A recent article in World Development by Musaddiq and Said (ungated, 2021) reports on the long-run effects of a conditional cash transfer (CCT) program in Pakistan that targeted adolescent girls and young women attending secondary schools. The paper reports some increases (modest, in line with similar evaluations) in secondary school completion and delays in early marriages for the women exposed to the intervention when they were secondary-school aged and some declines (again, modest) in the likelihood of their children being underweight or stunted. These effects on children are in line with other studies, like Cahyadi et al. (2020), which also find effects of CCTs on stunting in Indonesia, but take our knowledge a little further because the children are those of women who were exposed to a CCT not as pregnant women or mothers but as students.
However, I want to discuss something different about the mechanisms of impact in this short post. The authors investigate some potential mechanisms for the impacts observed for the women and their children, such as increased labor market participation and income, for which they cannot provide supporting evidence (using own data and the extant literature from Pakistan). Nor do they report effects on a smaller number of children or increased contraceptive use. They do have some suggestive evidence on increased awareness of health issues (knowledge of HIV/AIDS), which might be a contributing factor to improved child development. They do, however, show that each year of exposure to the program leads to a 0.6 percentage-point increase in the likelihood that their husband has completed secondary schooling or higher. It is this marriage market effect that I want to discuss, as it might be instructive for academics and policymakers alike…
Imagine that a number of young women with secondary education are in a room with more chairs than there are women. The chairs are such that you will have healthier children if you sit in them, but you can only do so with a secondary education certificate. Now, suppose that the number of chairs is fixed but I increase the number of women with the certificate, such that they outnumber the chairs. What happens now is that we have reached the maximum number of healthy children possible and some women will have to remain standing. As we increase the number of women without increasing the number of chairs, we might change the identities of women who sit down and the children who get lucky, but we do not increase the number of healthy children.
Now, if I am evaluating this intervention that increases the number of women with the certificate and do not consider the fact that the treated women are gaining access to the chair possibly at the cost of an equal number of women losing access to one, then I show a treatment effect that is illusory: child health has not improved overall, even though it has been shuffled around across households. Whether such negative spillovers are to be found among eligible women or potentially women who were too old (or maybe too young) to be eligible for the program, it is not possible to do so without a design to detect such effect and, more importantly, data on them. And, in the worst-case scenario, the policy implications might be overly optimistic and misleading. This type of spillover and general equilibrium effects that allow people to benefit from a program by jumping to the front of the queue are discussed in papers, such as Baird et al. (2018) and Crepon et al. (2013), among others.
Please note that the above scenario would go through if I did not insist that the women in the room all had secondary education but women with secondary education had an advantage in the game of musical chairs. As long as anyone sitting on a chair (getting married to a man with higher education) would have had healthier children, everything is the same. If there is something complementary about assortative matching regarding the production of child health, then increasing levels of secondary education would increase the average levels of child health with lower-educated women losing out in the marriage market.
So, even that type of a scenario requires us to think about the production function for child health and the implications of assortative matching (and therefore any program that causes it) on the distribution of future child health: average improvements may need to be weighed against tradeoffs between winners and losers.
There is, of course, another possibility, namely that the number of chairs is not fixed. What if the demands of the now more numerous highly-educated women in the marriage market for higher-quality husbands causes men to realize that they need to invest more in their own education when young, so that they don’t lose out in the market themselves? Then, the CCT program would have achieved something much more virtuous – by increasing education levels for everyone in the society, leading to all sorts of concomitant improvements, including in child health and development. In the context of Crepon et al. (2013) cited above, this would be equivalent to job training programs for youth causing an increase in the number of good jobs in France, rather than simply allowing trained people to replace untrained ones for existing jobs.
Are today’s CCT programs (or other interventions, like scholarships) effective enough to cause such general equilibrium effects? I have my doubts that they are: as Katy Bergstrom and I discuss in a new article on adolescent well-being in developing countries, most such programs increase education levels modestly and at a very high cost per treated household. We argue that innovations are needed to increase the effectiveness of these programs, perhaps especially in the area of targeting them better to those who would benefit the most. That’s perhaps the subject of a future post…
Join the Conversation