Hello everyone,
As a newcomer to this forum, I hope you’ll excuse me if my question has been addressed before. I'm currently engaged in estimating a spatial panel fixed effects (SDM) model using the xsmle command. The command I'm employing is as follows:
xsmle lny lnx1 lnx2 lnx3, wmat(Wn5) model(sdm) fe effects type(ind) vceeffects(sim, nsim(5000))
My concern is with the potential endogeneity of the variable x3, which represents regional GDP. In a non-spatial model utilizing xtivreg2, I've employed 'ln population' (lnz1) as an instrumental variable. The instrument appears valid based on various tests. However, I'm struggling to find a method for incorporating instrumental variables into a panel SDM estimation. Consequently, I resorted to a manual estimation approach as follows:
Step 1: xtreg lnx3 lnz1 lnx1 lnx2 i.yr, fe ro
Step 2:predict iv_lnx3, xb
Step 3:xsmle lny lnx1 lnx2 iv_lnx3, wmat(Wn5) model(sdm) fe effects type(ind) vceeffects(sim, nsim(5000))
I would greatly appreciate your insights on whether this approach is appropriate, or if there exists a more refined method to handle endogenous variables in panel SDM estimations.
Thank you very much in advance for your guidance and assistance.
As a newcomer to this forum, I hope you’ll excuse me if my question has been addressed before. I'm currently engaged in estimating a spatial panel fixed effects (SDM) model using the xsmle command. The command I'm employing is as follows:
xsmle lny lnx1 lnx2 lnx3, wmat(Wn5) model(sdm) fe effects type(ind) vceeffects(sim, nsim(5000))
My concern is with the potential endogeneity of the variable x3, which represents regional GDP. In a non-spatial model utilizing xtivreg2, I've employed 'ln population' (lnz1) as an instrumental variable. The instrument appears valid based on various tests. However, I'm struggling to find a method for incorporating instrumental variables into a panel SDM estimation. Consequently, I resorted to a manual estimation approach as follows:
Step 1: xtreg lnx3 lnz1 lnx1 lnx2 i.yr, fe ro
Step 2:predict iv_lnx3, xb
Step 3:xsmle lny lnx1 lnx2 iv_lnx3, wmat(Wn5) model(sdm) fe effects type(ind) vceeffects(sim, nsim(5000))
I would greatly appreciate your insights on whether this approach is appropriate, or if there exists a more refined method to handle endogenous variables in panel SDM estimations.
Thank you very much in advance for your guidance and assistance.
