Shrinkage Estimation of Cokurtosis Matrix for Portfolio Selection: An Empirical Study

Abstract

In this study, we propose a shrinkage estimation technique that incorporates a multi-factor model to estimate the cokurtosis matrix of asset returns. Our empirical analyses validate the effectiveness of this approach. The analysis results show that compared to other existing higher-order comoment estimation techniques, the multi-factor shrinkage method outperforms in constructing portfolios that consider higher-order comoment effects. Besides, the proposed method performs well in return data with different frequencies and rolling window lengths, demonstrating robustness in its application. This approach is more efficient in managing portfolio risk, providing significant investment value to investors looking to optimize the risk-return trade-off.

Keywords: Higher-order comoments, Multi-factor shrinkage estimator, Portfolio selection, Empirical analysis