A cool paper on the impact of maternal health on child health, by Leah Bevis and Kira Villa, is now forthcoming in the Journal of Human Resources. I’ve had the opportunity to see this paper presented by both Leah and Kira at multiple conferences over the last few years. It really is excellent work by two very talented economists.
The headline result is that a mother’s health impacts their child’s health throughout childhood. Thus, previous estimates of the transmission rate of maternal health on child health at a single point in time underestimate the full effect.
The authors start out by acknowledging that the relationship between parental health and child health is complicated. Here are some examples of complicating factors: (a) educational attainment tends to be correlated across generations, (b) parent education levels can influence child health, (c) parent income can influence child health, and (d) parental mental health—specifically measured by cortisol levels—can also influence child health. In this paper, the author’s primary aim is to disentangle these effects from the effect of parental health on child health.
To do this the authors instrument maternal health with random weather shocks around the time the mother’s birth and early childhood. Specifically, variation in the wind, rainfall, and temperature in the 1940s and 1960s in Cebu, a city in the Philippines, influences the health of mothers with children in the present.
One issue with this approach is that there are potentially many valid instruments and this can lead to a dimensionality problem when estimating the first-stage of the instrumental variable set-up. To solve this issue the authors use machine learning techniques to improve the efficiency of their first-stage estimation. Let me just pause and say that this contribution is so cool! There is a ton more detail in the paper on this analytical approach and I’d encourage anyone interested to check out those details.
The core finding is that “the effect of maternal health on child height-for-age z-score not only remains after birth, but actually grows stronger” through adolescence. This can be seen in the left panel in Figure 1 below. More specifically, a mother’s height transmits to their child’s height from birth through age 15. The persistence in this effect is striking and importantly is not found when estimating the transmission to the child’s weight (shown in the right panel in Figure 1 below). Instead, by the time the child is about 11 years old, their weight is no different from other children.
This leads to a critical follow-up question, what explains this persistence? One theory is that this relationship is entirely explained by a transmission process that occurs in utero. Thus, in this case, health at birth should explain these effects. An alternative theory is that less healthy mothers also tend to be poorer (in a variety of dimensions) and that access to resources in childhood, therefore, explains these results. As plausible as these two theories may seem they do not seem to fully explain the persistence found in the data. Even when controlling for socioeconomic status, parental inputs, and birthweight—as shown in the two figures below—we still see the persistence of the effect of mother’s health on child health through age 15.
This motivates further explorations. The sex of the child does seem to influence the strength of these results. As the figure below shows a mother’s health transmission is strongest for girls at around age 8 and is strongest for boys at around age 11. The authors highlight that these patterns, while not confirmatory of such a mechanism, are consistent with the idea that maternal health shapes the growth trajectories of children.
So, like any good study, this paper (a) shows us something we did not know before, (b) explores possible explanations for the novel finding, and (c) motivates future research aiming to better understand the effects. Again, and if you are not convinced yet, I’d really recommend reading this paper in full. It is excellent!