Spatial Dependence and Heterogeneity in Ten Years of Fertility Decline in Brazil

Knowledge, innovations and behaviors that influence fertility decisions spread over space through teaching, learning and imitative processes among people living in areas that are closer to each other. Using data based from the Brazilian censuses of 1991 and 2000 I investigate the spatial patterns of fertility in about 5500 municipalities, and evaluate the importance of socio-demographic variables accounting for these patterns. I do this by estimating and comparing the efficiency of three models: ordinary least squares (OLS), OLS with spatially correlated errors, and Geographically Weighted Regressions (GWR). The results indicate that the last model is the most appropriate in delineating the relationship between fertility and development shifts through time and space. That is, the importance of each variable depends on its spatial distribution. In particular, the level of fertility of neighboring municipalities is a strong predictor for local fertility rates, providing evidence that fertility has a multiplier effect mediated by distance. The results also indicate that past fertility is a significant determinant of how quickly fertility declines. After considering past fertility and spatial heterogeneity, improvements in education do not necessarily lead to fertility reduction.