Option Return Predictability via Machine Learning: New Evidence From China

Abstract

We extend the literature on empirical asset pricing to the Chinese options market by building and analyzing a comprehensive set of return prediction factors using various machine learning methods. In contrast to previous studies for the US market, we emphasize the uniqueness of this emerging market, investigate daily hedging strategies to construct delta-neutral portfolios, and identify the most important characteristics for return prediction. Short-selling restrictions in China's financial market diminish the effectiveness of spot hedging, whereas delta-neutral portfolios based on futures hedging deliver substantial improvements in both annual returns and Sharpe ratios. Machine learning models not only outperform the IPCA benchmark, but also demonstrate strong generalization ability when applied to newly issued option contracts. The out-of-sample performance remains economically significant after accounting for transaction costs.