Spatial dynamic panel data models with random effects

We develop a general space–time filter applied to panel data models in order to control for heterogeneity as well as both time and spatial dependence. Treatment of initial period observations is analyzed when the number of time periods is small. A second issue relates to a restriction implied by the filter specification on the space–time cross-product term that can greatly simplify interpretation of model estimates as well as the estimation procedure. An applied illustration of the method is provided using a Solow growth model. The application shows that the theoretical restriction implied for the cross-product term in our space–time filter specification is consistent with this particular dynamic space–time panel data set.