Search Results

Author: Pitt, Mark
Resulting in 1 citation.
1. Pitt, Mark
A Simple Correction for Fertility Selection
Presented: Minneapolis, MN, Population Association of America Meetings, May 2003
Cohort(s): Children of the NLSY79, NLSY79
Publisher: Population Association of America
Keyword(s): Birthweight; Child Health; Data Analysis; Fertility; Modeling, Logit; Modeling, Probit; Monte Carlo; Peabody Individual Achievement Test (PIAT- Math); Peabody Individual Achievement Test (PIAT- Reading); Statistical Analysis; Variables, Independent - Covariate

Permission to reprint the abstract has not been received from the publisher.

This paper sets out some methods for controlling and testing for fertility selection that are extremely easy to use and can be simply implemented without any knowledge of programming using only a half-dozen lines of code in standard software packages such as Stata or SPSS. These methods require that a random effects structure for regression errors as in Pitt (1997). One method is essentially a two-step estimator of Pitt's 1997 random effects selection model.. In the first step of this procedure a selection correction term (as in Heckman's two-step method) is calculated that does not rely on a distributional assumption to achieve parameter identification in the second stage. In particular, this paper demonstrates that successive Taylor series approximations of the woman-specific random effect can be easily calculated based simply on the parameters of a simple binary probit (or logit) model of fertility, and the actual fertility outcomes observed for each woman. As the number of potential births (time periods) gets large, this estimator converges to the true random effect. In practice, simulation experiments demonstrate the efficacy of this approach even for samples of women still early in the reproductive lives, and the substantially higher precision obtained as compared to the application of Heckman's two-step (or maximum likelihood) inverse Mill's ratio approach that does not make use of the panel nature of reproductive histories. Adding these estimated random effects as an independent variable in subsequent regressions of the determinants of child health (or schooling or other behaviors), controls for fertility selection. The paper provides simulation (Monte Carlo) results demonstrating the efficacy of this two-step method, compares it to Heckman's method, extends it to models with nonnormal errors and state dependence (lagged dependent variables), and presents results of its application to the determination of various measures of child health and cognitive achievement using the NLSY data set.
Bibliography Citation
Pitt, Mark. "A Simple Correction for Fertility Selection." Presented: Minneapolis, MN, Population Association of America Meetings, May 2003.