I have log transform my model already, choose per capita variables but multicollinearity is still a problem.
I plan to drop the variable which has the highest vif (over 200)... however, I read somewhere that I will be subjected to specification error/bias.
I need to address this one since...