The principle of pairs trading is remarkably simple. An investor finds assets whose prices moved together historically, open a trade by shorting the winner and buying the loser when the spread between them widens. The trade is closed when the spread converges. But while it may sound simple, the devil is in the details!

Over the years, pairs trading has become one of the most popular statistical arbitrage strategies. The strategy exploits temporary anomalies between prices of assets that have some equilibrium relationship. While methods may differ in sophistication, all implementations rely on the use of statistical analysis of historical prices to identify pair candidates with stable inter-relationships.

The main challenge in building such strategies is that, often, cointegration between two assets breaks down out of sample, making the trade a losing proposition.

Fundamental Similarity

In an attempt to solve the challenge of cointegration breakdown, investors can benefit from looking for pairs that have some degree of “fundamental similarity.” Typically, pairs trading programs are looking for cointegration relationships between stocks belonging to the same country and sector/industry group.

However, in a recent study, Deutsche Bank utilized a risk model to proxy fundamental similarity. Overall, they found that taking such approach significantly reduced divergence risk across their portfolio, and also improved the average return per pair.

Differentiating Between “Good” and “Bad” Divergence

Even though fundamentally similar stocks are more likely to move in tandem in the near future, there are no guarantees for such behavior. Considering any single stock, a large proportion of the price movement is driven by idiosyncratic risk, which could permanently alter the equilibrium relationship between a company pair.

The profits and risks from trading stock pairs are very much related to the type of information event which creates divergence. If divergence is caused by a piece of news related specifically to one constituent of the pair, there is a good chance that prices will diverge further. On the other hand, if divergence is caused by random price movements or a differential reaction to common information, convergence is more likely to follow after the initial divergence.

Considering Popular Sentiment

To test the effects of news on a pairs trading strategy, Deutsche Bank used two aggregated indicators derived from news and social media data measuring sentiment and media attention. Both of these indicators are available on FactSet. Specifically, using the two indicators, Deutsche Bank created a filter that would ignore trades where divergence was supported by negative sentiment and abnormal news volume. The figure from the report below, illustrates the pairs trading process with the news overlay.

Pairs Trading

Pairs Trading

Deutsche Bank’s key findings included:

  • Lower Divergence Risk: the percentage of non-converged pairs dropped by over a half from 15% to 7%
  • Higher Return: average profit per pair also increased from 2.3% to 2.8%, and the return distribution becomes more positively skewed
  • Significant P-Values: the increase in average returns is confirmed by significant p-values (<0.05) from the one-sided pairwise t-test

Pairs Trading

The figure below shows the results of the pairs strategies applied on the MSCI U.S. universe. As can be seen from the graph, the same conclusions can be reached, albeit the strategies have relatively lower returns in the U.S. The average return per pair under the benchmark strategy, the enhanced strategy using the risk model, and the final strategy with both risk model and news overlay are 0.2%, 1.6% and 1.9% respectively.

Pairs Trading

Overall, the research finds that applying a news analytics overlay can help differentiate between “good” price divergence (which is likely to converge) and “bad” divergence. More importantly, such ability provides significant improvements to the performance of a traditional pairs trading strategy, especially by reducing divergence risk.

This article was originally posted on RavenPack Blog.

Article By Peter Hafez, FactSet