Dear all,
I am trying to use a time-invariant IV (distance) for a time-variant variable (trade openess) in a panel setting (30 countries, 15 years). I am aware that including fixed effects is not an option as they are collinear with the IV. Including them would be possible by interacting the time-invariant IV with time. The first-stage results could then be interpreted as the effect's change over time. My first question is: Does this interpretation also hold for the second-stage results (in case I used this time-variant IV)? If not, how should one interpret the second-stage results when using an interaction variable as IV (distance x time)?
My second question is related to a random effects approach: Although not favourable due to strict assumptions on exogeneity, it should be possible to use a time-invariant IV for a time-variant variable in such a setting. However, I have read a paper in which the author includes fixed effects in a random effects approach for a very similar setting. (Lin; F. (2017): "Trade openess and air pollution: City-level empirical evidence from China", China Economic Review, Vol. 45, pp. 78-88, doi: https://doi.org/10.1016/j.chieco.2017.07.001 )
So far I was not able fo find adequate literature on this methodology (including FE in random effects model). Do you have literature recommendations on this?
I am looking forward to your answers. I hope that my questions are not too trivial as I know that some of the best econometricians are actively posting here.
Best,
Marius
I am trying to use a time-invariant IV (distance) for a time-variant variable (trade openess) in a panel setting (30 countries, 15 years). I am aware that including fixed effects is not an option as they are collinear with the IV. Including them would be possible by interacting the time-invariant IV with time. The first-stage results could then be interpreted as the effect's change over time. My first question is: Does this interpretation also hold for the second-stage results (in case I used this time-variant IV)? If not, how should one interpret the second-stage results when using an interaction variable as IV (distance x time)?
My second question is related to a random effects approach: Although not favourable due to strict assumptions on exogeneity, it should be possible to use a time-invariant IV for a time-variant variable in such a setting. However, I have read a paper in which the author includes fixed effects in a random effects approach for a very similar setting. (Lin; F. (2017): "Trade openess and air pollution: City-level empirical evidence from China", China Economic Review, Vol. 45, pp. 78-88, doi: https://doi.org/10.1016/j.chieco.2017.07.001 )
So far I was not able fo find adequate literature on this methodology (including FE in random effects model). Do you have literature recommendations on this?
I am looking forward to your answers. I hope that my questions are not too trivial as I know that some of the best econometricians are actively posting here.
Best,
Marius
