Skip to content
All library documents

Testing Stock Pairs for Cointegration with Engle–Granger and Johansen

Article Quant Q&A · Author: Alec Ric
Read the original

Summary

The document raises a methodological question about how frequently stock pairs are cointegrated. The author reports testing pairs formed from a sample of 50 stocks using Engle–Granger and Johansen approaches. The reported cointegration rates differ sharply: 45% with Engle–Granger and 96% with Johansen, compared with a cited book example reporting less than 5% for pairs.

The author describes the Engle–Granger procedure as fitting an OLS relationship and applying an augmented Dickey–Fuller test to the residuals, while expressing less confidence in the Johansen implementation. No test settings, sample period, pair selection process, significance thresholds, or treatment of multiple comparisons are given. Consequently, the reported percentages cannot establish how rare cointegration generally is; the discrepancy is a reason to check implementation and assumptions before interpreting the results or using the pairs in a strategy.

Key ideas

  • The document compares Engle–Granger and Johansen cointegration tests on stock pairs.
  • The reported rates are 45% and 96%, respectively, for the author’s sample.
  • A cited book example reports a much lower rate, motivating scrutiny of the methods.
  • The sample design and testing details are insufficient to infer general prevalence.

Tags

EquitiesPairs tradingStatistics
Full text
# How rare is cointegration in the case of stocks?


# How rare is cointegration in the case of stocks?












I did some cointegration tests on a 50 stocks sample, whith which I created pairs. I used the Engle-Granger approach and the Johansen approach, both of them gave me a high percentage of cointegrated pairs(45% and 96%). I'm fairly sure about the way I did the Engle-Granger test (OLS->get residual and test for stationnarity with ADF test), but a bit less for the Johansen test, which I modeled after an example in Jansen's book about machine learning. In the same book Jansen found less than 5 % of pairs were cointegrated.

Does anyone know how rare is cointegration ?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.