ST-metric Estimation of Factor Exposures

Abstract

Non-parametric methods are treasured in data analysis, particularly in finance. ST -metric is a new concept, introduced by Tulunay (2017). It offers non-parametric methods and a new geometric view to data analysis. In that paper, ST-metric concept has been applied to performance measures of portfolios. In this current paper, we purpose another ST-metric method for finding factor exposures in the fiv e-style-factors model. Here the style factors are value, size, minimum volatility, quality and momentum. The main idea is to find the factor exposures (weights) of the five-factors-model by minimizing the ST-metric between benchmark returns and the constructed factor model returns. We compare ST-metric method with Tracking Error method (TE-method) which is used for factor analysis of major indexes, decomposed into the style factors (tradable via Exchange Traded Funds (ETFs)) by Ang et al. (2018). We show that ST-metric method gives better estimation of the factor exposures (weights) than tracking error method, in general, and further how ST-metric values vary with respect to fluctuations. This explains the reason behind the e fficiency of the ST-metric method. We support this idea with empirical evidences.