Maternal Smoking, Misclassification, and Infant Health

Abstract:     
When a binary variable is misclassified, the measurement error is necessarily correlated with the truth. This observation has important implications for an instrumental variables (IV) framework in which the endogenous variable is a potentially mismeasured binary variable. Ignoring misclassification leads to attenuation in the first stage coefficients relating the endogenous variable to the instrument(s), and by extension to inflated second stage estimates of the causal effects of interest. In this paper, I propose an approach based on recently developed parametric methods for misclassification of a binary dependent variable that allows me to recover consistent estimates of the second stage coefficients. I then use this method to re-analyze the relationship between infant health and maternal smoking. When cigarette taxes are used as an instrument for tobacco use without correcting for measurement error in self-reported smoking, a conventional IV estimate of the effect of smoking on birth weight is only slightly smaller in magnitude than the OLS estimate for whites, but substantially larger for African Americans. Adjusting for misclassification leads to causal estimates that are substantially smaller in absolute value.

Keywords: Maternal smoking, birth outcomes, misclassification

JEL Classification: I12, H2, C39

working papers series