Model evaluation and causality testing in short panels: the case of infrastructure provision and population growth in the Brazilian Amazon

In this paper we examine the relationship between infrastructure growth and population growth in the Amazon using a panel of 293 municipalities over the period from 1975 to 1985. Contemporaneous cross-section analysis confirms a strong positive correlation between infrastructure and urban population but does not indicate direction of causality. Thus, we employ a modified form of the traditional Granger causality tests to suit the short time series that we have available. Based on out-of-sample forecasting tests we conclude that the empirical evidence supports the hypothesis that growing urban populations lead to more infrastructure development, rather than vice versa.