‘Growing’ social protection in developing countries: Lessons from Brazil and South Africa

Abstract

The rapid expansion of social protection in the South provides a rich diversity of experiences and lessons on how best to reduce poverty and ultimately eradicate it. Knowledge on how best to ‘grow’ social assistance, understood as long-term institutions responsible for reducing and preventing poverty, is at a premium. This article examines the expansion of social assistance in Brazil and South Africa, two of the middle income countries widely perceived to have advanced furthest in ‘growing’ social protection. It examines three aspects: the primacy of politics in explaining the expansion of social protection and assistance, the tensions between path-dependence and innovation in terms of institutions and practices, and the poverty and inequality outcomes of social assistance expansion. The article concludes by drawing the main lessons for other developing countries.

Keywords:

• developing countries
• inequality
• path-dependence
• politics
• poverty
• social assistance

1. Introduction

The rapid expansion of social protection in the South provides a rich diversity of experiences and lessons on how best to reduce poverty and ultimately eradicate it. In this article, social protection is taken to include three main components: contributory social insurance schemes covering life course and employment risks; non-contributory social assistance programmes addressing poverty and vulnerability; and labour market and employment policies, whether ‘passive’ or ‘active’.6 Rapid economic growth and improvements in basic services have contributed to the recent decline in global poverty, but direct assistance to the poorest households is increasingly acknowledged to be crucial in ensuring that these services reach them and that they can share in the growth. Knowledge on how best to ‘grow’ social assistance, understood as long-term institutions responsible for reducing and preventing poverty, is at a premium. A global debate on how best to design, focus and organise direct assistance, grounded on the diversity of country experiences, has been taking place among development researchers (Barrientos & Hulme, 2008; Grosh et al., 2008). This article discusses what lessons emerge from a comparative study of Brazil and South Africa, two of the middle income countries widely perceived to have advanced furthest in ‘growing’ social protection.

South Africa and Brazil provide an ideal context in which to examine the growth of social assistance. They are two middle income countries with persistent and high inequality, entrenched racial disparities and high poverty incidence. South Africa has a population of 50 million, a gross national income per capita of 10 360 US$ (PPP) and a life expectancy at birth of 52 years.7 Brazil has a population of 195 million, gross national income per capita of 11 000 US$ (PPP) and a life expectancy at birth of 73 years.

Landmark political change in the two countries – in South Africa the fall of apartheid in 1990 and the rise to power of the African National Congress (ANC), and in Brazil the ending of 20 years of dictatorship in 1985 – led to renewed ‘social contracts’. New constitutions, in 1996 in South Africa and 1988 in Brazil, opened the way to innovative social protection policies with an emphasis on social assistance. The term ‘social contract’ is open to a range of interpretations. Here we use it not in the classic sense of a compact between individuals and governments, as in Rousseau, Locke and Hobbes, but in the more restricted Rawlsian sense of an overlapping consensus, emerging in pluralistic societies with economic structures generating inequalities as to the equal distribution of primary goods and priority to the least advantaged (Rawls, 2001). This interpretation echoes European, but perhaps not Anglo-Saxon, definitions of the ‘social’ which include both institutions and political commitment to ensuring appropriate levels of well-being for the entire population (Leisering, 2003). Social protection systems in South Africa and Brazil have taken different trajectories. South Africa's has relied largely on social assistance, and especially transfers to vulnerable groups, and recent policy initiatives have extended this approach. Brazil's has traditionally relied on social insurance, but recent policy initiatives have given priority to the expansion of social assistance. As a consequence, the reform of social protection institutions and practices appears evolutionary in South Africa but more radical in Brazil. In this article we argue that key features of these reforms suggest the emergence of a new paradigm. A comparison of outcomes across the two countries brings to light an interesting set of issues. In both countries, studies suggest that poverty would be significantly higher in the absence of social assistance, but in Brazil social assistance also appears to have had a measurable impact on inequality.

