Related Experiment Videos
Birth outcomes and the effectiveness of prenatal care
1Department of Pharmaceutical Economics and Policy, University of Southern California, Los Angeles 90033, USA.
Objective:
To investigate pregnant women's self-selection effects on the estimation of birthweight production function. A particular emphasis is placed on assessing the effectiveness of prenatal care as a major medical input in the birthweight production function.
Data Sources:
Primary data compiled from birth and abortion certificates for the Commonwealth of Virginia in 1984. Several area-specific socioeconomic variables were also employed from the Area Resource File 1984; Supplemental Food Program for Women, Infants, and Children (WIC) Local Agency Directory; and the family planning clinics data by the Alan Guttmacher Institute (AGI).
Study Design:
Two types of self-selection effects are defined: selection effect due to sample censoring from the resolution of pregnancies as live births or induced abortions; and selection effect due to the use of prenatal care as an endogenous variable. Race- and location-specific birthweight production functions are estimated using models with and without correction for self-selection effects.
Principal Findings:
The self-selection effect in the resolution of pregnancies is race-specific, being significant for African American women. The effectiveness of prenatal care in birthweight production is underestimated substantially by the selection bias from the use of prenatal care, and overestimated by the selection bias from pregnancy resolutions. On average, the overall estimated effectiveness of prenatal care is over five times higher after controlling for the selection effects.
Conclusions:
Self-selection effects could be a very serious problem in measuring the effectiveness of birthweight determinants in general. The overall effectiveness of prenatal care, in particular, tends to be significantly biased downward without controlling for selection effects. The significance and scale of the bias depends crucially on specific data and cohorts of the population investigated.