Evaluating Pair Selection Methods: Accounting for Asymmetries & Non-Linearities

Abstract

Effective pair trading relies heavily on precise pair selection. Traditional approaches often assume linear and symmetrical relationships, which may not accurately capture real-world market conditions, potentially leading to suboptimal outcomes. This study evaluates both linear methods, such as the Augmented Dickey-Fuller (ADF) test for the price ratio approach and Engle-Granger cointegration, as well as non-linear methods, including AESTAR and Threshold Auto-Regressive (TAR) models, using data from the US (NASDAQ 100) and Pakistan (PSX) stock markets for the period from January to December 2023. The results indicate that non-linear pair selection methods generally outperform their linear counterparts, particularly in the US market. Specifically, the price ratio method with the AESTAR model proves most effective in achieving profitability, especially in emerging markets like the KSE-100. Conversely, cointegration methods, particularly when combined with the TAR model, show greater effectiveness in markets like NASDAQ, where long-term relationships between stocks are crucial. Overall, the study highlights that in both developed and emerging markets, price ratio methods utilizing non-linear tests offer superior performance in terms of profitability, risk management, and trading dynamics, positioning them as the preferred approach for pair trading strategies.