This paper compares four standard long-only risk-based portfolio selection models, namely minimum variance, minimum semivariance, minimum CVaR, and minimum maximal loss, within a common out-of-sample framework based on rolling estimation windows. The empirical analysis draws on three equity datasets, namely the FTSE 100, NASDAQ Composite, and S&P 500, covering the period 2015–2024. Portfolio performance is assessed in terms of return, volatility, drawdown-based measures, tail risk indicators, diversification, and turnover. The results reveal a relatively clear pattern across the three datasets. The minimum maximal loss portfolio tends to achieve the highest mean return, Sharpe Ratio, and Sortino Ratio, and often also exhibits favorable diversification properties. By contrast, the minimum variance portfolio emerges as the most stable strategy, systematically attaining lower volatility, lower realized CVaR, and lower turnover. Overall, the findings show that the choice of risk measure has a substantial impact on out-of-sample portfolio behaviour, highlighting important trade off among return, downside protection, diversification, and implementation costs.
Carleo, A., Congedo, M.A., Gheno, A., Mottura, C.D., Ricci, J.M. (2026). FROM VARIANCE TO WORST-CASE DOWNSIDE RISK IN AN OUT-OF-SAMPLE PORTFOLIO SELECTION. ANNALI DEL DIPARTIMENTO DI METODI E MODELLI PER L'ECONOMIA, IL TERRITORIO E LA FINANZA .... [10.13133/2611-6634/2053].
FROM VARIANCE TO WORST-CASE DOWNSIDE RISK IN AN OUT-OF-SAMPLE PORTFOLIO SELECTION
Alessandra CarleoSupervision
;Maria Alessandra CongedoVisualization
;Andrea GhenoConceptualization
;Carlo Domenico MotturaValidation
;Jacopo Maria Ricci
Writing – Review & Editing
2026-01-01
Abstract
This paper compares four standard long-only risk-based portfolio selection models, namely minimum variance, minimum semivariance, minimum CVaR, and minimum maximal loss, within a common out-of-sample framework based on rolling estimation windows. The empirical analysis draws on three equity datasets, namely the FTSE 100, NASDAQ Composite, and S&P 500, covering the period 2015–2024. Portfolio performance is assessed in terms of return, volatility, drawdown-based measures, tail risk indicators, diversification, and turnover. The results reveal a relatively clear pattern across the three datasets. The minimum maximal loss portfolio tends to achieve the highest mean return, Sharpe Ratio, and Sortino Ratio, and often also exhibits favorable diversification properties. By contrast, the minimum variance portfolio emerges as the most stable strategy, systematically attaining lower volatility, lower realized CVaR, and lower turnover. Overall, the findings show that the choice of risk measure has a substantial impact on out-of-sample portfolio behaviour, highlighting important trade off among return, downside protection, diversification, and implementation costs.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


