Information Theoretic Estimators of the First-Order Spatial Autoregressive Model

Information theoretic estimators for the first-order autoregressive model are considered. Extensive Monte Carlo experiments are used to compare finite sample performance of traditional and three information theoretic estimators including maximum empirical likelihood, maximum empirical exponential likelihood, and maximum log Euclidean likelihood. It is found that information theoretic estimators are robust to specification of spatial autocorrelation and dominate traditional estimators in finite samples. Finally, the proposed estimators are applied to an illustrative example of hedonic housing pricing.