Anyway, Jim makes two great points in the paper that are not widely appreciated by empirical macro people.
(1) OLS standard errors can be stupendously bad when the error terms are conditionally heteroskedastic, and White or Newey West standard errors are not always "good enough" either.
(2) Even if one only cares about inferences regarding the conditional mean, the efficiency gains from estimating GARCH effects via MLE vs. ignoring them via OLS can be extremely large.
Shout it from the rooftops!
4 comments:
Duh?
Huh?
Well, ideally you wouldn't want to fiddle with the distribution of the shocks (much of the model "features" come from there already) but rather get it somehow to be modeled. :-(
I think the result on the Taylor rule itself is worth mentioning.
A great article indeed and a very detailed, realistic and superb analysis of the current and past scenarios.
Post a Comment