Currency Correlations: The Dollar, Bond Yields and Commodities
Currency correlations measure how exchange-rate returns have moved alongside other market returns over a defined historical sample. They can help describe relationships between Forex pairs, the US dollar, government bond yields, gold, oil and stock markets. Correlation is not causation, however, and relationships can weaken, reverse or disappear when economic regimes change.
This guide explains the mathematics, the economic mechanisms and the limits of using historical correlations for analysis. It also distinguishes index construction from market behaviour and presents a disciplined approach to rolling correlations without turning them into automatic trading instructions.
What is correlation in Forex?
For the economic mechanisms behind relative exchange-rate movements, read What Moves Exchange Rates?.
Correlation is a statistical measure of the direction and strength of association between two variables. The familiar Pearson correlation coefficient ranges from −1 to +1: +1 represents a perfect positive linear relationship in the selected sample, −1 a perfect negative linear relationship and 0 no linear relationship. None of these outcomes proves one variable causes the other.
In markets, analysts ordinarily calculate correlation using returns or changes rather than the raw levels of two trending price series. Correlating two rising price levels can produce high apparent relationships even if the underlying daily return innovations are unrelated. Data frequency, missing observations, currencies, sample windows and outliers can materially change results.
The sample must also align calendars and timestamps. For example, daily FX closes and commodity futures settlement prices might be captured at different moments; matching same-calendar-day observations without checking the precise time can create misleading lead–lag relationships.
Pearson correlation: how the calculation works
For paired observations of returns x and y, the sample Pearson correlation is the sample covariance divided by the product of the sample standard deviations:
r = Σ[(xi − mean x)(yi − mean y)] / √(Σ[(xi − mean x)²] × Σ[(yi − mean y)²]).
For a simple hypothetical example, assume percentage returns on a currency pair across four comparable periods are +1%, −1%, +2%, −2% and the corresponding returns on a benchmark are −1%, +1%, −2%, +2%. The series are exact negatives: the Pearson correlation is −1 for these four observations. This illustrates the calculation, not a realistic claim about live markets; four periods are far too few for reliable inference.
One cannot conclude that a real series with a historical correlation of −0.7 will produce a −0.7 relationship next month. The coefficient measures one sample and is sensitive to regime changes, volatility clustering and the calculation window.
Pair quotation conventions can manufacture apparent sign changes
EUR/USD is dollars per euro, while USD/JPY is yen per dollar. A broad strengthening of the US dollar can push EUR/USD down and USD/JPY up, creating a negative association even if the euro and yen have not experienced identical independent shocks. Always identify which currency is first in each pair.
Reversing a pair changes the sign of log returns exactly, because log(1/S) = −log(S). For simple arithmetic returns, the opposite series is not exactly its negative for finite moves: the inverse return is a nonlinear transformation, although the approximation is close for small changes. This distinction matters when researchers compare reciprocal currency quotes quantitatively.
What does the US Dollar Index measure?
The ICE U.S. Dollar Index (often called DXY or USDX) measures the dollar’s value against a fixed basket of six currencies using a geometric weighting method. According to ICE’s index material, the basket is heavily weighted toward the euro; the index is not an equal-weighted measure of all global dollar activity.
| Basket currency | ICE USDX published weight |
|---|---|
| Euro | 57.6% |
| Japanese yen | 13.6% |
| British pound | 11.9% |
| Canadian dollar | 9.1% |
| Swedish krona | 4.2% |
| Swiss franc | 3.6% |
These are the fixed published USDX basket weights, not a claim about the proportions of current trade or holdings. ICE explains that the index reflects a specific historic six-currency benchmark; trade-weighted dollar indices published by other institutions differ in constituent currencies, methodology and rebalancing.
Because the euro has the largest USDX weight, DXY and EUR/USD commonly display a strong inverse association, partly as a consequence of mathematical construction. It would be inappropriate to treat the index and EUR/USD as wholly independent signals confirming one another.
Exchange rates and government bond yields
Government bond yields can influence currency pricing through changes in expected policy rates, real returns and perceived fiscal or credit risk. A rise in US Treasury yields may accompany dollar strength if investors revise US real-rate expectations higher. The same rise may coincide with dollar weakness if it reflects inflation risk, fiscal anxiety or a broader reduction in confidence.
Compare yield differentials, not only one country’s absolute yield. A two-year US yield rising from 4% to 4.25% while a comparable UK yield rises from 3% to 3.50% narrows the US–UK nominal differential from 1.00 to 0.75 percentage points, despite both yields rising. The resulting currency implication is conditional and depends on risk, inflation expectations and market pricing.
Bond yields are not homogeneous: overnight rates, two-year government yields, ten-year yields and inflation-linked real yields respond to different mixtures of policy expectations and premia. Choosing a convenient yield maturity after a currency move can create misleading hindsight explanations.
Gold and the US dollar
Gold is influenced by multiple factors, including real yields, reserve demand and positioning. See Gold Trading Fundamentals.
Gold is often quoted in US dollars, and its price can respond to real-rate expectations, investment flows, risk sentiment, central-bank reserve activity and currency changes. At times a stronger dollar coincides with lower dollar-denominated gold prices, but that is not a permanent law. Gold and the dollar can both rise during episodes of perceived financial or geopolitical risk.
Numerically, converting a dollar-denominated gold price to sterling involves the relevant GBP/USD rate. Suppose gold is hypothetically $2,000 per troy ounce and GBP/USD is 1.2500: the sterling equivalent is £1,600, ignoring fees and local pricing. If the dollar gold price stays $2,000 but GBP/USD falls to 1.2000, the sterling equivalent rises to approximately £1,666.67. Currency conversion alone altered the local-currency gold price.
