MEASURED · 24 SEPTEMBER 2026

Gold vs the dollar: did the relationship really break?

In 2024 gold and the US dollar both rose, and it was widely reported that their old inverse relationship had broken. We measured every trading day since 2014. It did not break — not in 2024, and not in any other year. What changed is how much else was moving gold.

The short version

What everyone says, and why it should work

Gold is priced in dollars. When the dollar strengthens, an ounce costs more in euros, yen and every other currency, demand from outside the US softens, and the dollar price tends to fall. So the rule of thumb is that gold and the Dollar Index move in opposite directions, and many traders keep the two charts side by side.

The mechanism is sound. The questions worth measuring are how strong the effect is, whether it is stable, and whether it is any use for predicting what gold does next — three separate things that usually get blurred into one.

How it was measured, and the data trap we nearly fell into

Our broker lists a symbol that looks like the Dollar Index. It is not: its price runs between 535 and 682, where the real index sits between about 89 and 114, and its daily moves correlate with the genuine index at only 0.55 — far too loosely to be the same thing. Any study built on it would have been wrong from the first number. The broker’s genuine index, meanwhile, only goes back to 2021.

So we rebuilt the Dollar Index ourselves from its published formula — a weighted basket of six currencies, 57.6% of it the euro — and checked the result against the broker’s real index on every day both exist. Daily returns matched with a correlation of 0.99845 and the average difference in level was 0.009%. That rebuilt series runs back to 2014-01-15, giving 3,905 trading days to compare against gold.

It is real, and close to one-for-one

Over the whole sample, gold and the dollar moved in opposite directions with a daily correlation of -0.433. Shuffling the dates 10,000 times never once produced a relationship that strong by chance.

The more useful number is the slope. For every 1% the Dollar Index moved, gold moved 1.04% the other way on average. That is about as close to one-for-one as markets get, and it is the figure worth remembering.

But it is only a fifth of the story

The dollar explained 19% of the day-to-day variation in gold. That is a genuine and sizable effect for a single variable, and it also means that on most days, most of what gold does has nothing to do with the dollar. Interest-rate expectations, risk appetite, physical demand and scheduled US data releases all move gold, and several of them move the dollar at the same time — which is part of why the two look connected.

It never broke — including in 2024

This is the part that contradicts what you have probably read. Measured over any rolling one-year window since 2014, the correlation was negative 100% of the time — never once positive, ranging only from -0.71 to -0.13. And it was negative in every single calendar year:

* Years in which gold and the dollar both rose. “Due to dollar” is what the dollar’s move predicts for gold at the measured slope; “everything else” is the remainder. The two combine, in log terms, to the year’s actual move.

Look at 2024. Both rose — gold +27.2%, the dollar +7.0% — and yet the daily correlation was -0.465, as negative as almost any other year. On the days the dollar rose, gold still tended to fall.

The two facts fit together once you separate them. Correlation measures whether the two move against each other day to day; it says nothing about which way each drifted over the year. The dollar’s rise was worth about -6.8% to gold in 2024. Gold rose +27.2% anyway, because something other than the dollar pushed it up by roughly +36.5%. The relationship held; it was simply outweighed.

The same thing happened in 2016 and 2019, and the right-hand column shows it growing: the part of gold’s move that the dollar does not explain was +36.5% in 2024 and +48.5% in 2025, far larger than earlier years. Central-bank buying and geopolitical demand are the explanations usually given. We did not measure either, so we are not claiming them — only that whatever it was, it was much larger than the dollar effect it sat on top of. How large it has grown once real yields are counted as well is measured in why is gold going up?

It shows up more clearly over longer periods

Day-to-day noise dilutes the relationship; over weeks and months it shows through a little more strongly. Over shorter three-month windows it did occasionally flip positive — 2.6% of the time, peaking at 0.29 — which is probably where the “it broke” stories come from. Over a full year, never.

What it cannot do: predict gold

This is the finding that matters if you trade. Everything above describes the two moving together, on the same day. If the dollar led gold — moved first, with gold following — you could watch the dollar and trade gold on it. It does not.

Today’s dollar move against tomorrow’s gold move correlated at 0.013, with a permutation p-value of 0.42: no relationship at all. Nor does gold lead the dollar (-0.026). We checked that the test could find a lead if one existed — a deliberately planted one-day lead was detected clearly — so this is a real absence, not a blind spot.

The practical reading: the dollar is useful for understanding a gold move after it happens, and nearly useless for anticipating one. By the time you can see the dollar’s move, gold’s matching move has already been made.

What this means if you trade gold

• Use it as context, not as a signal. A strong dollar day is a headwind worth knowing about. It will not tell you tomorrow’s direction.
• Expect it to be outweighed. At under a fifth of gold’s daily movement, the dollar effect loses to other forces most of the time — and in 2024 and 2025 it lost by a wide margin.
• Don’t mistake it for a hedge. Long gold and short dollar tend to gain on the same days, so holding both adds to one exposure rather than offsetting it. They are not the same trade — they share only about a fifth of their movement — but they will not protect each other.

For the other numbers that describe how gold actually trades — its daily range, costs, and how often a stop is hit by ordinary movement — see our gold trading statistics.

Limits

One broker’s prices, 2014-01-15 to 2026-09-24, daily closes. The Dollar Index is rebuilt rather than sourced from ICE directly, though it matches the broker’s index to 0.01%. Correlation is a linear measure and can miss relationships that only appear in extreme moves; the slope is an average across all conditions, not a constant.

The decomposition into “dollar” and “everything else” applies one average slope to every year, which is a simplification — the true sensitivity varies. And the “everything else” column is a residual: it measures how much of gold’s move the dollar does not account for, not what does account for it. The full data behind every figure is published with our other studies.

Common questions