Alpha forecasting in factor investing: discriminating between the informational content of firm characteristics

This paper applies a linear alpha forecasting framework to enhance commonly used factor investing strategies by taking into account the informational content and interaction effects of selected firm characteristics. To demonstrate conditions under which it is beneficial to deviate from equally weighted characteristics, we evaluate a comprehensive number of factor portfolios. We consider four single-factor portfolios with 14 different firm characteristics in total and a multifactor portfolio where all factors are included. Empirically, the strategies are analyzed with the S&P 500, the Stoxx Europe 600 and the Nikkei 225 index. In addition, we also examine the strategies’ performance in a simulation experiment and investigate the properties of the information coefficient estimates as a measure of the informational content. The empirical results are consistent with the simulation results, which reveal that the overall portfolio performance can be improved in well-defined factor models with a high dispersion among the mean information coefficients of the firm characteristics. In contrast, the naïve combination shows a comparable or better performance in factor models with a small dispersion in informational content between firm characteristics.