Abstract:
We determine the importance of long-term and short-term components of state variables for asset allocation decisions. The long-term and short-term decompositions are performed using a variety of filtering techniques. We allow for a flexible semiparametric form of the dependence of asset allocation decisions on state variable components. To account for short-sale restrictions, we extend the regular GMM moment conditions with the appropriate Lagrange-Kuhn-Tucker multipliers. Empirically, we find that investors can benefit from reacting differently to short-term versus long-term dynamics of state variables. The induced allocation decisions are implemented in an investment backtest. We find significant improvements in terms of out-of-sample Sharpe ratios and expected utilities for state variables such as the dividend yield and stock market trend.