Abstract This paper analyzes empirically the performance gains of using high frequency data in portfolio selection. Assuming Constant Relative Risk Aversion (CRRA) preferences, with different relative risk aversion levels, we compare low and high frequency portfolios within mean-variance, mean-variance-skewness and mean-variance-skewness-kurtosis frameworks. Using data on fourteen stocks of the Euronext Paris, from January 1999 to December 2005, we conclude that the high frequency portfolios outperform the low frequency portfolios for every out-of-sample measure, irrespectively to the relative risk aversion coefficient considered. The empirical results also suggest that for moderate relative risk aversion the best performance is always achieved through the jointly use of the realized variance, skewness and kurtosis. This claim is reinforced when trading costs are taken into account.

中文翻译：

---

在投资组合选择中使用高频数据的收益

摘要 本文实证分析了在投资组合选择中使用高频数据的性能收益。假设恒定相对风险规避 (CRRA) 偏好，具有不同的相对风险规避水平，我们比较均值方差、均值方差偏度和均值方差偏度峰度框架内的低频和高频投资组合。使用巴黎泛欧交易所 14 只股票的数据，从 1999 年 1 月到 2005 年 12 月，我们得出结论，无论考虑的相对风险规避系数如何，高频投资组合在每个样本外测量都优于低频投资组合。实证结果还表明，对于适度的相对风险厌恶，最好的表现总是通过联合使用已实现的方差、偏度和峰度来实现。