A1578
Title: Portfolio selection with complex network analysis
Authors: Roope Rihtamo - University of Turku (Finland) [presenting]
Joni Virta - University of Turku (Finland)
Harto Saarinen - University of Turku (Finland)
Abstract: Complex network analysis methods are used to derive optimal asset weights in a standard portfolio optimization problem. In the proposed empirical approach, different assets are represented by nodes that are linked together to form a network of assets. The links are based on similarity measures that are designed to capture the covariance structure of returns in the opportunity set. Different aspects of return similarity are addressed with measures motivated by financial theory as well as measures motivated by complex network analysis. New and enhanced measures of asset centrality are developed - a key aspect of portfolio diversification. The selected similarity measures and the selected investment opportunity set affect the structure of the formed network which, in turn, defines the optimal weights. Hence, the proposed methods offer novel extensions to the use of complex network analysis methods in financial economics. The proposed methods have straightforward interpretations rooted in theoretical asset pricing and can be easily implemented in various markets.
