The Use of Correlation Networks in Parametric Portfolio Policies

Abstract

Correlation networks reveal a rich picture of market risk structure dynamics. A rather compact and well-organized sector correlation network is indicative of a healthy market, whereas a widely spread sector correlation network characterizes a more fragile market environment. Intuitively, some characteristics of the correlation network can serve as natural measures of systemic risk. Pursuing an equity market timing strategy we document the predictive content of these measures to translate into a meaningful portfolio utility. Moreover, this result continues to hold when controlling for common predictors of the equity risk premium. Not only can correlation networks be useful as an aggregate market timing signal but also in navigating the cross-section of equity sectors. We especially document a significant outperformance of peripheral versus central equity sectors that cannot be explained by momentum or low volatility effects. Finally, we implement a parametric portfolio policy that comprises the complete information content of the sector network topology conditional on a given level of risk aversion.

Keywords: Correlation Networks, Parametric Portfolio Policies, Market Timing, Sector Allocation

JEL Classification: G11, G12, G14