [2] World Gold Council — ETF flow definitions and measurement example

The timestamp inside the fund matters too

SPDR Gold Trust’s June 2026 quarterly filing explains that its net asset value uses the LBMA Gold Price PM, subject to the filing’s fallback rules. That valuation reference is not automatically identical to a share price observed later in New York. The sponsor also notes that the bar list and website holdings can differ because settlement-date and trade-date reporting use different clocks.

This does not establish a problem with a fund. It establishes a need to label the comparison. Before describing a premium, discount or unexplained mismatch, match the fund, currency, observation time and valuation basis. An intraday share quote compared with an earlier gold reference can reflect a time difference rather than an unusual investor preference.

Source notes: [3] SPDR Gold Trust — June 2026 Form 10-Q, NAV methodology · [4] SPDR Gold Shares — official holdings and bar-list notes

What this can—and cannot—say about gold demand

My interpretation is that September demonstrates why valuation weakness should not be described automatically as abandonment of gold exposure. It does not establish why each buyer invested, whether those buyers will remain invested, or whether their demand will dominate the next price move. Monthly aggregates cannot identify the decision behind an individual transaction.

A stronger follow-up would compare successive releases on the same basis, record revisions and examine whether changes are broad or concentrated. It would also distinguish a hypothesis made before the next release from an explanation chosen afterward. A positive-flow month is useful evidence about the covered funds. It is not a forecast that prices must rise.

The next observation I would request

For the next update, I would ask for the complete reporting period, matching regional totals and the provider’s explanation of revisions before changing my conclusion. The Council’s data page identifies weekly and monthly updates; a page refreshed in October does not make every table an October month-end observation. No completed October monthly total is claimed here.

That discipline makes the article useful beyond one headline: follow the quantity, follow the valuation, and keep the reporting clocks visible. The result is a clearer demand assessment without pretending a single data series can explain the entire gold market.

Source notes: [5] World Gold Council — gold ETF dataset and update schedule