The performance of the Child Support Grant: Review and research priorities

Abstract

This article reports the findings of a review on the performance of the Child Support Grant (CSG) programme. It suggests that the CSG impact theory is sound, the programme has in general been implemented well and the limited research on impact is suggestive of it achieving its ultimate objectives of reducing child deprivation and promoting human capital development. It points to the vastness of the child poverty that remains to be addressed and raises the concern that in spite of young children being prioritised in the roll out of the CSG, child poverty incidence may be highest in children aged zero to four. Weaknesses in programme implementation are identified, as well as research priorities in three areas: the child poverty profile, CSG programme implementation and CSG impact.

Keywords:

• child poverty
• Child Support Grant
• child social assistance
• unconditional cash transfer
• child development

1. Introduction

The Child Support Grant (CSG) was introduced on 1 April 1998 as the primary social assistance measure in a package designed to enhance support for poor children. It was implemented following the recommendations of the Lund Committee, which was appointed in 1995 to review child social assistance in South Africa. The context surrounding the programme's introduction was one of optimism and urgency with respect to expanding support for poor children. There was strong advocacy from civil society for stepped-up spending to advance child rights, and promises in this regard were being written into social welfare policy and the Constitution. Fiscal austerity was also a feature of the policy environment and this shaped the CSG. The slogan from Cabinet for those working on policy issues was ‘reform, but reform within the existing envelope’ (Lund, 2008:30). The fiscal context explains why expanding the State Maintenance Grant was not seen as an option and it was phased out when the CSG was introduced.

This article presents a review that had the objectives of summarising findings from the existing research on the CSG programme's performance and identifying research priorities. Undertaken in 2010, the review applied the Rossi et al (2004) systematic approach for tailoring a social programme evaluation and was part of a larger research study that simultaneously analysed South Africa's child poverty profile. Whilst the review does not offer any new evidence on CSG implementation or impact, it still makes a useful contribution. By using the Rossi et al. (2004) approach it draws together the evidence base in a way that has not been done before and that offers a more comprehensive picture of the programme's performance than offered to date. The research priority offering is also unique.

Research on the CSG programme's performance remains important for three reasons. First, to provide knowledge about the value derived from the large amounts of resources allocated to the programme. Second, to provide information that can be used enhance the efficiency of child social assistance, which is all the more pertinent in light of the extent and depth of child poverty that remains in South Africa and the negative implications associated with failing to address child poverty, especially in the very early years. Third, to offer lessons to policy-makers in other countries about the power (and limitations) of the child-targeted cash transfer measure of social assistance. This motive needs to be understood in the context of the CSG being the largest cash transfer programme targeted at children in the world and being an unconditional programme, at least up until 1 January 2011.²

2. Background

2.1 Description of the Child Support Grant

Describing the theory of the programme is a critical first step when evaluating a social programme (Rossi et al., 2004). No policy document sets out the CSG theory. For the purposes of the review it was pieced together based on conversations with programme stakeholders and descriptions in the literature. Figure 1 presents the CSG programme impact theory.

Figure 1: CSG programme impact theory

A programme's organisational plan is articulated from the perspective of programme administrators and encompasses ‘the functions and activities it is expected to perform and the human, financial and physical resources required for that performance’ (Rossi et al., 2004:142). Central to it is the nature of the services it is supposed to deliver to which target population in order to produce the intended social outcome(s). The aim of the CSG programme from the perspective of programme administrators is to pay primary caregivers of children who qualify a cash transfer of the value set at that time on a monthly basis. Initially the benefit was R100, in April 2010 it was R220. At the time of writing it is R250. Annually, after April 2002 the CSG benefit increased on an annual basis guided by CPI (Consumer Price Index) inflation. In real terms, the value of the benefit increased markedly over 1998–2010. (Siebrits & Van der Berg, 2010).

Throughout the programme's history the aim has been to pay the benefit to primary caregivers (not biological parents) of targeted children. This design feature was informed by the fluid living arrangements of many children living in poor communities in South Africa. The initial child aged of eligibility for the benefit was zero to six years, but this cut-off increased over time as shown in Table 1.

Table 1: Child age of eligibility for the CSG

Date	Child age of eligibility
1 April 1998	0 to 6 (younger than seven)
1 April 2003	0 to 9
1 April 2004	0 to 11
1 April 2005	0 to 14
1 January 2009	0 to 15
1 January 2010	0 to 16
1 January 2011	0 to 17
1 January 2012	0 to 18

The programme was introduced with the following two-step selection procedure for targeting. First, certain categories of people were included by the use of three measures, namely: rural location; non-formal housing (both squatter shacks and traditional housing); and total income below a certain cut-off point (set at R9000 per year or R800 per month in 1995). Second, of those excluded (this should be read ‘included’) those above a certain level of means were again excluded (this level was set at R13 200 total income per year or R1100 per month in 1995). Late in 2008 the targeting mechanism was simplified and the means test adjusted upwards. The qualifying criteria became that the primary caregiver and his/her spouse's income be less than 10 times the current value of the grant (Peters & Williams, 2009). Supplying proof of identity of caregiver and child is another aspect of the eligibility requirements. Initially a birth certificate or identify document was required, but in May 2008 the government announced that affidavits would be accepted as alternative identification (Peters & Williams, 2009).

