Which eligible households get grants? Demographic correlates of receipt in South Africa

ABSTRACT

Since Apartheid, the South African government transformed and expanded the social grants system to improve the well-being of its vulnerable populations. Despite increased efforts, a sub-section of the grant-eligible population is not reached. Too little is known about the factors that contribute to grant receipt, especially for the household as a whole. This article examines the household and community characteristics associated with grant receipt among poor households in KwaZulu-Natal. We add to previous work by assessing grant receipt at the household level, examining receipt of the two major grants and analysing correlates in a multivariate framework. While associations with grant receipt are complex and varied, we find higher grant receipt (especially the Child Support Grant) among more disadvantaged households. We also find that characteristics across multiple domains are needed to best distinguish household grant receipt. We discuss theoretical implications for models of grant receipt and practical implications for improving grant access.

KEYWORDS:

• Social grants
• households
• grant receipt
• poverty
• demographic correlates

JEL CODES:

• I380

1. Introduction

Twenty years after the end of Apartheid, the well-being of South African families is still precarious. Twenty-three per cent of South Africa’s population lives below the national $1.25/day poverty line (World Bank, 2013) and this number rises to 50% in harder-hit provinces such as KwaZulu-Natal (KZN). Moreover, extreme poverty is concentrated among Black South Africans and differentially affects children. Sixty-three per cent of Black children live in ‘ultra-poor’ households earning less than R635 per month, while only 2% of white children live in comparable households (Hall, 2014a). In response, the South African government transformed and expanded social grants to bolster the well-being of South Africa’s most vulnerable populations. Research suggests social grants are effective at relieving poverty (Case et al., 2005; Agüero et al., 2006; DSD et al., 2011). However, they do not currently reach all intended beneficiaries. This article examines the characteristics associated with grant receipt among poor households in KZN. We seek to answer key policy and practice questions about how best to reach beneficiaries and theoretical questions about characteristics correlated with grant receipt.

1.1. Social grants in South Africa

Since the enactment of the Social Assistance Act of 1992, a policy priority has been the direct transfer of revenue to vulnerable groups outside the labour force, particularly children and older adults living in poverty, adults and children with disabilities, and children needing care due to parental illness, death, abuse or neglect. Five social grants (two targeted to adults and three to children) comprise the major share of the social grants system: the Old Age Grant¹ (OAG); the Disability Grant; the Child Support Grant (CSG); the Foster Care Grant; and the Care Dependency Grant. However, this article will focus on the two largest and most widely received grants, the OAG and the CSG. Eligibility requirements and monthly values of these grants are presented in Table 1.

Table 1. Rates of grant receipt in sample and grant eligibility criteria.

	Grant receipt in sample		Grant eligibility criteria
	Total eligible, N	Total receiving, N (%)	Age eligibility	Single maximum annual earnings	Married maximum combined annual earnings	Additional requirements/notes	Value of monthly payment
Old Age Grant	839	686 (81.7)	Age 60+	R44 880	R89 760	–	R1 140 (age 60 to 74); R1 160 (age 75 and up)
Child Support Grant	6 266	4 311 (64.9)	Age 0 to 18	R31 200	R62 400	May be received on behalf of up to six children per caregiver	R260

Note: Because a child may receive only one grant at any given time, the total eligible for child grants accounts for children receiving a grant and is thus ineligible for any other grant.

In terms of beneficiaries, the CSG and the OAG are the largest of the five primary grants, reaching over 11 million young South Africans and almost three million elderly South Africans, respectively (DSD et al., 2012; SASSA, 2013). While age and means tested, these social grants are not conditioned upon compliance with service use and are direct cash transfers rather than in-kind investments. This is similar to unconditional social welfare programmes offered in other Southern African countries (e.g. Mozambique, Namibia), but differs from others that condition receipt on desired behaviours (Richter, 2010).

