VIF is designed for detection of deleterious collinearity - it is related to the R-squared of of a predictor variable regressed on all the others , and that repeated for each predictor.

My understanding of variance decomposition has nothing to do with collinearity.

Did you mean to ask about condition index?

Kelvyn Jones First of all thank you very much for valuable response.

In eviews when Variance Decomposition is employed, it offers certain results and one of them as you said is 'Condition Index'. Eviews suggests to run Variance Decomposition to check multi-collinearity but at the end the main focus should be given to 'Condition Index' (ref. http://www.eviews.com/help/helpintro.html#page/content%2Ftesting-Coefficient_Diagnostics.html%23). If 'Condition Index' is the output of performing Variance Decomposition and if 'Condition Index' helps to detect Multicollinearity given a choice between VIF and 'Condition Index' which one would you prefer?

Zericho Marak OK I now understand that is the way the software works. In all these cases I know about both methods give equivalent results as of course they are highly related. As neither has a definite cut off, I think the choice does not matter much. Although I remember reading somewhere that in cases of extreme collinearity that the calculation of the VIF itself become unreliable but the condition indices flag the problem. You may be interested in these useful postings

https://statisticalhorizons.com/multicollinearity

http://blog.minitab.com/blog/starting-out-with-statistical-software/what-in-the-world-is-a-vif

https://pdfs.semanticscholar.org/ed1f/4466a0982f3e8de202de01ecceb473d11893.pdf

Article Confounding and collinearity in regression analysis: a cauti...

VIF.

Thank you Sir for your valuable response Hamid Mohsin Jadah

I always use VIF.

Thank you Sir for your answer Ibrahim Nandom Yakubu