2.2 Social assistance and child poverty context

The CSG was added to an existing array of social assistance programmes that were advanced by developing country standards and whose origins may be traced to the attempts to create a welfare state for whites in the apartheid era (Devereux, 2007). The main programmes in existence were the State Maintenance Grant, the pension, the War Veterans Grant, the Disability Grant, the Grant-in-Aid and the Foster Care Grant. The pension was the largest in beneficiary and benefit value terms and constituted the lion's share of social assistance spending (Van der Berg, 2002).

The period following the CSG's introduction, especially after 2000, saw rapid growth in social assistance beneficiaries and spending. This was driven by the CSG but other programmes, like the foster care grant, also saw rapid growth. By the late 2000s the CSG had become the largest programme in beneficiary number terms, and in April 2009 its beneficiaries constituted just over 65% of grant beneficiaries. Due to the relatively small magnitude of its benefit the CSG did not dominate social assistance spending by the end of the 2000s. However, its share of the budget had risen substantially and whereas in 2006/07 the CSG programmes share of the social assistance budget was 27%, by 2008/09 it was 31% (Siebrits & Van der Berg, 2010). Spending on grants as well as the share of the CSG in grant spending continued to rise after 2008/09. The 2010/11 budget provided for social assistance expenditure of R89 368 million, of which 38.1% was for pensions and 34.5% for child support grants (Siebrits & Van der Berg, 2010). Table 2 shows social assistance beneficiaries for the programmes in April 1998, 2003 and 2009, and Table 3 expenditure on them in 2006/07 and 2008/9. By early 2011, social grant expenditure had grown to around 3.5% of GDP (National Treasury, 2012).

Table 2: Cash transfer beneficiaries in South Africa: April 1998, 2003 and 2009

Grant	April 1998	April 2003	April 2009
State old age pension	1 697 725	2 009 419	2 324 615
War Veterans Grant	10 525	4 594	1 649
Disability Grant	660 528	953 965	1 404 884
Grant-in-Aid	9 183	12 787	–
Foster Care	43 520	138 763	487 510
Care Dependency	8 172	58 140	105 909
Child Support Grant		2 022 206	9 061 711
Total	2 429 653	5 808 494	13 386 278

Source: Van der Berg et al. (2009) for 1998 and 2003. National Treasury (2009) for 2009.

Table 3: Expenditure on social assistance programmes (Rand), 2006/07–2008/09

Grant	2006/07	2007/08	2008/09
State old age pension	19 527 000 000	21 289 000 000	22 889 023 222
War Veterans Grant	28 000 000	25 000 000	21 846 094
Disability Grant	14 099 000 000	14 261 000 000	15 281 021 331
Foster Care	1 996 000 000	2 851 000 000	3 391 528 413
Care Dependency	916 000 000	1 006 000 000	1 154 757 382
Child Support Grant	14 142 000 000	17 559 000 000	19 625 631 872
Total	50 708 000 000	56 991 000 000	62 363 808 317

Source: Supplied by a SASSA official to the author on 30 October 2009.

Note: SASSA did not supply data for two other grant programmes, which are smaller in budget and beneficiary number terms, namely the social relief of distress and grant in aid programmes.

In addition to the grants a range of other anti-poverty interventions aimed at promoting human capital development (e.g. provision of education and health services) and reducing deprivation/meeting basic needs (e.g. housing, water and electricity subsidies) were in place during the period 1998–2010 and remain today (Siebrits & Van der Berg, 2010). The array of social assistance measures that exist alongside the CSG makes assessing its impact all the more difficult.

When the CSG was introduced the details of the child poverty profile were just beginning to be drawn based on data gathered in the poverty hearings and the first household surveys that produced demographic and socio-economic data on the entire population. The results of the child poverty measurement studies based on the 1993 SALDRU household survey data, 2000 and 2006 October Household Surveys and 1996 Census, and which used the traditional measurement approach, produced different estimates of child poverty incidence depending on the assumptions made about intra-household resource allocation and costs of different household members in the adult equivalence scale (see Streak et al., 2009). However, they all confirmed that child poverty incidence was very high (see Streak et al., 2009). Moreover, they established the following characteristics of the child poverty profile: concentration of child poverty amongst the African and, to a slightly lesser extent, the coloured child populations; higher child poverty rates in rural than in urban areas; higher child poverty in households without wage income; a high correlation between child poverty and low level of education of the household head; over-representation of households headed by women in the poor child population; and large variation across the nine provinces, with particularly high incidence of child poverty in Eastern Cape and Limpopo (Streak et al., 2009).

A later measurement study that used the 2005/06 Income and Expenditure Survey (IES) (Streak et al 2009) identified another characteristic of the child poverty profile which is particularly important when reflecting on the CSG's performance. It found that child poverty, measured using the traditional approach, income per capita as the indicator and the P₁ incidence, P₂ depth and P₃ severity measures, was highest among children age zero to four. Barnes (2009), using the Community Survey 2007, also found child poverty incidence to be highest amongst children age zero to four.

