WIDER Working Paper 2014/127 Poverty, inequality, and prices in post-apartheid South Africa Arden Finn,1 Murray Leibbrandt,1 and Morné Oosthuizen2 October 2014 World Institute for Development Economics Research wider.unu.edu Abstract: Post-apartheid poverty and inequality trends have been the subject of intensive analysis, yet relatively little attention has been devoted to the impact of differential price movements on the measurement of poverty and inequality. This paper aims to tell the story of the evolution of both money-metric and non-money-metric poverty and inequality in post-apartheid South Africa, and to assess the effect of prices on this story. Our results show that inflation over the latter half of the 2000s has been anti-poor and that accounting for differential price movements dampens the measured improvements in poverty and inequality. Keywords: inequality, poverty, prices, South Africa JEL classification: E31, D63, I32. Acknowledgements: The authors would like to thank participants at the UNU-WIDER Conference on ‘Inclusive Growth in Africa: Measurement, Causes and Consequences’, held in September 2013 in Helsinki, Finland, and at the Microeceonometric Analysis of South African Data (MASA) Conference, held in November 2013 in Durban, South Africa, for their helpful comments and insights. 1SALDRU, University of Cape Town; 2DPRU, University of Cape Town; corresponding author: [email protected] This study has been prepared within the UNU-WIDER project ‘Reconciling Africa’s Growth, Poverty and Inequality Trends: Growth and Poverty Project (GAPP)’, directed by Finn Tarp. Copyright © UNU-WIDER 2014 ISSN 1798-7237 ISBN 978-92-9230-848-3 Typescript prepared by Sophie Richmond for UNU-WIDER. UNU-WIDER gratefully acknowledges the financial contributions to the research programme from the governments of Denmark, Finland, Sweden, and the United Kingdom. The World Institute for Development Economics Research (WIDER) was established by the United Nations University (UNU) as its first research and training centre and started work in Helsinki, Finland in 1985. The Institute undertakes applied research and policy analysis on structural changes affecting the developing and transitional economies, provides a forum for the advocacy of policies leading to robust, equitable and environmentally sustainable growth, and promotes capacity strengthening and training in the field of economic and social policy-making. Work is carried out by staff researchers and visiting scholars in Helsinki and through networks of collaborating scholars and institutions around the world. UNU-WIDER, Katajanokanlaituri 6 B, 00160 Helsinki, Finland, wider.unu.edu The views expressed in this publication are those of the author(s). Publication does not imply endorsement by the Institute or the United Nations University, nor by the programme/project sponsors, of any of the views expressed. 1 Introduction Trends in poverty and inequality during the post-apartheid period have been the subject of intensive analysis in South Africa. The widespread poverty and extreme inequalities prevalent at the time of the democratic transition represented one of the key areas of policy focus for the first democratic government, as well as one of the sets of outcomes against which its performance has often been judged. While little data on household incomes and expenditures existed prior to the transition, regular nationally representative household surveys—collecting detailed income and expenditure data—have been undertaken since the early 1990s by both Statistics South Africa and other institutions. The effect of prices on purchasing power is typically given only passing attention in the South African literature on poverty and inequality: incomes or expenditures are deflated by a scalar derived from some version of the Consumer Price Index (CPI) in order to make comparisons over time. One of the key gaps in this literature with respect to the analysis of trends in poverty and inequality is, therefore, the effect of differential price movements across the distribution. It is hoped that this paper will contribute towards filling this gap by more purposefully considering the impact of prices on estimates of poverty and inequality. The paper starts by briefly reviewing the received wisdom on post-apartheid growth and poverty well-being using secondary literature, income data, and non-money-metric sources. This narrative is that poverty has gone down over the post-apartheid period in such a way that growth has been pro-poor, but that inequality has remained stubbornly fixed at the very high levels characterizing the start of the post-apartheid period. Section 3 reviews the available nationally representative expenditure data covering the past almost 20 years, with a view to choosing appropriate datasets for the analysis. It also considers the relevant available price data. Adding value to this story is the central task of the fourth section of the paper, which assesses the sensitivity of poverty and inequality estimates to differential price movements. Section 5 is the conclusion. 2 Existing evidence on the evolution of post-apartheid well-being 2.1 The narrative South Africa’s economy has undergone substantial changes since the fall of apartheid and the first democratic elections in April 1994. Economic growth stagnated during apartheid due to sanctions on international trade and investment, uncompetitive local industries, rigid exchange controls, restricted skills development, and high levels of poverty and inequality (Aron et al. 2008). After the first democratic election, economic sanctions were dropped, labour restrictions were lifted, and policies were put in place to advance the interests of African workers, who had been marginalized for many decades. Since the first democratic election, South Africa has had stable macro management and, as shown in Table 1, South Africa’s economy has grown steadily both in real and per capita terms. 