Forex Pair Correlation vs. Currency Exposure: Why Co-Movement Is Not a Hedge

A quoted exchange rate is a price between two currencies. EUR/USD at 1.1000 means one euro is quoted at 1.10 U.S. dollars. Under a simple linear spot-style example, buying 10,000 EUR/USD units means being long 10,000 euros and short about 11,000 dollars at the opening rate. The dollar is one leg of the exposure, not the entire trade.

Selling a pair reverses those legs. A short GBP/USD position of 8,661.42 pounds at 1.2700 is approximately short 8,661.42 pounds and long 11,000 dollars. At those opening rates, the two positions below have nearly equal and opposite dollar legs:

• Long 10,000 EUR/USD at 1.1000: +10,000 EUR and about −USD 11,000.
• Short 8,661.42 GBP/USD at 1.2700: −8,661.42 GBP and about +USD 11,000.

This ledger is an exposure illustration, not a universal account statement. Deliverable currency exchange, rolling OTC forex, and CFDs can have different contract, settlement, margin, and reporting terms. Check what the quantity means for the actual product before treating it as a currency balance.

A common measure is the Pearson correlation coefficient between two return series. It ranges from −1 to +1: +1 means the observations in the sample line up in a perfect positive linear relationship, −1 means a perfect negative linear relationship, and 0 means no linear co-movement in that sample. A zero result does not prove independence or the absence of every other relationship.

In simplified notation, correlation is covariance divided by the product of the two return series’ standard deviations. The division removes the units and scales the result to a number between −1 and +1. That normalization is useful for comparing co-movement, but it discards the size of each series’ fluctuations. Two pairs can therefore have correlation +1 even when one typically moves only a fraction as much as the other.

Correlation also does not describe the currencies held by a trade. It compares the returns of two price series. Exposure depends on the pair direction, the number of base-currency units, the quote rate, and the product’s contract rules. The IG guide to currency-pair correlations explains the coefficient and discusses why using a second pair to hedge can leave exposure to the two non-dollar currencies.

A correlation number is incomplete without its inputs. State the two instruments, the quote direction, whether returns are simple or logarithmic, the observation interval, the sample window, and how timestamps were aligned. Daily closes taken at different local times can capture different news and liquidity conditions; a short-interval series can also contain missing or stale quotes.

Use changes in returns rather than correlating two price levels by default. Price levels can trend together for reasons that do not mean their next-period changes move together. With log returns, reversing a quote from EUR/USD to USD/EUR reverses the return’s sign exactly. Simple percentage returns also change sign when the quote is inverted, but their magnitudes are not exactly symmetric for larger moves. Put pairs in a consistent direction before interpreting the sign.

The window matters too. A six-week estimate answers a different question from a three-year estimate. FOREX.com’s correlation lesson presents historical pair estimates over several horizons; those figures describe its dated sample, not a current market reading or a stable property of the pairs. A Federal Reserve study of exchange-rate dependence examines how currency dependence can vary over time and considers business cycles and interest-rate differentials. It is research context, not a forecast for today’s trades.

Consider an invented four-observation sample. EUR/USD returns are +1%, +2%, −1%, and −2%. GBP/USD returns are +0.4%, +0.8%, −0.4%, and −0.8%. Every GBP/USD return is 40% of the corresponding EUR/USD return, so the Pearson correlation in this sample is exactly +1. The pairs move together perfectly in the sample, but not by the same amount.

Now assume a long EUR/USD position and a short GBP/USD position each have USD 11,000 of opening quote-currency notional. The first position is 10,000 euros at 1.1000; the second is about 8,661.42 pounds at 1.2700. On the first observation, a 1% EUR/USD rise contributes about +USD 110, while a 0.4% GBP/USD rise costs the short position about USD 44. The combined simplified price result is +USD 66, not zero. If both returns reverse, the combined result is −USD 66.

The arithmetic uses notional multiplied by the assumed percentage price change. It excludes spread, commission, financing, conversion, slippage, and any change in position size. It demonstrates a narrow point: equal dollar notionals plus a perfect positive correlation do not create an automatic hedge when return volatility differs. A coefficient measures direction and consistency of co-movement; it is not a one-for-one P&L offset.

<!-- learn:illustration -->

For the same opening positions, the USD legs are approximately −USD 11,000 and +USD 11,000, so their initial dollar exposure nearly nets to zero. The other legs remain: long 10,000 euros and short about 8,661 pounds. Their relative value can change. The pair-price example above shows one way that residual exposure can appear in P&L even while the starting dollar legs offset.

A useful first pass is a currency ledger. Write every open position as a signed quantity in each currency, then convert those amounts into the account currency using a stated rate and time. Add the legs by currency. This makes a shared USD exposure visible and prevents two similarly named pairs from being mistaken for two identical bets. If a position is partly closed, update the ledger for the units that actually remain.

The ledger is still not a complete risk model. Currency values move, units may be denominated differently from a platform’s lot display, and a provider can apply its own conversion and netting rules. See the forex lot-size and pip-value guide for unit conventions and the trade P&L reconciliation guide for translating actual fills and charges into account currency.

Portfolio variability depends on both return volatility and correlation. For two returns R1 and R2, a simple long-short return R1 − hR2 has variance σ1² + h²σ2² − 2hρσ1σ2. Here, σ1 and σ2 are the return volatilities, ρ is their correlation, and h is the relative position weight. Correlation is only one of the inputs.

In the invented sample, the second pair moves at 40% of the first pair’s rate. With equal dollar weights, the return difference is 1.0% − 0.4% = 0.6% on the first observation. That residual remains even though ρ is +1. A statistical weight estimated from past data may reduce historical variance under its assumptions, but it does not guarantee future protection. The estimate can change when the sample, volatility, quote direction, or market regime changes.

A correlation table also cannot show every risk that matters to the account. It does not by itself show a gap, a one-sided currency shock, a spread widening, or how positions behave when trading hours and liquidity differ. Model the currency legs and price sensitivities separately, then stress plausible moves that do not follow the recent average relationship.

Two positions can each require margin, incur separate spreads or commissions, and accrue different financing adjustments. The provider may calculate exposure gross or net, apply size tiers, or close positions under its own threshold and liquidation sequence. A position that appears offset in a spreadsheet may still use substantial available margin or be closed before the other leg can offset it.

Retail OTC forex is also not the same product everywhere. The CFTC’s forex advisory describes U.S. OTC forex risks and explains the dealer-counterparty context for that market. It is not a global rule for every CFD, deliverable exchange, or jurisdiction. Confirm the legal entity, product, order handling, margin terms, and customer protections that apply to the account. For related mechanics, see forex margin and leverage, spread versus commission, and rollover financing.

Before treating two pairs as a hedge, record the instrument and quote direction, actual unit quantity, entry rates, currency legs, account currency, and conversion method. For the return comparison, save the data source, timestamps, frequency, return definition, sample dates, and the volatility of each series alongside the correlation coefficient. A screenshot of a coefficient without these inputs cannot be reproduced.

Then calculate a simple scenario table: both pairs up, both down, one up while the other is flat, and a larger-than-recent move in either pair. Recalculate price P&L from each position’s quantity and price change, and list spreads, commissions, financing, conversion, and margin separately. Compare the result with the provider’s statement and contract terms, not just with an online correlation calculator.

This is a way to inspect exposures, not a signal to open or size a trade. If comparing accounts, check whether each provider offers the exact products and position modes you need, how it quotes the pairs, and how it handles costs, netting, and close-out. A correlation estimate can inform a risk review; it cannot replace the currency ledger or establish that a hedge will work.