Pair trading or statistical arbitrage is a popular quantitative trading strategy. Basically, we choose a pair of assets for trading, and when the assets’ prices diverge, we bet on them reverting back to the mean. The assets are often stocks, but they can be anything, ranging from commodities, interest rate products, to exchange-traded funds.
To choose a pair for trading, we often make a decision based upon,
- Correlation of assets,
- Cointegration of the underlyings, or
- Other methods.
A common question we often ask ourselves is which method is superior. We used to believe that using correlation or cointegration does not matter, as long as the pair mean reverts. However, article [1] shed some light on the matter. After conducting research, the authors concluded that choosing pairs based on cointegration resulted in superior returns,
Our results confirm the weak excess returns observed in the recent publications for the minimum distance method. The trades initiated following the stationarity of the price ratio are not able to generate, after transaction costs, large and significant excess returns. The key empirical fact revealed by this study concerns cointegration. After controlling for risk factors, transaction costs and data-snooping biases, cointegration-based pairs trading exhibits high and robust positive alpha. During a period of more than 10 years, even the least profitable parameterization dealing with cointegration delivers excess returns greater than 1.38% per month. Returns can rise up to 5% per month. Cointegration reduces significantly nonconvergence risk.
The article stated that cointegration reduced divergence risks, but did not explain why. This is an area for further research. Additionally, it also pointed out,
If pairs trading returns are not related to the equity premium, does the volatility/Vix index matter? Do and Faff (2010) regress pairs returns against market volatility over the same month but did not find any significant effect. A new direction could analyse the influence of conditioning openings to a certain level of volatility/Vix.
Based on our experience, we used to think that pair trading is implicitly shorting volatility, but as the article pointed out, this might not be the case. We look forward to new research results to learn more as to why this is not the case.
References
[1] N. Huck and K. Afawubo, Pairs trading and selection methods: is cointegration superior?, Applied Economics, 47:6, 599-613, 2015
Post Source Here: Cointegration or Correlation, Which Method is Better for Pair Trading?
No comments:
Post a Comment