Estimating a Spatial Autoregressive Model using fixed effects panel regression

[Petre Car]

Hello,

I estimated first using xsmle a SAR model for panel data.
y = lam * W* y + beta * x+ e

Then I multiplied W*y to create a spatial lag (of course, I normalized W, and kept zero entries on main diagonal). I then estimated the model:

y = lam * Wy + beta*x + e

using a fixed effects panel regression.

I get a different value for lam while beta is comparable (though different) across the two estimations. Am I doing something wrong?

[Sebastian Kripfganz]

Wy is an endogenous variable. The ML estimator computed by xsmle accounts for the endogeneity, the fixed-effects estimator does not. The latter therefore yields inconsistent estimates.

[Petre Car]

Thanks! I was requested by a referee to do this. Any reference for your comments (just to explain the differences between the two estimates in a footnote, though the referee should know about them)?

[Sebastian Kripfganz]

This should be covered by any textbook/handbook on the subject, e.g.

• Anselin L, Le Gallo J, Jayet H (2006) Spatial panel econometrics. In Matyas L, Sevestre P. (eds) The econometrics of panel data, fundamentals and recent developments in theory and practice (3rd edition). Kluwer, Dordrecht, pp.901-969

[Enver Sait Kurtaran]

Wy is an endogenous variable. The ML estimator computed by xsmle accounts for the endogeneity, the fixed-effects estimator does not. The latter therefore yields inconsistent estimates.

Is there a stata package for the spatial stochastic frontier analysis? I have done some research but couldn't find any info about that

[Bindya Sara Babu]

Hi
I am working on the spatial impact of climate variables on the agricultural variables of three districts of my state in India. I have my panel data available. I cannot create a weight. I cannot create a matrix. I need help. Please help