MEASURED · 2 OCTOBER 2026

Why is gold going up, or down? What actually moves the price, measured

Every explanation of gold’s move names the same suspects: interest rates, the dollar, the Fed, central banks. We took the ones with public data and measured each against the gold price, month by month, as far back as the data goes. Then we asked a harder question: how much of the run from $1,790 to over $5,000, and of the fall since January, do they actually explain?

The short version

• Gold still moves against real yields and the dollar, in the same month. Neither one told you anything about the following month.
• The rally since 2022 is not a rates story. Real yields rose 3.4 points. By the old rule, that and the dollar’s small rise were worth about −30% to gold. It rose 146%. In the sixteen years the rule was fitted on, it never missed by anything close to this.
• The fall since January is about half a rates story. Rising real yields and a firmer dollar account for −11% of the −23%.
• The Fed’s decision itself told you little. The last three cutting cycles were followed by big gains, the three before them were not.

What moves the gold price, measured

The table checks how gold’s monthly change lined up with the change in four candidates. Correlation runs from −1 (always moving opposite) to +1 (always together); 0 means no link. The figure under it is the typical gold move that went with a one-unit change.

The real yield is the 10-year Treasury yield after expected inflation, read from inflation-protected Treasuries (TIPS); the Fed’s daily series for it starts in 2003. It is the return you give up by holding gold, which pays nothing. The dollar here is the Federal Reserve’s broad index against the currencies the US trades with.

Monthly averages. The dollar column before 2003 starts in 1995, when the broad index begins. A correlation of −0.49 means real yields lined up with roughly a quarter of gold’s monthly movement; the rest came from something else.

Two things stand out. Real yields and the dollar are the only candidates with a steady, sizeable link, and both are still there since 2022: a one-point rise in the real yield went with gold about 9% lower, a 1% stronger dollar with gold about 1.8% lower. The Fed funds rate, the rate the Fed actually sets, had almost no month-to-month link before 2022 (correlation −0.09 and +0.04). And before 2003 the nominal 10-year yield had no link at all (+0.05 over 376 months).

None of these links gives a trading signal. Using month-end prices, last month’s change in the real yield correlated −0.06 with next month’s gold move over 283 months, and the dollar −0.02. Both are indistinguishable from zero. Our gold vs dollar study found the same on daily data: the dollar moves with gold, not ahead of it.

Why gold went up when rates said it should fall

To see how much the two real drivers explain, we fitted a simple rule on the 191 months from February 2006 to December 2021: gold’s monthly change against the change in the real yield and the dollar. It says a one-point rise in the real yield goes with gold 10% lower and a 1% stronger dollar with gold 0.65% lower, on top of an average drift of 6.3% a year. It explained 32% of gold’s monthly movement over those years.

Then we ran it forward on what real yields and the dollar actually did after 2021. Real yields went from −0.99% in December 2021 to +2.40% in August 2026, and the dollar ended +2.6%. The rule puts gold at $1,655 in August 2026, or $1,246 without its built-in drift. Gold averaged $4,411.

The split started in 2022, when real yields rose 2.35 points over the year, the biggest December-to-December jump since the series began in 2003: the rule fell and gold held still. Every year since, the gap widened. Is that just the rule’s normal error? Inside its own sixteen years, the furthest gold ever ran ahead of it over any 56-month stretch was 1.58 times, in the 56 months to August 2011, just before the 2011 peak. This gap is 2.67 times. We also rebuilt the rule’s misses 2,000 times, reshuffled in 12-month blocks: none produced a gap this large. Something other than rates and the dollar has been driving the price.

Central banks: the leading suspect, not a proven cause

The biggest documented change in the gold market over those years is official buying. According to the World Gold Council, central banks bought more than 1,000 tonnes of gold in each of 2022, 2023 and 2024, and 863 tonnes in 2025, against an annual average of 473 tonnes from 2010 to 2021. If a buyer that size was not pricing gold off yields, it would break exactly this kind of rule.

We did not measure this, and the timing alone cannot prove it. What the numbers do show is how much needs explaining: a gold price about 2.7 times what rates and the dollar account for. Anyone pricing gold off real yields alone since 2022 has been wrong by that much.

Does gold still follow interest rates?

Month to month, it does, but less tightly. Over the 36 months to December 2021 the correlation with real yields reached −0.77, its strongest; over the 36 months to August 2026 it was −0.25. The dollar link went the other way, to −0.43. So the short-term pull of yields has weakened while the long-term gap opened. That fits a large buyer that ignores yields, though it does not prove one.

