Using Zero-inflated Count Regression Models To Estimate The Fertility Of U. S. Women

Abstract

In the modeling of count variables there is sometimes a preponderance of zero counts. This article concerns the estimation of Poisson regression models (PRM) and negative binomial regression models (NBRM) to predict the average number of children ever born (CEB) to women in the U.S. The PRM and NBRM will often under-predict zeros because they do not consider zero counts of women who are not trying to have children. The fertility of U.S. white and Mexican-origin women show that zero-inflated Poisson (ZIP) and zero-inflated negative binomial (ZINB) models perform better in many respects than the Poisson and negative binomial models. Zero-inflated Poisson and negative binomial regression models are statistically appropriate for the modeling of fertility in low fertility populations, especially when there is a preponderance of women in the society with no children.