1.2. Who receives social grants?

Social grants can reduce poverty and improve physical health and cognitive and emotional outcomes for vulnerable South Africans, particularly children (Woolard, 2003; Case et al., 2005; Agüero et al., 2006; Armstrong & Burger, 2009; Richter, 2010; DSD et al., 2012). Recent research also indicates that social grants reduce HIV risk, particularly for children and adolescent girls (Cluver et al., 2013). However, although estimates vary, around 20% of eligible older adults do not receive the OAG (Samson et al., 2004) and about 25% of eligible children do not receive the CSG (Hall, 2014b). Thus, it is vital for research to examine the factors that account for variation in grant receipt and determine the characteristics that differentiate recipients from eligible non-recipients. This work is needed to expand social grant receipt to all eligible beneficiaries and maximise the utility of social grants as a poverty-reduction tool (Hernanz et al., 2004). The work is also needed to develop more coherent models of factors contributing to grant receipt and enable more inclusive empirical modelling of selection processes in future research.

The take-up of social assistance has received relatively little attention in both academic research and policy analysis (Hernanz et al., 2004). Early studies in South Africa focused on the most widely taken-up grant, the CSG, and identified administrative challenges (staffing and resource issues, problems accessing vital documents, misinformation) as barriers to grant receipt (Case et al., 2005; Eyal & Woolard, 2011). While attempts have been made to address these barriers, the Department of Social Development still cites lack of documentation as the biggest barrier to CSG receipt (Tiberti et al., 2013).

Other, largely descriptive, research in South Africa has sought to determine the demographic characteristics that distinguish CSG recipients from (presumably) eligible non-recipients. Delany et al. (2008) examined differences in CSG recipients versus presumed income-eligible non-recipients. Studies have also examined the demographic correlates of early receipt of the CSG (under age seven; Case et al., 2005; Zembe-Mkabile et al., 2012) and those of early application to the CSG (DSD et al., 2012) in selected provinces. Across all four studies, more disadvantaged households (in rural and informal settlements with higher household sizes and lower household resources) were more likely to receive the CSG. However, results diverged in terms of caregiver education and employment, with higher education and employment associated with greater overall receipt of CSG (Delany et al., 2008) and higher early application for the CSG (DSD et al., 2012), but lower receipt of the CSG for children ages under seven (Case et al., 2005). Take-up was also higher for a variety of other characteristics, including mother/father presence in the home and possession of vital documents (Case et al., 2005; Delany et al., 2008; DSD et al., 2012; Zembe-Mkabile et al., 2012).

Little recent research has examined the correlates of OAG receipt versus non-receipt; however, a study conducted in 1998 found that, similar to the characteristics that predict CSG receipt, more disadvantaged households were more likely to receive an OAG. Specifically, households with more members and with lower total incomes (excluding OAG income) were likely to receive more OAGs, and households with more young people (children aged zero to five, aged six to 15 and aged 16 to 24) were likely to receive fewer OAGs (Case & Deaton, 1998).

1.3. Gaps in the literature and the current study

Current studies of the factors distinguishing recipients from non-recipients are limited in several ways. First, they do not focus on grant receipt for the household as a whole. Recent work suggests that many households pool resources, particularly grant income, across members (Becker, 1993; Duflo, 2003; Case et al., 2005; Adato & Bassett, 2009) and that resource pooling needs to be taken into account when designing social grants in South Africa (Whitworth & Wilkinson, 2013). We address this limitation by examining receipt of the OAG and the CSG for the household as a whole. A study by Statistics South Africa (SSA, 2010) is the only prior study to our knowledge that has examined factors correlated with grant receipt at the household level. In this study, national data from the general household survey were used to examine demographic differences between households who received any of the five major grants compared with those that did not. While eligibility was not explicitly modelled, the sample was restricted to low-earnings households assumed to be grant eligible. Households receiving a grant had a significantly higher household size, dependency ratio (ratio of unemployed household members to all household members), unemployment ratio and illiteracy ratio, and a lower ratio of support from outside the household. However, they were also found to have a higher educational attendance ratio. Households receiving any grant also had a higher number of rooms in their dwelling and were less likely to have refuse removal or running water.

The second gap is that the vast majority of studies focus on the CSG, and little is known about what factors correlate with take-up of the OAG, the other major social grant. This is particularly surprising given that the value of the monthly CSG payment is lower than that of the OAG, and thus may be less influential in reducing poverty and improving well-being. Only one study to our knowledge has examined the correlates of receipt for a grant other than the CSG. Jelsma et al. (2008) examined a small number of demographic correlates of Disability Grant receipt among a convenience sample of Xhosa-speaking people with disabilities, finding that those receiving the Disability Grant were more likely to live in rural areas.

