Correlation and Cointegration

Core idea

Correlation measures co-movement over a chosen window, while cointegration tests whether a combination of non-stationary series has a stable long-run relationship. High correlation alone does not establish a mean-reverting spread.

Implementation

Relationship stability, residual diagnostics, and out-of-sample behavior should guide pair selection.

Primary risk

Structural breaks can invalidate both the fitted hedge ratio and the assumption of mean reversion.

Frequently asked questions

What is correlation and cointegration?

Correlation measures co-movement over a chosen window, while cointegration tests whether a combination of non-stationary series has a stable long-run relationship. High correlation alone does not establish a mean-reverting spread.

What is the main risk of correlation and cointegration?

Structural breaks can invalidate both the fitted hedge ratio and the assumption of mean reversion.