Year of Graduation
Conditional Density Model for Forecasting of Assets’ Risk Measures
Statistical Modelling and Actuarial Science
Value-at-risk (VaR) – one of the most common risk metrics, the advantages of which are intuitive sense and ease of calculation. VaR represents the maximum amount of loss at a given probability and time horizon. The aim of the work is to analyze the quality of volatility and VaR forecasts of three models – GARCH, EGARCH, ARCD-GARCH-SGED. The model of conditional density with the addition of skewness and kurtosis showed the best results. The models were compared using various criteria: information criteria, the value of the likelihood function, the loss functions for volatility, the number of VaR violations, the Lopez loss function. In addition, Friedman test was provided for the significance of model differences.