A final limitation is that scholars have relied primarily on descriptive statistics or univariate statistics (t tests, chi-square tests) to examine differences between recipients and eligible non-recipients. This approach disregards the fact that many of these factors are themselves interrelated and does not allow one to determine which factors are most associated with grant receipt when taking the others into account. This knowledge is essential to effectively target grants to eligible non-recipients.

By addressing these gaps, the current study seeks to generate usable knowledge about factors that differentiate households receiving grants. We capitalise on data from a community-representative sample of households in KZN. Because KZN is particularly hard-hit by both poverty and HIV/AIDS, we are able to examine correlates of grant receipt in an area of high policy relevance. Because our data were collected after the extension of the OAG to all adults aged 60 and older and the CSG to children aged 18 and under, we are also able to examine correlates of grant receipt under the latest age-eligibility criteria. We pose two research questions:

• In multivariate analyses, what household and community factors are associated with receipt of either the OAG or the CSG at the household level?

• In multivariate analyses, what household and community factors are associated with receipt of the OAG versus the CSG at the household level?

2. Method

2.1. Sample and procedure

Data for this article come from the ‘Sibhekelela izingane zethu’ (SIZE, ‘We Look Out for Our Children’) study conducted by the Human Sciences Research Council and New York University. The goal of the SIZE study is to assess child and family well-being in the context of HIV/AIDS and poverty. The study has two data collection waves, but the current analysis leverages household-level baseline data collected between November 2010 and May 2012.

A total of 1961 households with at least one child aged seven to 10 was sampled from 24 communities in the Msunduzi municipality of KZN province, whose population is 95% Zulu. This province was chosen for its general demographic representativeness of poor rural black South Africa, but it does not necessarily generalise to other grant-eligible households in the country.² The 24 communities were purposively selected such that 12 communities are classified as rural and 12 as peri-urban. Also, 12 communities are classified as having high rates of school matriculation, and 12 are classified as having low rates. Each community includes a school serving children seven to 10 years old, and was demarcated using information about the school’s catchment area, geographic boundaries identified by aerial maps and ethnographic mapping including transport routes to school and interviews with school principals (Van Heerden, 2015).

Aerial mapping was used to enumerate all households within each community. Households were then organised into clusters of 30 around a randomly-selected node and visited in a pre-determined order to find approximately 75 eligible households per community (number of households per community ranged from 58 to 121). Eligible households were defined as those which served as primary residences for at least one child aged seven to 10 years and whose members spoke isiZulu. Following a consent process, a knowledgeable household respondent aged over 18 years was interviewed about the household and its members, resulting in information about 14 157 individuals. Participants were compensated for their time with a food parcel valuing R30. Face-to-face interviews were administered in isiZulu by trained assistants using mobile phones to capture responses. The institutional review boards at New York University and the Human Sciences Research Council approved all study procedures.

2.2. Determining grant eligibility

Grant eligibility is determined by an age cut-off value, a means test and a set of other varying requirements for each grant (see Table 1). Each individual is eligible to receive only one grant at a time (however, caregivers can receive multiple grants on behalf of all biological or legally-adopted children and up to six non-biological children). Because of limited information in this data-set and precedent set by previous work (Booysen, 2004; Barnes & Noble, 2006), grant eligibility in this study was determined by age. Because our sample was selected for its high rates of poverty, we assumed that the vast majority of individuals in sampled households were income eligible. We verified this by asking respondents to indicate annual earnings for each household member. For the first half of data collection, the top earnings category included values both below and above the threshold for the most stringent means test. However, more detailed categories were included in the second half of data collection that distinguished eligibility for grants. In total, 0.06% of household members had earnings above the threshold for the most stringent means test. Because data collection was rolled out randomly, it is plausible to assume that 99.4% of household members in our entire sample are income eligible for grants. We therefore included all households in our analyses. However, we also re-ran all analyses excluding households that had at least one household member reporting earnings in the highest earnings bracket (using the less detailed and therefore more conservative categories available for the full sample). The results were substantively identical to the results presented in the following, with no differences above chance.

