Comment on the Identification Strategy in "Family Ruptures, Stress, and the Mental Health of the Next Generation", Part 2

The empirical methodology used by Persson and Rossin-Slater (2016) to estimate the causal effect of in utero exposure to stress on later outcomes has two sources of bias. First, they define the control group in such a way that biases their causal estimates and can lead to the finding of a significant relationship when there is none. Therefore, the causal effect that the authors claim to establish may in fact be spurious correlation from an error in their empirical methodology. Second, measurement error in the estimation of gestation age can exacerbate the magnitude of the bias caused by the misspecified control group. An additional concern is that the endogeneity of actual birth date may make little practical difference in this context, which undermines the author's claim of being the first to establish a causal effect.