--- title: Real Yields Stopped Predicting Gold Prices ---
Real yields used to be the gold trade. One number. The 10-year TIPS yield. Track it and you'd call most of the big moves in the metal for almost twenty years. Then the model broke. After the 2022 reserve-freeze precedent, the historical correlation between US real yields and gold prices collapsed from roughly negative 0.7 to near zero. It hasn't come back. This is structural, not noise.
The textbook model worked for two decades
From 2003 to 2021, you could trade gold with one data point. The 10-year US Treasury Inflation-Protected Security yield. Real yields fell, gold ripped. Real yields rose, gold sold off. Simple enough that hedge funds built whole desks around it.
The logic was clean. Gold pays no coupon. When real yields turn negative, holding gold costs nothing relative to holding government paper. When real yields go positive, you give up real return for sitting in a metal. So traders priced gold as the inverse of the 10-year TIPS yield. And for nearly twenty years, that priced everything cleanly. Bear Stearns blew up, real yields collapsed, gold doubled. COVID hit, real yields ran to minus one percent, gold cleared 2,000.
But starting in 2022, the trade stopped working.
The correlation snapped in 2022
Real yields rose from minus one percent to over two percent through 2023 and 2024. The biggest move in the real curve since 2008. By the old playbook, gold should have crashed thirty or forty percent. It didn't. It went up.
That's not a small miss. The correlation between weekly changes in 10-year TIPS yields and gold prices ran near negative 0.7 for the 2008-2021 window. Since early 2022, that same correlation has been roughly zero. Maybe slightly positive on some windows. The relationship that defined gold for a generation just stopped existing.
We've seen brief decouplings before. In 2013, real yields jumped during the taper tantrum and gold sold off harder than the model said it should. In 2019, both rose together for a few months. Each time, the relationship reasserted within a year or two. This one hasn't. Four years and counting.
Why? Three structural shifts hit at the same time.
Central banks rewrote the buyer base
The first shift. Before 2022, official sector gold demand averaged roughly 400 tonnes a year. Some years it ran higher, some lower. Most of the marginal price action came from ETF flows. When ETFs added 200 tonnes in a quarter, gold rallied. When they sold 200 tonnes, it puked.
After 2022, central banks started buying over 1,000 tonnes a year. Three years running. That's roughly two and a half times the prior baseline. World Gold Council data on net official purchases shows 1,082 tonnes in 2022, 1,037 in 2023, and 1,086 in 2024. Actually, when you add the unreported PBoC accumulation that surfaces in customs data, the true number for 2024 was probably closer to 1,300. The headline figures undercount.
These buyers don't care about TIPS yields. The People's Bank of China isn't running an opportunity-cost calculation against the 10-year. Neither is the Reserve Bank of India, nor the Turkish central bank, nor any of the dozen emerging market banks adding to reserves. They're rebuilding reserve composition. They buy at any price the market clears.
So you've got at least 600 extra tonnes of yearly demand that's structurally insensitive to real rates. Annual mine supply runs around 3,600 tonnes. So 600 tonnes is roughly 17 percent of new annual supply, taken out by buyers who don't watch a single chart you watch.
The IMF reserve composition data tells the same story from a different angle. Gold's share of total emerging-market FX reserves climbed from around 9 percent in 2021 to over 14 percent by late 2024. Developed-market central banks were already sitting at higher shares from legacy holdings, so most of the new buying has come from EM. China, India, Turkey, Poland, Singapore. The list keeps growing. And the buying happens whether real yields are at minus one percent or plus two percent.
Sanctions made gold a different asset
The second shift. February 2022, Russian central bank assets got frozen. Three hundred billion dollars of dollar and euro reserves became inaccessible overnight. Every reserve manager on earth watched that and ran the same calculation. If sanctions can hit reserve assets, what counts as a reserve asset?
Gold sitting in a foreign vault is fine. Gold sitting in your own vault is even better. It's the only major reserve asset where the issuer can't be told to freeze it.