We provide a comparative examination of these three key aspects: the primacy of politics in explaining the expansion of social protection and assistance, the tensions between path-dependence and innovation in terms of institutions and practices, and the outcomes of social protection expansion as regards poverty and inequality. This may seem an ambitious undertaking, but we aim to contribute substantially to the growing literature on comparative social policy in developing countries.

The article is organised as follows. Section 2 discusses the role of politics in the expansion of social protection in Brazil and South Africa, Section 3 examines institutional change and alternative paradigms, Section 4 assesses the outcomes of social protection policies for poverty and inequality in the two countries, and the final section discusses the main findings and draws out some lessons for other developing countries.

2. The primacy of politics

The main stimulus to the growth of social assistance in Brazil and South Africa has come from political change and democratisation. It is perhaps a general rule that the emergence and reform of social assistance institutions reflects political processes, but the more specific hypothesis to be examined here is that the quality of these political processes has a bearing on the quality and reach of social assistance. After two decades of dictatorship, Brazil's new 1988 Constitution was intended to mark the beginning of an era of progressive and egalitarian policies (Jaccoud et al., 2009). In South Africa, the fall of apartheid and the rise to power of the ANC were also expected to lead to progressive and egalitarian policies, with the new 1996 Constitution embedding, as in Brazil, citizen rights to social security (Lund, 2008; Seekings, 2008). The renewal of social contracts in both countries led to policy activism in social protection, and particularly social assistance. The next section considers the institutional changes brought about by this political activism. We consider three ways in which the quality of political processes has shaped social assistance: reach and scale, productivism and a focus on poverty.

In a context of democratisation, unmet need and large differences in opportunity and outcomes across the population become a powerful drive for the expansion of social assistance. Growing social assistance by extending the reach of existing programmes is an obvious response. In Brazil the initial focus of policy activism was on strengthening existing non-contributory pension programmes. In South Africa a review of existing grants led to the introduction of the Child Support Grant. These initiatives aimed to extend support for groups perceived to be especially vulnerable. The expanded reach and scale of these programmes reflected political imperatives.

While political conditions were propitious to the growth of social assistance, economic conditions were not. In the last two decades of the 20th century economic growth rates in the two countries were disappointing. Annual GDP growth rates in South Africa and Brazil from 1993 to 2008 were similar at around 3%. Fiscal responsibility and economic growth and development were key political imperatives emphasising productivism in social policy, the requirement that all public policy contribute to economic inclusion and growth. In both countries these political imperatives encouraged discussion on the effectiveness of social assistance and its contribution to growth and development. In South Africa discussion consolidated around a notion of developmental welfare that informed the 1997 White Paper for Social Welfare. Lund describes ‘developmental social welfare’ as reflecting ‘a commitment to overcoming inequity and racial discrimination’, seeking ‘to move away from curative services towards preventive programmes and towards linking welfare clients with opportunities for income generation’ and to be ‘inclusive of all citizens’ (Lund, 2008:13). In Brazil, and in parallel to the expansion of non-contributory pension programmes, policy activism at the municipal level focused on linking transfers to households in extreme poverty with investment in the human capital of children, a perspective informing the development of Bolsa Escola, a social welfare programme of the Brazilian government. In both cases the quality of political processes pushed towards a more productivist approach to social assistance.

In Brazil, soft conditions attached to transfers for households in poverty are seen as a means of coordinating the work of different public agencies and ensuring investment in children. Public perceptions of the effectiveness of Bolsa Escola and broad political support rely, to an important extent, on the social investment quality of the transfers (Lindert & Vinscensini, 2008). In South Africa, grants are perceived to facilitate investment in health and education even in the absence of associated conditions. Research on the wider effects of the grants supports this (Woolard & Leibbrandt, 2010).