Problems of incomparability across the available household survey datasets make establishing a trend in child poverty over 1998–2000 difficult. Comparison using, first, General Household Survey (GHS) 2000 and 2007 and, second, IES 2000 and 2005 and money metric indicators suggests a decline (see Streak et al., 2009). For example, using income per capita as the welfare measure and R350 per capita per month as the poverty line, the headcount is 76.8% in 2002 and 67.7% in 2007. Reports of child hunger from the GHS indicate a decline from 2002 to 2008 but increase between 2007and 2008, which is probably reflective of the impact of the international recession on the South African economy (Mabugu et al., 2010).

2.3 Conceptual and methodological considerations

Four conceptual considerations and associated methodological issues informed the analysis.

The first was the heterogeneity of the poor child population targeted by the CSG. More specifically, that within the child population targeted some children are at a particularly high risk of suffering due to poverty and in need of support. The particularly vulnerable groups include: children living in the poorest households; very young children (particularly those aged zero to two); and children without a biological caregiver in their household. The vulnerability of group two has been highlighted by the large body of research that now exists on the effects of malnutrition on child development outcomes and which shows that malnutrition is associated with irreversible brain damage and impaired cognitive and gross motor development (for an evidence overview see Behrman, 2012). The vulnerability of group three has been highlighted by research in African countries, including in South Africa (see for example Case & Ardington, 2006), which suggests that children living in poor households without biological caregivers may be discriminated against in spending on basic goods and schooling.

The second was the multi-dimensional nature of child poverty and the understanding that it is not children's lack of access to income that is the concern and that needs to be addressed via the CSG and other measures of support, but the range of deprivations associated with it (Noble et al., 2004).

The third was the potential for the CSG to have behavioural effects, of which the main ones raised in the literature are: the potential for it to encourage individuals to be less cautious about falling pregnant and increase fertility; the potential for it to affect labour supply; and the potential for it to affect household size/composition via affecting decisions about where to live. With regard to the labour supply effect, two arguments are put forward. One is that it is likely to reduce labour supply by making the price of leisure cheaper in recipient households. The other is that it may increase labour supply by facilitating migration in search of job opportunities. These effects not only point towards important avenues to be explored in the research on CSG impact. They also raise problems of endogeneity for researcher focused on teasing out the CSG's impacts using household survey data and indicators of child outcomes.

The fourth consideration was the problem of lack of knowledge about how resources are allocated inside households, the suggestion of the new models in this regard being that resources may be allocated unequally and who controls resources may matter for outcomes (see Wittenberg, 2001) as well as the empirical findings in this regard. With respect to the latter, the findings from research on the pension (see for example Ardington & Lund, 1995; Case & Deaton, 1998; Duflo, 2003) were particularly informative. These include that it is shared inside households, women spending it in a way that is more favourable for children, and that it has contributed to reducing child deprivation and enhanced access to schooling. The problem of not knowing how resources, including CSG income, are shared inside households and fungibility of money highlights the limitations of studies that have attempted to describe and quantify spending of the CSG transfer based on data gathered in household surveys from recipients on how they spend the transfer. They suggest that the findings of these should be treated with caution. The intra-household allocation problem also implies that studies which use the traditional poverty measurement approach to measure the impact of the CSG on child poverty are likely to produce inadequate results and should be treated with caution. This is because inherent in them are heroic assumptions about intra-household sharing and relative costs of individuals inside households (made in the adult equivalence scale).

These conceptual considerations raise the value of studies that try and measure the impact of the CSG on child deprivation and human capital development using household survey data and direct indictors of child outcomes. Even though use of direct indicators cannot shed light on the black box of intra-household resource allocation and the pathways of behavioural effects associated with the CSG and other transfers, they get around the problem of having to make assumptions regarding the first and take into account impact of behavioural effects.

3. Review method

The questions asked about the CSG's performance are presented in Figure 2. Their design was informed by conversations with stakeholders about the debates and key issues surrounding the programme, and the Rossi et al. (2004) hierarchical framework for organising questions asked in social programme evaluation used as a guide.³

Figure 2: Questions in the review of CSG programme performance

The search strategy considered published and grey literature as some of the higher quality studies have been government commissioned and not published. All research designs were included and methodological weaknesses factored in at the data analysis and synthesis stage.

The studies used in the review were a mix of quantitative (Cassiem & Kgamphe, 2004; Leatt, 2004, 2006; Samson et al., 2008; Case et al., 2005; Noble et al., 2005a, 2005b, 2008; Woolard et al., 2005; Barnes & Noble, 2006; Budlender & Woolard, 2006; de Koker et al., 2006; Makiwane et al., 2006; Wright, 2006; Budlender et al., 2008; Van der Berg et al., 2009), qualitative (Goldblatt et al., 2006; Hunter & Adato, 2007a, 2007b; Surender et al., 2007; Peters & Williams, 2009) and Q-squared method studies (Kola et al., 2000; Vorster et al., 2004; Rosa et al., 2005; Delany et al., 2008). Most provided evidence only on programme implementation but some also on impact. The quantitative and Q-squared method studies used a range of datasets, including: national-level data produced by Statistics South Africa (Census 2001, GHS 2003–2006, IES 2001&2005/06 and 2005/06); nationally representative data gathered by Delany et al. in late 2007; and area-specific data produced by universities and scientific institutions such as the KIDS 1993,1998 and 2004 data gathered by the Africa Centre in Hlabisa KwaZulu Natal and the HSRC South African Social Attitudes Survey. Administrative data supplied to the author by the South African Social Security Agency was also used.

