[Openspace] 2SLS

[Julia Koschinsky]

on behalf of Paulo Regis.
Julia

****

Date: Thu 29 Nov 14:09:04 CST 2007
From: "Paulo Regis" <pauloregis.ar at googlemail.com>

I would like to know if I am making a mistake when interpreting the endogeneity problem of the spatial lag model. Suppose you have 

Y=XB+r*WY+e

where WY is the spatial lag of the dependent variable and e is white noise. Obviously, there is a problem of endogeneity since WY introduces the error term of region j, with j=/=i. (ej) for the region i. My question is if the 2SLS is a good estimator for this equation. 

Operatively, I would think that further spatial lags of Y would be a good instrument (as well as spatial lags of X).


Cheers

Paulo