Pairs trading is a classic “market-neutral” trading strategy. Previously, we highlighted an article that claims that cointegration is a superior method for selecting pairs .
On the same topic, Reference  examined more pair selection methods. Specifically, it investigated the following approaches,
- Distance: Pairs are identified by using distance metrics. This is perhaps the simplest approach.
- Cointegration: Cointegration tests are applied to identify pairs and generate signals.
- Copula: Trading signals are generated by relative value drawn from estimating the joint probability distribution of returns.
- Time series: Focuses on generating trading signals by time series analysis. Often ignores formation period.
- Stochastic: Uses stochastic control theory in determining value and policy functions for this portfolio problem. Ignores formation period.
- Other: Experimental frameworks with less supporting literature. These approaches include machine learning and principal component analysis.
It concluded, contrary to Reference , that the best method for selecting pairs is the distance method. Cointegration is the second-best one only. Note that the results are applicable to the Finnish stock market.
This thesis confirms the assumptions made in previous literature about what types of pairs are suitable for pairs trading. When pair formation is limited to allow only pairs consisting of companies from the same industry, distance-based and cointegration-based selection favors pairs formed of different share classes of one company. An example of this is Orion Class A- Orion Class B. Such pairs provide more trading opportunities than other types of pairs (e.g. Stora Enso – UPM).
Although 2 articles are contradictory, we believe that:
- Simple methods work best. There is no need for more complex ones. Instead of using complex, advanced pair selection methods, one should focus on the strategy robustness, especially on minimizing divergence risks.
- Ideally the pairs should consist of different share classes on the same underlying. If different share classes don’t exist, then the pairs should consist of stocks in the same industries. If we generalize this to futures markets, then trading futures on the same commodities would make sense.
 N. Huck and K. Afawubo, Pairs trading and selection methods: is cointegration superior?, Applied Economics, 47:6, 599-613, 2015
 S. Kohvakka, Pairs trading revisited – the case of OMX Helsinki, Lappeenranta-Lahti University of Technology LUT, 2020
Have an answer to the questions below? Post it here or in the forum