We study the impact of market contagion on portfolio management. To model possible recurrence in the arrival of extreme events, we equip classic Poisson jumps with long memory via past-weighted randomization of the likelihood of their occurrences (Hawkes processes). Within this framework, we tackle the problem of optimal portfolio selection in terms of Expected Shortfall (ES). We use the generalized method of moments to estimate the model on three US stock indexes, representing three major sectors of the US economy. The moment conditions of the model are computed efficiently in closed form applying a novel technique. Given parameter estimates we maximize (at a monthly frequency in the period 2001–2016) the expected return subject to a constraint on the ES of a portfolio consisting of the three US sector indexes. We find that the weights of the optimal portfolio are significantly adjusted when the level of contagion is high. Finally, we perform an extensive out-of-sample back-test of the model’s ability to measure ES and find that the Hawkes jump-diffusion model outperforms two traditional models that are commonly implemented.

We consider a network of equity mutual funds characterized by different levels of compliance with Environmental, Social, and Governance (ESG) aspects. We measure the impact of portfolio liquidation in a stress scenario on funds with different ESG ratings. Fire-sales spillover from portfolio liquidation propagates from one fund to another through indirect contagion mediated by common asset holdings. The analysis is conducted quarterly from March 2016 through June 2018 using daily data from different sources at the fund and firm levels. Our estimation strategy relies on a network analysis where funds are not taken as stand-alone entities but are interconnected components of a unified system. We find evidence that the relative market value loss of the High ESG ranked funds is lower than the loss experienced by the Low ESG ranked counterparts in the time span with lower volatility. In the higher-volatility period there is not always a clear dominance of one class over another. Results are robust when controlling for size and for feedback effects, and for different model specifications. Our analysis offers new insights to both asset managers and policymakers to exploit the aggregate effect of portfolio diversification related to the system as a whole.