Can conditioning information add value to portfolio choice? An out-of-sample analysis

Using stochastic optimization, we derive an unconditional mean-variance efficient portfolio rule that optimally exploits the conditioning information from common predictors. This unconditional rule maximizes the unconditional mean-variance utility and Sharpe ratio across all dynamically-managed strategies. Out-of-sample analysis reveals that unconditional rules based on individual predictors generally surpass their conditional counterparts, although their out-of-sample performance exhibits significant heterogeneity. To address this, we propose a simple combination strategy that aggregates all the single-predictor unconditional rules. Empirical results show that this combination strategy delivers stable out-of-sample performance and, in most cases, outperforms sophisticated fixed-weight rules, highlighting the economic significance of incorporating conditioning information in constructing efficient portfolios.