DECLINING POVERTY IN SOUTH AFRICA THE ROLE OF SOCIAL GRANTS Presentation to a conference on social grants, Pilanesberg, 14 June 2007

DECLINING POVERTY IN SOUTH AFRICA THE ROLE OF SOCIAL GRANTS Presentation to a conference on social grants, Pilanesberg, 14 June 2007 Servaas van der Berg Department of Economics University off Stellenbosch INTRODUCTION In the past four years, there has been a substantial decline in children going hungry. The number of children whose parents reported that they have gone hungry in the previous year declined from 31% in 2002 to 23% in 2005 (see Figure 1). The reason for this is quite clear: The expansion of social grants, in particular the child support grant. As Figure 2 shows, social transfers (grants) per capita increased by about R500 between the year 2002 and 2005, in other words in the same period during which hunger declined. There is no doubt that the major cause of the decline in poverty, the decline in hunger that we have seen and that has been captured in the General Household Surveys, has been the result of this expansion of the social grants. Figure 1: 40% Frequency of children in household going hungry during the previous year 35% 30% 25% 20% 15% 10% 5% 0% GHS2002 GHS2003 GHS2004 GHS2005 Seldom Sometimes Often Always

2 Figure 2: Transfers per capita (in constant 2000 Rand) R 1 400 R 1 200 2 0 0 0 R a n d v a lu e s R 1 000 R 800 R 600 R 400 R 200 R 0 1990 1995 2000 20045 This paper looks at the impact of grants on poverty and on the poor, before turning to a discussion of the volume and extent of social grants in South Africa. Thereafter, we shall consider the targeting and incidence of grants in South Africa, before investigating the impact of the means test and finally discussing the incentive effects of grants. The paper ends with a short attempt at projecting what is likely to happen with regard to poverty and social grants in the next decade. GRANTS AND POVERTY TRENDS The large expansion of social grants since the new government came to power, and in particular in the last half a decade, has made available to households large sums of money in the form of social transfers, most of it going to the poor and most of it quite well targeted, as will be shown. This has had a major impact on poverty, and our research confirms that, after poverty had remained stubbornly high during the latter half of the 1990 s, it declined markedly over this period,. Despite an expansion of new jobs in the economy in the same period, the reduction in poverty cannot be linked so clearly to these additional jobs, mainly because there have been too few new jobs, and secondly because the more educated, rather than the poor, are first in the job

3 queue. This means that poor households have relatively seldom benefited from these new jobs. Of course, should the number of jobs keep on expanding, more jobs will be going to less educated people who are poor. But that was not the experience of the past few years, because of the great extent of unemployment even amongst the relatively better educated. The economy therefore have to grow for quite a substantial period at high growth rates before the less educated also start sharing in a major way in the benefits of additional jobs. This makes the role of social grants even more significant. But in the long run social grants cannot be the major poverty reduction tool in any society: What is necessary in the long run is sustained, substantial growth that increasingly draws more people into employment in the modern economy, thereby offering them full participation in the economy and society. Social grants are a way of alleviating poverty, of allowing people to improve their living standards, but grants cannot substitute for well-paying jobs for the majority of the population. To see what impact the expansion of the grants must have had on the reduction in poverty, it is useful to turn to Figure 3. This shows a cumulative distribution function for the year 2000 as derived from the income and expenditure survey of 2000 (IES2000). This cumulative distribution function, or CDF, shows the extent of poverty at any poverty line that one may wish to draw. For instance, if the poverty line is drawn at R3 000 per capita income (in 2000 Rand values), we can read off that about 47% of population would be in poverty. If one now considers that the social grants expanded by R500 per person in the period from 2000 to 2005, the likely impact in the absence of any other income distribution would have been to shift the CDF curve to the right, by R500 per person. The impact in Figure 3 is to reduce poverty to approximately 41% of the population in 2000, i.e. by 6% points, at the same poverty line. Thus, on the very conservative assumption that the grants were not well targeted at the poor population at all but just distributed approximately equally across the whole population, poverty must have declined quite substantially, in the absence of any other major changes in income distribution.. The data mentioned before about the reduction of children going hungry in South Africa confirm that this is indeed what has happened. So too does Figure 4, which shows a poverty headcount rate based on data from AMPS (the All Media and Products Survey). The 8 percentage point decline in hunger in Figure 1 and the 8 percentage point decline in poverty shown in Figure 4 is evidence that a poverty reduction of approximately this extent occurred in the last few years. As Figure 2

