The adverse consequences of maternal drug use on birth outcomes are a serious public health issue. Clinical data have clearly established that use of illicit drugs during pregnancy is significantly correlated with the incidence of low birth weight and a variety of other neonatal health ailments. Indeed, show that the dramatic rise in the aggregate rate of low birth weight among blacks in New York City between 1984 and 1988 is directly associated with the rise in illicit drug use. Moreover, as document, the economic consequences of prenatal illicit drug use are potentially enormous. Although well-controlled investigations have estimated the consequences and costs of prenatal illicit drug use in a clinical setting, the extent of the problem in the general population is not well known. National surveys such as the National Maternal and Infant Health Survey and vital statistics are a potential source of data on prenatal illicit drug use in the general population, but experience from clinical studies suggests that self-reported drug use seriously underestimates not only the prevalence, but also the potential magnitude of the consequences of prenatal illicit drug use.
The primary purpose of this paper is to investigate the potential ways that measurement error may bias estimates of the effect of prenatal illicit drug use on infant health. We confront two issues which are frequently skirted in the economics literature. First, we relax the standard assumption that measurement error is random. Second, we discuss measurement error in the context of binary variables. In addition, we suggest a method to correct for non-random measurement error with a binary indicator of illicit drugs which uses out-of-sample information to predict the probability of falsely reporting nonuse.
Our empirical results suggest that the estimates of the impact of self-reported prenatal illicit drug use on birth weight are biased upwards by measurement error a finding in direct contrast to predictions from a standard model of random measurement error..............
