Pairs trading is hardly applied to high-dimensionality datasets. The first step for its implementation, required to identify co-moving assets and usually based on cointegration-tests, comes with a remarkable computational burden. This paper compares seven different preselection measures that, while preserving the profitability for the pairs trading strategy, help in reducing the number of cointegration tests required to identify the most co-moving assets. Among the measures investigated, log-Prices Correlation and Sum of Squared Deviations are the ones providing the most significant results in terms of expected excess returns, whereby the latter is found to be less sensitive to periods of market turbulence.
 

---

Download from REPEC

Download from SSRN

---