Table of Contents
Executive Summary………………………………………………………………..7
ES1 Introduction…………………………………………………………………………….7
ES2 The rationale for means testing social security grants……………………………7
ES3 A review of international experience on means testing……………………………7
ES4 Alternatives to means testing………………………………………………………...9
ES5 Analysis of the child support grant…………………………………………………..9
ES5.1 Review of the means test……………………………………………………9
ES5.2 Micro-simulation evidence………………………………………….………10
ES5.3 Review of benefit levels………………………………………………….…11
ES6 Analysis of the foster care grant………………………………………………….…12
ES7 Analysis of the older persons pension, the disability grant and…………….……12
the care dependency grant
ES7.1 Review of the means test…………………………………………………..12
ES7.2 Micro-simulation evidence…………………...…………………………….13
ES7.3 Review of benefit levels…………………………………………………….13
ES8 Overall recommendations…………………………………………………………...14
Part 1 The context for targeting in South Africa……………………………16
1.1 Introduction……………………………………………………………………………....16
1.2 The rationale for means testing social security grants………………………………16
1.3 The benefits of targeting………………………………………………………………..17
1.4 The costs of targeting…………………………………………………………………..18
1.4.1 Exclusion error………………………………………………………………..18
1.4.2 Administrative costs………………………………………………………....19
1.4.3 Private costs……………………………………………………………….…19
1.4.5 Indirect costs………………………………………………………………....19
1.4.6 Social costs…………………………………………………………………..20
1.4.7 Political costs……………………………………………………………...….20
1.5 The Targeting Test………………………………………………………………….…..20
1.6 The theoretical underpinnings of targeting………………………………………..….21
Table of Contents (Continued)
1.7 General History of Social Assistance in South Africa…………………………….…25
1.7.1 History of the social pension……………………………………..….……...28
1.7.2 History of the Child Support Grant (CSG)…………………………………28
1.7.3 History of the Care Dependency Grant……………………………….…...29
1.7.4 Purpose and history of the Foster Child Grant (FCG)……………………30
Part 2 A review of the international experience on targeting………….…31
2.1 Introduction……………………………………………………………………...……….31
2.2 Unconditional social transfers………………………………………………………….31
2.3 Child-oriented programmes…………………………………………………….………32
2.4 Case studies on child-oriented programmes…………………………………………33
2.4.1 Child poverty in developed nations…………………………………………33
2.4.2 Mozambique………………………………………………………………..…34
2.4.3 Namibia……………………………………………………………………..…35
2.4.4 Kenya Cash Transfer for OVCs…………………………………………….36
2.5 Disability grants………………………………………………………………………….36
2.6 Case studies on disability grants………………………………………………………37
2.6.1 Mozambique (GAPVU)………………………………………………………37
2.6.2 Namibia…………………………………………………………...………..…38
2.7 Social pensions for older people………………………………………………………39
2.8 Case studies on social pensions………………………………………………………40
2.8.1 Lesotho……………………………………………………………………..…40
2.8.2 Nepal…………………………………………………………………………..40
2.8.3 Mauritius………………………………………………………………………41
2.8.4 New Zealand………………………………………………………………….42
2.9 Conditional cash transfers……………………………………………………………..43
2.10 Case studies on conditional cash transfers…………………………………………44
2.10.1 Brazil’s Bolsa Escola (now part of Bolsa Familia)……………………….44
2.10.2 Mexico’s Progresa (now Oportunidades)………………………………...44
2.10.3 Chile’s Solidario……………………………………………………………..44
2.10.4 Bangladesh Food-for-Education (now Cash-for-Education)…………...45
2.11 Public Works……………………………………………………………………………45
Table of Contents (continued)
2.12 Case studies on public works……………………………………….……………….46
2.12.1 India’s Employment Guarantee Scheme (EGS)…………………………46
2.12.2 Ethiopia………………………………………………………………………47
2.12.3 Argentina…………………………………………………………………….49
2.12.4 Malawi………………………………………………………………………..49
2.12.5 Indonesia…………………………………………………………………….50
2.13 Alternatives to means testing…………………………………………………………51
2.13.1 Verified means testing……………………………………………………...51
2.13.2 Proxy means testing………………………………………………………..51
2.13.3 Community targeting……………………………………………………….52
2.13.4 Categorical targeting……………………………………………………….52
2.13.5 Self-targeting………………………………………………………………..53
2.13.6 Combinations of targeting mechanisms………………………………….53
2.14 Conclusions…………………………………………………………………………….54
Part 3 A review of South Africa’s means tests……………………………..55
3.1 Introduction………………………………………………………………………………55
3.2 Analysis of the Child Support Grant…………………………………………………..55
3.2.1 Review of the means test for the Child Support Grant (CSG)…………..55
3.2.2 Assessment of the means test for the Child Support Grant……………..56
3.2.3 Micro-simulation evidence for the Child Support Grant………………….57
3.2.4 Review of benefit levels for the Child Support Grant…………………….59
3.3 Assessment of the Foster Child Grant (Foster Care Grant)………………………..60
3.4 Analysis of the Older Person’s Pension, the Disability grant……………………….61
and the Care Dependency Grant
