Family Ruptures, Stress, and the Mental Health of the Next Generation: Comment

Abstract

The empirical methodology used by Persson and Rossin-Slater (2018) to estimate the causal effect of in utero exposure to stress contains a potentially significant flaw. They define the control group in a way that may bias their causal estimates and can lead to the finding of a significant relationship when there is none. In this note, I describe the source of the bias and suggest an alternative specification of the control group. (JEL I12, J12, J13)

Family Ruptures, Stress, and the Mental Health of the Next Generation

“…We explored the correlation between treatment and first parity births in detail, and conclude that it is mechanically driven by differential seasonality in conceptions by parity that coincides with a seasonal pattern in relative deaths. We discuss this issue at length in online Appendix C. For this reason, all of our analyses include month of conception and parity fixed effects, and we show that our results are also robust to the inclusion of parity × month of conception interactions in online Appendix D. 28 A second, and related, concern for our identification assumption is that the death of a relative during pregnancy may cause an increase in miscarriages or fetal or infant deaths, leading to selection in our sample of surviving children. Moreover, there may be differential selection by parity, which could introduce the correlation between treatment and first parity that we see in online Appendix Table A1.…”

“…Thus, all of the regressions in online Appendix Tables A1 and A2 control for the relative's age and age squared. 28 The correlation between treatment and the likelihood of the mother being born outside Sweden is driven by a highly skewed distribution of relative deaths in the sample of children of foreign-born mothers that exhibits extra mass of relative deaths around 400-500 days post-conception (i.e., after birth). In online Appendix D, we show that our results are robust to dropping children of foreign-born mothers from our sample.…”

Childhood circumstances and young adulthood outcomes: The role of mothers' financial problems

“…A smaller but still considerable set of contributions has focused on potential effects on the children of the adults concerned. These have considered the intergenerational effect of parental job loss (Hilger, 2016; Oreopoulos, Page, & Stevens, 2008; Rege, Telle, & Votruba, 2011; Stevens & Schaller, 2011), parental income (Akee, Copeland, Keeler, Angold, & Costello, 2010, Bastian & Michelmore, 2018; Dahl & Lochner, 2012; Hoynes, Whitmore Schanzenbach, & Almond, 2016; Duncan, Magnuson, & Votruba‐Drzal, 2017, provide a meta‐analysis) and parental health shocks (Persson & Rossin‐Slater, 2018; Matsumoto, 2018).…”

Violence and Newborn Health : Estimates for Colombia

“…Yearly variables are linked with DHS on the year and month of birth: if the child was born after June, the value corresponds to that of the concurrent year; if the child was born before June, it corresponds to the previous year. 36 For instance, Mansour and Rees (2012) find effects for the third trimester,Valente (2015) for the second one, andCamacho (2008) and Quintana-Domeque and Ródenas-Serrano (2017) only for early-pregnancy.37 In the sample, only 154 cases (0.52%).38 Despite this, because gestational length is negatively associated with in utero shocks, there could be some bias introduced for correlated outcomes; for instance, mothers who are not exposed to stress can mechanically have higher birthweight because of the longer gestational length(Matsumoto 2018). In a robustness check, I try the alternative proposed by Matsumoto (2018) of using three full months for each pregnancy trimester, although the…”