SOS!! Stationarity in VAR and in its variables

[Voyant]

Hello my friends!

I have a VAR of 4 variables which is stationary (no roots outside the unit circle), however some of its variables are not stationary individually.

Is this a problem as far as the whole VAR is stationary?

I do not prefer differencing in this case because as some researchers mention I will lose valuable information of the data.

Please, give me a reliable answer.

Thanks in advance!

[trubador]

Please search the forum before posting your questions: http://forums.eviews.com/viewtopic.php?f=18&t=11275

[Voyant]

Please search the forum before posting your questions: http://forums.eviews.com/viewtopic.php?f=18&t=11275

Thank you very much!

I have already read this but I do want to have another opinion perhaps by the most recent users.

By the way, the link to the piece of work of Prof. Lutkepoh does not function.
What is it exactly?

In general I would prefer a further source such as a paper or a book that I can use as reference to my master thesis as it concerns this topic.

Please do provide me one if you know it!

[Voyant]

For example this paper

http://ashleymac.econ.vt.edu/working_papers/varsim.pdf

I think that this opposes the views and the specific references provided at the link you mentioned before.

Please, I need some reliable opinions!

[trubador]

I have already cited the book of Prof. Walter Enders, who also cites Sims, Stock and Watson "concerning this very topic". This is a very reliable source of information and these authors are among the most respected researchers that I know of.
And I would not recommend the "working paper" you've cited, since the approach of Toda and Yamamoto has been heavily criticized due to some methodological flaws.

[Voyant]

I have already cited the book of Prof. Walter Enders, who also cites Sims, Stock and Watson "concerning this very topic". This is a very reliable source of information and these authors are among the most respected researchers that I know of.
And I would not recommend the "working paper" you've cited, since the approach of Toda and Yamamoto has been heavily criticized due to some methodological flaws.

It seems that your response is valid and trustworthy.

Sincerely, thank you very much sir.

[Voyant]

Continuing from my last remark:

I would like to ask if using log in each of the 4 variables is recommended or not in order to make each individual variable and the whole VAR system more stationary?

I did use it and my results seem more correct empirically.

Therefore is this ok by theory?

Thanks in advance for any answer!

[trubador]

Taking the logarithm of variables may help ease the heteroscedasticity problem, but it does not ensure the stationarity of variance. If you know what you are doing, taking logs will not do any harm.