4. Child Support Grant programme performance: Findings from the research

The findings from the analysis for the questions on programme reasonableness are presented first followed by those on programme implementation and impact. Due to word limitations, those on CSG policy/design are not presented.

4.1 Impact theory reasonableness

The primary caregiver design feature of the programme, it was argued, implies that it is reasonable to expect that in most cases the benefit will find its way from recipient to the household in which the child on whose behalf it was received resides.

The following evidence was presented as suggesting that it is reasonable to expect that the transfer flowing into the household will be spent, at least in part, on the child: the evidence of significant consumption effects from child targeted conditional cash transfer programmes in countries such as Mexico and Brazil (see Fizbein & Schady, 2009); the findings from the research on the pension in South Africa that it is shared, spent much like other resources on basic goods and services, and that it reduces child deprivation and facilitates children attending school (see references above); and the finding from the literature on intra-household resource allocation that if women receive a transfer it is likely to be spent on goods that are important for child development (see references above).

At the same time it was pointed out that the research has highlighted the concern that in some cases the effects of the benefit on the targeted child may be diluted by it being used to support many individuals, and targeted children may not benefit equally.

4.2 Programme implementation

The main findings from the studies in relation to the questions on understandings and views about the CSG programme were as follows. In general, there is good knowledge about the grant and how to access it; however, there are a couple of weak knowledge areas including lack of understanding about the use of an affidavit as alternative form of identify in the application process. The CSG is understood as something to help meet basic needs of children as well as those surrounding them in the household. Mostly, people understand that the CSG is available to primary caregivers, not only biological parents. Stigma associated with male receipt of the grant may be a problem. In general feelings, towards the grant are positive but there are negative views. These include that it causes young girls to fall pregnant, facilitates girls migrating to cities and leaving their young children in the care of their grandmothers in rural households and resentment about the small value of the benefit relative to what is needed inside poor households to meet basic needs.

With respect the question on the reach of the CSG programme, the analysis found firstly that the number of CSG recipients and beneficiaries grew very fast and steadily from 2000 to 2010 and by April 2010 there were 5.4 million recipients and 9.4 million child beneficiaries. Secondly, that it is difficult to arrive at accurate estimates of the percentage of eligibles reached due to complications involved in using the available household survey data to calculate eligible primary caregiver and children who are eligible. However, the existing research has made clear that a very high proportion of children are eligible for the benefit, that the percentage of eligible children has increased over time, and that the percentages of eligible children reached, at both the national and provincial levels, increased steadily between 1998 and 2010. Tables 4 and 5 present CSG recipients and Tables 6 and 7 CSG beneficiaries for the years 1999–2010. Table 8 shows the percentages of the beneficiary targets set by the government for end of March 2003 and March 2006 reached.

Table 4: Number of CSG recipients by province, 1999–2004

Province	March 1999	January 2000	November 2001	March 2002	March 2003	April 2004
WC	2 713	6 437	87 285	107 001	168 329	208 934
EC	4 131	26 342	169 955	187 256	294 465	498 828
NC	2 052	6 646	22 660	25 937	37 971	53 839
FS	1 183	7 195	75 862	83 379	126 281	196 192
KZN	5 778	27 982	288 531	291 329	471 412	695 228
NW	1 206	14 375	112 089	120 929	165 838	259 760
Gaut	2 721	24 762	139 654	156 639	259 346	431 548
MPA	539	14 780	85 693	100 465	157 901	265 880
Limp	1 674	21 847	129 883	204 461	317 393	526 867
Total	21 997	150 366	1 111 612	1 277 396	1 998 936	3 137 076

Source: SOCPEN administrative data supplied to author by South African Social Security Agency.

Note: EC = Eastern Cape, FS = Free State, Gaut = Gauteng, Limp = Limpopo, MPA = Mpumalanga, NC = Northern Cape, NW = North West, WC = Western Cape.

Table 5: Number of CSG recipients by province, 2005–2010

Province	April 2005	April 2006	April 2007	April 2008	April 2009	February 2010^a
WC	252 304	288 176	307 016	321 097	348 791	396 763
EC	637 262	765 062	809 246	813 050	850 395	882 981
NC	66 319	75 333	80 763	110 535	120 266	128 415
FS	240 031	276 013	291 338	300 658	308 923	330 303
KZN	803 403	956 707	1 070 784	1 143 960	1 223 708	1 266 920
NW	315 948	371 041	405 552	394 993	406 498	427 703
Gaut	512 488	585 182	624 012	650 818	687 376	739 082
MPA	313 367	362 711	384 194	393 913	410 883	430 263
Limp	620 045	688 355	727 170	745 593	781 541	816 550
Total	3 761 167	4 368 580	4 700 075	4 874 617	5 138 381	5 418 980

Source: SOCPEN administrative data supplied to author by SASSA.

Notes: ^aData have been provided for the month of February not March, which is the month used in most other years, because this was the most recent data available at the time of writing.

