Can pairing strategies be even more profitable?

: a study on pairs trading for the U.S. Russel 2000 Index constituents

Abstract

Pairs trading is considered as a profitable Wall Street strategy. We analyse the influence of K-Means clustering using EPS and book-value per share ratios on pairs trading returns in the formation period, before coupling pairs, while using 3 different ranking methodologies which include the one proposed by Gatev et al. (2006). We use data from the Russel 2000 index from January 2003 to the end of June 2013, on 3.711 stocks. We find that a strategy based on the orthogonal regression approach outputs an impressive 4.35% statistically significant average excess monthly return and a Sharpe ratio of 3.23 if the K-Means clustering is applied and 20 pairs are selected, which compares to the Gatev’s strategy holding -0.54% without clustering for the same sample. We also find that the orthogonal regression method outputs on average 94 b.p. higher monthly excess returns than the average squared deviation methodology and that using the Pearson correlation coefficient for ranking pairs outputs permanently negative returns, independently of the clustering and number of pairs. Likewise, we find that the profitability of the pairs trading strategy is not dependent on the utilities sector.

Date of Award	2014
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Joni Kokkonen (Supervisor)