IMPLICATION OF HEAVY-TAILED DISTRIBUTION IN STOCHASTIC VOLATILITY (SV) MODEL

• Main Author: Puji Tristanti, Dwi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39707
• Institution: Institut Teknologi Bandung

Stochastic Volatility (SV) is one of the volatility models used to predict volatility of stock returns. In the SV model, volatility is assumed to follow Autoregressive (AR) stochastic process. Most of SV model is generated based on the normality assumption of their stock returns data. This assumption is empirically inappropriate since stock returns have higher kurtosis than normal distribution. In other words, stock returns data follow heavy-tailed distribution. Therefore, in this project the heavy-tailed distribution is also used as an assumption on the SV model. The type of heavy tailed distribution are Fr´echet distribution and Gumbel Distribution. Furthermore, the three distribution assumptions on the SV model are compared to determine the most appropriate distribution for SV model in stock returns. It is determined by the result of parameter estimation using Quasi Maximum Likelihood (QML) method and the result of volatility prediction. The result based on real data, reveal that the assumption of heavy-tailed distribution on SV model gives better result than normal distribution. While the most appropriate SV model for modeling volatility of stock returns is the SV model with Frechet distribution.