Skip to content
All library documents

Trading Cointegrated Pairs with a Negative Hedge Coefficient

Article Quant Q&A · Author: pbk
Read the original

Summary

The discussion explains how the sign of a cointegration coefficient affects the positions implied by a spread. With the stated spread defined as Stock A minus beta times Stock B, a negative beta makes the second term positive, so a long spread corresponds to holding both stocks long; a short spread reverses both positions. The questioner reports that the spread is stationary, the stocks are positively correlated, and a backtest and one live trade were profitable, but these observations do not establish that the strategy will remain reliable.

The answer considers the approach reasonable when it follows the spread equation, and gives a short ETF paired with another ETF as an example where a negative relationship can make same-direction positions sensible. It emphasizes that this construction may carry substantial tail risk: ordinary co-movement may break down in extreme markets, and holding two stocks long can amplify losses in a broad selloff. The discussion does not provide a formal sizing method, broader performance evidence, or a way to quantify that risk, so the example should be treated as an explanation of position direction rather than validation of a particular pair.

Key ideas

  • A negative cointegration coefficient can make a long spread require long positions in both assets.
  • Position direction follows the defined spread equation, rather than a universal long-short convention.
  • A backtest and one profitable live trade provide limited evidence about a pair strategy's reliability.
  • Same-direction equity exposure may create substantial losses during broad market declines.

Tags

EquitiesPairs tradingStatisticsRisk management
Full text
# Do i do a long/long or short/short of the trading pair when they have a negative cointegration coefficient?


# Do i do a long/long or short/short of the trading pair when they have a negative cointegration coefficient?












A pair of stock that I have been trading has a negative cointegration coefficient (Beta) that is statistically significant. When i want to long a spread, according to the spread equation below, I should long both stockA and stockB (vice versa for short spread). This pair has strong positive correlations that is statistically significant as well. The spread is stationary and hurst exponent is < 0.5.

Backtest shows the strategy is profitable and I traded it live once by longing the spread (long Stock A and long stock B) and was profitable on both legs in the pair, similar to what i see in the backtest. However, this deviates from the market neutrality principle of pair trading when you typical do a long/short combo for a pair of stock.

Am i doing the right thing by following the spread equation?

```
Spread = StockA - Beta*StockB = StockA - (-0.8)*StockB
```

## Answer by MrLCh (score 0)

https://quant.stackexchange.com/a/80439

In general it seems like what you are doing is reasonable to trade the negative cointegration. The extreme example would be thinking about trading a short etf against another etf. The cointegration could be negative and it would make sense in this scenario to go long both components. (As the one component is shorting the market).

Coming back to your case: There could be a large tail risk in trading this strategy. Maybe the cointegration explains the normal behavior quite well, but the behavior could deviate massively in extreme events. During a market crash typically all stocks tend to suffer and being long both stocks would increase your risk (although there may be stocks that profit from a harsh downturn and still move against the market).

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.