Portfolio stability assessment through multilayer networks

This paper presents a complex network model for assessing the stability of a generic portfolio composed of N stocks, which subsequently leads to the stability of optimal portfolios. We assume that each variable involved in the optimization problem is associated with one graph, having the considered stocks as the nodes. We also assume that nodes of different graphs may present links – hence, forming a multilayer network structure. The topological structure of the network/port- folio is assumed to depend on a clustering procedure applied over the nodes in the individual layers. Furthermore, in line with previous studies, we also assume that a modification of the communities in the layers may be responsible of the instability of the considered portfolio. On this, we introduce different devices for measuring the portfolios’ stability level. Empirical experiments based on the Dow Jones Industrial Average index stocks and the S&P500 index illustrate the applicability of the methodological proposal.