The Economic Case for the Spatial Error Model with an Application to State Vehicle Usage in the U.S

LeSage and Pace (2009) make an econometric case for the spatial Durbin model over, among others, the spatial error model. We make an economic case for the spatial error model because it captures spatial dependence more fully (i.e. beyond that which can be attributed to the dependent variables in neighboring units). Also, when faced with the choice between aggregate or disaggregated data the spatial error model or a related model (e.g. the seemingly unrelated spatial error model) should be fitted. This is to ensure that Wald tests of whole sets of coeffi cients against one another to establish if disaggregation is necessary are not invalidated. To illustrate the economic case which we make we extend the literature on the determinants of vehicle usage by modelling the spatial dependence of state travel for the U.S. over the period 1980−2008. On the basis of Wald tests, aggregate data on state vehicle usage is progressively disaggregated and spatial error models for travel on all twelve types of highway are fitted. In all cases the spatial dependence in the spatial error models is greater than or approximately equal to that in the spatial lag models, confirming that the former does capture any additional sources of spatial dependence.