Is prenatal care really ineffective? Or, is the 'devil' in the distribution?

Prenatal care should improve infant health, yet research frequently finds only weak effects. If there are two kinds of pregnancies, 'complicated' and 'normal' ones, then combining these pregnancies may lead prenatal care to appear ineffective. Data from the National Maternal and Infant Health Survey (NMIHS) offers compelling evidence. The standard 2SLS approach yields obviously bimodal residuals and frequently insignificant prenatal care coefficients. In contrast, estimating birth weights with a finite mixture model yields estimates revealing that prenatal care has a substantial effect on 'normal' pregnancies. Our Monte Carlo experiment confirms that ignoring even a small proportion of 'complicated' pregnancies can lead prenatal care to appear unimportant.