1 Table 1: South African macroeconomic trends, 1993–2012 GDP (ZAR GDP growth (%) GDP per capita GDP per capita million) growth (%) 1993 1,065,830 1.2 28,277 -0.9 1997 1,214,768 2.6 29,582 0.5 2001 1,337,382 2.7 30,024 0.8 2005 1,571,082 5.3 33,176 3.9 2008 1,814,594 3.6 36,392 2.3 2010 1,842,052 3.1 36,079 1.9 2012 1,954,303 2.5 37,476 1.5 Avg. 1993–2012 1,470,001 3.2 32,031 1.5 Source: Updated from Leibbrandt et al. (2010), South African Reserve Bank (2013). Over the same period, the schooling system transformed from one characterized by highly skewed spending across racial groups to one based on equitable government funding. School enrolment rates rose, though learning achievements remain very poor in previously disadvantaged schools (Van der Berg 2007). The new, young labour market participants have more education, on average, than their parents had a generation ago. Two in five young adults graduate with Matric certificates (which is the qualification awarded for those who pass a set of nationally set, standardized exams at the end of the final years of secondary schooling). Other countries, such as Brazil and India, have seen education gains translate into productivity and employment growth, and large decreases in poverty and inequality. Job creation in a dynamic labour market served as the key pathway through which these societies generated high social returns to improved education and second-round effects to social transfers. South Africa has not made similar gains. Over the post-apartheid period poverty has fallen only sluggishly. Eighteen years after the first democratic election, the share of people living below a US$2 per day poverty line has declined by no more than 4 percentage points from 34 per cent in 1993 to 30 per cent in 2008. These gains are often attributed to social policy reforms (i.e. a massive expansion of cash grant transfers) rather than economic development (Leibbrandt et al. 2010). Of equal concern is the fact that inequality has risen further from its very high levels under apartheid (Leibbrandt et al. 2010). Just as the labour market was the key intermediary in the successes in Brazil and India, so the unsatisfactory performance of the labour market sits centre-stage in South Africa’s disappointing development outcome. A total of 2.74 million jobs (net) were created between 1993 and 2008, of which 2.5 million were targeted at skilled labour, while unskilled workers lost a total of 770,000 jobs (net). Over the same period, unemployment rates more than doubled from 14 per cent in 1993 to a peak of 29 per cent in 2001, before declining to 23 per cent in 2008. By the time of the economic crisis in 2010, the unemployment rate had reversed to 25 per cent, using the narrow definition of unemployment (National Treasury 2011).1 If discouraged workers—who have stopped looking for work ‘because they do not anticipate finding any’ are included in this definition—the figure is substantially higher at about 32 per cent (Statistics South Africa 2012c). Of the total population of 4 million unemployed, 75 per cent are long-term unemployed and many young job seekers report having limited or no formal work experience, even at age 30 (National 1 It should be noted, however, that the narrow unemployment definition changed slightly with the introduction of the Quarterly Labour Force Survey (LFS) in 2008, affecting estimates of the unemployment rate. 2 Treasury 2011). The informal sector is small, with only 6 per cent of South Africans in self- employment. The supply of labour is therefore primarily directed at jobs in the formal sector. In general, the labour market has not had a positive impact on poverty because of the failure to pull individuals from poor households into employment. This unemployment situation worsened between 1993 and 2008, especially for those in the poorest households. The number of no-worker households has increased by 3 per cent in the last 15 years, pushing up the number of households relying on assistance, especially child grants, as their main form of income. Indeed, the improved aggregate poverty situation is due to increased support from social grants, and not from the labour market. Even in one-worker households, the poverty incidence remains high. Because of high living costs and the fact that many workers are in low-paid employment, the presence of an employed person in a household is not a guarantee of escaping poverty. The poverty impacts of pervasive unemployment are compounded by a social protection gap that exists for unemployed adults, as social cash grants target people who are not expected to be economically active: children, pensioners, and people with disabilities. This leaves unemployed adults deeply dependent on goodwill transfers from within their communities, placing a large care burden on communities and deepening poverty. Leibbrandt et al. (2010) go further to show that labour markets play a dominant role in driving inequality. Even though the average share of wage income in total income has remained constant at around 70 per cent over the post-apartheid period, wage income has contributed between 85 per cent and 90 per cent of the total inequality in household income over the years 1993, 2000 and 2008. In contrast, state transfers are shown to make up less than 1 per cent of the overall Gini coefficient. Reducing unemployment and creating a better-functioning labour market is the major economic and social challenge in South Africa, which is explicitly recognized by the South African government. Indeed, employment creation has emerged as a top policy priority of the ANC-led government. Its New Growth Path strategy aims to create 5 million jobs by 2020, with ‘the creation of decent jobs at the centre of its economic policy’ (Zuma 2011). In his 2011 State of the Nation Address, President Zuma (2011) declared year 2011 to be the year of job creation and announced the government’s intention to spend R9 billion on job creation. Despite this commitment and like many other countries around the world, there is a lack of solid evidence to back this commitment. 