This is the 17th in this year’s series of posts by PhD students on the job market
Cash transfers are a popular intervention to improve child nutrition. The bulk of the literature focuses on the effects of safety nets during the first 1000 days of life. In addition to early childhood, adolescence represents a critical period characterized by rapid biological and physiological growth, which increases nutritional needs. There is growing scholarly interest in the need for interventions during the first 8000 days (Bundy et al., 2018), which includes the crucial development stage of adolescence. This is because adolescent undernutrition and overnutrition have adverse long-term implications for adult morbidity (Lassi, Moin, & Bhutta, 2017), and other economic impacts, including lower labor productivity and wage penalty associated with obesity (Cawley, 2015).
In my job market paper, my coauthor Kira Villa and I examine the effect of the Child Support Grant (CSG), an unconditional means-tested cash transfer provided to age-eligible poor children in South Africa, on adolescent nutritional outcomes. CSG is one of the largest welfare programs benefiting millions of poor children for the past two decades. Specifically, we investigate the short-term effects of this safety net on a commonly used measure of adolescent nutritional status, Body Mass Index (BMI)-for-age z-scores, and its categories.
In South Africa, the prevalence of obesity among adolescents has massively increased in the last three decades. A lower, non-trivial proportion is also moderately and severely underweight. This double burden of malnutrition among children and adolescents is a public health issue in developing countries. Over the process of economic development in low and middle-income countries, the burden of obesity has shifted towards the poor (Popkin, Adair, & Ng, 2012). In this context, a question arises whether welfare programs can minimize food insecurity and improve health among the poor without unintended consequences such as increased obesity.
Identifying the causal effect of grant receipt
We use longitudinal data from the 2008, 2010, and 2012 waves of the National Income Dynamic Study (NIDS) in South Africa. The working sample comprises 6778 female and 6679 male adolescents aged 11-19 years. The outcomes include a continuous variable, BMI-for-age z-scores, and a categorical variable indicating nutritional status. Adolescents are defined as underweight if their BMI z-score < -2 standard deviations, normal weight if their -2 ≤ BMI z-score ≤ 1 standard deviation, overweight if their 1 < BMI z-score ≤ 2 standard deviations, and obese if their BMI z-score > 2 standard deviations from the median of the WHO reference population (WHO, 2007). Female adolescents are more likely to suffer from overnutrition whereas male adolescents are more likely to suffer from undernutrition.
To address the endogeneity of the CSG receipt, we employ a fuzzy regression discontinuity (RD) approach that exploits changes in the age-eligibility criterion as the source of exogenous birth-cohort variation in grant receipt. Since the CSG program began in 1998, the age-eligibility criterion increased from age 7 to age 14 by 2005 at irregular intervals. During the study period, the eligibility increased in 1-year increments, from 14 years in 2008 to 15 years in 2009, 16 years in 2010, 17 years in 2011 and to less than 18 years in 2012. These changes resulted in essentially extending eligibility for the same cohorts from age 14 to 18. In our identification strategy, we compare age-eligible cohorts who continue to receive the grant with age-ineligible cohorts who lost eligibility to the grant after they turned 14 years in 2008.
In the first stage, we predict the probability of grant receipt using a dichotomous variable indicating whether the adolescent is age-eligible (left of the cutoff) at the time of the survey (Figure 1). We use a flexible linear term of the running variable, age in months relative to eligibility. In the second stage, we regress BMI z-scores on predicted CSG receipt using a linear model. For nutritional status categories, we use a multinomial logit model.
The fuzzy RD approach assumes that the discontinuity in the outcome is attributable to a discontinuity in the treatment. We provide support for this assumption by demonstrating smoothness of individual and household covariates around the cutoff. We rule out possible sorting into the eligible group by showing that the density of the running variable does not exhibit a discontinuity at the threshold. We conduct falsification tests and show that our results are insensitive to the choice of polynomial of the running variable and alternative bandwidths using local polynomial approaches following Calonico, Cattaneo, and Titiunik (2014). The attrition rate, due to reasons other than ageing in- and out- of the sample, is not significantly different on either side of the cutoff and is uncorrelated with socioeconomic characteristics.
The effect of the Child Support Grant receipt
Our results indicate that CSG receipt increases BMI z-score by 0.324 standard deviations among female recipients relative to non-recipients (Figure 2). This effect is statistically indistinguishable from that of males. When we consider categories of nutritional status, we find that CSG receipt decreases the likelihood of being underweight by 7.18 percentage points (pp) and the likelihood of being obese by 6.07 pp among female recipients. Grant receipt increases the likelihood of being overweight by 12.2 pp among female recipients relative to non-recipients, and the effect is statistically different from that for males. The strong positive effects on the probability of being overweight may have negative long-term health implications as the prevalence of overweight is associated with mortality and co-morbidities (van Dam et al., 2006; Reilly & Kelly, 2011). However, most of the evidence is from developed countries and demonstrates larger economic costs related to obesity (Flegal et al., 2013; Cawley, 2015). There is scant evidence of adverse health outcomes of overweight prevalence in the context of developing countries, including South Africa.
Results from a quantile regression of BMI z-scores (Figure 3) indicate that the positive effects on female BMI are concentrated at the lower BMI z-score distribution up to the 60th percentile. The only significant effect of grant receipt for male recipients is an increase in BMI at the lowest end of the distribution. Although the effect on being underweight is noisy for male recipients, grant receipt improves the health of the extremely undernourished male adolescents.
Limitations of the study
The first-stage relationship between grant eligibility and grant receipt is almost the same for males and females. Due to unavailability of information on individual dietary intake and health behaviors, we cannot disentangle the causal pathways underlying the observed effects. We speculate that the effects of grant receipt on female nutritional outcomes may be driven by higher ex-post allocation of household resources in favor of female adolescents by female caregivers, to whom the grant amount is most often paid, and the higher biological sensitivity of females to external inputs during/post puberty.
Key Takeaways and Policy Implications
We find that cash transfers increase BMI and the probability of being overweight among female beneficiaries, and an increase in BMI at the lowest end of the BMI z-score distribution for males. Importantly, increased income from cash transfers alleviates extreme malnutrition (underweight and obesity) among female beneficiaries. Our results suggest that anti-poverty interventions can have additional benefits of improving the nutrition outcomes of female adolescents. Improved nutritional status during adolescence can generate long-term health gains and improved reproductive health. Welfare programs can contribute to reducing the burden of healthcare costs associated with extreme undernutrition and overnutrition.
Kritika Sen Chakraborty is a PhD candidate in the Department of Economics at the University of New Mexico.
Join the Conversation