Table 6: Number of CSG beneficiaries by province, 1999–2003

Province	March 1999	March 2000	September 2001	May 2002	31 March 2003
EC	5 670	55 717	206 394	277 939	405 815
FS	1 675	13 753	71 240	107 242	150 480
Gaut.	1 872	47 910	149 843	209 399	315 897
KZN	7 853	66 836	352 630	507 302	694 392
Limp.	2 384	53 815	159 989	301 289	456 882
MPA.	630	28 327	102 327	134 172	199 834
NW	1 662	31 792	125 176	166 849	206 421
NC	2 255	12 805	24 824	35 505	46 412
WC	3 576	10 951	89 268	155 962	204 534
Total	27 577	321 906	1 281 691	1 895 659	2 630 826

Source: Cassiem & Kgamphe (2004:2000) based on data supplied by provincial departments of social development.

Table 7: Number of CSG beneficiaries by province, 2004–2010

Province	April 2004	April 2005	April 2006	April 2007	April 2008	April 2009	February 2010
EC	747 838	1 032 201	1 379 325	1 482 450	1 481 128	1 575 528	1 653 020
FS	249 439	328 350	405 321	436 192	454 150	473 141	521 207
Gaut.	558 358	696 991	843 745	912 920	959 139	1 031 785	1 137 998
KZN	1 085 447	1 287 851	1 651 827	1 942 365	2 102 775	2 297 520	2 412 293
Limp.	740 937	955 630	1 175 221	1 249 443	1 274 823	1 364 184	1 443 223
MPA	371 463	468 987	598 787	641 068	657 534	696 614	742 944
NW	351 867	457 953	581 014	648 558	631 499	666 259	715 898
NC	71 703	94 830	116 887	128 658	182 225	202 630	221 052
WC	269 178	344 327	418 437	451 215	474 385	525 014	618 317
Total	4 446 230	5 667 120	7 170 564	7 892 869	8 217 658	8 832 675	9 465 952

Source: SOCPEN administrative data supplied to author by SASSA.

Table 8: Comparison of the number of child beneficiaries reached in March 2003 and March 2006 with the targets set by government

Province	31 March 2003			31 March 2006
	Target	Number reached	% reached	Target	Number reached	% reached
EC	780 000	405 815	52	1 306 204	1 379 325	105.5
FS	300 000	150 480	50	437 302	405 321	92.6
Gaut	90 000	315 897	350	932 345	843 745	90.4
KZN	600 000	694 392	115	1 555 213	1 651 827	106.2
Limp	600 000	456 882	76	1 064 417	1 175 221	110.4
MPA	210 000	199 834	95	506 263	598 787	98.0
NW	330 000	206 421	62	609 649	581 014	95.1
NC	30 000	46 412	154	134 752	116 887	86.7
WC	90 000	204 534	227	380 969	418 437	110.1
Total	3 030 000	2 630 826	86	6 927 114	7 170 564	103.5

Source: Author's calculations based on data supplied by SASSA and Department of Social Development and in the previous tables.

The evidence on the rapid growth in CSG beneficiaries over the period 2000 and 2010 and research findings on favourable targeting outcomes are arguably the most persuasive evidence of the programme being implemented well.

With respect to the characteristics of CSG recipients and beneficiaries, the research confirmed that most recipients have been women. Second, it has shown that in general the programme has been well targeted towards children living in households that lie at the bottom end of South Africa's income distribution and that targeting towards poor children has improved over time. However, whilst is suggests targeting may in general be classified as good, it has established that there are some errors of exclusion (estimated as in the region of 20%) and inclusion errors (around 15%). The research studies have also raised the concern that the following groups may not be sufficiently represented in the target population: children of non-biological primary caregivers; children with male primary caregivers, children in remote rural areas (of which a large proportion fall into the lowest deciles of the income distribution); and young babies. With respect to the latter, a couple of the research studies have suggested that application for the benefit may be delayed by the desire for caregivers to keep newborn children inside for a period following birth.

Only one study (Budlender et al., 2005) was identified with data on CSG means test costs. This showed administration and applicant costs as 11.7 % and 15.6% of the monthly benefit value, respectively. It argued that these costs should not be incurred, particularly by applicants, because they will be a high proportion of monthly income. However, the costing ignored cost savings associated with the income means test as a result of excluding children whose caregivers have incomes above the means test used to target the grant was raised. It was argued that it may be worth keeping the means test, and having the administration and applicants bear the costs associated with it in order to facilitate more resources flowing to support for very poor children instead of spreading them across more relatively poor children.

The main barriers to access that emerged from analysis of the data gathered in the research on CSG applicants' experiences of the application process were: difficulties in navigating the processes required to obtain necessary documents (IDs, birth certificates, affidavits); insufficient funds to finance travel to access officials who process applications; the desire to keep newborn children and their primary caregivers inside (which prevents travelling to apply for the benefit); and illness and poverty causing a level of incapacity amongst eligible primary caregivers, making it impossible to navigate the application process.

The analysis of the research studies on CSG implementation concluded that the main administration weaknesses to be addressed are: slowness in processing identity documents; lack of understanding in the target population and among programme administrators of the policy that allows an affidavit to be used instead of and ID and/or birth certificate in the application process; and lack of effective procedure to facilitate primary caregivers accessing the benefit immediately after the birth.

