Robust Replicating Portfolio via Risk-Averse Stochastic Optimization

Robust Replicating Portfolio via Risk-Averse Stochastic Optimization

• Autor
• Fernanda Maria Müller
• Co-autores
• Marcelo Brutti Righi , João Ramos Jungblut
• Resumo

• We propose a conditional framework for futures-based index replication that links a risk-averse stochastic program (RASP) with linear regression models. The approach nests a rich family of risk–score pairs—Expected Loss (EL), Mean plus Semi-Deviation (MSD), Expected Shortfall (ES), Expectile, and Maximum Loss (ML)—with Squared Error (SE), Absolute Error (AE), Quantile, Expectile, and Linear-Exponential Loss (LINEX). Using S&P 100 constituents to replicate S&P 500 futures over 2010–2022, with rolling windows and semiannual rebalancing, we compare 25 specifications under Monte Carlo resampling. Parsimonious, symmetric combinations (EL/MSD with AE/SE) yield the lowest absolute tracking error and risk. In contrast, aggressive RASP (ML with LINEX/Quantile) results in higher returns but higher absolute tracking error and risk. Results are robust across market regimes.

• Palavras-chave
• Robust measures; Replicating portfolio; Tracking error; Risk-averse.
• Modalidade
• Comunicação oral
• Área Temática
• Engenharia Financeira

Voltar Download