3.4.1 Review and assessment of the means tests for these grants…………..61
3.4.2 The scaling factor (ratio) between married and single individuals………68
and the rationale for different scaling factors
3.4.3 Micro-simulation evidence for the social pension…………………………71
3.4.4 Review of benefit levels for these grants……………………………….….73
3.5 The Social Relief of Distress Grant……………………………………………………73
3.6 Cost-benefit analysis of means testing social grants………………………………..74
3.6.1 Benefits and costs of means testing the social pension……………….…74
3.6.2 Costs and benefits of targeting the Child Support Grant………………..76
Part 4 Policy recommendations……………………………………………….86
4.1 Introduction………………………………………………………………………………86
4.2 Recommendation for automated updates of the means test………………….……86
4.3 Rationale for linking social grant values to a poverty index……………………..…89
4.4 Linking the poverty index and the means test…………………………………….…89
4.5 Key questions for subsequent projects……………………………………………....90
4.6 Conclusions………………………………………….…………………………………..91
References………………………………………………………………………….93
Appendix 1 Government Gazette, 22 February 2005 No. 27316………………..…….99
Appendix 2 Project Terms of Reference provided by DSD……………………..…...101
Appendix 3 Alignment of this report to the Terms of Reference………………..……103
Table of Figures
Executive Summary
Figure ES1.The evolution of the Child Support Grant benefit level…………………11
relative to inflation
Figure ES2. The evolution of the Foster Care Grant benefit level relative…………12
to inflation
Figure ES3. The evolution of the benefit levels for the Older Person’s……………..14
Pension, the Disability Grant, and the Care Dependency
Grant (CDG) relative to inflation
Part 1 The context for targeting in South Africa
Figure 1.1 Social protection and growth………………………………………………….22
Figure 1.2 A virtuous circle………………………………………………………………...22
Part 2 A review of the international experience on targeting
Figure 2.1 Child poverty in developed countries………………………………………...33
Part 3 A review of South Africa’s means tests
Figure 3.1The evolution of the Child Support Grant benefit level relative……….59
to inflation
Figure 3.2 The evolution of the Foster Care Grant benefit level relative………….61
to inflation
Figure 3.3 The evolution of the benefit levels for the Older Person’s……………..73
Pension, the Disability Grant, and the Care Dependency
Grant (CDG) relative to inflation
Part 4 Policy recommendations
Figure 4.1: Inflation rates from FY2005 to FY2006 by category……………………….87
Table of Tables
Part 1 The context for targeting in South Africa
Table 1.1: A framework for assessing the efficacy of targeting………………...…..…..21
Table 1.2: Impact of the social pension on labour force participation………………….23
Table 1.3 : Impact of the child support grant on labour force participation……………24
Table 1.4: Impact of child support grant on female labour force participation………...25
Part 3 A review of South Africa’s means tests
Table 3.1 Micro-simulation evidence for the Child Support Grant……………………..58
Table 3.2 The means test for the social pension (single people)……………………...69
Table 3.3 The means test for the social pension (married people)……………………70
Table 3.4 Micro-simulation evidence for the social pension (75% take-up)…………71
Table 3.5 Micro-simulation evidence for the social pension (90% take-up)………….72
Table 3.6 Key demographic statistics for the social pension…………………………..75
Table 3.7 Eligibility and receipt of the Child Support Grant by mothers…………...….80
Table 3.8 Characteristics of mothers by CSG eligibility and receipt…………………..82
Table 3.9 Take-up rates for the CSG 2003-2005 by age…...... ………………………..83
Part 4 Policy recommendations
Table 4.1 The impact of inflation adjustments on social grant spending………...…....88
Appendix 3 Alignment of this report to the Terms of Reference
Table A3.1 Alignment of the objectives to the text of the main report………...…….102
Table A3.2 Alignment of the deliverables to the text of the main report……………102
Table A3.3 Alignment of the elements of scope to the text of the main report….…102
Executive Summary
ES1. Introduction
This Executive Summary highlights the Economic Policy Research Institute’s key findings and recommendations from the research project commissioned by the Department of Social Development entitled the “Review of Targeting Mechanisms, Means Tests and Values for South Africa’s Social Grants (#SD13/2005)”. The analysis documents a rationale for means testing social security grants, reviews the international experience on means testing and discusses alternatives. Building on these research findings, the report analyses each of South Africa’s major social grants, provides a review of the associated means tests (if any), presents evidence from the micro-simulation of alternative means tests specifications, and evaluates the benefit levels. The report concludes with specific and actionable recommendations.