The same political imperatives were crucial in blocking proposals for a basic income in both countries (Matisonn & Seekings, 2003; Britto & Soares, 2010). Several factors explain why basic income proposals were rejected in both countries, but the fiscal implications (and the related implications for existing social assistance programmes), and an insufficiently productivist orientation, figured prominently in the associated policy debates. Proponents of basic income are often led to present it as a further extension (‘universalisation’) of social assistance transfers. They underplay the fundamental difference between basic income as a citizenship affirming intervention and social assistance as a poverty eradication intervention.

In sum, the impetus for the growth of social assistance in Brazil and South Africa has been the significant political change in both countries, leading to renewed social contracts. The quality of the political processes has been influential in determining the reach, orientation and objectives of social assistance. Our discussion now shifts to the social assistance institutions and the way they have evolved.

3. Institutional development

This section examines institutional developments, and especially the tension between path dependence and innovation.

3.1 Social protection systems in the 1990s

By the 1990s, South Africa had, perhaps uniquely among developing countries, well-developed social assistance programmes providing a range of means tested grants aimed at vulnerable groups. Social insurance provision, on the other hand, remained limited to employer retirement schemes. Among the social assistance grants, the Old Age Grant was dominant in terms of beneficiary numbers and budget (Van der Berg & Siebrits, 2010). Policy reforms in the early 1990s focused on eliminating discrimination in eligibility and benefit levels, both across racial groups and across provinces (especially in the former ‘homelands’ or Bantustans) (Lund, 2008).

In Brazil the most significant component of social protection was social insurance, providing earnings-related pensions and other benefits to workers in formal employment, especially in the public sector and for large employers in the private sector. The commitment to extending social security coverage in the 1988 Constitution led to the reform and expansion of two non-contributory pension schemes: the Previdência Social Rural, introduced in 1993, which provided social insurance benefits to workers in informal employment in agriculture, fishing and mining, and the Beneficio de Prestação Continuada, introduced in 1996, a means tested social assistance transfer to older people or people living with disabilities in very low income households in urban and rural areas. The design of the Previdência Social Rural reflected efforts to include informal workers in the social insurance system. Brazilian researchers describe this scheme as ‘partially contributory’ because on paper there are contribution requirements applying to these benefits. In practice, contribution requirements have never been implemented. The cost of the scheme has been financed largely through public transfers to the social insurance fund covering private sector formal workers. These two programmes now reach more than 10 million beneficiaries and provide a monthly transfer equivalent to a minimum wage (see Table 1 for current information on these programmes). At the turn of the century, non-contributory pension programmes dominated social assistance provision in both South Africa and Brazil.

Table 1: Summary information on main social assistance programmes, Brazil and South Africa 2010

Country/programme	Target population	Number of transfers (million)	Value of transfers (US$) (% of median per capita income)	Budget (% GDP)
Brazil
Bolsa Família	Households in poverty	12.8	If per capita household income is less than or equal to US$40, household transfer is US$23 plus US$18 for each child up to 3 and US$22 for 16/17 year olds up to 2. If per capita income is greater than US$40 and less than or equal to US$81, only child transfers apply. The range of transfers to a household has a minimum value of US$23 (5%) and a maximum value of US$178 (68%)	0.4
Beneficio de Prestação Continuada	Older or disabled people in poverty	3.4	US$318 (122%)	0.6
Previdência Social Rural	Long-term rural informal workers	7.8	US$318 (122%)	1.4
South Africa
Old Age Grant	Older people in poverty	2.5	US$230 (175%)	1.3
Disability Grant	Disabled people in poverty	1.3	US$230 (175%)	0.6
Child Support Grant	Children in poverty	9.4	US$53 (40%) per child	1.2
Other grants	Vulnerable groups	0.7	US$150 (115%) (Foster Care Grant only)	0.4
Sources: Woolard & Leibbrandt (2010), Barrientos (2011).
Notes: To enable comparison across the two countries, all monetary values are in purchasing power parity international US$. The conversion rates used are: 4.695 ZAR = 1 USD; 1.713 BRL= 1 USD. PPP conversion rates were accessed from EconStats. www.econstats.com/weo/V013.htm Accessed September 2011. The values are also given in brackets as a percentage of median per capita income.

