Asset class allocation and downside risk: does the investment horizon matter?

The main objective of this paper is to analyze within the Mean-Downside Risk (MDR)-framework the relevance of the investment horizon for deriving optimal US asset class allocations. The choice of this risk framework is motivated by its close connection towards the way investors perceive risk and the fact that it is much more general than the often used Mean-Variance (MV) analysis. Unlike the MV-studies of Levy and Gunthorpe (1993) and Thorley (1995) we do not assume returns to follow a random walk. Instead we use a vector autoregressive specification to model the historical time series such that short-term first-order auto- and crosscovariances are preserved. Different from the MV-paper of Lee (1990) is that we explicitly model the short-term first-order auto- and crosscovariances and that we consider more than two asset classes. Our simulation tests show the weight assigned to stocks to be positively related to the length of the investment horizon. This contradicts the MV-findings of Lee (1990), Levy and Gunthorpe (1993) and Thorley (1995). The relation appears to be most apparent for investors with a low downside-risk aversion. However, even investors whose downside-risk, aversion parameter goes to infinity end up with 11% in stocks for long horizons. This conclusion is based on various assumptions. In order to get insight in the robustness of our results we carried out an extensive sensitivity analysis with respect to the inputs. Only in the situation where we assume a form of investor behavior that is beyond the one that is