ES2. The rationale for means testing social security grants
The main rationale for means testing social security grants is to more economically deliver benefits to the poor. For a fixed budget, perfect targeting enables a social grant programme to deliver a higher level of benefits to poor households. Alternatively, for a fixed level of benefits, perfect targeting enables the programme to meet its objectives at the lowest possible cost. The assumption of perfect targeting, however, assumes a theoretical ideal. In practice, targeting is imperfect and always involves a significant basket of costs—administrative, private, social, economic and political. Extensive studies of targeting have documented the potential savings from more effective targeting as well as the significant costs imposed by targeting mechanisms.
A framework for evaluating targeting mechanisms—including means tests—starts with the baseline of universal distribution of grants. The poverty rate (or other targeting threshold) determines the potential gains from targeting. For example, a country with a 45% poverty rate potentially saves 55% of the cost of a universal programme. The framework requires the comparison of these potential savings with the total costs associated with targeting. While some costs are potentially calculable in rand terms (administrative and private costs and economic distortions), other costs (political and social costs) are harder to quantify.
Social grants are usually redistributive whether or not they are means tested or otherwise targeted. In South Africa’s case, the relatively progressive system of taxation ensures that the combination of financing and grant delivery leads to a significant redistribution of income. Economists frequently argue that universal social transfers provide the most efficient form of redistribution because there are no associated conditions that can distort people’s behaviour.
ES3. A review of international experience on means testing
South Africa’s social grants are distinguished internationally by their relatively unconditional character. The main alternative type of scheme commonly employed in developing countries involves more rigid conditionalities—in terms of health, education or employment. There are numerous advantages of unconditional cash transfers over alternative means of social protection. They have lower administrative costs, which mean they can potentially provide the most cost-effective solution to poverty reduction. Unlike subsidies, they do not distort the price system and unlike in-kind transfers, they enable households to decide on and meet their most critical needs. Additionally it is easier to predict the budget for cash transfers as the benefits are fixed and do not depend on unpredictable commodity prices. Transfers can also help to provide one more macroeconomic stabiliser, since the net benefits to the poor tend to increase during a recession or in the face of an economic shock.
Targeted conditional transfer programmes share the objective of breaking the ‘vicious circle of poverty’ by concentrating resources on increasing human capital accumulation among the children of the poor, and through them, to directly tackle persistent or chronic poverty. At the same time, there is considerable variation in the scope, design and objectives across the range of programmes.
Brazil’s Bolsa Escola and Child Labour Eradication Programme (PETI) and Bangladesh’s Cash-for-Education programme focus on a single cause of poverty: deficient school enrolment and attendance. Mexico’s Oportunidades, Nicaragua’s Red de Proteccion Social (RPS), Honduras’ Programa de Asignacion Familiar (PRAF), Colombia’s Familias en Acción, and Turkey’s Social Fund, have a broader range of components which address specific dimensions and correlates of poverty: household consumption, early childhood interventions, schooling and healthcare. Chile’sSolidario has the widest range of interventions and addresses the multi-dimensional causes of poverty. Differences in the design and objectives of these programmes reflect different views of the causes of poverty. The first group of programmes is premised on the belief that the primary cause of persistent poverty is education deficits among the poor, arising from poor school attendance or from the competing pressures of child labour. The programmes in the second group are based on the view that education deficits, together with deficits in parenting, primary healthcare, and nutrition in early childhood, are the main factors explaining persistent poverty. Chile Solidario maintains that poverty is intrinsically multi-dimensional, and regards the household as a whole, and not only the children, as the main agent of change.
Conditionalities are often based on the view that impoverished households are poor because of their behaviour. Conditionalities, in this view, create incentives to change their behaviour and lead them out of poverty. In many cases, however, poor households face structural poverty—which their behaviour cannot change. In these cases, enforcing rigid conditionalities risks the most perverse outcome possible – punishing the poorest who faced the highest cost to comply with the requirements, particularly when high quality schools and health care are inaccessible. A World Bank conference in Istanbul in 2006 reached the conclusion that there was no conclusive evidence that conditionalities themselves promoted human development outcomes.
