Estimating the Cost of Equity : Why Do Simple Benchmarks Outperform Factor Models?

We compare the accuracy of cost of equity estimates based on leading factor models to two simple alternatives: the asset mean and the market mean. The market mean proves to be a serious competitor to traditional implementations of factor models even if the underlying factor model is true. Pricing errors (alphas) that are negatively correlated with firm and industry market betas further improve the performance of the market mean relative to model-based approaches. We propose Bayesian estimators that address key weaknesses in standard plug-in approaches. These estimators outperform plug-in methods and the market mean in simulations and out-of-sample