A Comparative Analysis of Ability of Mimicking Portfolios in Representing the Background Factors

Our aim is to give a comparative analysis of ability of different factor mimicking portfolios in representing the background factors. Our analysis contains a cross-sectional regression approach, a time-series regression approach and a portfolio approach for constructing factor mimicking portfolios. The focus of the analysis is the power of mimicking portfolios in the asset pricing models. We conclude that the time series regression approach, with the book-to-market sorted portfolios as the base assets, is the most proper alternative to construct mimicking portfolios for factors for which a time-series of factor realisation is available. To construct mimicking portfolios based on the firm characteristics we suggest a loading weighted portfolio approach.