Mean-Variance Distance Based Stock Market Networks in Portfolio Optimization

Market networks have been used to describe relationships among market returns of assets. Although the economic relevance of market networks has been showed in many studies (e.g. Mantegna, 1999; Onnela et al., 2004, Lyocsa et al., 2012) applications had been rare. One of few exceptions is the study of Onnela et al. (2003) who showed that the set of equities selected by the Markowitz's mean-variance portfolio optimization technique can be related to the vertices on stock market networks, i.e. distant vertices (path length) are selected in the Markowitz's mean-variance portfolio as well. We propose a Black and Litterman (1992) type portfolio optimization model, where return estimates are derived from information given by the stock market network. Instead of usual correlations, the networks are constructed using extreme coexceedance and mean-variance distances between assets. The performance of our model is compared with standard Markowitz's mean-variance portfolio. This study has the potential to contribute to the existing studies on stock market networks, as it directly links stock market networks to portfolio optimization problems.