2.3. Measures

Items used in this study were adapted from larger national surveys (e.g. the National Income Dynamics Study) which have been cross-culturally validated and are reliable for use with native isiZulu speakers.

2.3.1. Outcomes

Primary respondents were asked whether each household member currently received the OAG or the CSG, resulting in a set of dichotomous indicators. These indicators were used to create the proportion of eligible household members receiving either the CSG or the OAG and the proportion of those receiving the CSG or the OAG individually. Table 1 presents the rates of grant receipt in the sample and Table 2 presents the sample characteristics (detailed further in the following).

Table 2. Sample characteristics.

Characteristic	N (%)^a	Mean (standard deviation)
Household characteristics
Household structure
Household head female	768 (39.2)
Proportion of female adults		0.30 (0.16)
Proportion of married adults		0.21 (0.31)
Average household age		24.12 (5.88)
Total household size		7.00 (3.1)
Dependency ratio		0.47 (0.15)
Proportion of highly-resident adults		0.94 (0.11)
Total children in the household		3.33 (1.77)
Mother absent	555 (28.3)
Economic characteristics
Household education level (average attainment of members)^b		4.55 (1.04)
Asset index (in quintiles)		3.00 (1.42)
Proportion employed adults		0.38 (0.28)
Household income (Rand/month)		14.52 (18.48)
Health
HIV-affected household	496 (25.3)
Proportion of deaths in household		0.30 (0.16)
Proportion of ill household members		0.18 (0.18)
Grant barriers^c
Any grant application problems	502 (25.6)
Any grant access problems	365 (18.6)
Community characteristics
In urban community	938 (47.8)
In high matriculation Community	963 (49.1)

^aPercentages calculated out of 1961 households in the sample.

^bAverage household education ranges from 0 = ‘no formal education’ to 5 = ‘Matriculation’: 4.5 is equivalent to ‘completed some secondary school’.

^cIndicates whether any barriers were experienced by the household as a whole.

2.3.2. Correlates

Correlates included household characteristics in four domains: household structure; socio-economic characteristics; household health; and barriers to grant access and application. The household structure domain included household head gender (1 = female), the proportion of female adults in the household, the proportion of married adults in the household, average household age, total household size, dependency ratio (proportion of household members of working age to total household members) and the proportion of highly-resident adults in the household (defined as spending 75% or more nights in the household in the past six months).

The socio-economic characteristics domain consisted of average education level of adult household members, a household-level asset index (created using a principal components index comprised of consumer durable indicators [Filmer & Scott, 2012] and categorised into asset quintiles) and the proportion of employed adults in the household.

The household health domain included whether the household was affected by HIV/AIDS (household member suffering from HIV or HIV-related death in the last two years), the proportion of deaths in the household and the proportion of ill household members.

The grant barriers domain consisted of two variables: whether or not any members of the household experienced any problems accessing and applying for grants.³

Using data from the 2001 South African Census and educational records, we also examined two community characteristics: urbanicity (rural/peri-urban) and matriculation rate (high/low).

2.4. Analytic plan

All analyses use the proportion of eligible household members receiving grants as the outcome variable. By doing so, we allow for more variation in patterns of grant receipt at the household level than a strict yes/no indicator would permit. Moreover, we account for the fact that households with a greater number of grant-eligible household members are more likely to receive a grant. Thus, our analyses indicate which factors are associated with a household having more of its eligible members receiving grants. All analyses employ multivariate techniques to determine which factors correlate most highly with the proportion of grant receipt while taking other factors into account. We ran a set of three multiple regression analyses predicting the proportion of eligible household members receiving the OAG, the proportion of eligible household members receiving the CSG and the proportion of household members receiving either the OAG or the CSG. Robust standard errors were used to correct for the clustering of households within communities.

3. Results

The results of regression models predicting grant receipt are summarised in Table 3.

Table 3. Household and community predictors of grant receipt.

