0
$\begingroup$

I am modeling liquidity variable, real money growth with real asset price returns. The former is I(1) and the latter is I(0). The objective is to see the predictive power and forecast. However, can we apply VAR if both the variables have a different order of integration?

Will using I(0) against differenced I(1) series be a correct approach? especially keeping in mind I need to check forecast power through rolling window and then forecast further?

Improve this question
$\endgroup$
  • $\begingroup$ There have been a number of similar questions on Cross Validated in the past. I have answered a number of them myself. Have you tried searching for them? Try e.g. I(0) and I(1) as keywords, add VAR to narrow down. Something like this will hopefully help. $\endgroup$ – Richard Hardy Jun 11 '19 at 12:48
  • $\begingroup$ Yes, I have seen them they were talking about checking cointegration, etc but I have only two variables. Will using I(0) against differenced I(1) series be a correct approach? especially keeping in mind I need to check forecast power through rolling window and then forecast further? $\endgroup$ – DeeDee Jun 11 '19 at 12:51
  • 1
    $\begingroup$ I suppose you will be fine. If the series are truly I(1) and I(0), then there really isn't much else you can do. What would not be fine is regressing I(1) on I(0) or the other way around. $\endgroup$ – Richard Hardy Jun 11 '19 at 13:39