Interpretation of the variable coefficient postestimation

[Afonso Rodrigues]

Dear all

I have run a ARDL estimation to see the impacts of the renewable and non-renewable energy sources on the economic growth. However, before estimating my model, I faced problems with multicollinearity since my VIF test presented high values for the CO2 emissions and the Non-renewable energy sources (NRES) variables.
Since my variables were already per capita, and since I didn't want to loose any of my variables, I have opted for dividing my CO2 variable for my NRES variable, generating, thus, a new CO2 variable.

So, my question is. Now that I have estimated my equation, should I interpret my new CO2 variable coefficient the same way I would if I didn't made any change? Or, in other hand, how should I interpret the coefficient since I have done that previous change?

I appreciate all the help in advance.

Best Regards
Afonso Rodrigues

[Phil Bromiley]

Dividing one colinear variable by another colinear variable seems like an odd solution to colinearity. I don't know ARDL, but generally models with lagged dv's suffer from potential endogeneity (especially if there is serial correlation in the errors).

Of course, since you have changed the variable, you need to interpret it differently. The parameter is the effect of a change in the ratio on the dv. However, in your case, you're holding a bunch of lags of the dv constant which can complicate the interpretation.