Testing for a Common Factor in a Spatial Autoregression Model

Three asymptotically equivalent tests for the presence of a common factor in a spatial autoregression model are presented. When such a common factor exists, the model reduces to the simple regression with spatially correlated disturbances. The tests can thus be used to examine the appropriateness of the latter specification. The procedure is illustrated by application to a model of Irish agricultural consumption discussed by O'Sullivan (1968), and Cliff and Ord (1973).