How it works
Correlation is computed from a sample of paired returns over a chosen window. Daily returns over 30 to 252 days are typical. The result is the Pearson correlation coefficient. It measures only linear relationship; non-linear dependence (such as same-direction moves with different magnitudes) is missed. Correlation also says nothing about causation.
Example
Spot gold and the US dollar index have a long-term correlation near -0.5: when the dollar rises, gold tends to fall, but the relationship is loose. EUR/USD and GBP/USD have correlation near +0.85: they move together most of the time, but not lockstep. AAPL and SPY have correlation near +0.7: Apple drives a lot of the index, so they share a large common factor. Equity-bond correlation has been positive in some decades and negative in others.
Why it matters
Correlation determines whether two positions diversify your risk or compound it. A portfolio of 10 stocks with 0.9 correlation has nearly the same risk as one position scaled to the same notional. Genuine diversification requires uncorrelated or negatively correlated assets. The catch: correlations spike toward 1.0 in market crises, so the diversification that worked in calm regimes evaporates exactly when you need it most. Stress-test position correlations under stress assumptions, not the long-run average.
