Uncovering the Benefit of High-Frequency Data in Portfolio Allocation

Abstract

In previous studies, high-frequency data has been used to improve portfolio allocation by estimating the full realized covariance matrix. In this paper, we show that strategies using high-frequency data for measuring and forecasting univariate realized volatility alone can already generate statistically significant and economically tangible benefits compared to low-frequency strategies. Most importantly, however, high-frequency data also allow us to separate realized volatility into different components and construct realized higher moments. Strategies using upside and downside volatility components as well as realized skewness are shown to reveal additional information and deliver incremental economic benefits over strategies using total realized volatility alone.

Keywords: high-frequency data, downside risk, asset allocation, realized moments, volatility forecasting

JEL Classification: C3, C5, G1