The dollar index and real yields:how much do they really move gold? (measured)

Almost every piece of gold analysis contains the line "gold is negatively correlated with the dollar index and real yields". True — but it is repeated so smoothly that many people treat it as a rule you can place orders on: see DXY tick up, short gold, and then get run over by a wave of central bank buying or safe-haven flow.

No mysticism. Three things only: take DXY and real yields apart and show the arithmetic, measure how large the correlations actually are on a year of real data, and then say when the relationship breaks down and how to use it on 60-second to 5-minute trades.

1. What DXY actually is: a weighted basket of six currencies

Many people assume DXY is a global gauge of whether the dollar is strong. It is only the dollar price against a basket of developed market currencies, at fixed weights:

Weights: ICE dollar index futures contract specification (public documentation, fixed since the euro was introduced in 1999, checked 2026-10-23).

Two implications that many overseas Chinese investors miss:

• DXY contains no renminbi and no emerging market currency at all. So a rising DXY does not mean a rising USD/CNY. If you watch gold from Singapore, Canada or Australia, your local price carries another layer of local-currency-to-dollar FX, and the two frequently diverge.
• The euro at 57.6% means DXY is mostly a European story. European energy prices and the ECB rate path move DXY more than US data does — one reason the correlation with gold is sometimes strong and sometimes weak.

2. Real yield = nominal yield − inflation expectation

Gold pays nothing. Hold it and that is the number on your statement, forever. Buy TIPS instead and you earn a real return after inflation — that is the real yield, and it is the opportunity cost of holding gold.

With three numbers side by side, here is what the formula looks like:

Note the 5bp gap between 2.81% and 2.76%: the two readings are a day apart, and the breakeven is fitted from the whole nominal-versus-TIPS curve rather than a simple subtraction. A gap that small shows the formula is good enough in practice, but do not expect it to be precise to two decimals.

Why not just use CPI? Because CPI is inflation that has already happened — the rear-view mirror. The bond market prices the inflation expectation for the next ten years, which is the question that actually matches whether to hold gold now. The violent reaction of gold on CPI day is, at root, the gap between realised inflation and expectation; the mechanism is broken out separately in CPI and gold: what the price does on inflation day.

3. The mechanism behind the negative correlation: one opportunity-cost chain

The whole logic is one sentence: real yields rise → the real return on TIPS rises → the relative cost of holding gold rises → gold comes under pressure. Reverse it: when real yields fall or turn negative, the zero coupon on gold stops being a drawback.

Over the past year the chain holds on an annual horizon:

• 2026-01-29: gold peaks at 5,318.40, with the 10-year TIPS real yield at 1.89%, the nominal yield at 4.24% and inflation expectations at 2.35%
• 2026-10-21: the real yield is up to 2.76% (+87bp cumulative) and gold is 4,150.00, -22.0% from the high; the 10-23 close was 4,007.90, -24.6% from the high

Direction matches the textbook exactly. But to conclude from this that every 10bp on the real yield moves gold by a fixed number of dollars will get you beaten up repeatedly intraday — on short horizons this chain is mostly noise. Let us measure it.

4. Measured: how strong is the gold-DXY correlation over the past year

Three things must be read out of that table:

• The negative correlation is real, but only moderate. -0.374 over twelve months is R² = 0.140. In plain terms: the dollar index explains 14.0% of the daily variance in gold and the other 86% comes from something else. Using it as your only basis is trading with your eyes shut.
• The transmission has an elasticity. Over the same period the standard deviation of daily gold moves was 1.747% and of DXY 0.322%, and the regression beta is about -2.03: a 1% move in DXY corresponds on average to a 2% move in gold, in the opposite direction. That 2x relationship is far more useful than the words "negative correlation".
• On daily frequency the real yield is actually less useful than DXY. -0.118 is clearly weaker than -0.374, and the reason is not mysterious: TIPS yields often move only 1–2bp a day and frequently not at all, so the series is sticky and the daily signal-to-noise ratio is poor, while DXY is continuously quoted with dense intraday information. Textbooks describe the long-run logic; traders need a tradeable frequency.

One more angle closer to live trading — days moving in the same direction:

The rolling 60-day correlation moves too

Correlation is not a constant. On a rolling 60-day window, across 442 windows over the past two years:

• The low was -0.768 (2025-07-07), a stretch when the dollar and gold were near mirror images
• The high was +0.152 (2025-02-18), when the relationship turned positive and the two rose and fell together
• The 2026-10-22 reading is -0.494, in the fairly strong band
• Only 18 of the 442 windows were positive, so the negative relationship is the norm — but it fails in phases

5. When the correlation breaks down

These three situations are where the relationship most often snaps. The test is simple: once the rolling 60-day correlation is above -0.3, stop using DXY to infer gold.

1. Risk-off: panic buys dollars and gold at the same time

In a genuine geopolitical or liquidity event, money rushes into both the dollar and gold, and both rise. Conversely, if the market only wants cash, gold gets sold too — 2026-01-30 is an extreme sample: gold fell from 5,318.40 the previous day to 4,745.10, -10.8% in one session, while DXY moved only from 96.28 to 96.99 (+0.74%). Explaining that 10.8% with DXY is like measuring an earthquake with a ruler.

2. Central bank buying: a buyer that ignores rates

Reserve allocation by central banks is a strategic decision that ignores the current real yield and the dollar index. Such buying is price-insensitive and irregular in time, and when it is heavy gold can rise while real yields are climbing. The World Gold Council publishes official-sector net purchases quarterly; for specific tonnage rely on that quarterly report — this article cites no unverified tonnage.

