Southern African Regional Poverty Network (SARPN) SARPN thematic photo
Regional themes > Children and education Last update: 2020-11-27  

 Related documents

The impact of unconditional cash transfers on nutrition: The South African child support grant1

Working paper number 39

Jorge M. Agьero2, Michael R. Carter3, Ingrid Woolard4

International Poverty Centre (IPC), United Nations Development Programme (UNDP)

September 2007

SARPN acknowledges the International Poverty Centre as a source of this document:
[Download complete version - 355Kb ~ 2 min (30 pages)]     [ Share with a friend  ]


Despite improvements over the last two decades, child malnutrition remains a serious health problem in developing countries, and is the main contributor to child mortality (World Bank 2006). For those children who survive, early childhood malnutrition contributes to the intergenerational transmission of poverty. In a striking study of Ecuadorian children, Paxson and Schady (2005) show that by age five, children who were likely inadequately nourished have already fallen well behind the cognitive development of their better nourished peers. Assuming that these disparities persist, the malnourished children would likely do less well in school, accumulate less human capital and enjoy lower adult earnings than their peers. We would expect that their own children would in turn repeat this inequitable cycle.

In an effort to break this intergenerational transmission of poverty (poverty that ‘lays its own eggs’, in the words of an informant quoted in the Chronic Poverty Research Centre Report (2004)), some middle-income countries have adopted cash transfer programs designed to bolster the nutrition, health and education of the children of poor families. At first glance, reliance on cash transfers to achieve these goals may seem somewhat surprising in light of the debate about whether nutrition responds at all to income increases amongst poor families. For example, Behrman and Deolalikar (1987) find an income elasticity close to zero for a sample of families in south India. While other studies find a positive elasticity, the issue is far from solved.5

There are, however, several critical differences between the new generation of cash transfer programmes and the market-generated income difference used to identify nutritional elasticities in the earlier literature. Mexico's Progresa programme, the best known of these cash transfer programmes, has two key design features that may mediate its nutritional impacts.6 First, cash transfers are conditional on the household meeting certain required behaviours: older children must attend school; and younger children must visit clinics for regular medical check-ups and nutritional monitoring (where among other aspects they are given nutritional supplements).

In addition to these conditionalities, Progresa cash transfers are also assigned to women. Unlike market-driven income increases, which may have been generated by increases in returns to assets owned by men, these targeted cash transfers have been designed to bolster the bargaining power of women, with the idea of giving more weight to their preferences because they are presumed to be more child-centric.7

While much of the impact evaluation literature on Progresa has focussed on schooling outcomes, there is evidence that Progresa has boosted child nutritional status (Behrman and Hoddinott 2005). However, because Progresa transfers were conditional (and included in-kind nutritional supplements), it is not clear whether these findings indicate non-zero income elasticities of nutrition, or simply the impact of transfer conditionality.

Unpacking the reasons behind this response is of more than academic interest. Aid agencies have noted that the heavy administrative burdens implied by transfer conditionality limit the ability of lower income African economies to implement programmes modelled on Progresa. Given that it is precisely these economies where malnutrition is most severe, understanding the importance of costly conditionality is important.8

This Working Paper aims to contribute to knowledge in this area by studying the impact of the South African Child Support Grant (CSG), which was first rolled out in 1998. Like Progresa and its sister programmes in Latin America, CSG cash transfers are targeted at women. Unlike those programmes, CSG transfers are unconditional, and come with no strings attached, nor with any in-kind transfers.9 Analysis of this programme thus promises a more sharply focused examination of the income elasticity of nutrition, at least in the context of income increases targeted at women.

However, in comparison to the Progresa program, evaluation of the CSG presents a particular methodological challenge. By randomly selecting rural areas to receive the cash transfer treatment, Progresa quickly became a popular program among academic analysts. In contrast, the CSG was a rolled out as a single, national programme, depriving analysts of purposefully randomized treatment and control groups. Alternative methods are thus needed to estimate the impact of the South African CSG.

One method would be to follow the current literature and use matching methods to evaluate CSG support as a binary treatment. Statistical problems aside, this approach would overlook the fact that the extent of CSG treatment (the 'dosage') varies significantly across the treated population. During the nutritionally critical first 36 months of life (which will be the focus of the analysis here), some children received CSG support nearly 100 per cent of the time, while others received only a month or two of support. The nutritional impact of the latter is likely negligible, while the impact of the former could be substantial. The analysis here will thus use variation in the extent of treatment to identify the impact of the CSG.

