Interconnectedness Risk and Active Portfolio Management

Abstract

Interconnectedness is an alternative risk concept that so far has earned little attention in the asset management academia and industry. In this paper, we show that this neglect is not justified, as interconnectedness risk (i) has only moderate or no connection to conventional portfolio optimization inputs and (ii) active investment strategies based on interconnectedness information outperform their conventional peers. Utilizing a multi asset dataset, we measure interconnectedness risk by the embeddedness intensity, i.e. centrality, of assets in a correlation network, a concept from graph theory. Using the most common centrality measures, we first conduct empirical similarity studies analyzing how different centrality scores relate to each other and to conventional portfolio optimization inputs. Next, we outline how centrality can be incorporated in a risk-based as well as in a risk-return-based framework. Out-of-sample performance studies of centrality-optimized portfolios prove their competitiveness.