3.2 Institutional reform and expansion in the 2000s

The 2000s saw rapid growth in the scale and scope of social assistance in the two countries. Table 1 provides summary information on the main social assistance programmes in Brazil and South Africa.

The introduction of a Child Support Grant in South Africa in 1998 was intended to address poverty and malnutrition in young children (initially those aged below seven) in low income households (Lund, 2008). In the following years the programme was gradually extended to older children (up to the age of 17). It now reaches over nine million children, who receive a monthly transfer equivalent to US$53, less than a quarter of the Old Age Grant. In Brazil the Bolsa Escola, introduced in a handful of municipalities in the mid-1990s and subsequently adopted as a federal programme in 2001, provided transfers to low income households conditional on children attending school. The effectiveness of Bolsa Escola encouraged a batch of direct income transfers to households in poverty. The Programme for the Eradication of Child Labour – introduced in some municipalities in the mid-1990s, at the same time as Bolsa Escola – was particularly effective. It provides direct income transfers conditional on children attending extended school hours. It was consolidated into Bolsa Família in 2004. In 2003 Bolsa Escola and four other social assistance programmes were integrated into Bolsa Família, which provides transfers to low income households conditional on schooling and health care utilisation. Bolsa Família reaches more than 12 million households in Brazil, providing transfers with values linked to household composition, a minimum of US$23 for households in moderate poverty, or in extreme poverty without children, and a maximum of US$178 for households with three or more children in extreme poverty.

From a mid-1990s baseline, both countries have expanded their social assistance significantly. Today, social assistance programmes reach over half of all households in South Africa and a quarter of all households in Brazil (Leibbrandt et al., 2010; Soares et al., 2010). The reach of the programmes exceeds the proportion of the population in extreme poverty. Old age transfers no longer dominate social assistance in either country, at least as far as beneficiary numbers are concerned. However, the different values of the transfers, with old age transfers being significantly larger than the Child Support Grant or Bolsa Família transfers, suggests that some age bias remains. (Examining the age bias in social assistance provision would require a separate article.) Before the recent expansion of social assistance, researchers in both countries raised this issue (Van der Berg, 2001; Camargo, 2004). Turra et al. (2007) note that in Brazil children were supported by private (intra-household) transfers whereas older people were supported through public transfers. The fact that transfers are shared within households suggests that in practice the situation was more complex. The expansion of assistance to children, albeit on a less generous basis, is likely to have reduced the age bias. In Brazil, the combined budget of the two non-contributory pension programmes is more than four times that of the Bolsa Família. In South Africa, the budget allocated to the Old Age Grant is approximately 10% higher than the budget for the Child Support Grant, with the latter reaching four times as many beneficiaries as the Old Age Grant (see Table 1).

3.3 Social assistance trajectories examined

The trajectories of Brazil and South Africa offer an excellent opportunity to study the dynamics of social assistance. Uniquely among developing countries, South Africa did not prioritise social insurance: medical aid and unemployment insurance have very limited coverage of the country's population. Brazil shows a very different trajectory, one that prioritised social insurance (Barreto de Oliveira & Iwakami Beltrao, 2001). Latin American policymakers expected that economic development would eventually lead to the formalisation of employment, enabling comprehensive social insurance coverage for the population. The perceived failure of social insurance institutions to reach workers in informal employment and groups in poverty has been the main driving force behind the expansion of social assistance. Equity considerations are also relevant as social insurance schemes are supported by large public subsidies that flow to better-off groups.

The orientation of social assistance in both countries was informed by a vulnerable-groups perspective. In line with traditional approaches, social assistance was intended to provide support to population groups facing significantly higher levels of vulnerability due to longer term factors, such as age or disability, which limited earnings capacity. The expansion of social assistance in Brazil and South Africa moves beyond that paradigm in at least three important respects: it focuses on households as opposed to individuals, it acknowledges that poverty and vulnerability are multidimensional, and it aims to progress towards more comprehensive, universalistic programmes.

