EViews.com

good day,
i want to know if it is necessary to test for multicollinearity for an ardl model? or am i just asking for trouble? i have done other diagnostic tests; cusum, heteroscedasticity,jacque bera, autocorrelation and they are as supposed to be. my model has one independent variable and 7 independent variables. the short run error term is negative and significant. of the long run coefficients 6 are significant and one is insignificant with a wrong sign. The problem is the simple correlations test for multicollinearity shows high levels of multicollinearity. i have also tried the coefficient variance decomposition test which i am afraid i may not be interpreting correctly; i have also attached the results output.

You are good to go. Ignore the multicollinearity, it is not a problem per se.

thank you so much Trubador

my supervisor does not agree with me that multicollinearity is not a problem. any advise how to fix for multicollinearity or convince him otherwise? According to him multicollinearity is not a problem to get around and should be fixed. i really need to be done with this before end of the week. i have used forum search.

Tough situation. I would not bother to convince him/her otherwise. At least read the discussion here for the sake of your understanding of the concept.
You should expect significant amount of correlation among variables if you are dealing with non-stationary series. However, ARDL is a robust and a dynamic method designed for that purpose and can therefore manage those cases successfully.