	Either grant			Old Age Grant			Child Support Grant
	b	SE	β	b	SE	β	b	SE	β
Household structure
Household head female	−0.003	0.009	−0.009	0.097**	0.035	0.157	−0.030	0.020	−0.041
Proportion of female adults	−0.021	0.027	−0.018	0.103	0.087	0.053	0.245**	0.063	0.112
Proportion of married adults	−0.077**	0.018	−0.122	−0.046	0.065	−0.038	−0.199**	0.034	−0.176
Average household age	−0.001	0.001	−0.036	−0.004	0.003	−0.080	−0.002	0.002	−0.026
Total household size	−0.0003	0.001	−0.004	0.0001	0.004	0.001	0.001	0.003	0.007
Dependency ratio	0.572**	0.033	0.440	0.015	0.128	0.008	0.257**	0.069	0.110
Proportion of highly-resident adults	0.052	0.032	0.030	−0.105	0.090	−0.044	0.083	0.072	0.027
Economic characteristics
Average household education	−0.023**	0.005	−0.121	−0.027	0.017	−0.088	−0.026**	0.010	−0.077
HIV-affected household	−0.0002	0.009	−0.0009	−0.018	0.031	−0.028	0.007	0.020	0.009
Asset index (in quintiles)	−0.012**	0.003	−0.088	0.019*	0.009	0.089	−0.025**	0.007	−0.101
Proportion of employed adults	−0.053**	0.016	−0.076	−0.129*	0.055	−0.106	−0.003	0.031	−0.002
Health
Proportion of deaths in household	−0.113^†	0.062	−0.040	0.086	0.164	0.018	−0.221^†	0.125	−0.044
Proportion of ill household members	0.073**	0.026	0.066	0.118^†	0.066	0.072	0.152**	0.055	0.077
Grant barriers
Any grant application problems	0.019*	0.009	0.043	−0.066*	0.027	−0.105	0.044*	0.018	0.054
Any grant access problems	−0.010	0.010	−0.021	0.039	0.030	0.055	−0.042*	0.021	−0.046
Community characteristics
Urban community	−0.031**	0.008	−0.079	−0.020	0.024	−0.034	−0.068**	0.016	−0.096
High matriculation community	−0.007	0.007	−0.018	−0.032	0.022	−0.057	−0.014	0.015	−0.020
R^2	0.326			0.074			0.092

Notes: b = unstandardized regression coefficient; β = standardised regression coefficient; SE = standard error.

^†p < 0.10. *p < 0.05. **p < 0.01.

3.1. Receipt of either the OAG or the CSG

Our set of correlates explained 32% of the variance in the proportion of all household members receiving either the OAG or the CSG. The pattern of significant correlates suggested that more disadvantaged households receive proportionally more grants. In particular, households with a higher proportion of eligible members receiving the OAG or the CSG had a more disadvantaged household structure (lower proportion of married adults and a higher dependency ratio); were more socio-economically disadvantaged (lower average household education, lower asset index, lower proportion of employed adults); had worse health (higher proportion of ill household members); were located in rural areas, which typically have less developed infrastructure and lower access to transportation; and reported experiencing at least one problem applying for grants. Because we operationalised the outcome as the proportion of all eligible household members receiving the OAG or the CSG, these associations are not simply due to households with these characteristics having more grant-eligible household members. Rather, these households have more of their eligible members actually receiving these grants. These associations were estimated using multivariate regressions, and thus they can be interpreted as the unique association between each factor and grant receipt after adjusting for all other factors in the model. The dependency ratio had the strongest relationship to grants (standardised regression coefficient β = 0.44), indicating that a major driver of higher grant receipt is having more children and older persons in the household. This could be because vulnerable household members are drawn to households with more grant receipt (Klasen & Woolard, 2009).

3.2. Receipt of each grant

3.2.1. Old Age Grant

Our set of correlates explained only 7% of the variance in OAG receipt and only four factors were significant. Households with a female household head and a lower proportion of employed adults, but a higher asset index and no reported problems applying for grants, had more of their eligible members receiving the OAG. These findings suggest a mixed pattern of advantage (higher assets) and disadvantage (female-headed household and lower proportion of employed adults). Of the significant correlates, female household head had the strongest relationship with grant receipt (β = 0.16), suggesting that there is something about the structure or nature of female-headed households that leads older members to seek out and receive the OAG at a higher rate than otherwise similar households.

