Agree with colinwang, another thing you might consider is for variables that might be collinear is to try them both in different models (i.e. one model with one choice and then another model with the alternative collinear choice) if they are truly collinear you should get the same coefficients - if however they are not, then the coefficients will be different and you may find that having them both adds to the performance and economic intuition of the model