What Actually Moves the Price of Gold?

The dollar: the most consistent correlation

Gold is priced in US dollars, so when the dollar strengthens against other currencies, gold becomes more expensive for buyers in India, China, or the eurozone, and demand tends to soften — and vice versa. This is the simplest mechanism of the four, and it's also the one with the steadiest statistical footprint.

Over the past year, the broad trade-weighted dollar index has moved with gold at a daily correlation of -0.51 (as of the reading through September 11, 2026). A correlation is a number between -1 and 1 that measures how closely two things move together day to day; -1 means they move in perfect opposite directions, 0 means no relationship. A -0.51 is a real, moderate relationship — not proof, but the steadiest of the four we can measure. It holds up across the 21-day, 63-day and 252-day windows (-0.56, -0.38 and -0.32 for the narrower DXY dollar index), which is unusual: most macro relationships come and go. Right now the broad dollar index sits at 118.21 (September 11, 2026), down about 1.1% over three months and near the 12th percentile of its one-year range — a soft-dollar backdrop that, mechanically, removes one headwind rather than adding a tailwind.

Real interest rates: weak daily, stronger over six to eight weeks

The textbook driver of gold is the "real interest rate" — the yield on an inflation-protected government bond (in the US, a TIPS, or Treasury Inflation-Protected Security) after accounting for expected inflation. Because gold pays no interest or dividend, the logic goes: the more a safe bond pays you in real terms, the more it costs you to hold gold instead. The US 10-year real yield sits at 2.60% as of September 14, 2026 — up 90 basis points (0.90 percentage points) over the past year and in the 99th percentile of the last five years, a genuinely restrictive level by historical standards.

Here's the honest complication: on a plain daily basis, gold's correlation with the 10-year real yield is only -0.12 to -0.32 depending on the window measured — much weaker than the dollar relationship. But measured on 60-day changes instead of daily wiggles, the same relationship shows up much more clearly, at -0.52 for the 10-year real yield and -0.63 for the 5-year. The reason is scale: real yields grind slowly, so a one-day correlation can't detect a relationship that plays out over weeks. Judging the real-rate channel by its daily correlation alone would be a mistake — it's active, but on a slower clock than the dollar.

Positioning: the futures market is already crowded

The third driver isn't a macro variable at all — it's how one-sided the futures market has become. The Commitment of Traders (COT) report, published weekly by US regulators, shows how many futures contracts speculative traders (as opposed to commercial hedgers like miners and jewelers) hold net long or short. As of the report dated September 8, 2026, non-commercial traders held a net long position equal to 56.4% of total open interest — in the 96th percentile of both the past one and three years, and up sharply (+14,020 contracts) over the prior four weeks.

That reading doesn't predict a price direction. What it does describe is fuel: when almost everyone who wants to be long is already long, there are fewer new buyers left to push the price further on the same news, and a larger pool of existing longs who could unwind if the story turns. It is a measure of crowding, not a forecast.

Central bank buying: real, but invisible to correlation

The fourth driver is the hardest to quantify because it doesn't show up in a daily price series at all. Central banks buy physical gold in the open market or from producers, generally to diversify official reserves away from the dollar. News flow tracked over the past four months shows China's central bank added roughly 20 tonnes to reserves in August 2026, its largest monthly purchase since 2023 and a 22nd consecutive month of buying, alongside a BofA estimate reported September 12, 2026 that central banks' cumulative purchasing potential could exceed 20,000 tonnes. Namibia was also reported (September 15, 2026) to be building its first-ever gold reserves.

This demand is structural — it doesn't trade in and out day to day the way futures positioning does, so it can't be captured in a correlation coefficient. But its footprint may already be visible indirectly: a regression of gold's price level against real rates and the dollar together, run over the past three years of trading, explains only about 21% of the variation in price (and is flagged internally as statistically unreliable at the price-level, since real rates show a fallacious positive relationship in that window). Even the daily-change version of that same model — the statistically valid one — puts real rates and the dollar together at an R² of just 0.17, meaning they jointly account for under a fifth of day-to-day price variation. The unexplained majority is where things like central bank buying, safe-haven demand and term-premium effects live, even though we cannot put a clean correlation number on them.

Ranking the four, and what the ranking can't tell you

Read the table as a ranking by how directly each driver can be measured, not by how important it is: the dollar has the steadiest statistical signature, real rates need the right time window to show up, positioning tells you about crowding rather than direction, and central bank buying is likely large but sits almost entirely in the 83% of price variation the other two don't explain.

That last row is the most important line in this article. A correlation, even a strong one, describes how two things moved together in the past — it is not a mechanism you can trade on, and MacroGold's own backtesting has found no measurable edge from these relationships alone. None of this tells you what real rates, the dollar, positioning or central bank reserves will do next, and nothing here is a reason to buy or sell anything. What the data does show is which levers exist and roughly how much of the recent move each one can plausibly account for — the rest is genuinely unexplained, which is itself useful to know before trusting any tidy story about "why gold moved today." For readers who want to track when the inputs to this — CPI, PCE, FOMC decisions, COT releases — actually land, MacroGold's economic calendar lists the release dates.