Predictability of Stock Returns: An application of presentvalue state-space models to the German Stock Market

Abstract

In my thesis, I introduce a state-space representation of the present-value model to analyze predictability in the aggregated German stock market. The proposed model uses the information contained in annualized price-dividend ratios and realized dividend growth rates and defines relations to the latent state variables in the form of expected returns and expected dividend growth rates. I apply the Kalman Filter to generate estimates of the model parameters using a conditional Maximum Likelihood Estimation. The corresponding optimization problem is solved via an adjusted version of the Simulated Annealing algorithm. The final model produces good estimates for dividend-growth rates, while it lacks quality in terms of the estimation of stock returns.