International experience demonstrates that public works programmes are more likely to succeed when the link between poverty and unemployment is exceptionally strong. Since most programmes offer only short-term employment, they are more appropriate for transient rather than chronic poverty. It is also important that the value of assets produced through public works substantially offset the cost of the programme. Public works can be an expensive way to deliver social protection. The net income gains to the workers from the Trabajar programme in Argentina have been estimated to be just a quarter of the benefits paid by the government. Many programmes cost the equivalent of between $250 and $500 dollars to deliver $100 of net benefits to a household. Due to their high implementation costs, unless programmes generate substantial production-related benefits, they are unlikely to deliver social transfers in a cost-effective manner. For chronic poverty and deep structural unemployment, short-term public works have limited impact.
ES4. Alternatives to means testing
South Africa’s approach to targeting can be classified as “verified means testing”, whereby the grant recipient directly reports the key targeting variable (income), and procedures are implemented to corroborate this information. Alternatives to verified means testing include unverified means testing, proxy means testing and universalism. Historical and international experience documents the substantial challenges developing countries face in implementing an effective verified means test. South Africa’s means tests now rank as some of the developing world’s most effective targeting mechanisms, demonstrating considerable improvement over the past five years. Nevertheless, it is relatively costly and continues to yield significant errors of inclusion and exclusion. However, many of the alternatives — proxy means testing, geographical targeting, community-based targeting, and others — are inconsistent with a rights-based approach. The main feasible and cost-effective alternative to consider is a more universal approach to targeting.
ES5. Analysis of the Child Support Grant
ES5.1 REVIEW OF THE MEANS TEST
The means test for the Child Support Grant is the most complicated of all the targeting mechanisms employed by the Department of Social Development, distinguishing among urban and rural households as well as the materials used in the construction of the caregiver’s dwelling. The urban/rural distinction aims to reflect the greater prevalence of poverty and the higher cost of living in rural areas.
The means test for the Child Support Grant reflects a strong theoretical understanding of the profile of poverty in South Africa. The focus on the means of the caregiver rather than the entire household takes into account the challenges of intra-household resource allocations. The utilisation of poverty proxies—informal dwellings and rural locations—reflects the nature of poverty in South Africa. In practice, however, the resulting complexity challenges the administrative capacity of the implementing institutions. Even Statistics South Africa has abandoned the task of identifying the urban/rural distinction at a detailed geographic level. The main strength of the means test today is that the Department of Social Development has significantly relaxed the strictness of its application. While this has led to the extension of the grant to households that do not strictly qualify, most of these are nonetheless poor households—and would qualify based on the standards of poverty established in 1998.
A major shortcoming of the means test for the Child Support Grant is that the income thresholds have not been adjusted since the programme’s inception. The threshold for rural (and some urban) households is R1100 per month, while that for urban households in formal dwellings is R800. Since 1998 the cost of living for poor households has risen approximately 72%.
ES5.2 MICRO-SIMULATION EVIDENCE
EPRI simulated the social and fiscal impact of a number of alternatives to the existing means test for the Child Support Grant. The first scenario addressed the most salient problem with the existing test—the fact that the income thresholds for grant eligibility have not been updated since the programme’s inception. If these thresholds were adjusted for the inflation rate facing the poorest households, the 2005 urban threshold of R800 would increase to approximately R1300, and the rural threshold from R1100 to approximately R1750. Applying these new thresholds would result in an estimated 27% increase in eligibility in 2005—approximately another 1.3 million children would become eligible.
Of these 1.3 million newly eligible children, approximately 800,000 already receive the grant, although they are not strictly eligible. As a result, the adjustment of these thresholds for inflation would lead to an increase of approximately 500,000 new grants—an 8% increase over existing levels, at a cost of approximately R1.2 billion per year.
Another incremental improvement in the means test involves the elimination of the urban/rural distinction. This aspect of the means test reduces the transparency of eligibility and frequently delays the determination of entitlements. Eliminating the distinction and adopting the higher inflation adjusted thresholds would lead to an increase in eligibility of approximately 3.5%, but only a 2.4% increase in actual grants paid, since many of the newly eligible currently receive the grant. The additional cost of the change would be approximately R250 million.
At a broader level, the question arises of whether or not a means test is justified for the Child Support Grant. With the inflation adjustment and the elimination of the rural/urban distinction, more than three out of four of South Africa’s children qualify for the child Support Grant. This reduces the theoretical potential savings from targeting to less than 25% compared to the cost of a universal programme. This substantially increases the likelihood that the multiple costs of targeting—in terms of administrative and private expenditures, economic distortions, stigma, and political costs—exceed the realised benefits. The micro-simulation of the elimination of the means test indicates an increase in take-up by an additional 3.9 million children, at a cost of R9.3 billion.