4 shows, this coincides with the expansion of the social grants; our own research using the AMPS data, the Labour Force Surveys and the Income and Expenditure Surveys confirms that this reduction in poverty was mainly driven by the expansion of the grants, rather than by employment expansion. Figure 3: Cumulative % of population below each income level 0.1.2.3.4.5.6 Cumulative distribution function for IES2000 plus effect on poverty of R500 per capita grant Poverty line R3 000 per capita Poverty reduction 0 1000 2000 3000 4000 5000 Per capita income CDF before grant expansion Grant increment of R500 per capita distributed equally would shift distribution to the right, thus reducing poverty substantially. (Grants are actually well targeted.) CDF after grant expansion Thus we may conclude safely that poverty has indeed declined, and that social grants played a major role in this regard. Social grants are an important source of income particularly for the rural poor. Such grants are the best targeted of all social spending programmes, when one compares them for instance to education or health spending. They are generally well targeted at the rural poor, and there are only small leakages to the non-poor: 76% of spending on social grants goes to the poorest two quintiles of households, who represent about 50% of the population. The poorest 40% of households received only 4.7% of pre-transfer income, but they received 7.8% of post transfer income, as a result of the impact of these transfers, or social grants.

5 Figure 4: Poverty headcount rate based on AMPS 54% 52% 50% 48% 46% 44% 42% 40% 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Social grants also have a significant influence on household formation: it empowers old people, retaining them in the community and making them less dependent on their children and keeping them at the centre of households rather than in old age homes. Poor children often join or remain in pensioner households, so grant income also offers them a major source of support in the absence of enough jobs. This has, however, a variety of effects. It is perhaps one of the reasons why there is less migration to urban areas and a smaller informal sector in South Africa than one would usually expect in a middle income country with such high levels of unemployment. (Kingdon & Knight 2004; Klasen & Woolard 2002). Social grants also have important impacts on savings behaviour and private provision for retirement. The poorest in South African society are those who now have neither social grants, nor employment. The child support grant was party intended to cover this gap, and may have done so quite successfully as far as that is possible in a country with such very high levels of unemployment. THE SOUTH AFRICAN SOCIAL SECURITY SYSTEM, ITS SIZE AND GROWTH Social security systems provide protection against risks of income loss due to contingencies such as old age, unemployment, disability, or injuries sustained at work. They are thus a form of

6 consumption smoothing. They redistribute income between generations, amongst the insured according to risk and vulnerability, and across time, which is the consumption smoothing function. The main parts of the South African social security system are: Occupational (social) insurance Social assistance (grants) Informal insurance (social networks, the extended family) Means tests tie social insurance and grants together and make the social security system indeed a system that is linked and integrated. In South Africa the enduring gap in the safety net arises from the fact that there is such large unemployment. South Africa social assistance, or social grants, are very large in scale for a developing country. There are 9 million recipients in a population of about 46 million, i.e. every 5 th member of the population receives a social grant. About 3.5% of GDP is spend on social grants. This amount is larger than the GDP of almost half the countries in the world, including some 35 African countries. The money spent on grants now even exceeds Zimbabwe s GDP, after its recent economic collapse. The good targeting of the social grant system arises from the fact that it is means tested, that the programme has been expanded to the rural poor to a far greater extent than in any other middle income country, and that there are only small leakages to the non-poor. The main components of the social assistance programme in South Africa are social pensions, disability grants, and then child support grants or CSG s, which have expanded rapidly in recent years. The origins of the South African social assistance system should be sought in the apartheid era attempt to create a welfare state for whites. A racially differentiated social assistance system evolved from the 1920 s. Over time, grants were increasingly extended to other race groups, even though initially at much lower benefit rates. This discrimination in the benefit levels was gradually reduced from the late 1970 s onwards, and ironically it was eliminated in 1993, i.e. just before the new government came into power. Figure 5 shows that the grants were unified in terms of benefits levels by 1993, through a reduction in the real value of the maximum social pension received by whites and an increase in that received by blacks, so that all groups ended

