xsmle: Spatial Durbin Model with correlated random effects

[Lom Newton]

Dear Statalisters,


I want to implement a Spatial Durbin Model with correlated random effects. Is it even possible? If it is, how do I specify it? Is this correct:

yit = α + ρ1Wyit + ρ2Wybari + β1Xit + β2Xbari + θ1WXit + θ2WXbari + θ3Ci + λt  + μi + Ɛit

I know for the "ordinary" SDM with random effects, this would be:

xsmle y X, wmat(W) model(sdm) re vce(cluster id) nolog

So if it is possible to implement the SDM with correlated random effects, would it be proper to run the following 'xsmle' command?

xsmle y X, ybar Xbar wmat(W) dmatrix(W) model(sdm) durbin(ybar X Xbar ) effects re vce(cluster id) nolog

Can someone please help me out on this?

Thank you

[Lom Newton]

Hello Statalist users,

I really need help on estimating a Spatial Durbin Model with correlated random effects specification. The type of model I am looking has been done in this paper https://www.recercat.cat/bitstream/h...pdf?sequence=1.

Does anyone know of a Stata routine that can implement it? By including the means of time varying covariates, would xsmle with option re or spxtregress with option re sarpanel be appropriate for such a model?

I need help please.

[Sebastian Kripfganz]

ybar should not be part of the equation. Eventually, a correlated random effects model is really just adding time averages of the regressors to the model. I do not see a reason why this should not work with an SDM.

[Lom Newton]

Thank you Sebastian Kripfganz for letting me know ybar should not be part of the equation. My main issue is that, as done in the paper I referenced, I want to also examine if the individual effects of a unit also have spillover effects. I will give it a try and see.