The characteristics and evolution of the Brazilian spatial urban system: empirical evidences for the long-run, 1970-2010

In this paper we use spatial analysis and spatial econometrics methods to assess some empirical issues on the size distribution of the Brazilian Urban System. The main novelty is the long historical period of analysis which includes all the demographic censuses from 1940. More specifically, we describe the spatial distribution of the Brazilian cities, as well as its temporal evolution; we test test Zipf’s Law in its original and spatial version, and we use Markov Chains analysis in order to shed some light on the dynamics of the cities within the urban system. We introduce spatial dependence in both Zipf’s law estimation and Markov chain framework in order to capture the influence of the geographical environment on the relative position and mobility of the cities within the urban system. Our estimates for the Pareto coefficient were quite different for OLS and spatial models suggesting the inconsistency of OLS estimates in the non-spatial models. For the full sample of municipalities the Pareto coefficient is much smaller than 1, which features a polarized urban structure. Municipalities with urban population above the average show a less concentrated urban structure and seem to confirm Zipf law. Traditional Markov chain analysis indicates a high stability and low mobility inter-class over time confirming that, with rare exceptions, radical changes in urban population size are uncommon. Results for the Spatial Markov matrix show that municipalities with more populous neighbors are more likely to grow.