This document discusses estimating asset price volatility using generalized autoregressive conditional heteroskedastic (GARCH) models. It begins with an introduction to modeling stock return volatility and the assumptions of non-constant variance. It then presents the GARCH model for estimating variance in the univariate case. Next, it discusses estimating the GARCH model parameters using maximum likelihood estimation. Finally, it discusses extending the GARCH model to the multivariate case to simultaneously estimate the volatilities and correlations of a portfolio of stocks.