Testing for Spatial Autocorrelation in a Fixed Effects Panel Data Model

This paper derives several Lagrange Multiplier statistics and the correspondinglikelihood ratio statistics to test for spatial autocorrelation in a fixed effectspanel data model. These tests allow discriminating between the two main typesof spatial autocorrelation which are relevant in empirical applications, namelyendogenous spatial lag versus spatially autocorrelated errors. In this paper, fivedifferent statistics are suggested. The first one, the joint test, detects the presenceof spatial autocorrelation whatever its type. Hence, it indicates whetherspecific econometric estimation methods should be implemented to account forthe spatial dimension. In case they need to be implemented, the other four testssupport the choice between the different specifications, i.e. endogenous spatiallag, spatially autocorrelated errors or both. The first two are simple hypothesistests as they detect one kind of spatial autocorrelation assuming the otherone is absent. The last two take into account the presence of one type of spatialautocorrelation when testing for the presence of the other one. We use themethodology developed in Lee and Yu (2008) to set up and estimate the generallikelihood function. Monte Carlo experiments show the good performance ofour tests. Finally, as an illustration, they are applied to the Feldstein-Horiokapuzzle. They indicate a misspecification of the investment-saving regressiondue to the omission of spatial autocorrelation. The traditional saving-retentioncoefficient is shown to be upward biased. In contrast our results favor capitalmobility.