Long-term strategic asset allocation: An out-of-sample evaluation
We investigate the out-of-sample performance of realistic strategic asset allocation models by analyzing plug-in and decision-theoretic approaches with restricted and unrestricted portfolio weights. This paper makes three major contributions. Firstly, this paper refines ex-isting numerical techniques for the calculation of dynamic strategies. Secondly, these refined techniques are used to test myopic, constant proportion and dynamic strategies empirically.Finally, we propose a shrinkage prior with superior performance. The empirical results show that myopic, constant proportion and dynamic strategies could lead to very unstable results unless shrinkage estimators are applied. Dynamic strategies outperform myopic strategies only when using shrinkage. Certainty equivalence returns increase to more than 10% annually since shrinkage-based strategies avoid extreme events and result in less variable portfolioweights. Shrinkage turns bad models into good models and good models into great models.Finally, incorporating parameter uncertainty improves performance when restricting portfolio weights.