4.3 Programme impact

4.3.1 Achievement of proximal outcome objectives

The research on the CSG has found that, as anticipated, in most cases⁴ primary caregivers and beneficiaries reside in the same household (see for example Case et al., 2004 ; Woolard et al., 2005; Delany et al., 2008). It was argued that this implies the CSG transfer has, in most cases, been reaching households in which the children on whose behalf recipients receive it live. The findings from the research on targeting outcomes (see above) of small inclusion error and that the households into which the CSG has been transferred are primarily those that lie in the bottom portion (lower three or four deciles) of South Africa's income distribution were cited as suggesting that in most cases these children and caregiver pairs will have been those targeted.

The findings of studies that have estimated the size of the CSG's contribution to household income were also presented as evidence of the CSG achieving its proximal outcome objective one. This includes, for example, the finding of Van der Berg et al. (2009) that CSG income contributes at least 20% of household income in the bottom two deciles and around 10% to the next two deciles. Another kind of evidence that was raised as lending support to achievement of this proximal objective was the explanations of recipients gathered in focus groups about what the CSG means to household resources (see for example Vorster et al., 2004; Hunter & Adato, 2007b; Surender et al., 2007; Delany et al., 2008). The statements below are illustrative:

My wife and I depend on this grant, which we get for our child. I am not working, but we get the paraffin because of her … When I come home at night, even if I have not found work, she is able to put food on the table. (Adult from CSG recipient household, cited in Surender et al., 2007:18)

It does make a difference; we depend on the grant. It's not like nothing comes at the end of the month because we are not working … we are able to buy and sell small things like sweets so that when the CSG runs out before the end of the month, you have something to fall back on, you can buy paraffin and make lunch for the child when going to school. (Adult from CSG recipient household, cited in Surender et al., 2007:18)

With respect to evidence of the CSG achieving its proximal objective two, of increasing spending on basic goods and services for targeted children inside recipient households, it was argued that qualitative research on how the grant is spent (see Hunter & Adato, 2007b; Surender et al., 2007; Delany et al., 2008; Peters & Williams, 2009) are suggestive of success. The qualitative data indicate that the CSG benefit is shared inside recipient households and spent mainly on basic goods (such as food and clothing), basic services (such as fuel and transport) and meeting child education and care costs. The explanations below are illustrative:

I get the grant for two kids … Their father is not working so it supports all of us. (Recipient cited in Delany et al., 2008:34)

I get the grant for two kids. I pay transport, crèche, mealie meal and electricity. (Recipient cited in Delany et al., 2008:34)

We are able to buy food, though it does not last … You can buy meat and make lunch for the child when going to school. (Recipient cited in Surender et al., 2007:18)

The finding from qualitative research (see Vorster et al., 2004; Hunter & Adato, 2007a, 2007b) that it is mainly women who control over spending the CSG benefit was also raised as evidence that the CSG has been spent, at least in part, on poor children inside recipient households.

4.3.2 Achievement of ultimate outcome objectives

A handful of quantitative studies were identified that have estimated CSG impact using direct indicators of child outcomes. The results of these, which are suggestive of it achieving its ultimate objectives, are summarised directly below.

Case et al. (2005) used data collected by the Africa Centre in Hlabisa KwaZulu-Natal in the early 2000s to estimate the impact of CSG receipt on school enrolment among children aged six to eight. They find that ‘children who receive the grant are significantly more likely to be enrolled in school in the years following grant receipt than are equally poor children of the same age’ (Case et al., 2005:467). CSG receipt in 2002 was found to be associated with an 8.1% increase in school enrolment amongst six year olds and a 1.8% percentage point increase among seven year olds. For both six and seven year olds, the association between grant receipt and school enrolment emerges as significant (Case et al., 2005:479). To ascertain whether it was receipt of the CSG that led to this better school enrolment (not some other factor such as particularly committed caregivers), enrolment of older siblings when they were the same age as the children for whom this positive school enrolment effect is found is considered using earlier data collected by the Africa Centre. The authors found no significant association between school enrolment for siblings and the indicator that a maternal sibling would in future receive the CSG. This study therefore provides suggestive evidence of the CSG having the positive effect of enhancing access to schooling, although it may be earlier enrolment facilitated by the grant rather than increased enrolment that it picked up.

Budlender & Woolard (2006) explored the association between CSG receipt and child access to education using cross-tabulations from GHS 2004 and KIDS 2004 and econometric analysis using a micro-simulation model. Their analysis based on GHS 2004 focused on children aged six to eight and that based on KIDS 2004 on children aged seven and eight. Their tabulations showed association between CSG receipt and school enrolment. The size of the significant positive effect shown was small. However, this is to be expected in light of the high enrolment rates that exist in South Africa. These authors also found that enrolment of children who are not direct CSG beneficiaries but live in households of children who are direct CSG beneficiaries was higher.

Woolard et al. (2005)⁵ used state-of-the-art statistical techniques, KIDS 2004 and the child development outcome indicator height for age to estimate the effect of the CSG on child development. Their analysis found that large doses of CSG treatment early in life (within the first 36 months of life) significantly boosted child height. Their analysis is limited to children aged zero to three as this is what the data permit. To measure impact the authors took advantage of the slow programme rollout, which created exogenous variation in the extent of CSG treatment and allows for use of the continuous treatment method of Hirano & Imbens (2004). In their analysis, effects are insignificant for children who receive the transfer for less than 50% of their 36-month window. No gender differences are found in the size of the effect. Their analysis raises once again the importance of reaching children who need support early.

