Mean Reversion in Time Series and Cointegrated Pairs
Summary
The article introduces mean reversion through time-series concepts, distinguishing trend, cycles, seasonality, and irregular movements. It explains the intuition of buying when price falls below an average and selling when it rises above one, then relates the premise to stationarity: a series with stable statistical properties can be analyzed for deviations from its mean. It also mentions differencing or detrending as ways to make some non-stationary data stationary, while noting that non-stationary series may not exhibit the reversion a strategy assumes.
For pairs trading, the article proposes testing whether two securities are cointegrated by regressing one price series on another and applying an Augmented Dickey-Fuller test to the residuals. If the residual spread is stationary, the basic trade is to short the relatively expensive asset and buy the cheaper one. A Pepsi and Coca-Cola example is reported as not cointegrated under the stated test conditions. The discussion is introductory: a test result alone does not establish profitability, and the article gives limited attention to parameter choice, transaction costs, structural changes, or risk controls.
Key ideas
- Mean reversion describes a tendency for sufficiently large price deviations to move back toward a historical average.
- Stationarity is presented as a useful property for identifying series that may revert around a stable mean.
- Differencing and detrending can transform some non-stationary series into stationary ones.
- A pairs approach can test whether regression residuals are stationary using an Augmented Dickey-Fuller test.
- For a cointegrated pair, the described position is long the relatively undervalued security and short the overvalued one.
- A stationarity test does not by itself establish that a trading strategy will be profitable.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.