This paper estimates the determinants of child mortality and child health allowing for the possibility that samples of children are choice-based, reflecting prior selective fertility and mortality behavior. Parameter identification is the most serious practical problem in controlling for fertility and mortality selection. Identification is achieved by imposing a random-effects structure on the error correlation matrix for the set of fertility, mortality, and health behaviors. Fertility selection is found to be statistically significant in the estimation of the determinants of mortality in all 14 Sub-Saharan DHS data sets studied, and fertility and mortality selection is found to be significant in the determination of child height in Zambia. Nevertheless, most parameters are little changed when selection is accounted for.

Mark Pitt is a professor of economics at Brown University. This paper benefited from the comments of seminar participants at Brown, Harvard, Michigan, Penn, and Yale, from the excellent research assistance of Deon Filmer, from discussions with Bent Sørensen and Pedro Gozalo, and from the comments of an anonymous referee. This research was supported by NICHD grant HD 22096. The data used in this article can be obtained from Demographic and Health Surveys, Macro International, Inc., Suite 300, 11785 Beltsville Drive, Calverton, MD 20705-3119.

© 2002 by the Board of Regents of the University of Wisconsin System

US ISSN 0022-166X

Return to JHR Home Page