Direct answer
Pair correlation is most active when trading sessions overlap and overall market liquidity is higher. In practical terms, this usually means the hours when major trading centers are simultaneously open, because more participants can trade and the same macro/news drivers can affect multiple pairs at the same time.
Mechanism and definition
Pair correlation is a statistical measure of how two return series move together. For currency pairs, a “return series” is typically computed from price changes over chosen time intervals (for example, 1-minute, 15-minute, or hourly changes). Correlation is then calculated over a moving window.
Why sessions matter: when more markets are open at once, trading activity and the number of active orders increase. Higher liquidity often makes price discovery more efficient, and more instruments respond to common information (for example, broad risk sentiment, central-bank-related headlines, or changes in global funding conditions). When the same information hits multiple currency pairs during the same clock hours, their returns can become more synchronized—raising observed correlation.
A simple non-real-time model to check the idea is:
- pick a currency-pair pair (or two pairs you want to study),
- compute returns on consistent intervals,
- calculate correlation separately for different session time windows (for example, “one center open only” versus “overlap hours”). If correlation is “most active” in your data, the overlap windows should often show higher (or at least more stable) correlation than isolated hours.
Evidence and example (independent verification method)
Consider two pairs whose movements are influenced by shared factors. You can test which session windows show stronger co-movement without using live data claims:
- Assumptions: use one fixed definition of returns (same interval and same timezone), and use the same correlation window length across sessions.
- Example setup: compute correlation for the same pair returns over (a) non-overlap hours and (b) overlap hours.
- Interpretation: if the overlap-hours correlation magnitude (absolute value) is consistently larger, that supports the idea that session overlap is where pair correlation becomes more noticeable.
This approach also helps you distinguish two effects:
- “More trading” can change the measured correlation because there are more return observations.
- “Different drivers” can change correlation because the underlying relationships between currencies can shift.
Limitations and failure modes
Pair correlation is not guaranteed to peak during any specific session overlap, and it can fail in several ways:
- Correlation regime changes: relationships can weaken or invert when macro drivers diverge between currencies, even during overlap hours.
- Volatility and tail moves: correlations can be dominated by a small number of extreme moves (for example, sudden news), which makes results sensitive to the chosen time window.
- Market microstructure effects: execution costs and spread changes can affect returns differently across pairs, especially when liquidity thins.
- Different time alignment: incorrect timezone handling or mixing intervals (like using 5-minute returns for one window and 15-minute for another) can distort comparisons.
- Historical non-persistence: even if overlap hours show higher correlation historically, that does not establish future behavior.
Verification and next question
To verify the “most active” sessions for Pair Correlation in your own study, recalculate correlations using the same return definition, the same lookback window, and separate time windows representing (1) a single session open versus (2) major overlap hours. Then check whether the correlation level is higher, more stable, or both.
Next question to consider: which currency pairs share the same underlying drivers you expect (for example, both heavily influenced by the same macro cycle), because correlation is about shared movement, not about clock time alone.