3. Dollar squeeze: liquidity before everything

Both March 2020 and September 2022 produced dollar squeezes: funding markets ran short of dollars, institutions were forced to sell anything sellable for dollars, and gold and the dollar moved together with the traditional negative correlation completely absent. Such windows usually come with wider cross-currency basis and abnormal gold lease rates, and their signature is short, violent and large.

4. The three seconds around a data release

At the moment payrolls, CPI or the FOMC print, quotes gap and spreads widen, and no correlation holds for a while. The mechanism and how to avoid the damage are in the Fed, FOMC and gold.

6. How to use this on a 60-second to 5-minute decision

Conclusion first: it is a filter, not a signal source. And before you use it you have to work out the cost — otherwise you can be right on direction and still lose.

Step one: work out the cost hurdle (most people skip this)

These numbers all come from the XAUAD quote feed, sampled 2026-10-23 23:27 Beijing time:

The current spread is USD 0.658. The percentage is spread ÷ median range. With gold at 4,000, a 1-minute median range of USD 1.81 is about 0.045%.

These numbers are brutal: on a 1-minute timeframe the spread eats 36% of your median range. If you are trying to grab a small piece at 1-minute resolution, cost takes more than a third, and your win rate has to be well above 50% just to have positive expectancy. Five minutes is better at 18%, but still not casual-trading territory.

Step two: use DXY as a direction filter, not an entry signal

1. Read DXY direction on the 5-minute chart. If DXY is clearly strengthening on 5 minutes and you want to be long gold, that trade is against the macro filter — not forbidden, but halve the size.
2. Check the filter is valid. With the rolling 60-day correlation between -0.3 and -0.8 the filter works; once it is above -0.3 or turns positive (18 times in the past two years) the filter is off, and you switch to something else — risk sentiment, central bank headlines, technical levels.
3. Convert to expected magnitude. Beta is about -2.03, so a 0.1% DXY move implies roughly 0.2% in gold, which at 4,000 is USD 8. Compare that with a 1-minute median range of USD 1.81 — by the time DXY has moved meaningfully, the 1-minute window in gold closed long ago. On short timeframes the job of DXY is to confirm direction, not to forecast magnitude.

Step three: only trade when all three stack

One more execution detail: do not chase with market orders on a 60-second timeframe. The average 1-minute move is only 0.028%, and a single 0.658 spread is 0.0165% — more than half the average move. Rest a limit order at a pullback level and the cost is controlled; use a market order and you are paying a certain cost for an uncertain gain, which compounds into negative expectancy. For how spreads convert into money per lot see the gold P&L calculator; for the cost structure by timeframe see short-term gold tactics.

There is also the session layer: the same ~USD 0.66 spread is 26.8% of the 1-minute median range between 20:00 and 24:00 Beijing time, but 68.6% between 04:00 and 08:00. Right direction with doubled cost is the most common way these trades die — for when it is cheaper to trade, see the best hours for gold: three sessions measured.

7. Three common misconceptions

• Treating DXY as USD/CNY. There is no renminbi in the DXY basket. CNY gold = USD gold ÷ 31.1035 × FX, and the FX leg is an independent variable that often offsets the other one. For today reading see today gold price.
• Treating the nominal yield as the real yield. Nominal 5.13%, real 2.76% — the 2.37 point gap is inflation expectations. Get that one number wrong and your direction can invert completely.
• Treating correlation as causation, and as a constant. Over the past two years the rolling correlation has swung between -0.768 and +0.152. It is not a law of physics, it is a statistical regularity for a period, and it changes without warning.

For the everyday reading of the dollar index and Treasury yields, see the dollar index and Treasury yields: how they affect gold; this article focuses on quantifying the correlation and turning it into order rules, so the two do not overlap.

If you care more about how to read the real-yield line itself — the difference in definition between TIPS and breakeven inflation, the comparison of daily / 5-day / 20-day correlations over 247 trading days (-0.140 / -0.418 / -0.611), and the four-step filter from 60 seconds to 5 minutes — see real yields and gold: reading the 10-year TIPS yield.

8. Frequently asked questions

Q. If the dollar index rises, must gold fall?
A. No. Over 252 trading days the measured correlation was -0.374, the two moved in the same direction on 90 days (35.7%), and DXY explains only 14.0% of the daily variance in gold. Fine as a filter, useless as a prophecy.

Q. Why look at TIPS rather than CPI for the real yield?
A. CPI is inflation that has already happened; the TIPS yield is the real return the market is pricing for the next ten years right now. On 2026-10-21: nominal 5.13%, TIPS 2.76%, and the 2.37 point gap is inflation expectations.

Q. Is gold more correlated with real yields than with the dollar index?
A. Daily data over the past year does not support that: -0.118 versus -0.374. TIPS moves are too small and too sticky day to day, so noise dominates. But over a year the direction still holds: the TIPS real yield rose from 1.71% on 2025-10-23 to 2.64% on 2026-10-22 (+93bp) while gold (COMEX GC=F daily close) fell from 4,145.60 to 4,007.90 (-3.3%); measured from the 2026-01-29 high of 5,318.40 the decline is -24.6%. Do not confuse the two measures.

Q. What correlation counts as valid?
A. On a rolling 60-day window, treat |ρ| below 0.3 as broken. Only 18 of the 442 windows over the past two years were positive, and the current reading (2026-10-22) is -0.494.

Q. Is the dollar index any use for 60-second trading?
A. Only as a direction filter. With the spread at USD 0.658 against a 1-minute median range of USD 1.81, cost takes 36% before you start, so the marginal value of a directional read is limited.