7 roughly at the same per capita grant level as was historically received by coloured and Indians, in real Rand terms. Thus it is ironic that South Africa, despite apartheid, ended up with a system of social grants that is far more progressive than seen in any other developing country in world. R 6 000 R 5 000 Figure 5: Real value of maximum annual social pension (1990 Rand) Whites Coloureds Indians Blacks R 4 000 R 3 000 R 2 000 R 1 000 R 0 1975 1980 1985 1990 1995 2000 Figure 6 shows the numbers of the different forms of grants. There were just over 5.6 million child support grants, followed by 2.1 million social old age pensions, (including a small number of war veterans pensions) in April 2005. The third most common grant was the disability grant, with 1.3 million such grants being paid. Figure 7 shows the budgeted values of grants for the year 2006/2007. The total amount budgeted for grants spending in this year was R57 billion, which was almost $10 billion at the then reigning exchange rate. In this case, most money is spent on the old age pension (R21.4 billion), followed by the child support grant (R16.6 billion) and the disability grant (R15.5 billion). Other grants are relatively small.

8 Figure 6: 6.0m 5.63m 5.0m 4.0m 3.0m 2.0m 2.10m 1.31m 1.0m 0.0m 0.09m 0.26m 0.02m O ld age pensions Disability grants C hild support grants C are dependency grant F oster care grants G rants in aid Figure 7: Budgeted grants by value, 2006/07 (Total R57bn, ±US$9.5bn) R25bn R20bn R15bn R21.4bn R15.5bn R16.6bn R10bn R5bn R0bn R2.4bn R1.0bn Old-age pensions Disability grants Foster care grant Care dependency grant Child support grants Figure 8 shows that social assistance spending in South Africa, at 3.5% of GDP in 2006, is even high when one compares it to social assistance spending in western Europe at the height of the welfare state, in 1980. Only Denmark of the countries shown here had a higher social

9 expenditure ratio than is presently the case for South Africa. Even though one would expect such countries to have lower levels of expenditure, given smaller poverty levels, it is nevertheless instructive that countries that pay such a lot of attention to dealing with welfare issues generally spent less on social assistance than is the case in South Africa. Figure 8: Social assistance spending ratio (% of GDP)(Western Europe 1980, SA 2006) 5.0% 4.5% 4.0% 3.5% +-3.5% 3.0% 2.5% 2.0% 1.5% 1.54% 1.0% 0.5% 0.0% Belgium Austria Italy Switzerland Germany Norway Netherlands UK Finland France Sweden South Africa 2006 Denmark TARGETING OF SOCIAL GRANTS It is customary to think of targeting of social grants in terms of errors of exclusion and errors of inclusion (or leakage). Errors of exclusion occur when people who are poor do not get access to benefits. Errors of inclusion, on the other hand, occur when there is a leakage of funds meant for the poor to those who are not poor. This arises from the way in which targeting mechanisms function. There are a number of alternative targeting mechanisms used for grants or alternative subsidies meant for the poor: Means testing is generally used in South Africa for social grants.

10 Categorical subsidies or transfers are also sometimes linked to as indicator targeting. So, for instance, grants for pregnant woman and young children that were introduced at the time of the transition to democracy in South Africa. Though pregnant woman and young children are not all poor, the fact that these services are subsidised does mean that many poor people benefit from it, although there would be errors of exclusion and errors of inclusion involved in this. With geographic targeting, certain geographic areas are targeted because they house many of the poor. To some extent one can say that the child support grant has an element of this, as has the South African housing subsidy. A fourth way of targeting is self targeting, for instance in the Working for Water project, where low wages are paid to people willing to work on certain projects. Only the poor are usually willing to work at such low wages, with the result that it targets itself: only the poor are the beneficiaries. Targeting administration also has costs attached to it; these can be as high as 3% to 8% of the actual value of the grants. Such costs rise with attempts at improving targeting accuracy. There are further costs attached to targeting such as incentive cost including moral hazard; introducing targeting often leads to changes in behaviour of potential beneficiaries that are suboptimal from society s viewpoint. An example may be if people were to stop working so that they can claim an unemployment grant. A further cost of targeting is stigma. In many Anglo Saxon countries, where means testing is common, some people do not want to be seen as being on welfare, therefore they refuse to apply for such a grant, even though they would qualify for it. In South Africa, in some school feeding programmes, food is sometimes only provided to the poor; some poor children refuse to take part because of the stigma associated with it. In such cases, it is better to consider making the grant available to everybody, rather than to exclude some through the stigma effect. Finally, there are also political economy costs associated with targeting. For instance, the political support for a grant targeted only at very poor people may be less than that of a grant reaching a larger group, as is the case with the South African old age pension. One ways of judging the targeting of grants is to use concentration curves. A concentration curve is derived in a similar fashion as a Lorenz curve, i.e. one first arranges the population from

