engleski

Parametric Forecasting of Value at Risk Using Heavy Tailed Distribution

This paper deals with modeling volatility of returns of Pliva stocks on Zagreb Stock Exchange for Value at Risk forecasting. Volatility reaction and volatility persistence are measured using asymmetric GARCH process. Croatian capital market characteristic is absence of intensive reaction on "good" information. But it is confirmed that Pliva stocks volatility on Croatian capital market are under dominant influence of "bad" information. If the data are heavy tailed, the VaR calculated using Normal assumption differs significantly from Student's t-distribution. The fact that kurtosis and degrees of freedom from Student's distribution are closely related is used in estimation procedure of GARCH model. The complete procedure of Value at Risk forecasting for Croatia is established with assumption that returns follows extreme value distribution, precisely Student's t-distribution with non-integer degrees of freedom. The optimization problem is solved by FinMetrics module of S-Plus package.

Value at Risk; asymmetric GARCH process; leverage effect; volatility persistence; heavy tailed distribution; non-integer degrees of freedom

nije evidentirano