Regional analyses have a long tradition in fertility research. However, it is controversial to what extent spatial disparities and in particular urban-rural differences are related to variation (1) in compositional population characteristics or (2) regional opportunities. The evidence on how specific factors are relevant and interacting is still limited. We consider Germany for two reasons: New census data allow for in-depth spatial analyses; these seem promising as Germany is a federal country that exhibits high regional heterogeneity regarding aspects such as labour market conditions, urbanisation, housing, and - after ambitious family policy reforms - child care.

The main objective of this paper is to investigate the relevance of regional contexts for understanding variation in fertility, while controlling for individual-level aspects. For this we conduct multilevel analyses using individual-level data from the German 2011-Census enhanced by district-specific aggregate data providing information on compositional and contextual characteristics. As dependent variable we use the child-women ratio based on information on children younger than three years in the household. Further, we explore the role of spatial autocorrelation through Local Indicators of Spatial Association and Bayesian Hierarchical Models.

First findings suggest that fertility is signi?cantly higher in districts with a high percentage of Catholics, a more traditional economic structure and a relative surplus of men. For the urban-rural divide we identify several underlying determinants and their interaction: It is influenced by the composition of individual level characteristics such as migration background, living arrangement and education. Further, a low population density and a greater stock of large dwellings for families are associated with higher fertility. Our analyses provide evidence for both compositional and contextual effects. We also detect a high level of explained variance on the district level suggesting that regional contexts matter. Further we demonstrate how the relevance of regional opportunities varies by educational level.