Diversification and Systemic Risk of Networks Holding Common Assets

This paper focuses on studying the influence of diversification on systemic risk for networks, which consist of a set of companies holding common assets. Diversification prevents initial company failures, but permits financial contagions due to companies’ overlapping asset portfolios. We comprehensively study diversification’s effect on systemic risk by proposing not only three different measurements of systemic risk, but also an algorithm to increase a Poisson-random network’s diversification in simulations. The numerical results show that whether risk is measured in terms of the number of failed companies or the total assets’ loss of value, diversification and systemic risk exhibit an inverted U-shaped relationship. The mechanism to generate such a relationship is further analyzed by combining the tradeoff effect of diversification on both the initial risk and risk contagion. Moreover, we find that the inverted U-shaped relationship between diversification and systemic risk is more significant when the assets correlate less. Finally, we propose new methods for measuring general networks’ level of diversification, for which average degree of company nodes divided by number of companies is affirmed to be more appropriate via simulations performed on different sized random networks and a further analysis of the 2007 U.S. commercial banks balance sheet data. The mechanism to generate the relationship between diversification and systemic risk—as well as their measurement methods—provide valuable references for networks’ risk control and structure optimization.