3.2.3 Child Support Grant

Nine household and community factors were significantly associated with receipt of the CSG, but these accounted for only 9% of the variance. Households with a higher proportion of eligible members receiving the CSG had a more disadvantaged structure (higher proportion of female adults, lower proportion of married adults, higher dependency ratio); were more socio-economically disadvantaged (lower average household education, lower asset index); had a higher proportion of ill household members; were located in a rural area; and reported experiencing at least one problem applying for grants (but no problems accessing grants). The proportion of married adults emerged as the strongest correlate of CSG receipt (β = –0.18). Thus, households characterised by more overall disadvantage than other eligible households have members who seek out the CSG at a higher rate.

4. Discussion

Previous investigations into the correlates of grant receipt in South Africa have typically examined receipt by individual household members, focused largely on the CSG and utilised univariate tests of statistical significance. The current study investigated grant receipt at the household level and examined factors that distinguish household receipt of either the OAG or the CSG, as well as the OAG and the CSG individually, among a sample of poor households in KZN. We also examined correlates of grant receipt in a multivariate framework, thereby accounting for the inter-correlations between characteristics and assessing which factors are most related to grant receipt. In so doing, we sought to provide critical information to effectively expand outreach efforts to target social grants to all eligible beneficiaries and enable more accurate empirical modelling of selection processes in future research.

We first examined the characteristics associated with the proportion of household members receiving either the OAG or the CSG. The pattern of significant associations suggests that households which have a more disadvantaged household structure, are most socio-economically disadvantaged, have poorer overall health and are in rural areas have a significantly higher proportion of household members receiving grants. Thus, among a sample of poor households, what distinguishes those receiving more grants is comparatively more disadvantage across these domains. Because grant receipt was operationalised as the proportion of eligible household members receiving grants, it is not that the most disadvantaged households simply have more eligible household members. Rather, eligible members in these households receive grants at a higher rate than eligible members in other households.⁴ Overall, our analysis provides important corroborating evidence that South Africa’s social grants system is largely reaching the most disadvantaged households. While this is no doubt positive, it is important to remember that our entire sample is living in poverty and is income eligible to receive social grants. Our results indicate a need to ensure that all disadvantaged households are informed about social grants and can access them. This is especially pressing considering the demonstrated benefits of grant receipt for well-being and HIV-risk reduction. In addition, given the cross-sectional nature of our data, the fact that members of the most disadvantaged households receive more grants could indicate that grant values are not currently high enough to lift households into a better economic state. However, it could also be that these households have higher income from grant receipt and therefore attract more vulnerable household members who could have lived somewhere else (Klasen & Woolard, 2009).

The pattern of significant associations for the CSG also suggested a story of disadvantage. Households with more female adults, fewer married adults, higher dependency ratio, lower average education and assets, and more ill household members received proportionally more CSGs in our multivariate framework. This is consistent with previous work suggesting that comparatively more disadvantage is related to higher CSG receipt (Case et al., 2005; Delany et al., 2008; DSD et al., 2012; Zembe-Mkabile et al., 2012). However, it also provides additional evidence that lower education is associated with greater CSG receipt, while controlling for a large number of potentially confounding covariates. Second, the strength of associations suggests that the proportion of married adults in the household is a key factor distinguishing households in which more eligible members receive the CSG.

Interestingly, the constellation of significant correlates that predicted OAG receipt were somewhat dissimilar from the correlates of CSG receipt, particularly in the sense that more households with more (rather than less) assets had a higher proportion of their eligible members receiving the OAG. This is in contrast with previous research finding that OAG receipt was linked to lower household income (Case & Deaton, 1998). However, because this association is correlational and the monthly monetary value of the OAG is high, it could simply be that household assets are higher in households where proportionally more members receive the OAG. In addition, having a female household head emerged as the biggest predictor of OAG receipt, followed by a lower proportion of employed adults and fewer problems accessing the grant. Taken together, this pattern of significant associations suggests that households headed by females (perhaps grandmothers) with fewer employed members have more of their eligible members receive the OAG, and that, in general, these households experienced fewer problems accessing grants, at least compared with the CSG. These findings are interesting in that they suggest these two grants have their own set of unique associations with household and community characteristics, probably due to do their distinctive features. That the significant predictors of receiving either the CSG or the OAG largely mirror the predictors of the CSG and not the OAG is also interesting, and is evidence for the greater prevalence of CSG receipt in these households.

