Hamad Mahar completed his PhD in economics from Trinity College Dublin and is currently on the job market. More details about his research can be found at his website.
Click here for Hamad’s job market paper Elevator Pitch.
Introduction:
One billion children suffer some form of violence each year and nearly 3 out of 4 suffer from negligence or abuse at the hands of caregivers and parents. In Sub-Saharan Africa, rates of Child Neglect and Abuse are the highest in the world (UNICEF, 2020). There has been a wide array of policies designed to curb such practices from micro interventions such as cash transfers to families (Saha et al, 2018) to macro interventions such as the outright banning of social practices that harm children (Camilotti, 2016). The evidence on the efficacy of macro policies working is mixed ranging from finding no effect to short and medium-run positive effects (Camilotti, 2016; Alfano, 2020).
Part of the reason why macro policies may not work is that they are applied in settings where local customs are embedded and hard to change (Fouka, 2020). 170 counties in the world have signed up for some form of child protection orders, however, there is no clear consensus in the literature if these policies have improved the lives of children. It may be that granting legal protection to children may have unintended negative effects and backfire in an environment where the prevalent social norms are embedded and thus hard to influence (Acemoglu and Jackson, 2017).
In my job market paper, I estimate the effect of granting formal legal protection to children on investments in children by exploiting a natural experiment. In 2003, Nigeria passed an ambitious act known as the Child Rights Act (CRA). The CRA was described as a “milestone” (Philip and Ajanwachuku, 2018), it regulated the behaviour of parents towards their children by guaranteeing protection for children against mistreatment and abuse in a variety of different contexts such as providing adequate healthcare. Parents could be subjected to strict actions such as fines or more extreme punishments such as prison sentences. Over 12 years, 26 states passed the legislation with differential timing due to differences in state machinery and debates around the passing of the law, the 11 northern states have yet to pass the legislation.
Identifying the causal effect of child rights on children’s health:
I rely on data from four waves of the DHS (2003, 2008, 2013 & 2018). The pooled sample yields 94,053 children who are less than 5 years old and were born between July 2003 and December 2018. Since the CRA regulates different aspects of child welfare, I examine its effect on the anthropometry of the child and their vaccination status. The choice of these variables is driven by what they represent, investments in preventive healthcare (vaccinations) and investments in general child wellbeing (anthropometry). To investigate the effect of the CRA on anthropometry, I use the z-score methodology issued by the WHO. For the vaccination status of the child, I use those responses verified by the enumerator using a vaccination card.
In the TWFE DiD design, identification depends upon the existence of parallel trends between the treated and untreated groups. To begin with, I consider the raw data trends during the pre-period for anthropometry and vaccination between children in the CRA and non-CRA states, these descriptive graphs show similar trends for children as a function of their age in months. Next, I present a number of event studies namely the Sun and Abraham (2021) estimator, Gardner’s (2021) two-stage estimator, Borusyaks et al (2022) robust/efficient estimator, and Clément & D’Haultfoeuille (2021) estimator with heterogeneity. In the pre-period coefficients are statistically indistinguishable from zero. In the post period, the coefficients become statistically significant after T3 for the main results of height for age.
Lastly, I employ Rambachan and Roth’s (2022) “honest” differences-in-differences approach to relax the parallel trends assumption and conduct a sensitivity analysis on what the allowable deviations are from parallel trends in the pre-period. In essence, the honest DiD approach estimates the amount of non-linearity permissible while still rejecting the null hypothesis, the breakdown value or M as it is referred to. In simple terms when M=0 only linear violations from parallel trends are allowed, while increasing the value of M permits greater deviations from linearity. In my setting imposing linear parallel trends yields a positive and statistically significant estimate for my main anthropometric result regarding height for age (i.e.: when M=0), and this is the case when increasing non-linear violations up till M=0.20.
Main result of the CRA on anthropometry
The introduction of formal protection for Children in Nigeria led to a reduction in stunting, children born under the new policy had a positive Z score gap of 0.25σ to 0.5σ (depending on the specification) for their height relative to their age compared to children in the control states. This positive score represents an improvement in the long-term well-being of children under 5. In terms of other interventions and contextualising the magnitude of these findings, there are a vast number of studies conducted on the effect of nutrition on height for age Z scores. It is important to note that Height for Age (HForA) is a reflection of a variety of factors which include genetics, disease, nutrition, and environmental influences (Perkins et al, 2016). My HforA estimates from my main specifications (TWFE) and the Sun and Abraham (2021) estimator are broadly in line with the effects found in the literature from nutritional and combined nutrition and information interventions.
I conduct a heterogeneity analysis by the ethnicity of the caretaker (mother), I find these positive effects on children’s anthropometry are driven by smaller ethnic groups. I also conduct a heterogeneity analysis based on the belief in witchcraft of the ethnic group that the child belonged to. Belief in witchcraft can be seen as a proxy for belief in superstition, given the accusations against Children in Nigeria of being “Child Witches” (Snow and College, 2017), the effect of the CRA on the anthropometry (Height for Age) of the child is higher for children that belong to an ethnic group which has a higher than median belief in witchcraft compared to those groups that do not, this positive effect on children belonging to ethnic groups with higher than median belief in witchcraft can be viewed as suggestive evidence of the CRA combatting adverse social norms.Notes: Table 2 reports estimates for the effect of the CRA on various anthropometric measures using the estimation shown in Equation 1; State Policy Change is a dummy taking value 1 if Child i resides in state that introduced the CRA; Born After takes the value 1 if the child is born after the policy changed using the month & year of birth of the child. The samples are panel A is for the whole of Nigeria, Panel B limits the treated states to only those states that border the northern states and Panel C only compares outcomes for the Southern states.; all parameters are estimates based on the pooled sample of DHS; standard errors are reported in parentheses and are clustered at the state level; each child contributes 1 observation; ***, **, and * indicate significance at the 1%, 5% and 10% levels. Sample consists of children between July, 2003 and December, 2018 who are under 5 years of age at the time of the interview.
Limitations of the study
The improvement in Anthropometry due to the CRA could be driven by demand or supply side factors. The point estimates fluctuate minimally with the inclusion of two broad supply-driven controls. A more nuanced measure of capturing supply-driven effects could be to look at the intensity/frequency of vaccinations campaigns that took place during the period of study. Such data, unfortunately, is not publicly available. In terms of looking at the demand of healthcare by parents, I investigate if the CRA led to an increase in health-seeking behaviour (i.e.: if they travelled to a health facility in the last 12 months) by parents of children born under the protection of the CRA compared to those that were not. I find a positive correlation of 6.4% for this behaviour. Naturally, it is possible that consulting a traditional practitioner was involved when the child became unwell. Still, unfortunately, this data is not available within the DHS survey to formally test this hypothesis.
Implications for policy and research
The evidence presented in this paper suggests that the introduction of formal protection to Children in Nigeria led to a reduction in stunting. These findings may be of interest to policy makers in as far as they shed further light on how exactly parents in low-income countries make decisions regarding their children in environments characterized by adverse social traditions. The results regarding stunting, in particular, illustrate how changes in the environment can influence parental behaviour. The size of the Child Rights effect on child welfare illustrates the importance of cultural and secular factors in determining how parents treat their children. This insight is important for anyone interested in designing policies aimed at improving child welfare. The results presented in this paper may also serve as an indication of the importance of child protection in low-income settings.
Feature Image Create in Canva
