Estimating Nonlinearities in Spatial Auregressive Models

Mardi | 2009-12-01

Nicolas DEBARSY – Vincenzo VERARDI – Amine LAHIANI

In spatial autoregressive models, the functional form of autocorrelationis assumed to be linear. In this paper, we propose a simple semiparametricprocedure, based on Yatchew’s (1997) partial linear least squares, that doesnot impose this restriction. Simple simulations show that this model outper-forms traditional SAR estimation when nonlinearities are present. We thenapply the methodology on real data to test for the spatial pattern of vot-ing for independent candidates in US presidential elections. We …nd thatin some counties, votes for  » third candidates » are non-linearly related tovotes for  » third candidates » in neighboring counties, which pleads in favorof strategic behavior.