Budlender et al. (2008) explored the effect of the CSG (and pension) on child hunger, illness, and school enrolment using survey data collected by Geospace and HSRC in 2005/06. Their analysis showed a statistically significant impact of the pension on hunger, nutritional outcomes, and illness, but not on school attendance or grade repetition. The analysis for the CSG found an impact on enrolment rates, but no association with reduced reported illness or better nutritional outcomes.

Samson et al. (2008) used the propensity score matching technique and data from GHS 2002 and 2004 to measure the impact of the CSG on child outcomes using hunger and school enrolment as the indicators (pre-school and early grades). Their study found positive statistically significant effects for all indicators.

The downward trend in reports of child hunger in the GHS datasets (see above) was raised as additional suggestive evidence of the CSG programme achieving its objectives of reducing child deprivation and promoting human capital development. Whilst it cannot be asserted with certainty that the faster decline in child hunger in CSG recipient households is due to the inflow of income into poor households associated with the CSG, as Van der Berg et al., (2009) have argued, this does seem likely, particularly when one reflects on the growth of the CSG budget relative to that of other transfers.

4.3.3 Child Support Grant behavioural effects

No evidence was found to confirm the views identified in some qualitative studies that the CSG has increased fertility. One quantitative study (Makiwane et al., 2006) on the association between CSG receipt and fertility was identified that found no association. However, methodological limitations limited the drawing of firm conclusions from it.

The impact of the CSG on labour supply in recipient households emerged as an under-researched area. Two qualitative studies (Surender et al., 2007; qualitative component of Delany et al., 2008) and one quantitative study (Noble et al., 2008) offered findings in this regard. The qualitative research suggests that the CSG has not reduced the desire of adults to work. To the contrary, the qualitative studies found that the mere suggestion of such a relationship causes recipients to become indignant and feel insulated. The nature of the responses, illustrated directly below, are probably reflective of the difficult circumstances adult caregivers find themselves in and the feeling that those who would suggest such a behavioural effect from such a small transfer do not understand their suffering:

There's no way you won't want to work in order to live on R190 a month. When you work, you earn more than that. Yes, we are hungry, we are used to poverty, but there is no way you won't work only to depend on R190. By the time the R190 comes, your child needs a multitude of things from milk to shoes. You buy shoes and other small things, after that it's finished. (CSG recipient cited in Surender et al., 2007:39)

A grant, my sister, is a last resort. After you have tried all avenues. So its not gonna change once you start getting a grant … that there are not jobs. (CSG recipient cited in Surender et al., 2007:39)

The quantitative study (Noble et al., 2008), which was based on a module in the South African Social Attitudes Survey, also found no evidence of the CSG reducing labour supply. Instead it suggests that CSG and other grant recipients desperately want employment.

Whilst little evidence was found on the labour supply effect associated with the CSG, a research programme on the labour supply effects of the pension was identified. This warrants a brief description here for its implications for the future research priorities relating to the CSG. Bertrand et al. (1998) have found based on the 1993 SALDRU dataset that the pension reduced the labour supply of prime aged adults, an effect concentrated amongst men. They did not consider migrants in their analysis. Posel et al. (2006), who repeated these authors' analysis but took into account migrant effects, found a positive association. They explain the positive labour supply response to the pension by its bankrolling of migrants and the availability of grandmothers in the households from which migrants move to serve as caregivers of children left in rural households. Ardington et al. (2007) also found that, based on more recent longitudinal data from the African Centre Surveillance Site in Hlabisa, the pension was conducive to labour supply and that this was due to migration in search of work being facilitated by the pension paying the costs of migration (as well as grandmothers being available to take care of children left behind by migrants). Klasen & Woolard (2008), who like Posel et al. (2006) and Ardington et al. (2009) include how the pension affects household structure in their analysis, find, contrary to these authors, that the pension may have reduced employment prospects of some individuals inside pension recipient households, and in particular amongst youth. Their study is based on a range of national-level data sources. They argue that the adverse effects of the pension may be explained by: young members of households with limited prospects of finding employment due to mass structural unemployment and who are not entitled to any cash transfer from government choosing to remain in pension recipient households located in rural areas to share in the pension; and in so doing individuals place themselves far from job and skills development opportunities.

No evidence was identified in the review to draw conclusions about the CSG's effects on decisions about where to reside, migration and household structure/composition.

5. Research priorities

The research priorities that emerged may be grouped into three areas: the child poverty profile, CSG implementation and CSG impact.

