Systemic risk from overlapping portfolios: A multi-objective optimization framework

We present a multi-objective portfolio optimization framework that accounts for both systemic risk arising from overlapping portfolios and individual risk. To address non-convexity in the objective function, we introduce an Evolutionary Search algorithm that enables efficient exploration of the solution space. Applying our framework to EBA data on sovereign exposures, we find that minimizing systemic risk results in highly concentrated and diverse portfolios, adding empirical evidence to a growing literature on the ambiguous effects of diversification on systemic risk. In contrast, individual risk optimal allocations exhibit high portfolio diversification and homogeneity. By characterizing a set of Pareto frontiers, we identify a trade-off between the two risk components. Even a small preference for minimizing systemic risk leads to optimal portfolios on the frontier that differ significantly from the observed ones, suggesting potential inefficiencies in actual portfolio structures. • We propose a portfolio optimization framework balancing systemic and individual risk. • We develop a heuristic algorithm to address non-convexities in the problem. • We show that optimizing systemic risk generates concentrated portfolios. • Pareto frontiers reveal a trade-off between individual and systemic risk. • Our empirical application suggests inefficiencies in EU banks’ sovereign portfolios.