AI Summary & Key Takeaways
AI InsightThe results indicate that the asymmetric PARCH volatility specification performs the best in terms of both in-sample and out-of-sample forecasts.
📌 3 Key Takeaways
Abstract
This paper has attempted studying the twin issues of asymmetry/leverage effect and excess kurtosis prevalent in Indias stock returns under alternative volatility specifications as well as conditional distributional assumptions. This study has been carried out using daily-level data, based on Indias premier stock index, BSESENSEX, covering Indias post-liberalisation period from January 1996 to December 2010. Apart from lag returns, three other variables viz., call money rate, nominal exchange rate and daily dummies have been used as explanatory variables for specifying the conditional mean. Three alternative models of volatility representing the phenomenon of leverage effect in returns viz., EGARCH, TGARCH and asymmetric PARCH along with standard GARCH have been considered for this study. As regards the assumption on conditional distribution for the innovations, apart from the Gaussian distribution, two alternative conditional distributions viz., standardized Students distribution and standardized GED for capturing the leptokurtic property of the return distribution have been considered. Further, comparisons across these models have been done using forecast evaluation criteria suitable for both in-sample and out-of-sample forecasts. The results indicate that the asymmetric PARCH volatility specification performs the best in terms of both in-sample and out-of-sample forecasts. Also, the assumption of normality for the conditional distribution is not quite statistically tenable against the standardized GED and standardized Students distribution for all the volatility models considered.