Robust Asset Allocation for Long-Term Target-Based Investing

4 This paper explores dynamic mean-variance asset allocation over long horizons. This is 5 equivalent to target-based investing with a quadratic loss penalty for deviations from the target 6 level of terminal wealth. We provide a number of illustrative examples in a setting with a 7 risky stock index and a risk-free asset. Our underlying model is very simple: the value of the 8 risky index is assumed to follow a geometric Brownian motion diffusion process and the risk-free 9 interest rate is specified to be constant. We impose realistic constraints on the leverage ratio and 10 trading frequency. In many of our examples, the mean-variance optimal strategy has a standard 11 deviation of terminal wealth less than half that of a constant proportion strategy which has the 12 same expected value of terminal wealth, while the probability of shortfall is reduced by a factor 13 of two to three. We investigate the robustness of the model through resampling experiments 14 using historical data dating back to 1926. These experiments also show much lower standard 15 deviation and shortfall probability for the mean-variance optimal strategy relative to a constant 16 proportion strategy with approximately the same expected terminal wealth. 17 JEL Classification: C63, G11 18