In this paper, we study the effects caused by the presence of outliers on the identification and estimation of GARCH models. First, we derive the asymptotic biases of the sample autocorrelations of squared observations and their effects on some popular homoscedasticity tests when uncorrelated...