11 poorest to richest, and then derive the curve. In the case of the Lorenz curve, cumulative income is shown on the Y-axis while cumulative population share is shown on the X-axis. A Lorenz curve further from the diagonal implies more inequality. In the case of the concentration curve, the cumulative benefit or subsidy or grant is shown on the Y-axis, whilst the cumulative percentage of the population is shown on the X-axis. If the concentration curve lies above the Lorenz curve, such spending is redistributive. However, it is only where the concentration curve also lies above the diagonal where one can say that the concentration curve is strongly equity enhancing. In such a case, we can say that the concentration curve shows targeting towards the poor, or that it is per capita redistributive. Figure 9 shows a Lorenz curve and a concentration curve for social grants for 1997. The Lorenz curve lies below the diagonal as it always does, but the concentration curve lies above the diagonal. Whilst it is not possible for the bottom 40% of the population, for instance, to earn more than 40% of the income, it is indeed possible for them to receive more than 40% of the spending on social grants. In this particular case, applying to 1997, the bottom 40% of the population gets approximately 70% of the spending on social grants. Figure 10 shows different concentration curves; here the concentration curve for social grants lies much above all the other curves. This shows that spending on social grants is far better targeted at the poor, who get more than their share of overall spending on such grants, whilst this degree of targeting is seldom experienced for other social expenditure programmes. The table in Figure 11 shows the concentration indices for different programmes for both 1995 and 2000. A concentration index is calculated very much like a Gini index, but its value is negative when the concentration curve rises above the diagonal. As can be seen in this table as well as in Figure 12, the concentration indices for social grants in both 1995 and 2000 were negative, and considerably below that for other social expenditure items. Put differently, spending on social grants was far better targeted at the poor than spending on other social programmes. Figure 13 supports this conclusion, by showing that social spending on the poorest 40% of households approximately 50% of the population exceeded three-quarters of overall spending on social grants. This is much more than the share of these four deciles of the population in the same two years from other social programmes.

12 Figure 9: Lorenz curve & concentration curve for social grants, 1997 Cum. % of Income & Spending 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Cumulative % of Population Figure 10: Concentration curves: All spending 2000 Cumulative % of service 100% 90% 80% 70% 60% 50% 40% School education 30% Tertiary education Health (net) 20% Social grants 10% Housing Total 0% 0% 20% 40% 60% 80% 100% Cumulative %of population (arranged from poorest to richest)

13 Figure 11: Concentration indexes by programme, 1995 and 2000 School education Tertiary education Health: Hospitals Health: Clinics Social grants Housing Total 1995-0.016 0.484-0.014-0.103-0.434-0.018-0.057 2000-0.104 0.497-0.057-0.132-0.431 0.007-0.120 Change 1995-2000 -0.088 0.013-0.043-0.029 0.003 0.025-0.063 Figure 12: 60% 40% 1995 2000 Concentration indices by sector, 1995 & 2000 0.484 0.497 20% 0% -20% -0.016-0.104-0.045-0.082-0.018 0.007-0.057-0.120-40% -60% -0.434-0.431 School education Tertiary education Health (net) Social grants Housing Total

