Interpreting Negative Cointegration Coefficients in Pairs Trading
Summary
The document asks whether a negative coefficient in a cointegrating relation changes how a stock pair should be traded. Its example uses two beverage stocks and a spread formed by regressing one price on the other. It compares the sign of the coefficient with proposed buy and sell signals based on the spread’s distance from zero.
The accepted response says cointegrating coefficients may be positive or negative, but points to an external explanation rather than deriving a hedge rule. Another response proposes reversing both legs’ entry directions when the coefficient is negative. The discussion does not establish that rule, define the spread convention or signal precisely, or assess hedge ratios, stationarity, costs, or risk. Treat it as a question about interpreting the sign, not as a complete or validated trading strategy.
Key ideas
- A cointegrating coefficient can be negative.
- The coefficient sign affects the relationship between the two prices in the spread.
- Trading directions depend on how the spread and its signal are defined.
- The document does not provide a derivation or empirical test for its proposed entry rules.
Tags
Full text
# Can the coef be negative in cointegrated stocks?
# Can the coef be negative in cointegrated stocks?
I'm searching for cointegrated stocks using the Python CointAnalysis library. While computing stock prices on a 5 mins time frame, I found that the stocks MNST (Monster Beverage Corp.) and KDP (Keurig Dr. Pepper Inc.) are cointegrated. The computed spread formula is:
```
MNST - (-1.74 * KDP + 112.84)
```
So, my strategy is based on this article:
`BUY {coeff} KDP` and `SELL 1 MNST` if the spread is one Standard Deviation over 0
`SELL {coeff} KDP` and `BUY 1 MNST` if the spread is one Standard Deviation below 0
But as the coefficient `-1.74` is negative, `KDP` shares quantity is always twisted:
`BUY -1.74 KDP` which is equivalent to `SELL 1.74 KDP`
`SELL -1.74 KDP` which is equivalent to `BUY 1.74 KDP`
So, my strategy becomes:
`SELL 1.74 KDP` and `SELL 1 MNST` if the spread is one Standard Deviation over 0
`BUY 1.74 KDP` and `BUY 1 MNST` if the spread is one Standard Deviation below 0
Is it possible to have a negative coefficient in a cointegration formula?
Does it make sense to go in the same direction on both stocks in cointegration?
## Answer by Sane (score 1, accepted)
https://quant.stackexchange.com/a/80220
In the context of cointegration, the sign of the coefficient in the cointegrating vector can be either positive or negative.
Check the following answer: https://stats.stackexchange.com/questions/566682/can-the-cointegration-coefficient-be-negative
## Answer by Begoodpy (score -1)
https://quant.stackexchange.com/a/76116
Google Bard helps find an explanation for the concept of inverse relationship. When the coefficient is negative, I just have to inverse the direction of the entries on both stocks, not just one. Then if the coeff is `-1.74`, the entry signal becomes:
`SELL 1.74 KDP` and `BUY 1 MNST` if the spread is one Standard Deviation over 0
`BUY 1.74 KDP` and `SELL 1 MNST` if the spread is one Standard Deviation below 0Shown 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.