If an H.VAR equation is used to model the variance, and a zero-mean model is fit to the dependent variable, then PROC MODEL incorrectly returns the error:

ERROR: No parameters to estimate

and only the Model Summary table is produced by the procedure.

In other words, you cannot specify:

y = 0;
h.y = &ltexpression for variance model>

There is no circumvention for this problem. You must include a parameter for the intercept term in the model for the dependent variable, such as:

y = mu;
h.y = &ltexpression for variance model>