Abstract
Discrete-time hazard models for first- and higher-order births are estimated from DHS data for 22 countries in Sub-Saharan Africa for the 1990s. It is found that, among women with the same number of completed years of schooling, fertility decreases with increasing average educational level in the neighborhood (DHS cluster). Such a significant community effect, net of urbanization, has never been documented before. When the aggregate variable is left out of the model, some of its effect is captured by the individual variable, but only a small part. One may therefore be led to conclude that investments in education are much less powerful in reducing fertility than they really are. For example, a simulation based on a model including only individual education reveals that average fertility for these 22 countries would have been 0.52 lower if education were expanded from the current level in the region to that in Kenya, which is the most advanced country in this respect. However, according to a model where also average education is included, the drop would have been as large as 1.00, of which 0.38 is the purely individual effect. A considerable aggregate contribution is estimated even when several potential determinants of education are entered into the models.