That single fact changed gold's price input. Before 2022, gold competed with dollars and Treasuries as a store of value. The cost of holding gold relative to dollars was real yields. That was the math. After 2022, gold also has to be valued for the optionality of being uncensorable.
How do you price that optionality? Nobody knows. But it's clearly worth something north of zero. And it's a value that goes up, not down, when real yields rise. Because rising real yields usually come with a stronger dollar, which is exactly what reserve managers are trying to hedge against in the first place.
So we now have a buyer base bidding for gold because of a property the old model didn't price. Real yields can rise and gold still bids, because the demand isn't coming from cost-of-carry math.
ETF flows decoupled from price
Here's the part that confused a lot of macro traders. Through most of 2023 and 2024, gold ETF holdings actually dropped. Western ETFs sold tonnes. SPDR Gold Shares saw outflows for stretches even as spot ran from 1,800 to over 2,500. By the old playbook, ETF redemptions should have driven price lower. They didn't.
That's because the new marginal buyer wasn't operating in the ETF market. Central banks buy bars directly, not through ETF flow. They source from the LBMA market, from refiners, from other central banks. So you could have a Western ETF flow gauge flashing bearish and price still bidding, because the central bank window was open and hungry.
This breaks a second tool macro traders relied on. ETF flows used to lead price by a few weeks. Now they're a coincident or lagging signal. Watching them as your primary read is going to get you offside more than it'll get you positioned correctly.
The same logic applies to managed-money positioning on COMEX. The speculative net long used to mark cycle tops with reasonable consistency. Through 2024 and into 2025, speculative net long ran at moderate levels while price kept printing new highs. The reason is the same. Speculators trade the futures market, central banks settle in physical, and the physical bid was running the show.
What this means for trading gold now
Most macro funds spent 2023 and 2024 underweight gold because their factor models said overweight didn't make sense at 2 percent real yields. Some had outright short positions in gold versus TIPS as a relative-value trade. That trade got run over. Hard.
The lesson isn't "real yields don't matter anymore." They still matter at the margin. When real yields collapse, gold still tends to bid harder. When they spike, you'll get some pressure. But real yields are no longer the dominant input. They're maybe one of four or five inputs. The model needs more variables now.
Useful inputs for gold pricing in the current regime:
- Central bank reported and unreported buying. PBoC adds quietly off-tape, the official figures undercount
- Sanctions and reserve-freeze risk perceptions, with credit default swaps on vehicles holding reserves as a proxy
- Real yields, but at a much smaller weight than the old model assumed
- Dollar trend and DXY level versus a basket
- ETF flows, but as a sentiment read, not a flow driver
Trying to call gold off TIPS alone is like trying to call oil off OPEC quotas alone after 2014. The model worked once. Then the structure shifted.
How Kunkel Capital reads gold without TIPS as a guide
Our gold framework starts from structure. Where does the multi-month Elliott Wave count put the asset? Is gold in a wave 3 impulse or a wave 4 consolidation? That's the primary read.
Then we overlay positioning. We watch the commercial net short on COMEX, the gold ETF flow trend, central bank reported purchases on monthly lag, and refiner premiums in Shanghai. The TIPS yield is a sanity check, not a driver.
We've been positioned long gold for the entire run off the 2022 low. Not because real yields told us to. The wave count called for a five-wave impulse from that low. The official sector bid added a structural floor that meant pullbacks would be shallow. So we played the channel.
That's the work. Build a model that prices the asset based on the buyer base that actually moves it, not the buyer base that used to. Reserve managers in Beijing and Ankara don't track real yields. So neither does the price, most days.
Want to see where the wave structure currently sits and where the next major support comes in? Kunkel Capital maps the full multi-year count for gold and silver in our research letter. Same framework that called the 2,000 base, the 2,400 breakout, and the current 5-wave structure still developing into its terminal leg.
