search

xuyiqing / gsynth

Generalized Synthetic Control Method
Other
132 stars 40 forks source link

Issue with control varaibles #39

Open WilliamCMU opened 4 years ago

WilliamCMU commented 4 years ago

Hi, I'm having issues with including control variables in gsynth. These controls are industry fixed effects. I am receiving the following error message:

Error in gsynth.default(formula = NULL, data = data, Y = Yname, D = Dname, : Variable "total.SICCD.f1381" is unit-invariant. Try to remove it.

Does this error imply that this method can not include fixed effects beyond unit and time specifications? The variable it is flagging is a dummy variable reading 1 for firms in a certain industry and 0 otherwise. There are around 40 of these dummies. If I remove this variable it sends the same message for the next variable of this kind.

Thanks, WC

xuyiqing commented 4 years ago

Dear William,

In theory, it is allowed. I think the error occurs because of perfect collinearity of the some of the dummy variables you added. Note that adding many more fixed effects this way will render the algorithm extremely slow.

Best, Yiqing

On Tue, Mar 17, 2020 at 11:04 AM WilliamCMU notifications@github.com wrote:

Hi, I'm having issues with including control variables in gsynth. These controls are industry fixed effects. I am receiving the following error message:

Error in gsynth.default(formula = NULL, data = data, Y = Yname, D = Dname, : Variable "total.SICCD.f1381" is unit-invariant. Try to remove it.

Does this error imply that this method can not include fixed effects beyond unit and time specifications? The variable it is flagging is a dummy variable reading 1 for firms in a certain industry and 0 otherwise. There are around 40 of these dummies. If I remove this variable it sends the same message for the next variable of this kind.

Thanks, WC

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/xuyiqing/gsynth/issues/39, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB2PKGFTNGW5DZ3AVDJVS3TRH63SJANCNFSM4LNWG43Q .

-- Yiqing Xu

Assistant Professor Department of Political Science Stanford University http://yiqingxu.org/

jefeerzhang commented 4 years ago

You should not put id variable and dummy industry variables in your model at once, that will cause Multicollinearity problem. In my opinion, some papers said they control industry fixed effect, that means they don't control individual fixed effect。