Currency correlations and choosing which pairs to watch
Calculate a correlation coefficient from returns, understand why currency pairs move together, and build a small watchlist without duplicate exposure.
What you will learn
- How to calculate a Pearson correlation coefficient from daily returns, step by step.
- Why currency pairs are correlated: shared currencies, shared drivers and the way pairs are quoted.
- How and why correlations change over time, and how to monitor them with a rolling window.
- How to build a focused watchlist of four to six pairs with different drivers.
Explanation
Measuring total open risk across correlated trades ("portfolio heat") was covered in the course "Risk Management" (lesson "Correlation and Portfolio Heat"). This lesson goes one step earlier: understanding where currency correlations come from and using them to decide which pairs deserve your attention.
The correlation coefficient
The Pearson correlation coefficient (r) measures how closely two series move together, from −1 (always opposite) through 0 (no linear relationship) to +1 (always together). Formula: r = Σ[(x − x̄)(y − ȳ)] ÷ √[Σ(x − x̄)² × Σ(y − ȳ)²], where x and y are the returns, x̄ and ȳ their averages, and Σ means "sum of".
Always use returns (percentage changes), not price levels. Two prices that both trend for a year can show a high correlation even if their daily moves are unrelated.
Three sources of currency correlation
- Shared currency: EUR/USD and GBP/USD both contain the dollar. When the dollar moves broadly, both move.
- Quote direction: EUR/USD and USD/CHF are usually strongly negatively correlated because the dollar is the quote currency in one and the base in the other. If you invert USD/CHF into CHF/USD, the sign flips.
- Shared drivers: AUD, NZD and CAD often react to commodity prices and global growth; JPY and CHF often react to risk-off moods (covered in the course "Macro", lesson "Risk-On vs Risk-Off").
Correlations are not stable
A correlation measured over 60 days may look very different over the next 60 days. Central bank divergence, a local crisis or an intervention can break a relationship. A rolling correlation recalculates r every day over a fixed window (for example the last 20 and the last 100 trading days). If the short window differs strongly from the long one, the relationship is changing.
Practical rules of thumb
- |r| above about 0.7: treat the two pairs as largely the same trade (adjusted for direction).
- |r| between about 0.3 and 0.7: partly related; check which driver they share.
- |r| below about 0.3: weak linear link over that window.
These thresholds are conventions, not laws. Correlation only measures linear co-movement and says nothing about cause.
Building a watchlist
Rather than scanning 28 pairs, group them into "driver blocks" and pick one or two pairs per block: a dollar block (EUR/USD or GBP/USD), a yen block (USD/JPY or a yen cross), a commodity block (AUD/USD or USD/CAD) and a European cross (EUR/GBP or EUR/CHF). Prefer pairs whose sessions match your available time (lesson 2) and whose costs are reasonable (lesson 1).
Worked example
Hypothetical example: six daily returns in percent. Pair A (EUR/USD): 0.40, −0.20, 0.30, −0.50, 0.10, 0.20. Pair B (GBP/USD): 0.30, −0.10, 0.40, −0.40, 0.00, 0.10. Six data points are far too few for real decisions; they only show the method.
- Averages: x̄ = 0.30 ÷ 6 = 0.05; ȳ = 0.30 ÷ 6 = 0.05.
- Deviations A: 0.35, −0.25, 0.25, −0.55, 0.05, 0.15. Deviations B: 0.25, −0.15, 0.35, −0.45, −0.05, 0.05.
- Products summed: 0.0875 + 0.0375 + 0.0875 + 0.2475 − 0.0025 + 0.0075 = 0.465.
- Squares summed: A = 0.575; B = 0.415.
- r = 0.465 ÷ √(0.575 × 0.415) = 0.465 ÷ √0.238625 ≈ 0.465 ÷ 0.4885 ≈ 0.95.
Interpretation: in this sample the two pairs moved almost as one. A long EUR/USD and a long GBP/USD at the same time would be close to one larger short-dollar position. On a watchlist, one of the two is usually enough unless you have a specific reason to compare them.
Common mistake
Practice activity
On paper: (1) Three planned trades: long EUR/USD, long GBP/USD, short USD/CHF. Express each as "long X / short Y" and decide what the combined exposure is. (2) Over 20 days, r(AUD/USD, NZD/USD) = 0.45; over 100 days it was 0.85. What does this suggest?
Show solution
(1) Long EUR/short USD; long GBP/short USD; short USD/long CHF. All three are short the dollar, so together they are largely one short-dollar bet, roughly three times the size of one trade if the correlations stay high.
(2) The usual strong link has weakened recently. Something specific to Australia or New Zealand (for example, different central bank expectations) may be driving one of the pairs. Do not assume they will move together.
Limitations: correlations from a short window are noisy; use at least several months of daily returns and recheck regularly.
Key takeaways
- Calculate correlation from returns, not price levels; r ranges from −1 to +1.
- Currency correlations come from shared currencies, quote direction and shared economic drivers.
- Correlations change; compare a short and a long rolling window.
- Build a watchlist with one or two pairs per driver block to avoid duplicate exposure.
Next step
In the next lesson you will see how central bank decisions, forward guidance and interventions move currencies.
Self-check
Answer the questions, then check your answers. You can retry as often as you like.
Automatic self-check for your own revision. Your answers are checked in your browser and are not seen or assessed by us. It is not an exam and not a qualification.
Your progress
“Complete” only means you marked the lesson as done. It is not a test of your understanding. Progress and notes are saved only in this browser on this device — not in an account.