Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

Preterm birth rates have risen 20 percent since 1990, with 12.7 percent of births in 2005 delivered at less than 37 weeks of gestation (Hamilton et al. 2006). The Institute of Medicine (IOM 2007) estimates the cost of preterm births at over 26 billion dollars in 2005, with a disproportionate share going to the two percent of babies born prior to 32 weeks. The IOM also indicates that the risks for preterm infants include not just the immediate mortality and health complications, but also include longer term developmental issues. Moreover, these risks also apply to infants born closer to term.

The goal of this research is to explore the factors that contribute to preterm births and whether changes in these factors can account for the increasing incidence. Particular attention is paid to whether the weathering hypothesis accounts for the changes. This notion, that older African American women have relatively poorer birth outcomes, has found some empirical support in research on birth weight outcomes (See Rauh et al., 2001).

The data used in our analysis is based on births in Georgia over the period from 1994 to 2004. The increase in preterm births in Georgia, which mirrors that of the nation, has occurred in later preterm births between 32 and 36 weeks of gestation, with no change to slight declines for the very early preterm births. Preliminary results confirm previous empirical findings that black and low-educated mothers have higher rates of late preterm births. However, we find that the rate of growth over the period is much higher for more educated and Caucasian mothers. Interestingly, after controlling for other human capital and health characteristics, the effect of being in an older age group appears to be similar across racial groups, thus not lending support to the weathering hypothesis as an explanation for the racial differences in preterm birth rates.

The estimation strategy we employ considers both conditional probabilities of preterm births and quantile regressions. Previous work on birth weight (see Abrevaya and Dahl 2007, for example) indicates that conditional mean regressions do not provide much information on the effect of covariates in the lower tail of distribution of birth weights. It seems that the same argument could be made here, and that estimations of conditional quantiles for gestation length could provide more useful results.

This paper utilizes Vital Statistics birth records from the State of Georgia for the period 1994 to 2002 linked with three sets of state administrative records and the Public Use Microsample of the Census (PUMS). The first two, the Employer File and the Individual Wage File, are compiled by the Georgia Department of Labor for the purposes of administering the state's Unemployment Insurance (UI) program. The third data set contains Welfare Recipient Data from the Georgia Department of Human Resources. All the data used in the analysis are highly confidential and strictly limited in their distribution.