Adjusting for Nonnumeric Responses Regarding Ideal Number of Children: Evidence From Sub-Saharan Africa

David Shapiro , Pennsylvania State University

Women’s preferences for numbers of children indicate demand for numbers of children. Preferences are indicated by responses to a question about the ideal number of children; but mean values of responses cannot take account of nonnumeric responses. Women who provide such responses are disproportionately likely to have never attended school, and women who never attended school almost invariably have the highest fertility rates of any education group. Hence, their absence from calculation of mean ideal number of children introduces a downward bias. In this paper, with a sample of 31 countries, we use data on mean ideal number of children from women with numeric responses to impute predicted values for women with nonnumeric responses, based on numerous characteristics. These imputed values allow one to have an estimate of mean ideal number of children based on all women, and these estimates are superior to those conditional on a numeric response.

See paper

 Presented in Session 116. Innovations in Measurement for Fertility, Family Planning, and Sexual and Reproductive Health