Comparing the Performance of GARCH Family Models in Capturing Stock Market Volatility in India

Abstract

In recent times, the prediction of stock market volatility has emerged as a central focus in the domain of financial econometrics. This paper presents an empirical analysis aimed at modelling the volatility of the Indian stock market, particularly focusing on the NSE NIFTY 50, by utilizing various GARCH models. The investigation explores the volatility of stock returns, considering the daily closing prices, and examines the influence of two external factors: Crude oil prices and the INR/USD exchange rate. The inquiry employs data encompassing the period from January 1, 2012, to December 31, 2022, for all three variables. The manuscript delves into an array of univariate GARCH models, encompassing both symmetric and asymmetric models, and assesses their performance by utilizing metrics such as the Akaike Information Criterion, Schwartz Bayesian Information Criterion, and Log Likelihood. To assess the predictive accuracy of these models, statistical error measures such as Mean Squared Error, Root Mean Squared Error, and Mean Absolute Error are employed. The findings strongly suggest that the EGARCH model is the most effective in predicting the variations of the NIFTY index. Furthermore, the research highlights the significant impact of exchange rates and crude oil prices in relation to the volatility of the stock market in India.