Quantification of Risk and Return for Portfolio Optimization : A Comparison of Forecasting Models

Crucial for asset allocation and portfolio management is the estimation of volatility and covariance structure of asset returns. Even the simplest Markowitz portfolio construction requires estimates of the mean returns, variance and covariances. Various studies show that the calculation of efficient frontiers can be very sensitive to the estimation of these input parameters. In the literature, there exist a plethora of models promising sophisticated forecasts of risk and return, but how best to obtain reliable and robust estimates is still an issue under investigation. The purpose of this study is to apply a variety of popular techniques for obtaining ex ante estimates of the risk and return of portfolios. Our primary objective is to investigate whether forecasts obtained from different model specifications have an actual effect on optimal portfolios and, most important, whether more sophisticated and well-specified models lead to superior out-of-sample performance