A trader friend told me a while back that gold had become impossible to analyze, and his evidence was honestly decent. The model that worked for over a decade, gold up when real yields fall and down when they rise, broke around 2022 and never apologized. He is right that the model broke. He is wrong that gold became unreadable, because gold has only about four drivers that show up in the data, they take turns being in charge, and you can usually tell which one is driving within a few minutes of checking. The mistake is treating one driver's correlation as a law of physics when it was only ever a description of who the marginal buyer happened to be at the time.
The four drivers worth tracking
Real yields come first because for most of the 2010s they were the whole story. Gold pays no coupon and no dividend, so the cost of holding it is the real return you give up somewhere else, and the cleanest proxy for that is the 10-year TIPS yield. When real yields fall, the penalty for holding a sterile lump of metal shrinks and gold tends to rally. When they rise, the penalty grows and gold tends to sell off. From roughly the financial crisis through 2021 the inverse relationship was tight enough that you could almost trade gold as a leveraged bet on real rates without embarrassing yourself.
The dollar works through two channels at once. Gold is priced in dollars, so a weaker dollar makes it cheaper for everyone who earns euros, yen, or rupees, which lifts demand mechanically. The dollar is also a competing safe asset in its own right, so money rotating out of it needs somewhere to land. Both effects usually point the same way, which is why gold and the dollar index historically move inversely, though the relationship wobbles, and the wobbles are the informative part.
Central banks are the slow structural bid. Reserve managers buy gold for reasons that have nothing to do with TIPS yields: diversification away from the dollar, insulation from sanctions, sometimes plain domestic politics. They buy in size, they report with a long lag, and they are famously insensitive to price. For most of the 2010s they were a steady but modest presence that nobody really needed to model, which stopped being a safe assumption later.
ETF flows are the fast Western money. Funds like GLD publish holdings daily, so you can watch tonnage move almost in real time. ETF investors are the most rates-sensitive and momentum-driven of the four groups, so flows usually amplify the rates story rather than lead it, but the daily data makes them the easiest driver to observe directly.
Who was in charge, and when
The rough history is worth keeping in your head. The 2000s were mostly a dollar story, a long dollar bear market plus emerging-market income growth, and gold ground higher for most of the decade. From the financial crisis through about 2012 the baton passed to real yields and ETF flows together: QE crushed real rates, Western investors piled into the newly convenient gold ETFs, and GLD briefly became the largest ETF in the world in 2011.
Then 2013 produced the cleanest rates-driven move on record. The Fed hinted at slowing its bond purchases, real yields spiked, ETFs disgorged hundreds of tonnes, and gold had one of its worst years in decades. The same logic ran in reverse from 2018 into 2020, when real yields ground down to deeply negative levels and gold made new all-time highs. Up to that point the textbook held together fine.
Starting in 2022 the relationship flipped. Real yields went from deeply negative to their highest levels in over a decade, about as violent a rates headwind as gold has ever faced, and the textbook said the metal should have been demolished. Instead it chopped sideways for a while and then made new all-time highs. The explanation shows up in the flow data. After Western governments froze Russia's central bank reserves in 2022, official gold buying roughly doubled from its prior pace and stayed elevated, while Western ETFs bled holdings almost the entire way up. The marginal buyer switched from a rates-sensitive fund manager to an emerging-market reserve manager who does not care what TIPS yield he is forgoing, because what he wants is an asset nobody can freeze, and the forgone yield is the fee for that. Correlations follow the marginal buyer, so the rates correlation went quiet and the price followed the new bid.
The five-minute checklist
When gold makes a move I want to understand, I run through this in order.
- Check the 10-year TIPS yield over the same window. If gold and real yields moved opposite each other, the move is rates-driven, and the real question becomes what repriced rate expectations, a data print, a central bank meeting, a growth scare.
- Reprice gold in euros, yen, and at least one emerging-market currency. If the move mostly disappears outside dollars, you are looking at a dollar move wearing a gold costume, and you should be analyzing the dollar instead.
- Check ETF tonnage. GLD reports daily and the World Gold Council aggregates the rest. A rally with heavy inflows and quiet rates is Western flow chasing momentum, which is real while it lasts and just as real on the way out.
- If gold is grinding higher against rising real yields, a firm dollar, and flat or negative ETF flows, the bid is structural, almost certainly central banks or Asian physical demand. You will not confirm it for months because official buying data lags badly, but the price behavior is the tell: steady, indifferent to macro headlines, absorbing every dip.
The expensive failure mode is shorting a flow-driven rally with a rates argument. Anyone who made the textbook short against rising real yields in 2022 and 2023 learned that being right about rates and wrong about the marginal buyer still loses money, sometimes a lot of it. The quieter failure mode runs the other way, assuming the structural bid is permanent. Central banks change behavior slowly, but they do change, and if that bid fades while real yields are still high, the old correlation is sitting there waiting to reassert itself.
How I actually use this
I gave up on valuing gold a long time ago. There is no cash flow to discount, so every fair-value model I have seen is a correlation study wearing a suit. The question I find more useful is who the marginal buyer is in the current regime and what would make them stop. Rates regimes end when the rates story changes, dollar moves end when the dollar story changes, and structural regimes end slowly and only show up in lagged quarterly data, which means you should hold views about them more loosely than the daily commentary does.
The checklist will not tell you where gold goes next. It tells you why gold went where it just did, and in my experience that is worth more, because knowing the active driver tells you which data to watch and which arguments to skip. When someone hands you a rates thesis in the middle of a central-bank-driven regime, you do not need to argue. Pull up the TIPS chart, note whether the correlation has been alive or dead over the past few months, and spend your attention somewhere it pays.