Gold at records — what central-bank buying means for a 60/40
Official-sector demand has changed who sets the gold price. Why the model holds a 10–20 % gold sleeve as a structural position rather than a bet, and what that does to a portfolio that used to be called 60/40.
Dear reader,
Gold set a new nominal record on 19 August and, more quietly, a new record in real terms — above the January 1980 peak once you adjust for four decades of inflation. Most commentary treated this as a story about fear, or about rates, or about the dollar. We think it is mostly a story about who is buying, and that this changes how a long-horizon investor should think about the size of a gold sleeve in a portfolio that used to be called 60/40.
Who sets the price now
For most of the past forty years, gold's marginal buyer was a private investor: a jeweller in India, a retail ETF holder in the US, a hedge fund with a view on real yields. Those buyers are price-sensitive and rate-sensitive, which is why the old model — gold falls when real yields rise — worked as well as it did from 2006 to 2021.
Since 2022 the marginal buyer has been the official sector. Central banks bought more than 1,000 tonnes a year in 2022, 2023 and 2024, roughly double the average of the previous decade, and the central-bank gold statistics's quarterly data suggest 2025 and the first half of 2026 have continued at that pace. The buyers are concentrated: China, Poland, Türkiye, India, Singapore, the Czech Republic. Their motive is reserve diversification away from assets that can be frozen, not a view on the next Fed meeting. They do not sell when real yields rise thirteen basis points. They do not sell much at all.
This is why the correlation between gold and real yields, which was reliably around −0.6 for fifteen years, has drifted toward zero since 2023. The old relationship is not dead — private buyers still matter at the margin — but it is now one force among two, and the second force does not read the FOMC statement.
What that does to a 60/40
The classic 60/40 portfolio rests on one assumption: that when equities fall, bonds rise. That assumption held beautifully from 1998 to 2021 and failed badly in 2022, when both fell together because the thing driving markets was inflation rather than growth. A portfolio needs an asset that does well in that third state — falling growth expectations and rising inflation — and there are not many candidates. Commodities broadly, some real estate, and gold.
Gold is not in the portfolio because we expect it to go up. It is in the portfolio because it is the asset most likely to be up in the scenario where everything else is down.
That is a structural argument and it leads to a structural weight. In the model, gold's band is 10–20 % in Balanced and never below 8 % in any profile, including Growth. It moves within that band with the regime, but it is never sold to zero, because the scenario it hedges does not announce itself in advance. This week's brief explains the top of the band; the floor is the more important number.
Is 20 % a lot?
It sounds like a lot, so here is the arithmetic. Gold's annualised volatility over the past twenty years is about 15 %, similar to equities. Its correlation with equities over the same period is close to zero, and with bonds slightly negative. A 20 % sleeve therefore contributes roughly the same amount of portfolio risk as a 20 % equity sleeve would, while diversifying rather than adding to the two big exposures. In a risk-off regime, when the equity sleeve is already at the bottom of its band, that trade-off is attractive. In a risk-on regime — expanding liquidity, anchored inflation — it is not, which is why the model sells gold down to 10 % in that state, and why it did exactly that in 2024.
What the sleeve does not do is improve returns in a normal year. Over the 2010s a 60/40 with gold underperformed a plain 60/40 by roughly half a point a year. That is the price of the insurance. Whether it is worth paying depends entirely on what you think the distribution of the next decade looks like — and on whether you can hold a position that lags in good years without abandoning it. Most people cannot, which is the real argument for putting the weight into a rule and letting the rule hold it.
The "what if it is a bubble" question
We get this every week gold makes a record. Three honest answers. First: yes, official-sector buying can slow, and the price would fall if it did; the 2013 episode, when gold fell 28 % in a year, is the template. Second: that is why the weight is banded and capped, not why the weight is zero. A 20 % sleeve that falls 28 % costs the portfolio 5.6 points, which is a bad quarter, not a catastrophe, and the regime model would very likely have moved to the bottom of the band before the worst of it because such a fall would coincide with expanding liquidity and easing geo-risk. Third: the "bubble" framing assumes gold has a fair value that the price has departed from. It does not. It has a set of buyers, and the composition of that set has changed.
How to hold it
For most readers a physically-backed ETF is the practical route; the largest ones hold allocated bars in London vaults and charge 0.1–0.4 % a year. Coins and small bars carry 3–8 % dealer spreads and are better thought of as a permanent holding than a rebalancing instrument. Mining equities are not gold: they are equities with a gold-price sensitivity, higher volatility and their own operational risks, and the model does not count them inside the gold sleeve. Readers rebalancing by hand will find the gold leg on the order-ticket export, priced in whole units against the model weight.
DFII10 against the gold reference price PM fix and taking a 90-day rolling correlation. The official-sector purchase series is the central-bank gold statistics quarterly, listed in our data sources.How our model allocation changed this week
| Asset | Last week | This week | Δ | Reason |
|---|---|---|---|---|
| Equities | 45 % | 45 % | 0 | Regime unchanged: neutral → risk-off (leaning) pending |
| Bonds (total) | 28 % | 28 % | 0 | No change |
| Gold | 17 % | 17 % | 0 | Mid-band; record price does not change the rule |
| Bitcoin | 6 % | 6 % | 0 | No change |
| Cash | 4 % | 4 % | 0 | No change |
Balanced profile shown. No allocation changes this week; the meter fell from 41 to 39, inside the two-week confirmation window.
Rulebook Wealth Editorial. The brief is written from the week’s published data. Every chart cites a public series and every model figure carries the version that produced it. Corrections are appended with a date, never silently edited. Data sources · Editorial policy
Want to check this brief’s charts yourself? Every series identifier we cite is listed in the data sources, and Pro accounts can export the allocation, bands and caps as JSON or CSV from the portfolio builder.