Direct Portfolio Weight Estimator : Mitigating Specification Risk with Realized Utility

Estimation noise is a well-known issue in empirical portfolio modelling. Estimated weights are known to have huge standard errors and bad predictive quality, which often results in an inferior out-of-sample portfolio performance compared to simple alternatives. Most of the recent literature concentrates on the improvement of covariance matrix forecasts, which would hopefully result in better portfolio performance. However, the proposed models often suffer from the curse of dimensionality, such that the forecasting error still dominates the theoretical gain. In this paper, we propose a direct weight estimator (DWE), which explicitly takes into account forecasting risk and avoids the over-parametrization problem by forecasting a one-dimensional portfolio measure directly. We define a forecasting error based on realized measures and optimize for a weight vector which results in a more precise forecast in terms of the forecasting error variance and at the same time is not far from the optimal portfolio solution. The proposed method is shown to outperform commonly used approaches in both simulation and empirical studies