As noted in the discussion above, social assistance grants in South Africa target specific categories of people facing deprivation and vulnerability, and old age grants dominate. To an extent the Child Support Grant can be interpreted as a linear extension of the grant system to another vulnerable group. Current proposals for the expansion of social assistance have also drawn attention to other groups that are currently left out, such as the unemployed who are unable to draw unemployment insurance, and the youth (Van der Berg, 2001; Møller, 2010). Research has amply demonstrated that transfers are not only shared within households but often lead to a re-arranging of households' labour resources (Ardington et al., 2009) and to changes in intra-household relations and decision making (Møller & Sotshongaye, 1996). Grants are therefore best understood as transfers to households in poverty that are channelled through particular individuals, for example the means tests used to establish eligibility also take account of other household members' income and assets. In Brazil, research has also established the fact that non-contributory pensions are shared within households, with implications for intra-household relations (Lloyd-Sherlock, 2006). A household focus is explicit in Bolsa Família. Whereas Bolsa Escola targeted families with children of school age and low incomes, Bolsa Família covers all households with per capita income below US$81 (about one fifth of the minimum wage in 2012). The focus on households is justified on the grounds that transfers are intended to facilitate exit from poverty by supporting agency, which is, by and large, located at the household level. The extent to which an implicit or explicit household focus on social assistance generates different outcomes is one area where further research is needed.

Linkages between transfers and service provision distinguish Bolsa Família from the grants in South Africa. Bolsa Família includes soft conditions on health utilisation and schooling but in fact institutional development in Brazil has gone further with the establishment of a Unified Social Assistance System combining a range of service provision and intermediation through reference centres. These centres are able to coordinate a range of public sector programmes and services and tailor them to the particular needs of individual households. There are two types: Centros de Referência em Assistência Social and Centros de Referência Especializada de Assistência Social, which deals with more complex cases. In South Africa, linkages between transfers and service provision are more limited and there is greater reliance on the income effect of grants and the availability of relevant services from other agencies. A normatively polarised debate about the desirability of adding conditions to existing grants has reduced the scope for assessing the potential gains from linking transfers with services and intermediation (Barrientos, 2011; Lund, 2011).

In both countries there has been a marked expansion in the reach of social assistance. There are grounds for arguing that this constitutes progress towards the development of social assistance programmes that are more inclusive and all encompassing of households in poverty. The term ‘universalisation’ is often over-used and mis-used in the social protection literature. In the context of social assistance, however, it appropriately describes the capacity of programmes to assist all households and individuals in poverty, regardless of personal characteristics. Bolsa Família constitutes a clear shift in paradigm towards universalising social assistance, as the only condition on eligibility is the household's income level. Reach is important for the effectiveness of social assistance. In both countries, social assistance reaches beyond households in extreme poverty and increasingly covers households that are not in poverty but are vulnerable to falling into poverty.

The legal institutionalisation of social assistance in the two countries merits a brief discussion. In Brazil, entitlement to non-contributory pensions is enshrined in the constitution, with the implication that provision is – at least on paper – not limited by the government's fiscal position. Bolsa Família, on the other hand, does not have the same legal position, and entitlements are subject to government discretion, including budgetary controls. In South Africa, the constitutional right to social assistance is dependent on budgetary conditions and the specific form of provision is at government discretion. Constitutional recognition of social assistance entitlements in Brazil can be advantageous in protecting non-contributory pension provision from government discretion, but it can also be disadvantageous if it prevents innovation and change in provision and also if it ensures different treatment of otherwise similar households in poverty depending on whether they include older persons. Acknowledging a general principle that is left to governments to implement, the constitutional approach in South Africa appears to achieve a better balance between the needs to protect rights, to enable innovation and to ensure fiscal sustainability. The use of the courts as an instrument for the expansion of social assistance is present in both countries, but to date it has been more successful in South Africa, for example making the age of entitlement to the Old Age Grant the same for men and women (Seekings, 2008). In Brazil the main issue considered by the courts is whether the government's interpretation of minimum benefit levels for non-contributory pensions, at the minimum wage, is appropriate in the case of beneficiaries living with severe disabilities.