Our findings have a number of implications for practice and future theoretical and empirical work. First, our results suggest that the CSG reaches the most disadvantaged households, but not eligible households with only slightly less disadvantage. In particular, households with a higher dependency ratio (more children and older adults) had proportionally more eligible members receiving the CSG or the OAG, or the CSG alone. Why would this be? One reason could be economies of scale – the marginal costs of applying for (and picking up) grants for additional household members are lower once you have already done so for one household member. Theoretically, this implies households make complex cost–benefit analyses in deciding which grants to apply for. While this has been acknowledged in literature on benefit receipt in high-income countries (Currie, 2004; Hernanz et al., 2004), it has received less focus in low and middle-income countries. Practically, outreach efforts should target households with fewer children and older people to inform them of the benefits of applying for grants. For the CSG, households with more married adults could be targeted for additional outreach, and the earnings eligibility threshold for married caregivers could be reconsidered.

Second, the findings suggest that households receiving the OAG might be marginally less disadvantaged than households receiving the CSG. Although households with proportionally higher OAG receipt had a lower proportion of employed adults and were more likely to have a female household head, they were more advantaged in other ways. They had higher assets (although this could be due to the higher monetary value of the OAG versus the CSG) and did not suffer as much disadvantage in their household structure, educational characteristics or health. The OAG may be reaching more of its eligible beneficiaries (even those who are slightly better off) because it is easier to access than the CSG. Indeed, the maximum earnings cap is 30% higher for the OAG than the CSG. Even though all our household members are earnings eligible for the CSG, the more relaxed earnings thresholds of the OAG might make it feel more accessible and ubiquitous. The monetary value of the OAG is also considerably higher than the CSG. Again, the better-off households in our sample may be making strategic cost–benefit analyses regarding grants and decide that the OAG is worth pursuing while the CSG is not (Currie, 2004; Hernanz et al., 2004).

Third, our results also suggest that efforts to improve access to each grant need to be grant specific. By analysing correlates of grant receipt in a multivariate framework, we were able to identify the most influential correlate for receipt of each type of grant. Our analysis suggests that efforts to improve take-up of the OAG should target male-headed households, perhaps through informational campaigns directed towards men. On the other hand, attempts to improve access to the CSG would be better focused on targeting households with relatively less disadvantage. Efforts could also target households with more married adults and consider changing the earnings eligibility threshold for married caregivers.

Fourth, a notable finding from our analyses is that the correlates most highly associated with grant receipt in a multivariate framework were not limited to economic characteristics, but fell across multiple domains including household structure, household health, barriers to grants and community characteristics. Most previous research has focused on the economic targeting of social grants and on supply-side barriers to grant receipt. Our analysis suggests instead that, among a sample of poor households, a broader array of household and community characteristics are needed to differentiate recipients from non-recipients. Future work should focus greater attention on these other characteristics. In addition, practical efforts to improve grant receipt could target households with similar constellations of structural, economic, health and community features. Indeed, research in high-income countries suggests that individuals take up benefits more often when similar others receive those (Aizer & Currie, 2004). Efforts to target households which share similar structural characteristics could be a way to increase the perception that certain grants are for households like ‘theirs’.

Finally, it is important to note that the proportion of variance our correlates explained varied across models. While they explained a good amount in receipt of either grant, they explained much less in receipt of the OAG and the CSG individually. Therefore, other aspects of households and communities may also be influential in predicting grant receipt. Indeed, research on the CSG suggests possession of the child’s birth certificate and presence of the father in the community are also associated with CSG receipt (DSD et al., 2012). In addition, research shows that people living in poverty are more risk averse and less likely to forego current monetary rewards for higher future rewards (Haushofer & Fehr, 2014). Thus, people may not be willing to spend money on transport to the granting office or to obtain documentation, even if it means receiving grant money later on. What implications the psychological processes associated with poverty have for grant take-up have yet to be considered.

