Pre-selection in cointegration-based pairs trading

Abstract

The paper compares the final profitability of a cointegration-based pairs trading strategy when pairs of stocks are pre-selected by means of seven different measures. Some of the measures considered have been extensively used in the pairs trading literature, while others represent a novelty in this type of application. We find that pre-selection matters, since the excess returns remarkably vary, in terms of both average and variability, depending on the metrics used. Differences in profitability by pre-selection metrics are retrieved even after considering commissions and cut rules, market impact, a stricter definition of the Spread reversion to the equilibrium and alternative cointegration tests. Besides, the pairs trading profitability is found to be heterogeneous across the different pre-selection metrics also in terms of exposure to the systematic stock-market risk factors.