To recap, this section focused on differences and similarities in the nature of institutional changes in social assistance expansion in South Africa and Brazil. Path dependence in social assistance is stronger in South Africa, because of the prominent role of grants and their income effects. In Brazil institutional innovation has been strongest in the expansion of Bolsa Família. A shift in the focus of social assistance – towards households, towards linking transfers and services and towards more comprehensive reach – can be observed in both countries. The shift is explicit in Brazil's Bolsa Família. Differences in legal institutionalisation across the two countries have implications for the future evolution of social assistance.

4. Poverty and inequality outcomes

This section discusses poverty and inequality trends in the two countries and assesses the contribution of social assistance.

4.1 National poverty and inequality outcomes

We begin by sketching aggregate trends in poverty and inequality. Figure 1 provides estimates of the poverty headcount rate and the Gini coefficient for the two countries. It shows a marginal drop in poverty incidence in South Africa between 1993 and 2008. Brazil, on the other hand, managed to make significant inroads into poverty incidence between 1993 and 2008 as the poverty headcount fell by half, with a sustained decline thereafter. Brazil and South Africa are among the countries with the highest inequality in the world. Estimates of the Gini coefficient for the two countries suggest divergent trends, as inequality increased in South Africa but fell in Brazil by as much as 10%, with most of the decline being between 2000 and 2008. The changes in inequality in the two countries are relatively small and from a very high base, but they do show divergent trajectories.

Figure 1: Poverty and inequality trends: South Africa and Brazil, 1993 to 2008

Sources: Leibbrandt et al. (2010) for South Africa; Ferreira & Leite (2009) and Soares et al. (2010) for Brazil.

Notes: Poverty lines are R515 (R2008) for South Africa and R100 (September 2004 reais) for Brazil; approximately US$96 PPP and US$68 PPP respectively.

This raises the question of why Brazil has achieved large reductions in poverty and a significant reduction in inequality at the same time that South Africa has achieved only small reductions in poverty and experienced a marginal rise in inequality. In the context of this article, the interest is in whether the expansion of social assistance has played any part.

Research in both countries points to the type of growth process as the primary factor. Critics argue that in South Africa economic policy has emphasised growth in high-tech industries, perhaps at the expense of employment creation (Nattrass & Seekings, 2001), but others point to rigidities in the labour market. Persistently high unemployment has been a feature in South Africa, while in Brazil high commodity prices and generally favourable economic conditions have enabled a significant reduction in unemployment, from 12% in November 2003 to 5% in November 2011. Together with the expansion of social assistance, this has translated into strong income growth among the bottom 40% of Brazil's population. Assessing the types of growth in the two countries would need a longer and more detailed analysis than is possible in this article, but this brief discussion underlines the significance of the issue.

Redistribution, especially through the expansion of social assistance, is important too. Leibbrandt et al. (2010) simulate, for 2008, the potential impact on poverty of South Africa's grants by comparing the poverty headcount estimated from reported household income with a measure of household income excluding income from grants. This approach does not take account of behavioural responses to changes in income, for example labour supply responses. In the labour supply case, estimated poverty would have been around six percentage points higher (60% compared to the 54% shown in Figure 1). The impact of the grants on the Gini coefficient is marginal: in 2008 the Gini would have been 0.03 higher when estimated on a measure of household income excluding grants. In Brazil, researchers using a similar methodology compared the poverty headcount for household income with and without social assistance transfers and found a strong poverty reduction effect. Soares et al. (2010) find that Bolsa Família and its antecedent programmes are associated with around 40% of the reduction in the poverty headcount in the 2000s, and can be said to be solely responsible for the reduction in extreme poverty in the same period (from 10% to 5%). The same study estimates that Bolsa Família and non-contributory pensions are associated with around one third of the 10% reduction in the Gini coefficient in the same period. These estimates are based on a decomposition of the Gini coefficient by income source.

