The continuous evolution of financial markets highlights how quantitative financial risk management has become a key tool in investment decisions, capital allocation, and regulation. Although several methods have been proposed to estimate the risk of an investment in capital markets, Value-at-Risk (VaR) and Expected Shortfall (ES) can be considered the standard measures of market risk, as they are used both for internal control of financial institutions and for regulatory purposes. In this direction, the choices of modelling and estimation methods for VaR and ES play a critical role. Nowadays, a variety of possibilities is available. For instance, there are models belonging to the class of parametric, semi-parametric, and non-parametric methods. Moreover, among the class of parametric models, there are several error distributions that could be considered. Also, some models allow for the use of variables mixed at different frequencies. To mitigate the impact of these sources of uncertainty, we propose a forecast combination strategy by adaptively weighting the pool of most accurate predictors based on the Model Confidence Set (MCS) results. The empirical analysis suggests that combinations of VaR and ES forecasts lead to higher predictive accuracy over a wide range of competitors.
Adaptive combinations of tail-risk forecasts
Alessandra Amendola;Vincenzo Candila
;Antonio Naimoli;Giuseppe Storti
2023
Abstract
The continuous evolution of financial markets highlights how quantitative financial risk management has become a key tool in investment decisions, capital allocation, and regulation. Although several methods have been proposed to estimate the risk of an investment in capital markets, Value-at-Risk (VaR) and Expected Shortfall (ES) can be considered the standard measures of market risk, as they are used both for internal control of financial institutions and for regulatory purposes. In this direction, the choices of modelling and estimation methods for VaR and ES play a critical role. Nowadays, a variety of possibilities is available. For instance, there are models belonging to the class of parametric, semi-parametric, and non-parametric methods. Moreover, among the class of parametric models, there are several error distributions that could be considered. Also, some models allow for the use of variables mixed at different frequencies. To mitigate the impact of these sources of uncertainty, we propose a forecast combination strategy by adaptively weighting the pool of most accurate predictors based on the Model Confidence Set (MCS) results. The empirical analysis suggests that combinations of VaR and ES forecasts lead to higher predictive accuracy over a wide range of competitors.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.