A modified hurdle model for completed fertility

In this paper it is argued that models for completed fertility have to take into consideration that childless couples and couples with an only child are qualitatively different from couples with two or more children. Indeed, these differences may be the cause of the underdispersion that characterizes completed fertility data. An empirical illustration using Portuguese data suggests that accounting for the qualitative difference between having zero, one, or more children leads to considerable improvements over a model of the type generally used to describe this sort of data.