These figures need to be treated with caution, as the methods used to generate these estimates of the impact of social assistance on poverty and inequality do not take full account of the influence of complementary policies (Souza, 2011; Saboia, 2012). In Brazil, for example, non-contributory pensions, minimum pension guarantees in contributory schemes and wage bargaining in the informal sector are all benchmarked to the minimum wage, the value of which rose by around 50% in real terms from 2002 to 2008 (Saboia, 2009). All in all, social assistance appears to have important effects on poverty in South Africa and larger effects on both poverty and inequality in Brazil.

4.2 Old age poverty and vulnerability

The poverty estimates discussed above represent snapshots of the population below the poverty line, but changes in households' poverty status over time are also important. It is of interest to consider how social assistance helps to facilitate exit from poverty. To look into the dynamics of poverty status, we make use of panel data from comparative surveys of older people and their households collected in 2002 and 2008 in the two countries.8 The surveys focus on older people and their households and consider explicitly the role of pension provision. Figure 2 reports on changes in poverty status for survey respondents using reported income and a measure of income excluding pension income (Barrientos & Mase, 2012). The differences in poverty, estimated with and without pension income, provide an upper bound of the poverty reduction potential of pensions.

Figure 2: Estimates of poverty headcount with and without pension income for a panel sample of older households in selected locations in Brazil and South Africa, 2002 and 2008

Source: Study on ageing, well-being and development. Sample of households with older persons: 700 households in Brazil (metropolitan Rio and Ilhéus) and 615 in South Africa (ZA) (Eastern and Western Cape), visited in 2002 and 2008. www.sed.manchester.ac.uk/research/ageingandwellbeing/index.htm Accessed February 2012.

It is important to keep in mind that the panel samples are not nationally representative. The data were collected in the Eastern and Western Cape provinces of South Africa and in metropolitan Rio and Ilhéus in Brazil, and in low income locations in both countries. The interest here lies in the changes in poverty status for the older households sampled. The information in Figure 2 makes three points. First, and as noted in the discussion above, excluding non-contributory pension income from household income leads to higher headcount poverty in both samples. Although it is to be expected that lower household income would worsen household welfare, the point is that non-contributory pension income is important enough to make a difference to the poverty headcount. Second, non-contributory pension income might not be sufficient to prevent persistent poverty, understood in this case as older households observed to be poor in 2002 and 2008. In the absence of social assistance for older people and their households, persistent poverty among them would be significantly greater. Persistent poverty in the sample would be close to 50% higher for the South African households and almost twice as large for the Brazilian households. Third, non-contributory pension income makes an important contribution in helping households to exit poverty. In the absence of pension income, the proportion of households in poverty in the South Africa sample in 2002 that would have exited poverty would have fallen from 25% to 10%. In the case of the Brazilian households, 65% of those in poverty in 2002 exited poverty, but this share would have been reduced to 53% if non-contributory pension income had been excluded. Social assistance support for older people is not as important for the Brazilian households sampled as it is for the South African households sampled.

This brief discussion of changes in the poverty status of older people and their households serves to underline the role of social assistance in explaining changes in poverty status over time. Furthermore, the comparison across the two countries also demonstrates the significance of complementary policies supporting exit from poverty. The effectiveness of social assistance in securing poverty reduction and equity is strengthened by complementary policies encouraging employment and basic service provision.

5. Conclusions

It is important to ensure that economic and social development reaches all. Countries in the South face significant challenges in achieving this goal, not least from globalisation, limited fiscal resources and accumulated social debt. As a means to achieve social justice and eradicate poverty, growing social protection is a global imperative. There is much to be learnt from the experiences of Brazil and South Africa, two middle income countries with considerable achievements in social protection. This article has provided a comparative perspective, and we conclude by underlining the main lessons learnt.