The question of which children are deprived, in what sense as well as the similarities and differences between the compositions of child poverty based on indirect monetary measures and various direct measures of deprivation emerged as a first priority in the child poverty profile area. The question of how access to the CSG and other key services and programmes as well as the quality of key services delivered to children with different characteristics varies was another. A third is research to enhance understanding about the contours of deprivation experienced by children aged zero to four. In this regard it needs to be established whether children aged zero to four are still the most in need; and if so, whether this is this is simply due to their living in the largest and poorest households, or whether it is something to do with the way they relate to the CSG. Furthermore, is delay in application for and receipt of the CSG amongst caregivers of very young children still a problem? Or is the CSG being received by primary caregivers of very young children who are then being discriminated against in the distribution of the CSG within poor households? It is difficult to think about appropriate methods for addressing the causal question of why children aged zero to four are still the poorest and most in need (at least in the resource deprivation sense). A useful step in this direction may be to study the characteristics of children aged zero to four, such as the size of the household they live in, the employment status of their caregivers, and their access to poverty relief programmes, as compared with children in older age cohorts.

The priority questions that emerged in the area of CSG implementation are as follows. Are primary caregivers of children without biological parents and very young children (below age one) still under-represented amongst CSG recipients, and if so why and what can be done to address this? Are primary caregivers of some of the poorest children in very remote areas still finding it difficult to access the grant, and if so why and how can they better be reached? Is there still lack of understanding that affidavits may be used in lieu of birth certificates and ID books to access the grant and what it means in practice for how to navigate quickly through the application process? How can this be addressed? Is stigma associated with male caregivers receiving the grant a problem, and if so how can it be addressed? How has the requirement that children attend and be enrolled in school affected access to the CSG? How is the CSG integrated with other measures of support for poor children?

With respect to impact on child deprivation and human capital development, a message from the review is the value of additional qualitative research for understanding impact, which should focus on the questions of what the grant means to members in recipient households, who controls spending of it, how it is spent and how it affects decision-making. Regarding quantitative research in this area the review flagged the importance of research focusing not only on size but also composition of impact. A randomised control trial would be most valuable. Yet this is unlikely due to cost and ethical issues. As a second-best strategy, researchers should continue building the evidence base using econometric techniques, cross-sectional data, longitudinal data and direct child well-being indicators, where possible doing sub-group analysis to enhance understanding of how benefits are shared. Such composition impact information is needed to generate information that can be used to enhance equity in the distribution of benefits derived from the CSG.

The review raised the need for future research on the CSG to pay attention to behavioural effects, which could alter household structure and/or income earned from wages and thereby affect the impact of the CSG on child poverty. This is not only due to its relevance for thinking about how the CSG may affect child poverty and child well-being as more broadly defined in the future, but also for its relevance in decisions about the merits of increasing the value of the CSG or opting for some other measure as a means to offer additional support for poor children. It is clear that a better understanding of child poverty and the impact of the CSG requires a better understanding of the household, its boundaries, the relationships within it, as behavioural responses of household members to grants are crucial. However, existing data, methodologies and theory offer only limited possibilities for further advances in this regard.

6. Conclusion

The review leaves one with a sense of the limitations of the social assistance instruments at our disposal to address child poverty. The CSG was selected as the primary measure to alleviate child deprivation after careful consideration of alternatives. The review suggests it was an appropriate choice. Whilst there are concerns about exclusion and possible inequitable distribution of benefits, it has achieved excellent reach, is implemented well and has reduced child deprivation. Yet child poverty remains extensive and deep, and very young children may still be among those most in need of income support. The programme is still probably the most cost-effective measure for providing support to poor children. However, in the context of the large expansion in the social assistance budget (driven by the CSG), need for fiscal prudence, and possibility for the CSG to yield perverse behavioural effects if the benefit value is raised, raising the value of the CSG is an unlikely or even unwise option. This highlights the critical role of economic growth and job creation, and enhancing implementation of the CSG and other programmes that provide support to poor children and families in the fight against child poverty.

Acknowledgements

The author would like to acknowledge that contribution of Professor Servaas van der Berg and Professor Johan Burger from Stellenbosch University for the contributions they made to the design of the review. In addition, the author would like to thank Professor van der Berg for his input into the analysis.

Notes

²The issue of whether to define the CSG as a conditional or unconditional programme is complicated. This is because whilst from mid-1999 to February 2010 it officially had no behavioural conditions, up until early 1999 there were some conditions. Moreover, at times officials have, when processing applications, imposed illegal conditions (Goldblatt et al., 2006). Furthermore, in February 2010 government announced that from 1 January 2011 caregivers of CSG beneficiaries need to ensure that children for whom they are in receipt of a grant are enrolled in and attend school. However, for the period 1998–2010, the programme is best described as unconditional.

³There are three questions some may argue are missing. First, whether the CSG is sufficiently developmental. This was not asked because the vagueness of the ‘developmental’ concept makes it difficult to address and it is at odds with what we should expect the programme to achieve. The CSG was designed as a social assistance measure to help alleviate the deprivations associated with child poverty, not as a programme to address lack of employment/insufficient adult income earning capacity. The latter problems, which underpin child poverty, arise from structural issues in the economy and need to be addressed by other measures implemented in conjunction with the CSG. Second, how the CSG is integrated with other measures of support, which was not asked because it would have broadened the review scope beyond the resources available. Finally, CSG cost-effectiveness, which was not asked because no studies provide data to answer the question.

⁴Although some of the qualitative researchers' studies (see for example Hunter & Adato, 2007a) have highlighted exceptional child living arrangement cases, which may cause the benefit not to flow from recipient to child.

⁵Also published as Aguero et al. (2007).