14 Figure 13: 60% 40% 1995 2000 Concentration indices by sector, 1995 & 2000 0.484 0.497 20% 0% -20% -0.016-0.104-0.045-0.082-0.018 0.007-0.057-0.120-40% -60% -0.434-0.431 School education Tertiary education Health (net) Social grants Housing Total Figure 14 shows that different social grants were not equally well targeted in 2000. The best targeted was the social old age pensions followed by disability grants, while social child support grants and maintenance grants were less well targeted. However, one should consider that the CSG programme was not yet fully rolled out in 2000, thus it may not have reached poorer and particularly rural areas as well as it should in this year; this the situation may since have changed considerably for the better. Figure 15 shows the geographic distribution of different grants. From this it can be seen that compared to the population distribution, social grants spending was far better targeted to the rural population, who received 56% of social grants and only contained 42% of the overall population. Other forms of social spending did not reach rural areas as well. Table 16 shows that the poorest decile of the population received a considerable proportional increase to their income from social grants in 2000. In contrast, the richer deciles received a proportionally smaller increment.

15 Cumulative % of service 100% 90% 80% 70% 60% 50% 40% 30% Figure 14: Concentration curves for various social grants, 2000 SOAP, war vets 2000 Disability 2000 CSG & maintenance 2000 20% All Grants 1995 10% All grants 2000 0% 0% 20% 40% 60% 80% 100% Cumulative % of population (arranged from poorest to richest) Figure 15: Geographic distribution of social spending & population, 1997 100% 80% 60% 40% 20% 35% 31% 34% 24% 25% 14% 28% 26% 30% 33% 33% 25% 28% 30% 32% 65% 37% 38% 43% 51% 56% 42% 42% Metropo litan Other urban Rural 0% Housing School education Tertiary education Water Health Social security Population All social spending

16 Figure 16: Cumulative density curves for total population, children (0-6, 7-14), pension aged and narrowly unemployed, 1995 % of individuals below each income level 70% 60% 50% 40% 30% 20% 10% Under 7s Children 7-14 Narrow unemployed Total population Pension aged 0% R 0 R 500 R 1 000 R 1 500 R 2 000 R 2 500 R 3 000 R 3 500 R 4 000 Per capita income of household containing individual The cumulative density or cumulative distribution curves shown in Figure 17 are interesting. The income per capita of individuals is here derived by allocating to them the income per capita of the household in which they reside. These curves show, firstly, that there is less poverty amongst old people than there is amongst the population as a whole. In the case of pensioners, a smaller proportion of them fall under any poverty line one can draw, compared to the overall population. Put differently, there is so-called stochastic poverty dominance irrespective of the poverty line chosen, the pension age have less poverty than the population as a whole. This is undoubtedly the result of the great reach of the old age pension, which makes it an important source of income even in relatively poor rural areas. In contrast, in 1995, when this Figure was derived, the children in the age groups 0 to 6 and 7 to 14 experienced greater poverty than the population as a whole, and even than the narrowly unemployed. However, it is impossible to distinguish between children 0 to 7 and children 7 to 14 in terms of this graph, so it is clear that the nature and extent of poverty in these two groups did not really differ. Therefore, it made sense that after the introduction of the child support grant, it was extended from the younger age group to the older group of children as well, with major results in terms of poverty reduction. It is noticeable also that the narrowly unemployed, although poorer than the population as a whole, are not as

17 badly off as children, who more often find themselves in poorer households, than was the case for those narrowly unemployed. (Note that this was, however, before the introduction of the CSG.) Figure 17: Cumulative density curves for total population, children (0-6, 7-14), pension aged and narrowly unemployed, 1995 % of individuals below each income level 70% 60% 50% 40% 30% 20% 10% Under 7s Children 7-14 Narrow unemployed Total population Pension aged 0% R 0 R 500 R 1 000 R 1 500 R 2 000 R 2 500 R 3 000 R 3 500 R 4 000 Per capita income of household containing individual THE MEANS TEST AND ITS OPERATION Means testing is essential for ensuring that funds target the poor rather than the less poor part of the population. Means testing has always been a part of the South African social grant system, and the major social grants are linked to explicit means tests. The first question one should ask when discussing means testing is whether it serves as a means of inclusion or a means of exclusion. It the means test is set at a high level, such as the case with the social old age pension, it is largely used to exclude a small segment of the population with a fairly high income from benefiting. In South Africa s case, approximately 80% of those who qualify by age for old age pensions do receive such pensions. In contrast, if means testing is to be used to focus only on the bottom part of the population, i.e. to target a relatively small group of very poor people, it could be much more difficult because at such low income levels the poor cannot always be clearly distinguished from the less poor. This is linked to how incomes are usually distributed, i.e. they tend to follow a log normal distribution, i.e. there is a lot of bunching of incomes around low

