Improving statistical arbitrage investment strategy: Evidence from Latin American stock markets

This paper analyses the profitability of pairs trading strategy in Latin American stock markets through a PCA approach with a multi‐factorial model. We propose two criteria for selecting the optimal thresholds by using moving training‐trading windows. By applying the methodology in six countries with 338 stocks in total for the period 2013–2017, we found that this strategy outperforms the markets' Sharpe ratio by 1.55 points on average. In addition, by using a correlation matrix, we found that the largest eigenvalue is clearly dominant and that its associated eigenportfolio presents a co‐movement with the market while the number of dominant components is contrary to market volatility.