The relationship between early child nutrition and schooling - PDF document

The relationship between early child nutrition and schooling outcomes: Evidence from South Africa’s first longitudinal national household survey Daniela Casale School of Economic and Business Sciences University of the Witwatersrand daniela.casale@wits.ac.za September 2017 DRAFT Abstract Much of the research investigating the determinants of schooling outcomes in South Africa has focussed on concurrent characteristics of the child’s environment, and parti cularly deficiencies in the schooling system. This paper draws attention to the early childhood period and explores the relationship between poor nutrition, reflected in childhood stunting and obesity, and subsequent schooling outcomes. The work draws on the data from the recent release of four waves of the National Income Dynamics Study 2008 - 2015, South Africa’s first large-scale longitudinal household survey. This survey presents the unique opportunity to examine the effects of early child health on human capital accumulation at the national level, in a country beset with persistent inequality in schooling outcomes. The results from a series of household fixed effects estimations suggest that children who were stunted in the early period go on to complete significantly fewer years of schooling by age 14 than otherwise comparable children. This is because they enter school at a later age and are more likely to fail the grade, therefore progressing more slowly through the schooling system. In contrast, no discernible relationship is identified between obesity status and subsequent schooling outcomes in young children. 1

1. Introduction The poor performance of South African learners on average coupled with large inequality in schooling outcomes, has been well-documented in the post-Apartheid period and is one of South Africa’s most pressing development issues (Branson and Lam, 2009; Timaeus et al , 2013; van der Berg et al ., 2016 ) . Much of the recent work on the determinants of schooling outcomes in South Africa has focused on concurrent characteristics in the child’s environment, such as the household’s socio -economic status or deficiencies in the schooling system itself (Timaeus et al , 2013; van der Berg et al ., 2016). The objective of this paper is to draw attentio n to the early period in the child’s life, a period that is increasingly being recognised in the economics literature on human capital formation as being critical for the child’s long -term development (Cunha and Heckman, 2007). More specifically, this paper focusses on the relationship between poor early childhood nutrition, reflected in either stunting or obesity, and the subsequent schooling outcomes of young children. Undernutrition in early childhood, which manifests in low height-for-age or stunting, is predicted to affect educational attainment for a number of reasons. From a physiological perspective, a lack of nutrients may cause structural damage to the brain, especially in the first few years of a child’s life when the rate of formation of neural connections is highest (Morgan and Gibson, 1991). In addition, children who are poorly nourished may lack energy, affecting the way they engage with their environment and ultimately how they learn. Parents and teachers may also treat children who look smaller differently from other children their age, perhaps delaying school enrolment or challenging them less (Behrman, 1996). Poor nutrition resulting in obesity has been hypothesised to influence children’s educ ational attainment through cognitive function because of a potential deficiency in micronutrients. Further, being overwe ight or obese can affect children’s self -esteem, mental health and social interactions, which in turn can influence their performance at school (Caird et al , 2011). If a poorly nourished child starts school later or is more likely to repeat grades, then the child will complete fewer years of schooling or will enter the labour market later (or both). Even if the child completes the same number of years of schooling as his/her peers, entering the labour market later will result in a significant reduction in lifetime earnings (Alderman et al , 2009). Therefore understanding the implications of child health for educational attainment is 2

important particularly in the South African context of large and persistent inequalities in human capital accumulation, labour market success, and consequently household income. While there has been no work in South Africa on the relationship between child obesity and schooling outcomes, research on stunting or low height-for-age suggests negative causal pathways from undernutrition in early childhood to other human capital outcomes (Casale, Desmond and Richter, 2014; Yamauchi, 2008). However, this work is based on small- sample, region-specific data from the early 1990s. Casale et al. (2014) found that children who suffered from early malnutrition, indexed by stunting at two years, did worse on cognitive tests at age five than non-stunted children, using the Birth to Twenty Data from 1990 to 1995, an urban birth cohort study conducted in Johannesburg. Yamauchi (2008) provided evidence of a positive relationship between nutrition in the pre-primary years (measured by height-for-age) and later schooling outcomes using the KwaZulu-Natal Income Dynamics Survey (KIDS) from 1993, 1998 and 2004, a longitudinal survey of African and Indian households living in one of South Africa ’ s nine provinces. The release of the first four waves of the National Income Dynamics Study (NIDS) from 2008 to 2015, presents the unique opportunity to examine the implications of early child health on human capital accumulation at the national level and for a more recent cohort of children. NIDS is South Africa’s first large -scale, nationally representative household survey panel that collects detailed information both on the health status of children, including anthropometric data, and on their progression through the schooling system. 1 The empirical work in this paper uses these data to estimate the relationship between poor nutrition in early childhood, measured by either stunting or obesity, and later educational outcomes, namely age at first enrolment, whether the child had failed the grades enrolled for, and number of grades completed. The anthropometric data are drawn from Wave 1 (2008) of the panel when the sample of children were aged 0 to 8 years, and matched to information on the schooling outcomes from Wave 4 (2014/2015) of the panel, when the children were aged 7 to 14 years. Multivariate regressions controlling for household fixed effects are estimated to account for household level unobservables, and to try to identify causality in the data. 1 The 1993 Project for Statistics on Living Standards and Development (PSLSD) also collected data on both anthropometric and education outcomes for a nationally representative sample, but only at the cross-section. 3

Overall, the results from the analysis thus far suggest that poor nutrition in early childhood, and specifically undernutrition, has a negative effect on the educational attainment of young children. Compared to children who did not exhibit linear growth retardation, children who were stunted were found to have: 1) enrolled later for Grade 1; 2) completed fewer years of schooling, and 3) been more likely to fail the grades they had enrolled for in the preceding years. These findings should be considered in light of the most recent national estimates on the prevalence of stunting in South Africa, which suggest that even in 2012 prevalence was likely to be around 27% for children aged 1 to 3 years (Shisana et al. 2013: 212). The remainder of this paper is structured as follows. Section 2 gives a brief description of the data used in the empirical work. Section 3 describes the child health outcomes from Wave 1, and specifically the anthropometric data used to proxy for poor nutrition. Section 4 provides both a descriptive and regression analysis of the relationship between child nutritional status in Wave 1 and educational outcomes in Wave 4. Section 5 summarises the findings and includes a discussion of various ways in which the NIDS data can be used to extend this work. 2. A brief note on the data NIDS, conducted by the Southern Africa Labour and Development Research Unit at the University of Cape Town is South Africa ’ s first nationally representative longitudinal panel of individuals. Wave 1 collected information on approximately 28 000 individuals living in 7 300 randomly sampled households in 2008. In addition to household and adult questionnaires, a separate child questionnaire was administered to collect detailed information for all children who were resident in the sampled households and who were aged 0 to 14 years (9605 children). In subsequent waves, information was collected in the child questionnaires also on children born to or adopted by female continuing sample members (CSMs) 2 , and children co-resident with CSMs at the time of the interview (or temporary sample members). However, because the anthropometric data collected in the Wave 1 child questionnaire are linked to the educational outcomes collected in the Wave 4 child questionnaire, by definition, the sample of interest becomes restricted to children who were 2 These are resident members of the original Wave 1 households (including children) and any children born to or adopted by female CSMs in the subsequent waves. 4