Title: Asset Pricing Model with Functional Principal Component Analysis
Authors: Bo Li - Beijing International Studies University (China) [presenting]
Zhenya Liu - Renmin University of China (China)
Shixuan Wang - University of Reading (United Kingdom)
Yifan Zhang - Renmin University of China (China)
Abstract: This paper proposes a functional principal component analysis (fPCA) estimator for the characteristic-proxied risk factor in factor models. It first projects individual stocks into the space with a fewer factor structure on univariate characteristics. Then, extracting the statistical factors by fPCA. We empirically verify that the second statistical factor is a reasonable estimator for the characteristic-proxied risk factor. Moreover, we propose using the eigenfunction of the second statistical factor as weights to construct the fPCA factor (mimicking portfolio). Further empirical study confirms that the fPCA factors substantially improve the pricing performance of the multi-factor model for anomalies, and their mean-variance efficiency portfolios achieve remarkable out-of-sample Sharpe ratios, which are higher than 2 with the momentum factor.