Capturing risk Premia in the Taiwanese market: A characteristic-free approach

This study uncovers a characteristic-free approach to capturing risk premia in the Taiwanese market. We find that, even without relying on predefined firm characteristics, purely statistical factors can effectively identify systematic risk premia. By using the most liquid Taiwanese companies each week from 2007 to 2023, our empirical results show that the Risk Premium-Principal Component Analysis (RP-PCA) method could effectively capture market risk premia with annualized Sharpe ratios as high as 0.82; in comparison, the equal-weighted benchmark in the same period has only a Sharpe ratio of 0.54. We also consider two approaches to imposing short-selling constraints: direct enforcement of non-negativity on RP-PCA weights and convex non-negative matrix factorization (convex-NMF). The direct method still delivers a Sharpe ratio of 0.78, whereas the convex-NMF approach reduces it to 0.49. The factor regression results indicate that RP-PCA factors are positively associated with momentum and negatively associated with short-term reversal, suggesting that RP-PCA captures price dynamics across different time horizons.