some help would be much appreciated. I have basic knowledge of econometrics (did a 2nd year introductory unit at uni). My question involves panel data and is:
An error term consists of the effects of time-invariant unobservable effects (e.g. managerial skill, cultural influences), individual-invariant unobservable effects, and unobservable effects that are both time-variant and individual-variant (e.g. hetereogenous effects of technological advancement amongst different countries).
If I first difference my model I remove the time-invariant unobserved effects. Keeping the time-variant and individual-variant component of the error term is ok.
How do I remove the time-variant effects? Econometric books say to use the difference from the mean method, but I don't think that is appropriate? Or do you just not remove it, and incorporate a time trend variable into your econometric model? (e.g. eviews "@trend" command).
Hopefully you guys can help!