We must acknowledge several limitations of this study. First, this is a cross-sectional examination of the correlates of grant receipt. While we believe many characteristics are unlikely to be influenced by grant receipt, some – especially household assets – may be. Thus, we can only assert that these characteristics are associated with grant receipt, and causal inferences should be avoided. Future research leveraging longitudinal data is needed to clarify the direction of associations. Second, we examined a wide range of household and community demographics, but there are probably other important factors distinguishing grant recipients from non-recipients that were not considered here. Finally, our sample is limited to Zulu households in the Msunduzi municipality in KZN. While the households in our study represent a key demographic target for social grants, the extent to which these results are generalisable to other areas is unknown. While the households in our study represent a key demographic target for social grants, they do not necessarily generalise to other grant-eligible households across the country.

Despite two decades of implementing progressive social grant benefits to improve the well-being of vulnerable populations in South Africa, the legacy of Apartheid remains a threat. Given the number of eligible individuals failing to receive the grants to which they are entitled, the goal of the current research was to determine what household and community factors are associated with grant receipt. We add to previous work by conceptualising grant receipt at the household level, examining receipt of the two major grants both simultaneously and individually and analysing correlates in a multivariate framework. Our findings suggest new avenues for improving grant take-up, including increased focus on the disadvantaged (as opposed to most disadvantaged) households, tailoring theoretical accounts of grant receipt and targeting efforts to specific grants, and leveraging information about a variety of household characteristics to better understand patterns of grant receipt.

Acknowledgements

The authors would like to especially thank Lucie Cluver for helpful comments on earlier versions of this article and the families for taking part in the data collection and sharing their experiences.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Lucia C Knight http://orcid.org/0000-0001-9938-6887

Additional information

Funding

This research was supported by the Eunice Kennedy Shriver National Institute for Child Health and Human Development [grant number 5R01HD055137] (PI: JL Aber).

Notes

1 Also known as the Grant for Older Persons, the Older Person Grant or the Old Age Pension.

2 By design, our sample is 100% Zulu and rural/peri-urban. Because of data limitations, it is difficult to ascertain how our sample compares with a national sample of grant-eligible households. However, we were able to compare some characteristics of our sample with households from the National Income Dynamics Study (NIDS) 2012, which we assumed to be grant-eligible based on income eligibility criteria. Reports of household earnings in the past month were comparable (NIDS, R1743; SIZE, R1452), as were the number of households living in a house or brick structure (NIDS, 62%; SIZE, 71%). However, household members had slightly higher education, on average, than NIDS households (‘completed some secondary school’ compared with ‘completed some primary school’).

3 The majority of households reported that their members experienced no barriers (see Table A1 in Appendix A).

4 To further explore this finding, we examined whether rates of grant receipt were indeed higher in households in the lower asset quintiles. Results indicated that grant receipt was higher in lower asset quintiles, but only moderately so. This suggests that – at least as it relates to grant receipt – household disadvantage is best conceptualised across multiple domains and includes household structure, household health and location as well as economic resources, as we have done here.

Appendix A

Table A1. Reported barriers to social grant receipt.

	N (%)
Difficulties in applying for a social grant
None	1 455 (74.2)
Administrative problems	98 (5.0)
Misinformed about necessary documentation	97 (4.9)
Application process too complicated or time consuming	119 (6.1)
Staff disrespectful and unsympathetic	119 (6.1)
Opening times not convenient	8 (0.4)
Long queues	271 (13.8)
Staff try to get bribes	50 (2.5)
Other	7 (0.4)
Difficulties accessing a social grant
None	1 596 (81.3)
Lack of documentation	104 (5.3)
Shortage of money for transport	207 (10.6)
Applicant too sick/disabled to apply	12 (0.6)
Cost of getting grant too high	7 (0.4)
Stigma/embarrassment	12 (0.6)
Too difficult to get to office	7 (0.4)
Involves too much time away from work or home	68 (3.5)
Other	1 (0.1)

Note: The primary household respondent was asked whether any member of the household had experienced any of these difficulties. Percentages are calculated out of 1961 households in the sample.