2.2 Trends in money-metric poverty and inequality As already noted, most of the analysis of poverty and inequality in post-apartheid South Africa has used income as the welfare measure. Studies using household consumption spending per capita have generally been restricted to one or two points in time (for example see Klasen 1997). Leibbrandt et al. (2010) use household income per capita to track changes in inequality and poverty between 1993 and 2008, and include a short section on the comparability of income and expenditure in the datasets that were used for analysis. The authors conclude that income and expenditure track each other closely in the 2008 first wave of the National Income Dynamics Survey (NIDS), but are significantly different in the 1993 Project for Statistics on Living Standards and Development (PSLSD) data. Leibbrandt et al. (2012) find that the Gini coefficients are the same in 2008 (0.66) whether measured by adult equivalized income or expenditure, but are very different (0.61 compared to 0.51) in the 1993 data. It is understood that the expenditure data in 1993 are not as reliable as the income data, thus motivating the focus on an income-based comparison. Before presenting findings based on expenditure data, we briefly present some of the quantitative analysis that has been undertaken in support of the above narrative using income data from national household surveys. Figure 1 shows three post-apartheid real income per capita densities 3 as an example of extensive empirical work that has been undertaken on the distribution of income (Fedderke et al. 2003; Hoogeveen and Özler 2006; Simkins 2004; Van der Berg et al. 2006, 2008). It provides a representative snapshot of the weight of evidence that has been marshalled in support of the above narrative.2 We begin by considering changes across the entire income distribution between 1993 and 2010. In Figure 1, the income distributions for 1993, 2000, and 2010 are all plotted on the same set of axes. A poverty line is inserted on the graph as a reference point. It is a cost-of-basic-living poverty line developed by Hoogeveen and Özler (2006). This means that we have a lower poverty line of ZAR573 per person per month and an upper poverty line of ZAR1,056 per person per month in real 2010 Rands. The lower poverty line of ZAR573 is superimposed on the graph. The graph shows that the distribution of income shifted rightwards at almost all points between 1993 and 2010. This is in line with the generalized Lorenz curves presented in Figure 2 (Panel B) which show that average real income increased for the population as a whole over the period. At the bottom of the distribution, the major shift took place between 1993 and 2000, with relatively little movement between 2000 and 2010. This pattern is reversed as we move up the distribution (but remain below the poverty line) where we see that there was a significant rightward shift from 2000 to 2010. There is evidence of a significant rightward shift at the very bottom of the distribution and poverty dominance analysis confirms a reduction in poverty. However, this shift does not represent a dramatic decrease in poverty. According to the poverty head count ratio—simply the proportion of the population living below the poverty line—the poverty rate at the lower poverty line stood at 56 per cent in 1993 and remained steady at around 54 per cent for the later years in our analysis. The reduction in poverty incidence using the upper poverty line also stands at 2 percentage points—from 72 per cent in 1993 to 70 per cent in the late 2000s. The rightward shift at the bottom of the distribution is reflected by consistent decreases in the poverty gap rate, which gives us a broad measure of the depth of poverty in society. The main driver behind increasing incomes at the bottom of the distribution is the rapid expansion of the government social support programme. The importance of state grants in raising these incomes is highlighted in Leibbrandt et al. (2010), who note that in 1993 one-fifth of households were beneficiaries of state grants, while in 2008 this proportion had climbed to one-half and Leibbrandt and Levinsohn (2011), Bhorat and Van der Westhuizen (2011), and Woolard and Leibbrandt (2011) show clearly that social grants reduced both poverty and inequality. 2 This section is based on Finn et al. (2013a). 4 Figure 1: Distributions of income 1993, 2000, and 2010 4 . Poverty line R573 (2010 Rands) 3 . y nsit2 e. D 1 . 