18 income levels. It is therefore very difficult to draw a poverty line that runs through the middle of this bunching of incomes and to separate those who are very poor from those who are slightly less poor, as has been attempted with the child support grant. For this reason the child support grant means test is quite difficult to operate and there are probably many errors of inclusion as well as errors of exclusion, plus a large degree of discretion in application of this grant, which makes the means test problematic. Figure 18 shows the operation of the means test for social old age pensions in the year 2004. The means test formula reads that the value of the benefit equals 1.15 times the maximum grant minus 0.5 times private income derived from other sources. Of course, this is subject to the fact that the maximum grant cannot be larger than the set maximum, which was R9 120 in 2004. In addition, there is a provision that when the grant falls below 40% of the maximum, there will be no grant payable, as it is considered too expensive to administer such a low level grant. This implies that the dotted line in Figure 18 is the curve from which the rest of the means test can be derived. The maximum level is where the first kink in that graph occurs, and the 40% cut-off is where the second kink occurs. There is another proviso that if assets of the individual exceed 30 times the maximum benefit, that is R273 600 for a single person, no such a grant will be paid. It is crucial where the centre of the lognormal distribution lies. Until recently, it was probably below the threshold value of where the sliding scale started to operate, as very few black people had retired with private retirement income. This is changing, however, as changes in the 1970 s had brought more black people into the social insurance net, so that many of them are now retiring with some level of private pension. That means that more and more people are retiring who have incomes somewhere where the sliding scale is suppose to be operating, making application of the means more difficult and accurate information as to people s actual private incomes essential. For this reason there is a strong case to be made to also include the top 20% of the retired population in benefits from the social pension and then to take back some of this income through the tax system. This means in practice that the social old age pension then becomes a universal pension. Because South Africa is so close to that situation, and because of the increase in administration costs associated with trying to deal with a differentiated pension for different people along the sliding scale, it may be useful to again consider this. The Smith committee as well as the Mouton committee and the National Consultative Retirement Forum, all

19 considered the option of eliminating the means test for the old age pension, but all of them again came to the conclusion that it was best to retain it, as a number of pensioners was growing more rapidly than the economy was, and consequently there would have been strong pressure on the fiscus as a result of the expansion of the number of beneficiaries. However, the growth rate of the economy has improved and as a result of that it may again be worth considering the abolition of the means test for old age pensioners in the present conditions. Figure 18: Operation of the means test, 2004 Social grant per annum R 12 000 R 10 000 R 8 000 R 6 000 R 4 000 R 2 000 R 0 R 0 R4 000 R8 000 R12 000 R16 000 R20 000 Private income per annum Grant max = R760x12=R9120 p.a. No grant paid if private income exceeds R13 680 p.a. or if assets exceed 30x9120 =R273 600 for single person Slope of line = 0.5 (R20 more income reduces pension by R10) 1.15*MaxGrant-.5*PrivateIncome Social pensions therefore act as a major source of income, particularly for the rural poor but such pensions also affect household formation. A number of studies in South Africa have shown that old people tend to be an important source of household formation. (Klasen & Woolard 2002; Bertrand & Mullainathan 2000; Case & Deaton 1998; Case, Fin & McLanahan, 1999; Duflo 2000). Households in South Africa often form around income, when not all adult members of the household have access to own sources of income. Therefore children who are poor often remain in pensioners households or even join such pensioner households. This has an impact on choice of residence (urban or rural location), labour market participation, and functioning of the informal sector. On the other hand, it brings a significant source of security for the unemployed, who benefit from being attached to pension households. The poorest households are those with neither social grants nor employment.

