Association for Public Policy Analysis & Management

The Earned Income Tax Credit (EITC) provides refundable tax credits to low-income earners. With expenditures of over 63 billion dollars and 26 million recipients in 2013, the EITC is considered a highly successful anti-poverty program designed to encourage employment among low-income families. In a recent paper, Hoynes, Miller and Simon (2015), henceforth HMS, use the expansion of the EITC in 1993 to test whether income transfers improve infant health. The question is fundamental to U.S. social policy. First, preterm birth and low birth weight births are the most important predictors of infant mortality. Second, there is a clear inverse gradient between adverse birth outcomes and socio-economic status. Third, the long-term effects of low birth weight on adult health appear significant.

HMS find that the EITC lowers the rate of low birth weight among primarily single black women with no more than a high school education. The EITC has little impact on the birth outcomes of whites and Hispanics. In this paper we question their findings. Despite the aggregation of over 32,000,000 births into 47,000 cells across year, states and demographics, the reported improvements in birth outcomes are identified by variation in low birth weight over 8 years and 3 family sizes or 24 observations. Specifically, HMS use a difference-in-differences (DD) estimator to compares changes in low birth weight from before to after the 1993 expansion of the EITC between women with one or more children to women with none. We show that differential trends in low birth weight prior to the EITC violate the parallel trend assumption. We also show that failed placebo tests undermine others. We contend that HMSs’ associations are confounded by the waning of the crack epidemic. Data from New York City birth certificates show differential exposure to cocaine by race and parity and its link to low birth weight between 1983-1998. We follow recent work that uses the change in homicide rates to identify the onset of the crack epidemic. We show that the inclusion of national homicide rates eliminates any association between the EITC and birth outcomes. We conclude that identifying small, causal effects of a national policy at single point in time is exceedingly challenging.