Disparities in Birthweight by Maternal Educational Attainment and Race/Ethnicity: An Illustrative Example of a New Modeling Approach

Michael S. Pollard, RAND

This paper takes an in-depth look at a new way of modeling the major underlying cause of infant mortality and childhood morbidity – low birth weight. The paper illustrates the quantile regression technique and its benefits over current methods (almost exclusively based on statistical models of the mean using OLS regression, or dichotomous measures of "low birthweight" using logit or probit regression) by focusing on the association between maternal education and subsequent birth weight using the 2001 U.S. Natality Data Set containing nearly 4 million observations. After establishing the complex relationship between maternal educational attainment and birth weight, we then further consider how maternal education is associated with birth weight outcomes within 14 race/ethnic groups. Preliminary results indicate that the common analytical techniques would have obscured the strong association between maternal education and birthweight at the most vulnerable portion of the birthweight distribution.