20 The means test for social old age pensions has an incentive effect on the retirement age, as few people now find it worthwhile to remain in the labour force beyond the inception age for the old age pension; it has an impact on own provision for retirement (e.g. private pensions), as many people have little incentive to provide for own retirement, if the net result will only be that they receive less from the state; it has a major impact on the preservation of pensions when people leave their jobs and on their choice between whether they should belong to a pension or to a provident fund; it may have an impact on asset creation, as certain assets are less visible and therefore less likely to exclude people from benefiting from the old age pension; and it has, as mentioned before, a major impact on household formation. Regarding disability grants, although the means test operates in exactly the same way, the pressure for application of this means test falls on the application of the disability criteria rather than on the means test itself. This means test would in any case have to be retained, even should the old age pension system become universal, as the same conditions do not apply to disability grants. The means test for child support grants is based on a two-step procedure: First, the rural and squatter population and those with an income below R800 per month, are selected; Of this group, those with an income of above R1 100 are excluded. This means that people in rural and squatter areas receive the grant if their income is below R1 100, whilst people in urban areas receive the grant if their income is below R800 per month. There is thus a bias towards the rural and squatter population. Also, because a child support grant is given to the caregiver, this has implications for who would benefit from the grant, and may also influence where and with whom children reside. In many cases, children are left with grandmothers or relations in rural areas whilst mothers search for jobs in urban areas. Given the availability of grants, mothers may now more often prefer to keep children with them and therefore to themselves receive the grant in the urban areas, with the result that more children may migrate to the urban areas. Whether this is better for children s welfare or not is ambiguous.

21 It may be better to be with their mother than remaining with grandmother or relations in rural areas, but if the mother does not have the time to give children the care they need, this may be to their detriment. TRENDS AND PROSPECTS As was indicated before, poverty has declined substantially after the year 2000. Also, as was indicated, social grants were largely responsible for this decline. Even though more jobs were created, those jobs did not have the required poverty impact, as the poor were at the back of a very long job queue and therefore could very often not share in the benefits of such additional jobs. Thus social grants played an important role, but their very large magnitude now makes it difficult to expand them without running into severe budgetary constraints. In terms of future prospects, unemployment will probably decline strongly as a result of the slowing labour force growth, and continuation of the expansion in employment of recent years as a result of economic growth and a relatively healthy employment elasticity. Despite this, distribution will not improve much, as the beneficial impact on distribution of more jobs and rising wages in a tightening job market may be counteracted by the skill shortage, that will work to worsen wage inequality and thus to worsen overall income distribution. Despite stubbornly high inequality, poverty is likely to decline fairly strongly, unless income distribution worsens even much more, or unless a growth collapse prevents such poverty decline. In the circumstances, thus, the need for social grants may not again grow as strongly as in the recent past, but there will still be a strong need for such grants even in the most optimistic scenario. However, it is equally likely that the expansion of social grants will be limited by fiscal constraints. Thus the major changes in the social grants in the coming decade will probably be in the consolidation of the grant system. The changes recently mooted by the government may be an important precursor to such consolidation.

22 BIBLIOGRAPHY Bertrand, Marianne; Miller, Douglas; & Mullainathan, Sendil. 2000. Public policy and extended families: Evidence from South Africa. NBER Working Paper 7594. Cambridge, USA. National Bureau for Economic Research. March. 46pp. Case, Anne & Deaton, Angus. 1998. Large cash transfers to the elderly in South Africa. Economic Journal 108(450): 1330-61 Case, Anne; Lin, I-Fin; & McLanahan, Sara.. 1999. How hungry is the selfish gene? NBER Working Paper 7401. Cambridge, USA. National Bureau for Economic Research. October. 31pp. Duflo, Esther. 2000. Grandmothers and granddaughters: Old age pensions and intra-household allocation in South Africa. NBER Working Paper 8061. Cambridge, USA. National Bureau for Economic Research. December. 36pp. Kingdon, G. and J. Knight. 2004. Unemployment in South Africa: The nature of the beast. World Development 32(3), March: 391-408 Klasen, Stephan & Woolard, Ingrid. 2002. Surviving unemployment without state support: Unemployment and household formation in South Africa. IZA Discussion Papers 237. Institute for the Study of Labor (IZA), Bonn