The case for gold in a portfolio is usually made with a correlation figure and left there. The figure is not wrong; it is close to meaningless on its own. This edition tests the diversifier claim properly, because a position sized on a misunderstood statistic behaves unpredictably at exactly the wrong moment. Educational content only — not investment advice.
Average correlation hides the whole point
Gold's long-run correlation with developed-market equities sits near zero. That number is an average of two very different regimes, and averaging them destroys the information: in ordinary conditions the relationship is weak and directionless, while in acute stress it is unstable in both directions before frequently turning strongly negative.
Anyone allocating on the basis of the average is implicitly assuming the relationship is stable. It is not. The useful question is conditional — how does gold behave given a specific type of shock — and that question has a more honest and more limited answer.
The crisis pattern: two phases
The pattern that recurs across liquidity events has two phases. Phase one: gold sells off with everything else, because in a margin call the most saleable asset gets sold first and gold is extremely saleable. Investors who bought gold expecting day-one protection have repeatedly been surprised here, and it is not an anomaly — it is a mechanical consequence of being liquid.
Phase two: as the policy response arrives and real yields fall, gold re-rates and typically outperforms while equities are still repairing. The protection is real but it is lagged, and a holder who sells during phase one to stop the pain converts an insurance policy into a realised loss.
What it has and has not protected against
The record is reasonably clear on the categories. Gold has historically done its job against currency debasement, sharply negative real yields, sovereign credit stress and domestic banking failure. It has done its job poorly against equity drawdowns driven by rising real yields, and it is not a reliable short-horizon hedge against a general inflation print, for the reasons Edition 06 set out.
Stated plainly: gold is insurance against monetary and sovereign failure modes, not against market volatility in general. Those are different products, and the more expansive claim is what generates disappointed holders.
Sizing and rebalancing make the claim real
A diversifier only diversifies if it is rebalanced. If gold is allowed to run without trimming, it becomes the dominant risk in the portfolio and the diversification argument quietly inverts. If it is trimmed on every strong month, it is being traded rather than held.
The workable middle is a written rule set in advance: a target weight, a tolerance band around it, and a stated frequency for checking — with rebalancing executed in the liquid sleeve so the physical core is never disturbed. The value of writing it down beforehand is that the rule is authored in a calm month and executed in a violent one.
A worked example, without a recommendation
Take a simple illustration rather than a prescription. A portfolio holds a target weight in gold with a tolerance band either side, split between a physical core held in custody and a liquid sleeve used for adjustment. Gold has a strong quarter and drifts above the upper band; the rule trims back to target, and the trim is executed entirely in the liquid sleeve so the custody arrangement, its bar list and its cost basis are never touched. Equities then fall sharply while gold falls with them in the first week; the rule says nothing, because a joint drawdown inside the band is not a trigger. Weeks later gold re-rates through the upper band and the rule trims again — this time selling into strength while the rest of the portfolio is repairing.
Nothing in that sequence required a forecast. The behaviour that made the diversifier claim real was mechanical: a written band, a designated sleeve for execution, and no discretionary intervention during the phase-one drawdown that historically causes holders to abandon the position.
The three failure modes we see most
Sizing to a story. A position sized on the assumption of immediate crisis protection is almost always too large, because the holder expects it to offset equity losses in the same week. Sizing to the narrower, lagged claim produces a smaller and considerably more durable allocation.
Confusing the vehicles. Physical metal in custody, an ETF holding and a futures position have different liquidity, cost and counterparty characteristics, and they are not interchangeable within a rebalancing rule. Rebalancing a physical core incurs spread, shipping and assay friction that a liquid sleeve does not; treating them as one line item is what makes rules unworkable in practice.
Measuring in the wrong currency. A holder whose liabilities are in one currency and who measures gold in another is measuring two things at once. The diversification result changes materially depending on which currency the return is expressed in, and the honest version of the exercise states the base currency before quoting any correlation.
Where gold sits against the other defensive assets
The diversifier question is comparative, not absolute. Long-dated government bonds protect against growth shocks and fail against inflation shocks, as 2022 demonstrated to anyone who had treated them as an all-weather hedge. Cash protects nominal capital and loses purchasing power at exactly the moment gold is doing its job. Trend-following strategies frequently profit from sustained drawdowns but not from sudden ones, and they carry manager and strategy risk that a bar does not. Index put options provide precise, immediate protection at a recurring premium that compounds against you in calm years.
Gold's distinctive property in that set is that it has no counterparty, no premium decay and no manager, and that its failure mode — long stretches of flat or negative real return — is uncomfortable rather than catastrophic. It is not the best hedge for any single scenario. It is the only one that requires nobody else to remain solvent, which is why it survives in preservation portfolios that hold all of the above.
The cost of holding insurance you do not use
Any honest version of this argument has to price the waiting. Gold pays nothing, so the carry is negative in real terms whenever real yields are positive, plus custody and insurance costs on physical, plus the management fee on a fund holding, plus the bid-offer spread and any premium over spot on coins and small bars. Over a decade in which the metal does nothing, those costs are the premium paid for the option — and they are the reason a position sized for a scenario that does not arrive is genuinely expensive rather than merely idle.
This is also why sizing and the claim have to match. A modest allocation whose costs are tolerable across a barren decade can be held through one; a large allocation justified by an expectation of same-week crisis protection cannot, because the holder eventually abandons it for reasons that have nothing to do with the original thesis. The discipline is to size the position so that being wrong for years is survivable and uninteresting.
How to check this yourself
The claims above are reconstructable from public data. Take a long history of the spot price in your own base currency, a broad developed-market equity index, and a series for inflation-adjusted long-dated government yields. Compute rolling correlations rather than a single full-period figure, then split the sample by the direction of real yields. The conditional pattern described here — weak and directionless in ordinary conditions, unstable then frequently negative in stress, poor when real yields are rising — should be visible without any modelling sophistication. If it is not visible in your data, trust your data over our summary and tell us.
What would falsify the claim
We would revise this framework if gold failed to re-rate through a sustained period of sharply negative real yields, or if it began providing reliable same-week protection against equity drawdowns — which would suggest the liquidity mechanism described in phase one had stopped operating. Either observation would change the sizing conclusion, not just the commentary.
The honest summary
Gold is not a hedge for everything and it does not work on day one. It is a long-horizon claim on the failure of monetary arrangements, with a lag, priced in a market with its own plumbing. That is a narrower claim than the marketing and a considerably more defensible one.
On the programme
Allocation, sizing and rebalancing rules are worked through in the allocation-in-practice session on 10 October, using real portfolios rather than model ones.