Cointegration for Pairs Trading: Stationary Spreads and Tests
Summary
The article introduces cointegration as a way to find a stationary spread from non-stationary asset prices. If two price series share common long-run trends, a weighted combination may remove those trends; the resulting spread can fluctuate around a stable mean and serve as a basis for mean-reversion strategies. It stresses that this long-run relationship differs from correlation, which describes co-movement in returns over shorter intervals. High correlation alone does not establish cointegration, and cointegrated assets can show weak or changing return correlations.
For estimation, the article outlines the Engle–Granger approach: regress one price series on another, then test whether the residual is stationary with an Augmented Dickey–Fuller test. It also names the Johansen method for estimating relationships involving multiple assets and describes simulating a pair using stationary return and spread processes. These techniques require appropriate statistical assumptions and do not guarantee that a relationship will persist or produce profitable trades. The provided text cuts off during its discussion of test application, so some details are missing.
Key ideas
- Cointegration means a weighted combination of non-stationary price series can be stationary.
- A stationary spread can support a mean-reversion trading hypothesis.
- Cointegration describes a long-run price relationship, while correlation measures return co-movement.
- Engle–Granger tests regression residuals for stationarity; Johansen methods can handle multiple assets.
- A detected relationship does not guarantee persistence or trading profits.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.