I am trying to fit a Vector Autoregression model to forecast GDP growth Rate.

I have 2 series, monthly GDP growth rate and a monthly economic indicator. For the monthly GDP growth rate, the latest growth rate I can use is from the previous quarter. For the monthly economic indicator, I am able to get it near real time.

Meaning, to forecast 2016-Apr GDP growth, I can only use 2016-Mar or earlier GDP data and 2016-Apr or earlier economic indicator data. Likewise, to forecast 2016-May GDP growth, I can only use 2016-Mar or earlier GDP data and 2016-May or earlier economic indicator data.

As the latest economic indicator contains a lot of information on the GDP growth, I would like to include it in the VAR model.

My question is, is it okay to fit the VAR model in such a way where the 2 series came from different time interval (i.e. fit model with GDP data from t-3 to t, and economic indicator data from t-2 to t+1)?

And is it appropriate to forecast GDP by simply using 1-step ahead forecast?

Using the above model specification, I am able to get decent forecast performance. I am more concern if I am using VAR model inappropriately or violated anything statistically.

  • $\begingroup$ This is an interesting question, so I did some digging. Apparently the answer is not trivial. There is a nice literature review at the beginning of Research Working Paper 11 (2011) produced by the Kansas City Fed. $\endgroup$ – shadowtalker Jun 29 '16 at 13:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.