An Evaluation of Conditional Multi-Factor Models in Active Asset Allocation Strategies: An Empirical Study for the German Stock Market

This paper examines the out-of-sample performance of asset allocation strategies that use conditional multi-factor models to forecast expected returns and estimate the future variance and covariance. We find that strategies based on conditional multi-factor models outperform strategies based on unconditional multi-factor models, and do better than a passive buy-and-hold strategy. However, a strategy that uses the sample mean as a return forecast is superior. We also find that the estimation of the covariance matrices based on the conditional and unconditional multi-factor models does not improve the performance of the active asset allocation strategy relative to the incorporation of the historical covariance matrices. These results are fairly robust to different estimation approaches, as well as to the impact of transaction costs and the consideration of upper and lower bounds for the portfolio weights.