Choices under uncertainty and the investment horizon
研究发现，均值-方差投资规则在短期有效，但长期会错误建议增持债券；而均值-变异系数规则在长期更符合预期效用，中期则需用随机占优或直接最大化预期效用。
Abstract Mean–Variance (M–V) is the most popular investment rule employed by both practitioners and researches. For a short-planned investment horizon, this rule generally conforms with expected utility paradigm. However, for a relatively long horizon, generally more than one year, the distributions of returns become positively skewed, hence the M–V rule loses ground. As the horizon increases, by the M–V rule one needs (mistakenly) to increase the weight of bonds in the stock–bond portfolio, e.g., almost 100% in bonds for a 30-year horizon. Expected utility maximization recommending almost 100% in stocks for long horizons. This, gap is of crucial importance, because life expectancy is increasing, implying longer investment horizons. For long horizons the mean–coefficient-of-variance rule conforms with expected utility, as the distributions are close to log-normal. For intermediate horizons one should employ stochastic dominance or direct expected utility maximization.