Chapter 12 - Forecasting volatility in the stock market data using GARCH, EGARCH, and GJR models

Abstract

Most of the stock market data are high-frequency data showing nonlinearity, asymmetry, and chaotic behavior due to sharp variations in prices over time. Volatility explains these sudden variations in prices over time and is related to conditional variances that reveal some important facts about stock market returns. So, volatility modeling is required to investigate these significant facts of stock market returns. Autoregressive conditional heteroskedastic (ARCH) models are powerful tools for modeling and estimating conditional variances and volatility in stock market prices. The present study deals with volatility prediction of stock market returns using different variants of Generalized Autoregressive Conditional Heteroskedastic (GARCH) models such as Exponential Autoregressive conditional heteroskedastic (EGARCH) and Glosten-Jagannathan-Runkle (GJR) models with Gaussian distribution and Student’s t-distribution. The input data for the study consist of stock market data of the daily and weekly returns of the BSE 100 S&P stock index series from 2009 to 2019. After modeling the daily and weekly returns with GARCH, EGARCH, and GJR models, the volatility of price returns is forecasted for the out-of-sample period. The performance of the proposed models has been evaluated by error statistics that compare the values of original volatility with the predicted values. The results reveal that the out-of-sample volatility forecast with the EGARCH model tends to generate more accurate results with Student’s t-distribution when compared to GARCH and GJR models.