Forecasting Volatility in Stock Market Using GARCH Models

Abstract

Forecasting volatility has held the attention of academics and practitioners all over the world. The objective for this master's thesis is to predict the volatility in stock market by using generalized autoregressive conditional heteroscedasticity(GARCH) methodology. A detailed explanation of GARCH models is presented and empirical results from Dow Jones Index are discussed. Different from other literatures in this field, this paper studies forecasting volatility from a new perspective by comparing GARCH(P,Q) model with GJR-GARCH(P,Q) model and EGARCH(P,Q) model. GJR-GARCH(P,Q) model turns out to be more powerful than GARCH(P,Q) model due to catching some leverage effects successfully. This makes our prediction more reliable and accurate. This paper also shows that both GARCH(P,Q) model and GJR-GARCH(P,Q) model are good choices for dealing with heteroscedastic time series.