Why is gold going down? The fall since January 2026

Gold set its record London afternoon price of $5,405 on 29 January 2026. On 29 September 2026 it was $4,163.40, −23.0%. Over the same days:

• The real 10-year yield rose from 1.89% to 2.91%, a rise of 1.02 points.
• The broad dollar rose 2.5% (to 25 September 2026, the latest the Fed has published).
• On 16 September 2026 the Fed raised its target range by a quarter point, to 3.75% to 4%, its first hike since July 2023.

By the 2006-2021 rule, those yield and dollar moves are worth −11.3% to gold. The actual fall was −23.0%, so rates and the dollar explain about half of it. The other half is the same unexplained part that drove the rally, now going the other way. How this fall compares with every other 20% drop since 1971, and how long those took to recover, is in gold price history.

Does gold go up when the Fed cuts rates?

It is a common line in gold commentary. We took every turn in Fed policy since February 1994, when the Fed began announcing its decisions: the first cut after a run of hikes, and the first hike after a run of cuts. Before 1994 the target moved in steps as small as 1/16 of a point and was not announced, so from 1983 to 1989 there is a “first cut” nearly every year.

Monthly average gold price in the month the change took effect, against the average 6 and 12 months later. The target is the upper end of the Fed’s range since December 2008.

After first cuts, gold was higher a year later four times out of six, a median of +11.4%. After first hikes, three out of six, a median of +3.1%. Across all months since 1994, gold was higher a year later 64.6% of the time, a median of +6.4%.

The last three cycles fit the belief: after the cuts of 2007, 2019 and 2024 gold rose +16%, +31% and +43% over the next year. After the three before them (1995, 1998 and 2001) it moved −1%, −8% and +6%. Six cycles is far too few to call a rule: random 12-month stretches of gold’s own history matched or beat the average after cuts 30% of the time. And the September 2026 hike, 3.75% to 4%, is too recent to judge.

Why is gold so expensive?

Because its price has risen much faster than prices in general. In August 2026 gold averaged $4,411 an ounce. After inflation that is higher than in 98.9% of months since 1971, above the January 1980 average ($2,906 in August 2026 dollars) and the 2011 peak ($2,616). By the rates-and-dollar rule it would be near $1,655, so most of what makes gold expensive is the part that rule cannot explain.

A high price is not, by itself, a sign that a fall is coming. When we tested gold’s inflation-adjusted price as a warning sign in is gold a good investment?, random price paths matched the result about half the time. How gold has done against inflation over every holding period is in is gold a good hedge against inflation?

What this means if you trade or hold gold

• Watch real yields and the dollar, not the Fed headline. They are what gold reacts to in the same month. The Fed funds rate itself had almost no link before 2022.
• Don’t expect them to tell you where gold goes next. Neither last month’s yield move nor last month’s dollar move said anything about next month’s gold.
• Don’t value gold off yields alone. A rule that explained a third of gold’s monthly moves for sixteen years was wrong by 2.7× within five. Relationships in markets drift, so size positions for the case where yours stops working.
• Treat “the Fed is cutting, buy gold” as a story. It worked in the last three cycles and not in the three before.
• For intraday XAUUSD, the yield-moving releases matter. What gold did around US jobs reports is in news trading on gold.

Where gold stood against the stock market through all this is in gold vs the S&P 500. This page is history, not a forecast or personal advice.

How we measured it

Gold: the World Bank’s monthly average price (the “Pink Sheet”), the same series as our gold price history, and the LBMA afternoon price for the record and the fall since it. Rates and the dollar: from the Federal Reserve via FRED: the 10-year TIPS yield (DFII10), the 10-year Treasury yield (DGS10), the effective Fed funds rate, the broad dollar index (the older DTWEXB to 2005, then DTWEXBGS, joined at January 2006), and the Fed’s target rate and upper bound. Daily series are averaged by month to match the gold price.

The rule is an ordinary least-squares fit of gold’s monthly log change on the change in the real yield and the dollar, February 2006 to December 2021. The check on its miss resamples its own monthly errors in 12-month blocks, so runs of misses stay together. The Fed check compares the average 12-month change after each kind of first move with 2,000 sets of random 12-month windows from gold’s history since 1971. Monthly averages smooth single days, so daily moves were bigger than shown here.

Common questions