This does not establish a causal price forecast. It shows why the investor’s base currency matters and why dollar-denominated commodity returns should not be compared casually with non-dollar FX returns.
Oil, the dollar and commodity-exporting currencies
Energy markets affect currency fundamentals through export revenues, imports, fiscal receipts, inflation and corporate investment. A net oil exporter may benefit from higher oil prices under some circumstances, while a net importer may face a more expensive energy bill. But policy, hedging, production, capital flows and international risk appetite can alter the link.
The BIS Bulletin 74 (2023) documents an important historical exception to the familiar negative dollar–commodity relationship: during the studied period, commodity prices and the dollar sometimes rose together, reflecting overlapping shocks and structural changes. This is direct evidence against treating an inverse correlation as a fixed rule.
Risk sentiment, stocks and currencies
Cross-asset correlations often change across risk regimes. Compare Risk-On Versus Risk-Off Markets.
Stock-market returns sometimes move alongside currencies perceived as sensitive to economic growth or financial conditions. Other currencies may benefit from safe-haven demand during stress. But such categories are imperfect: the same currency may behave differently depending on funding trades, domestic interest-rate changes or policy responses.
Foreign exchange and equities can also interact mechanically through international portfolios. A US equity rally may attract foreign capital, but foreign buyers can hedge their dollar exposure. A fund that increases currency hedging while buying US shares can produce FX transactions different from a fund that remains unhedged.
Correlation versus causation: three analytical tests
- Mechanism: identify the economic channel linking the variables, such as expected real rates, foreign-asset flows or commodity export revenues.
- Timing: compare when the supposed driver changed with when FX repriced, using consistent market timestamps.
- Alternatives: check whether a third factor—such as global risk aversion—could explain both series.
A high correlation is not automatically useful for forecasting. Two series can move together because they both react to a common shock; predicting one from the other is difficult if the shock becomes known to both markets simultaneously.
Rolling correlations and regime changes
Rolling correlation recalculates the statistic over a moving window, such as the preceding 60 daily observations. Short windows respond faster to recent changes but can be noisy; longer windows are more stable but may conceal regime changes. An analysis must disclose its return definition, horizon, sample size, handling of holidays and data source.
For an illustrative comparison, consider a 60-day rolling correlation between EUR/USD daily returns and the change in the US two-year government yield. The variable for yields should be expressed consistently (for example, daily basis-point changes), rather than mixed arbitrarily with the level of the yield. The sign and magnitude can differ across monetary-policy regimes.
Cross-asset analysis without double-counting evidence
Suppose DXY is rising, EUR/USD is falling and a dollar-weighted basket of major currencies also rises. These may reflect much of the same underlying dollar move, rather than three independent macroeconomic signals. Similarly, real yields and nominal yields can respond to the same policy surprise. Analytical confidence should not increase merely because closely related indicators agree.
One way to reduce double-counting is to group evidence by independent mechanism: monetary-policy expectations, economic growth, financing flows, positioning, valuation and market liquidity. Ask whether each observation contributes new information or repeats an already considered factor.
Hedging and portfolio implications
Correlation estimates should be evaluated alongside leverage, liquidity and drawdown. See Portfolio Trading Risk.
Historical correlations are commonly used in risk models. Portfolio risk depends on asset volatilities, position sizes and the covariance matrix. If all exposures become positively correlated during market stress, a portfolio thought to be diversified can experience simultaneous losses. Correlations estimated in calm markets may understate this risk.
Even a historically negative correlation between a currency and a commodity cannot be relied upon as a complete hedge. Transaction costs, derivatives basis, mismatch in exposure amount and non-linear payoffs can leave significant residual risk.
A robust correlation-research checklist
- Write down exactly which currency quotes and instruments are being compared.
- Use consistent returns, units and calendar timestamps.
- State sample period, frequency, number of observations and missing data handling.
- Compare multiple windows and separate materially different market regimes.
- Consider outliers, structural breaks and non-linear relationships.
- Investigate the possible common driver, rather than merely quoting the coefficient.
- Distinguish research descriptions from actionable predictions.
Key takeaways
Currency correlations are sample-dependent descriptions, not enduring economic laws. DXY’s composition creates structural relationships with some currency pairs. Bond yields, gold, oil and equities can influence FX through real economic and financial channels, but the sign of any association can change. A professional analysis makes data definitions and uncertainty explicit.
Frequently asked questions
Does DXY always move opposite EUR/USD?
The index’s large euro weight creates a strong mechanical association, but other basket currencies and market effects prevent a fixed one-for-one relationship.
Is a correlation of −0.8 a guaranteed hedge?
No. It describes a past sample; future relationships, volatility and execution conditions may differ.
Do higher US Treasury yields always strengthen USD?
No. The cause of the yield move and relative conditions elsewhere matter.
Does gold always fall when the dollar rises?
No. Both can rise during certain risk and policy regimes; the BIS documents important changes in the dollar–commodity relationship.
Should analysts use price levels or returns?
Returns or appropriately defined changes are usually preferable for studying market co-movement, because correlating trending price levels can be misleading.
Sources and further reading
- ICE, Currency Indices and USDX.
- ICE, U.S. Dollar Index Composition.
- BIS Bulletin 74, Commodity Prices and the Dollar (2023).
- BIS, Financial Conditions in a Changing Global System (2025).
Important notice: General education only, not a market forecast or personalised investment recommendation. Past correlations may break down and leveraged currency positions can result in substantial losses.