I am trying to use VAR for an undergraduate class but keep having degrees-of-freedom issues. I am using quarterly data, have 3o years of data, and have 10 time series.

I want to look at the effects of monetary policy shocks. But we learned that this method produces a price puzzle and that one way to fix the price puzzle is to include commodities prices.

My problem is that this seems like it treats a symptom and not a problem. I thought it might be a better idea to include a broad set of variables instead of just that one. It is justifiable to include lots of information but not justifiable to cherry-pick information, I think.

Anyone know how to include more variables in that set?


1 Answer 1


My recommendation is that you read through this paper from Bernanke, Boivin, and Eliasz. This paper introduced something called factor-augmented VAR. The gist is that you can use principle component analysis to do exactly what you're after. This method allows you to include hundreds of time series in the conditioning information. And FYI, this does allow one to solve the price puzzle with solid economic rationale instead of the ad hoc inclusion of a single time series.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.