The RATS Software Forum

Is it acceptable to extract the residuals from a Vector Autoregression (VAR) model and use them as inputs for the estimation of the BEKK GARCH model?

People have been known to do that---a combined model has a large number of parameters and can be hard to estimate successfully. Ideally, you would want to do the combined model but it isn't always feasible. You would only do that if the GARCH part of the model were the main interest and not the mean-level dynamics.