1. Volatility clusters
2. Heavy tails
3. The GARCH model provides a simple parametric function that can be used to describe the volatility evolution.

Please refer to the chapter of "Conditional Heteroscedastic Models" in <<Analysis of Financial Time Series>>

另外,相对于ARCH而言,GARCH还有一个parameter parsimony的优点.比如,如果单
用ARCH模型,你可能要用ARCH(5)或者更高阶,但是用GARCH的话,GARCH(1,1)可能跟ARCH(5)的效果差不多,这样一来,你就可以少用3个参数,那么估计起来也更简单.因为GARCH模型实际上可以看成是ARCH无穷模型.短处是不能刻画非对称信息.