The first lesson concerns the primacy of politics. In the cases of Brazil and South Africa, landmark political change provided a renewed impetus for social protection and especially assistance. The growth of social assistance in the two countries has its roots in renewed social contracts, embedded in new constitutions redefining the right to social protection. As we have noted, the political processes had a significant effect on the extension of social assistance in the two countries, ensuring a focus on poverty reduction and equity. Research and policy discussions on social assistance have focused on design issues, but it is crucial to understand the political processes shaping the growth of social assistance. In the South, growing social assistance is part and parcel of democratisation processes shifting attention to poverty and equity. The quality of these processes shapes the reach and orientation of social assistance.

We next considered institutional changes. Tension between path dependence and innovation provided a focus for the discussion. In terms of their social protection system, South Africa and Brazil started off from entirely different points. Since the 1950s, Brazil's social insurance has been the dominant component of its social protection system, whereas in South Africa social assistance has played that role. This explains why the growth of social assistance in Brazil represents a more radical break with policy priorities, and why path dependence is stronger in South Africa. A key lesson for developing countries here is the need to pay attention to social assistance within social protection systems. For low income countries with weak social insurance institutions there are strong advantages in growing social assistance first. Recognising the difficulties in extending social insurance to workers in informal employment and groups in poverty, Brazil and Latin American countries in general are placing a stronger policy priority on the development of social assistance. In Brazil and South Africa, the reach of social assistance goes beyond the population in poverty to include groups vulnerable to falling into poverty. In developing countries with high poverty incidence, social assistance is a core policy area.

We noted the emergence of a paradigm shift as regards social assistance institutions. Three aspects are apparent: a growing focus on households as opposed to individuals, a productivist or developmental perspective that links transfers to human capital investment and economic inclusion, and a more inclusive perspective replacing the vulnerable groups approach that dominated developments in the previous century. This shift in perspective is ‘work in progress’, but Bolsa Família best represents the changes that are under way. The evolution of this paradigm shift is of great significance for other developing countries.

In both countries, social assistance makes a strong contribution to poverty reduction and equity. At the national level, overall trends diverge across the two countries. Poverty and inequality reduction trends are strong in Brazil, but in South Africa poverty reduction is marginal and inequality shows a rising trend. This suggests that the type of growth makes a difference to the impact of redistribution through social assistance on poverty and inequality. In Brazil a combination of factors is responsible for the strong gains in poverty reduction and equity: inclusive growth, the rise in real terms of the minimum wage, which affects social assistance and social insurance transfers and wage bargaining in the informal sector, and redistribution through social assistance. This finding is also supported by analysis of longitudinal data on older households. Old age social assistance transfers are associated with lower poverty and lower chronic poverty, but also exit from poverty. These effects are stronger for the Brazilian households sampled because of complementary policies. Perhaps the main lesson emerging from examining the poverty and inequality outcomes in the two countries is the need to combine redistribution through social assistance transfers with inclusive growth and employment policies.

The main lessons, in a nutshell, are as follows:

•	In the South, renewal of social contracts and democratisation will shape the growth of social assistance as a means of reducing poverty and inequality.
•	The growth in the reach of social assistance is important, but so is a shift in paradigm, with a stronger household focus, a productivist or developmental perspective and greater inclusion.
•	In both countries social assistance has made a strong contribution to poverty reduction and equity. However, the stronger outcomes in Brazil emphasise the fact that social assistance is likely to be most effective when accompanied by, and embedded in, inclusive growth and employment policies.

Notes

⁶Terminology varies across countries. In South Africa, the term ‘social security’ is used to cover social insurance, social assistance and public works programmes, and the term ‘social welfare’ to describe social assistance transfers.

⁷PPP stands for purchasing power parity, and is a measure of the monetary value of a similar basket of goods in different countries in international US$.

⁸See details of project at www.sed.manchester.ac.uk/research/ageingandwellbeing/index.htm