ASYMMETRIC DISTRIBUTION IN MOD-VOLATILITY MODEL FOR RISK PREDICTION

• Main Author: WIDI LESTARI (NIM: 10114072), EVI
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/27023
• Institution: Institut Teknologi Bandung

Risk prediction is an important role in diverse fields especially in Finance and Insurance. Theoretically, assume a distribution give significant effect for risk prediction. Asymmetric distribution as the model conditional distribution used to capture potential skewness and heavy tails of the loss return series. Aside the distribution assumption, volatility is the main component in risk prediction. Good volatility model must be able to capture the empirical facts of loss return and volatility such as kurtosis and autocorrelation function, volatility clustering, volatility persistence, also asymmetric of volatility. GARCH model and SV model is developed to accommodate this characteristic. In this Final Project, the model is illustrated by applying it to loss return of NASDAQ and NYSE stock market indices for generating one step-ahead forecasts of Value-at-Risk and Expected Shortfall and backtesting method are applied to accuracy these forecasts. Results show that Mod-GARCH model and Mod-SV model are good enough to accommodate the empirical properties of loss return and volatility based on analytic approach. From the numerical simulation can be concluded NS-GJR-GARCH(1,1), NS-T-GARCH(1,1), NS-ST-GARCH(1,1), S-GARCH(1,1), and SV-AR(1) model with asymmetric distribution form as the conditional distribution well accommodate the empirical properties of loss return and volatility also gives more accurate risk prediction than NS-GARCH(1,1) model. It indicates that asymmetric distribution, threshold, transition function, and stochastic component in model can improve the accuracy of the risk prediction result.