Buying the Drawdown

The Divergence

The World Gold Council's demand data for the second quarter puts both figures in the same table. Official institutions took 288.9 tonnes, up from a revised 56.5 tonnes in the first quarter and 177.9 tonnes a year earlier, with Poland and China leading. Exchange traded products went the other way, at minus 44.8 tonnes.

The price context matters. Gold reached $5,589.38 an ounce on 28 January 2026. By 1 September it traded at $4,369.19, a fall of 21.8 per cent from that peak, and it stood at $4,400.58 on 9 September, still 20.9 per cent higher over twelve months. The LBMA quarterly average of $4,506.29 was 8 per cent below the first quarter and 37 per cent above the second quarter of 2025. So this was not a case of official buyers chasing a rally. They accumulated a record volume during a substantial drawdown.

The private selling has a straightforward explanation and the Council gives it: weaker prices, upward revisions to inflation and interest rate expectations, and a firmer dollar. Gold pays nothing. When the yield available on a liquid alternative rises, the price a rational investor will pay for a non-yielding asset falls. On 28 August, after the Federal Reserve chair's Jackson Hole remarks moved the market towards pricing a rate rise, gold fell about 3 per cent in a day. Private investors behaved exactly as portfolio theory says they should.

The consumer leg confirms the same reading. Jewellery consumption fell to 278.2 tonnes, the lowest quarterly volume since the pandemic and 17 per cent down year on year, as high prices constrained affordability. This is not a market gripped by debasement panic. It is a market in which one specialised buyer with a non-financial objective is absorbing metal that price sensitive buyers no longer want at these levels.

Insurance Procurement, Not a Price Call

The interpretation that follows is the one most commentary gets wrong in both directions.

A reserve manager is not maximising a risk adjusted return over a five year horizon. They are acquiring an asset that carries no counterparty, cannot be frozen by a foreign jurisdiction and does not depend on the solvency of any government. The yield forgone is the premium on an insurance policy. When the policy becomes more attractive, the buyer pays the premium regardless of the spot price, and a lower spot price simply means the same premium buys more cover. Reading record official purchases as a bullish price signal confuses an insurance decision with an investment decision.

Why the policy has become more attractive is not mysterious, and the reasons are largely institutional rather than monetary. Reserve assets were demonstrated, over the past four years, to be contingent on the political relationship between the holder and the issuing jurisdiction. Once that is established, every reserve manager outside the issuing bloc has a mandate question to answer, and the answer is a larger allocation to an asset with no issuer. The record purchase quarter is the visible output of that reasoning across several dozen institutions at once.

It is worth noting what the same data does not show. The International Monetary Fund has been explicit that when gold surpassed United States Treasuries as a share of official reserves in 2025, that development was driven almost entirely by gold price valuation effects rather than by reserve managers selling Treasuries to buy bullion. Separately, research published by the Federal Reserve Bank of New York in September 2026 argues that the aggregate decline in the dollar's reserve share reflects the actions of a small number of large holders rather than a systematic global shift, with roughly equal numbers of countries increasing and decreasing dollar holdings. That analysis has a significant gap of its own, since four countries including China and Russia are absent from the complete data set and are implied to account for a two percentage point contribution on their own. Both the official reassurance and the reset narrative are arguing past a data set that is incomplete by construction.

An allocator does not need to resolve that dispute. The observable fact is sufficient: a large group of institutions with no return objective is buying an asset that produces nothing, in record size, during a drawdown. They are paying for the absence of a counterparty.

What the Long Record Shows

The reason a private investor should not simply copy the trade is the return record, which is unusually well constructed and not flattering.

The UBS Global Investment Returns Yearbook, compiled by Dimson, Marsh and Staunton, covers 126 years across 35 markets. Since 1900 the real dollar gold price has risen 5.2 fold, an annualised real return of approximately 1.3 per cent. United States equities returned 6.6 per cent a year in real terms over the same period, long bonds 1.6 per cent and Treasury bills 0.5 per cent. One dollar in equities in 1900 became $3,296 in real terms by the end of 2025. The same dollar in gold became $5.20.

The inflation hedging claim fares no better. Of the 28 years since 1900 in which inflation exceeded 3 per cent, gold returns were negative in 13. That is close to a coin toss.

There is a serious qualification and it should be stated rather than buried. Over the 54 years since the collapse of Bretton Woods, annualised real gold returns were considerably higher, at 4.7 per cent in dollars, 5.8 per cent in sterling and 4.3 per cent in Swiss francs. An investor who treats the post 1971 monetary regime as the relevant sample has a materially stronger case, and if one believes that regime is now changing again, the sample argument cuts in their favour rather than against. That view depends on a judgement about politics, not an inference from data, and it should be labelled as such in any allocation document.

The Lesson for Other Hard Assets

The transferable insight is about what a hard asset is being asked to do.

Gold does one job well: it removes counterparty and jurisdiction risk. It does not produce income, it has no operating leverage, and over investment horizons that matter to a living person it has been an unreliable inflation hedge. At a sensible weight, for the reason central banks hold it, that is a perfectly respectable allocation. As a growth asset it has 126 years of evidence against it.

Every other real asset has to justify itself on different grounds, and the current rate environment sets the bar. The ten year inflation indexed Treasury yield stood at 2.44 per cent in the Federal Reserve's H.15 release of 18 August 2026, and the thirty year nominal at 5.31 per cent, the highest since June 2007. An illiquid real asset now has to clear that real return after every cost of ownership and then pay the holder something further for the fact that the position cannot be exited at a screen price. An asset held purely for scarcity clears neither test.

What does clear it is contracted cash flow, provided three things are true: the obligation is written and the obligor's credit can be examined; the operating cost base is allocated explicitly rather than assumed away; and the residual value assumption is stressed rather than flattered. A marine charter asset, an infrastructure concession or a long duration inflation linked credit can each satisfy those conditions or fail them. The category does not decide it. The contract does.

The central banks are right about what gold is for. They are not making a statement about what a portfolio should hold, and their record quarter is not the buy signal it has been reported as.