Dear All,

I estimate a model like the following spatial model (SARAR):

y=rho*Wy+beta*X+u
u=psi*u+eta

where W is a spatial distance matrix, X is a set of covariates and u is the error term.

The parameter rho is positive and significant, while the parameter psi is negative and significant. I know that the interpretation of coefficients needs caution (I use estat impact for that). As far as the spatial coefficients are concerned, the positive rho parameter in my view is an evidence of a positive spillover from an observation to the other, but I have not a clear idea about the interpretation of the psi parameter. Does it mean that observations with an high y are close to observations with a low y? I am not entirely convinced about this interpretation. Any suggestion?

Thanks in advance.

Dario