While the continuous treatment estimator of Hirano and Imbens (2004) opens the door to analysis of the extent of treatment among the treated, it still requires strong orthogonality assumptions. The critical statistical identification assumption is that the extent of treatment is unrelated to unobserved factors that themselves can affect child health and nutrition.

While this assumption may seem hard to sustain for a voluntary programme that had a single national eligibility date, we show how information on effective program rollout can be used to derive a caregiver ‘eagerness’ measure that allows analysts to control for otherwise unobservable, and confounding, characteristics. Conditional on ‘eagerness’ and other covariates, we argue that exposure to the CSG depends (randomly) on the interaction between the child's birth date and the effective programme rollout for the child's locational and temporal cohort.

Exploiting this eagerness variable, our analysis of the impacts of early-life CSG as a continuous treatment case uncovers economically and statistically significant effects for large dosages of CSG support. These estimates show that effects are insignificant for children who received CSG support for less than 50 per cent of their 36 month window. These results are robust to the inclusion of cohort and cluster variables meant to control for differences in the supply of health-related public goods and other time and locational differences.

Finally, in an effort to get an understanding of the possible economic value of these nutritional gains, we project forward in time using best estimates from the literature concerning the impact of adult height on wages. Adaptation of these estimates to the South African reality suggests that the present value of early CSG support is 1.5 to 2 times as large as the direct cost of that support. These findings in no way imply that there are not further gains from CSG support later in childhood. Indeed, they suggest that such further gains are quite likely. However, within the confines of this study, we have not been able to estimate their magnitude.

The remainder of this Working Paper is organized as follows. Section 2 provides background description of the South Africa Child Support Grant. Section 3 presents descriptive statistics on the programme, discussing key measures and identification strategies that are available. Section 4 presents the methods and results from using continuous treatment effects. Section 5 attempts to infer the lifetime economic value of the nutritional impact identified in Section 4, and Section 6 concludes the paper.

  1. We thank for their helpful comments and suggestions Eduardo Zepeda and Rafael Ribas of the International Poverty Centre, who served as the internal peer reviewers of this Working Paper, and Tom Kelly, of the Millennium Challenge Account Corporation, Washington D.C., who served as the external peer reviewer. We also thank DfID and USAID for financial support. Please address correspondence to Jorge Agьero at We would also like to thank Laura Schechter, Steven Helfand and seminar participants at the University of Wisconsin-Madison, the American Agricultural Economics Association, the University of Colorado-Denver, the University of Florida-Gainesville, the University of California-Riverside, the Centre for the Study of African Economies, PEGNet and the North American Summer Meetings of the Econometric Society.
  2. University of California, Riverside.
  3. University of Wisconsin, Madison.
  4. University of Cape Town.
  5. For example, see the studies by Subramanian and Deaton (1996), Bouis and Haddad (1992) and Behrman, Rosenzweig, and Foster (1994). Haddad, Alderman, Appleton, Song, and Yohannes (2003) provides a review of the literature over the last 20 years.
  6. Other Latin American cash transfer programmes include the Bolsa Escola programme in Brazil (Arends-Kuenning, Fava, Kassouf, and de Almeida 2005), the Social Protection Network programme in Nicaragua (Maluccio and Flores 2005) and PRAF in Honduras (Glewwe and Olinto 2005).
  7. A number of the key studies on the income elasticity of nutrition drew on data generated as part of the International Food Policy Research Institute’s (IFPRI’s) commercialization of agriculture studies. In most cases, the IFPRI studies concerned communities where technological and other changes enhanced returns to male assets, raising the issue as to whether low estimated nutritional elasticities represented immutable family preferences or simply a re-weighting of child nutrition by unfriendly male preferences (see the discussion in the summary volume by von Braun and Kennedy (1994)).
  8. On the basis of their ex ante analysis, Kakwani, Soares, and Son (2005) suggest that without school attendance conditionalities, cash transfers in Africa will not increase school attendance. They do not, however, speak to the question of the necessity of conditionalities to boost nutrition.
  9. In a recent paper, Araujo and Schady (2006) evaluate the impact of an unconditional cash transfer programme in Ecuador. However, unlike the CSG, programme administrators in Ecuador did emphasize the importance of school enrolment when beneficiaries signed up for the programme, leading some households to believe that enrolment requirements existed.

Octoplus Information Solutions Top of page | Home | Contact SARPN | Disclaimer