Spatial Dynamic Panel Model: An Application to Environmental Violations in Brazil

This chapter presents a review of the Spatial Econometrics literature, which deals with interactions of economic units in space, traditionally using cross-sectional data. In recent years, a great deal of theoretical effort has been made to deal with data from spatial observations that are repeated throughout time, named panel data models. These models take advantage of the more information available, which helps to fix some selection problems existing in the cross-sectional models, less collinearity, and higher variation in the data. The spatial panel data model can include dynamic effects to investigate the spatial dependence and serial correlations simultaneously. This kind of model considers variation in space and time and treats the serial dependence among spatial units; the spatial dependence over time; specific non-observable space–time effects; and the endogeneity in one or more regressors other than the spatial or time lag in the dependent variable This chapter presents a new method to estimate this kind of model, called Spatial Dynamic Limited Information Maximum Likelihood (SDLIML), when some exploratory variables are endogenous. A Monte Carlo simulation shows that this estimator performs better than the alternative Generalized Method of Moments (GMM) in all situations. We include an illustration section, showing how to perform a small version of the Monte Carlo experiment. In the end, an application using Brazilian data on environmental violations is also presented. Evidence of a pedagogical deterrent effect associated with the value of fines is also found.