Correlation is a statistical measure of how two variables move together within a defined sample. Positive correlation means they tended to move in the same direction, negative correlation the opposite direction, and a value near zero indicates little linear co-movement in that sample.
How it works
For market returns, the commonly used Pearson correlation scales covariance by the volatility of each return series. The result ranges from −1 to +1.
The estimate depends on data frequency, return definition, sample dates, and window length. Rolling correlations recalculate the statistic as the window moves through time.
Why it matters
Correlation is useful for describing diversification and changing relationships, but it does not establish causality or guarantee that a hedge will work in the next shock.
Macro regimes can change the sign of relationships. Research documented U.S. equity-government bond return correlation moving from the negative pattern typical of much of the prior two decades to positive after mid-2021.
A simple market example
A stock-bond relationship measured over 30 days can differ from a 1-year estimate because each window contains different shocks and policy regimes.
Common mistakes
Treating a historical correlation as a permanent economic law.
Claiming one asset caused another to move simply because their returns were correlated.
Frequently asked questions
Does correlation imply causation?
No. Co-movement alone does not identify the mechanism or direction of causality.
Why does rolling correlation change?
New observations enter and old ones leave the sample, while volatility and underlying economic drivers can also change.
Can a negatively correlated hedge fail?
Yes. Correlation can rise sharply during a new regime or stress event, and execution/liquidity can also change.
Educational content only. Definitions describe common market usage and may vary by jurisdiction, instrument, or institution.