What Moves Gold: Real Yields and the Dollar

Gold pays no coupon, earns nothing and has no central bank setting its price, so explanations for its moves fill the gap. Most of them come down to four readings: real yields, the dollar, buying by central banks, and how speculators are positioned in the futures market. This post takes each in turn: what it is supposed to do, when it has not, and why the four work better as a checklist than as a model. It makes no forecast, and nothing in it is a recommendation to buy or sell gold.

Real yields and the opportunity cost of holding gold

A real yield is a nominal yield minus expected inflation. The market’s own measure is the yield on inflation-indexed Treasuries: FRED publishes the 10-year one as DFII10, from the Federal Reserve Board’s H.15 release. FRED’s 10-year breakeven inflation rate, T10YIE, is the gap between the ordinary 10-year Treasury yield (DGS10) and DFII10: in every daily row since 2010 the three agree to the two decimals FRED publishes.

The mechanism is opportunity cost. Money in gold is money not in an inflation-protected bond. When the real yield rises, holding gold means giving up more real income; when it falls below zero, the bond itself loses purchasing power and the comparison flatters gold. That is why gold has tended to move against real yields, and why a rise in nominal yields matters for gold mainly when inflation expectations do not rise with it.

It is a tendency, not a law. In each of 2023, 2024 and 2025, DFII10 averaged higher than in any year from 2012 to 2022, and gold’s average price still rose in each of those years, on the World Gold Council’s annual averages. Buyers that do not price gold against a Treasury yield can outweigh those that do; central banks, two sections down, are one such group.

Meaterm’s macro tab charts the nominal Treasury yields, the 10-year among them, but not DFII10 or the breakeven; both are on FRED.

The dollar, and when the inverse breaks

Gold is priced in dollars. When the dollar strengthens, the same ounce costs more in every other currency, and the dollar price tends to adjust down; when the dollar weakens, the reverse. So gold often moves against the dollar, and part of that link is plain arithmetic: some of what looks like a move in gold is a move in the unit it is quoted in.

Which dollar index you watch matters for reading the numbers. Meaterm’s macro tab charts the Federal Reserve’s nominal broad dollar index (FRED series DTWEXBGS, from the H.10 release), weighted by trade against a broad basket of currencies, not ICE’s six-currency index usually quoted as DXY. Their levels differ, so compare the direction, not the figures.

The inverse breaks when both are in demand at once. In 2024 the broad dollar index ended the year higher than it began, and gold’s average price rose too. When both are bought as havens, or by buyers with reasons of their own, the arithmetic link can be outweighed for a long time.

Central-bank buying: structural demand, reported late

Central banks hold gold as a reserve asset: it is not a liability of any other government or issuer, and it diversifies reserves held mostly in other currencies. Their decisions follow reserve policy set over years, so this demand is structural rather than cyclical. It does not need a falling real yield or a weaker dollar, and it does not necessarily stop when either turns.

It also arrives late. The World Gold Council’s country-by-country reserve figures are compiled from the IMF’s statistics, run about two months behind and hold only publicly reported changes. Its quarterly Gold Demand Trends, the usual reference for the total, comes out after each quarter ends and adds an estimate for buying that was not reported; for 2025 it put more than half of the year’s central-bank buying in that unreported part. By the time the number is known, the buying it describes is months old. Read it as the backdrop the other readings move against, not as something that changes week to week.

“Inflation hedge”: right over decades, wrong over months

The long-run case for gold as an inflation hedge is that it is nobody’s liability and cannot be issued at will, so over long spans its price has tended to rise with the general price level. The spans are long, and the starting point matters: bought at a peak, gold has taken decades to catch up with consumer prices.

Over months the link is weak. In June 2022 the twelve-month rise in US consumer prices was the largest since November 1981, on FRED’s consumer price index (CPIAUCNS), and gold’s average price for 2022 was little changed from 2021. What moved in the same year was the real yield: DFII10 went from below zero in January to well above it by December, while the 10-year breakeven ended the year lower than it began. An inflation-hedge story would have expected gold to rise that year; the real-yield story would not have.

Gold positioning: Managed Money in the Disaggregated report

Gold is a physical commodity, so the CFTC reports gold futures in the Disaggregated Commitments of Traders report, not in Traders in Financial Futures. Meaterm’s public COT pages read Managed Money for gold: registered commodity trading advisors, registered commodity pool operators and unregistered funds, hedge funds included, trading futures on behalf of clients. The file is the Disaggregated futures-only dataset 72hh-3qpy, and the row is GOLD (CFTC contract code 088691), COMEX’s 100-troy-ounce contract: of the rows whose name contains “GOLD”, the one with the deepest open interest once micro and nano contracts are dropped.

Every long has a short: when Managed Money is net long, the offsetting positions are held by the report’s other categories, including producers, merchants, processors and users that hedge a physical gold business, swap dealers hedging swaps they have written, and traders too small to report. TFF vs Disaggregated COT reports has the full definitions.

The reading is the same as for every market on those pages: the net position, its change from last week, and its rank against gold’s last 156 weekly net positions. Once the full 156 weeks exist, a rank of 90 or above on a net long prints Crowded long, and a rank of 10 or below on a net short Crowded short; what “crowded” really means covers the other two labels. That describes how stretched Managed Money is against its own three-year history. It is not a signal: crowded positioning has stayed crowded for months, and the figure describes the Tuesday before a Friday release, so it is at least three days old when it is published. Gold’s contract page, /cot/gold, has the figure, the CFTC row behind it and the last 12 reports; gold’s row is also on the COT report page and in the weekly archive.

Reading the four together: a checklist, not a model

Each reading updates on its own clock, and a checklist makes that visible:

Then ask which readings agree. When they point the same way, the account you give of gold is simpler. When they disagree, the disagreement is the useful part: it tells you which account you would have to believe, and what would have to change for it to be wrong. Neither case tells you what gold does next.

There is no score here, no weights and no combined number, on purpose. Each link has broken for years at a time, and a model that adds them up would hide the moments when one of them stops working. For the rates side of the same backdrop, the 2s10s spread post reads the yield curve the same way, and what “crowded” really means covers the positioning rank in detail. How the positioning figure is computed is on the methodology page.

Sources: FRED (Federal Reserve Bank of St. Louis) series DFII10, DGS10, T10YIE, DTWEXBGS and CPIAUCNS, read on 27 October 2026; gold’s annual average prices and central-bank demand from the World Gold Council’s Gold Demand Trends; positioning from the CFTC’s Disaggregated futures-only file.

Open the Meaterm terminal free → Charts, a macro dashboard, CFTC COT positioning and fundamentals from SEC filings in one place. Free plan, no card; Pro is €9.99 per month.