The Efficiency of Linear Value Models for Stock Selection Across Regions and Sectors

The top quintile portfolio of a linear model that equally weighs value factors Book Value/Price, Cash Flow/Price, Dividend/Price, Earnings/Price and Sales/Price and one 9-month price momentum factor has outperformed the MSCI Developed World benchmark by 18% per year since 2001 and by 13% since 1996. Factors can be ranks or values scaled by their monthly standard deviation. This model performs for all regions and sectors, and especially well for small caps and emerging markets. Across universes the average annual outperformance of top to bottom quintile is of the order of 20% since 2001 with information ratios around 2.The bulk of the performance stems from the factors themselves. Performance improvement of the order of 10% is achieved by optimizing model architecture like dynamically adapting factor weights or ranking within sectors. Risk characteristics improve more and information ratios improve by up to 30%. Factors using analyst's estimates underperform factors based on reported data in periods of crisis and have to be used with caution.Simple portfolio construction schemes that keep turnover below 10% monthly allow to implement these performances by feasible equal weight portfolios. As such this simple system draws a bottom line of performance that should be achieved by quantitative investment systems