This article attempts to identify, from among the family of Generalised Autoregressive Conditional Heteroscedasticity (GARCH) models, the model that best describes the Indian stock market volatility by (a) building volatility models using the traditional GARCH models that accounts for asymmetry; (b) selecting a suitable model by nesting through Box-Cox transformation, a family of GARCH models. Our results confirm the stylised fact of the presence of leverage effects in the stock market. But contrary to popular belief, our results show that (a) it is the smaller shocks that affect the returns in the Indian stock market and dominate the news impact curve than the larger shocks; (b) nesting exercise has narrowed down to two entirely different sets of models that could describe equally well the returns data of the Indian stock market; but overall results indicate that a non-linear model that uses the conditional standard deviation with an exponent that accounts for smaller shocks may be preferable to explaining time-varying volatility in Indian stock returns data. (c) Another feature typical of the Indian stock market but unexplored so far, the non-trading days, has been found to be accounting for a sizable portion of the return variance, contributing almost one-fourth as much to volatility as any trading day.
