A New Asymmetric GARCH Model: Testing, Estimation and Application

Since the seminal work by Engle (1982), the autoregressive conditional heteroscedasticity (ARCH) model has been an important tool for estimating the time-varying volatility as a measure of risk. Numerous extensions of this model have been put forward in the literature. The current paper offers an alternative approach for dealing with asymmetry in the underlying volatility model. Unlike previous papers that have dealt with asymmetry, this paper suggests to explicitly separate the positive shocks from the negative ones in the ARCH modeling approach. A test statistic is suggested for testing the null hypothesis of no asymmetric ARCH effects. In case the null hypothesis is rejected, the model can be estimated by using the maximum likelihood method. The suggested asymmetric volatility approach is applied to modeling separately the potential time-varying volatility in markets that are rising or falling by using the changes in the world market stock price index.