0 0 2 4 6 8 10 12 Log of real household income per capita 1993 2000 2010 Source: Finn et al. (2013a, from own calculations using PSLSD 1993, Income and Expenditure Survey [IES] 2000 and NIDS wave 2 2010). The expansion of government grants was not complemented by a reduction in the unemployment rate. The labour market is by far the most important factor to consider when decomposing poverty (see Leibbrandt et al. 2010). While the expansion of state support has helped to lower poverty, the persistently high levels of unemployment have prevented poverty reduction on a substantial scale. Decomposing poverty rates by the labour market status of household members emphasizes the crucial role of finding employment in reducing poverty. In 1993, almost 90 per cent of individuals living in a household where nobody had a job were living below the poverty line. This reduced somewhat to around 80 per cent in the period under study, but it remains very high. In fact, almost half of all the poor in the country live in a household where not one person is employed. This is in contrast to the poverty share of those living in households with two or more workers, which stands at around 17 per cent. Decomposing poverty by different groups reveals some interesting trends. Leibbrandt et al. (2010) find that the decrease in poverty in post-apartheid South Africa is driven mainly by a fall in the poverty incidence among Africans, and particularly African males. Poverty rates for this group fell from 66 per cent to 60 per cent, while the corresponding figures for African females are 72 per cent and 68 per cent. Despite these changes, the African share of overall poverty remained constant at 93 per cent in 1993, 2000, and 2010. This far outweighs the African share in the overall population, which is close to 80 per cent. A great deal of rural-urban migration took place in South Africa in the period under study. Our data reflect that the share of urban residents in the population rose from 49 per cent in 1993 to 60 per cent in the late 2000s. As a result of this movement, the urban share of total poverty rose from 30 per cent to about 43 per cent. That said, the poverty rate in rural areas was higher than in urban areas for any choice of poverty line. 5 We now move now to a discussion of inequality. South Africa has been recognized for a long time as having among the highest levels of inequality in the world and, of all countries that have reasonably good survey data, the only countries with similar levels of inequality are a handful of comparable countries from the two ‘extra-high’ inequality regions of the world, namely Latin America and Southern Africa. In panel A of Figure 2, we plot three corresponding Lorenz curves. In panel B, we do the same but with generalized Lorenz curves. The former gives a graphical measure of income inequality while the latter provides a graphical measure of social welfare through its inclusion of both inequality and mean income. The Lorenz curves suggest a high level of inequality. The richest 20 per cent of people earn about 70 per cent of the total income, and the second richest about 20 per cent of total income. Thus the poorest 60 per cent together only earn about 10 per cent of the total income in the population. This is approximately true regardless of which dataset is being used, and is exceptionally low by international standards. The primary observation is that the distributions do not vary much with time. In this case, the 2000 graph lies slightly below 1993, and the 2010 distribution almost perfectly overlaps with 1993. The big picture conclusion is that inequality has remained mostly stable and stubbornly high over the post-apartheid era (see Leibbrandt et al. 2010; Van der Berg 2011). Whereas Lorenz curves are unaffected by the mean of the income distribution, the generalized Lorenz curves of panel B are shifted up by mean income. If everyone in a society earned twice as much as they previously did, the new generalized Lorenz curve would rotate upwards, whereas the corresponding Lorenz curve would remain unchanged. What we observe from panel B is that the 1993 distribution is always below the 2000 distribution, which in turn is always below the 2010 distribution. Thus, panels A and B together reflect a society with stable inequality but with rising mean incomes amounting to an improvement in aggregate welfare over this time period. However, an increasingly pressing policy focus has developed as to why South Africa’s inequality seems to be so stubbornly persistent. Some of the evidence points to the emergence of a small but well-paid black professional class.3 Some researchers have emphasized the importance of unemployment and earnings.4 A third line of thinking has considered the high rates of return to tertiary qualifications in conjunction with wide variations in the quality of primary and secondary schooling.5 3 Hoogeveen and Özler (2006) find increases in inequality between 1995 and 2000, and attribute this mostly to increases in inequality among the African subpopulation. They also observe that the returns to education increased during this time period, particularly for Africans with high levels of education. See also Van der Berg and Louw (2004). Leibbrandt and Levinsohn (2011) support the contention that the share of within racial group inequality has risen over the post-apartheid period, but caution that the between group component remains exceedingly high by international standards. 4 See Leibbrandt and Levinsohn (2011) for decomposition work supporting this argument and Leibbrandt et al. (2010) for a review of the literature on this issue. 5 See, for example, Van der Berg (2009), Branson and Leibbrandt (2013a, 2013b) and Pellicer and Ranchhod (2012). 6 Figure 2: Lorenz curves 1993, 2000, and 2010 1 Panel A e m8 o. c n of i on .6 rti o p o r p e .4 v ati ul m u.2 C 0 0 .2 .4 .6 .8 1 Cumulative proportion of population 1993 2000 2010 e 000 m 2 o Panel B c n of i n 00 o 5 rti 1 o p o r p ative 1000 ul m u c d 0 e 0 al 5 c s n a e M 0 0 .2 .4 .6 .8 1 Cumulative proportion of population 1993 2000 2010 Source: Finn et al. (2013a, from own calculations using PSLSD 1993, IES 2000 and NIDS wave 2 2010). 7 Figure 3 follows on to provide a representative snapshot of the empirical work that has been undertaken to understand the drivers of these changes in the densities. It shows the share of income sources in total household income by income quintile in 2008. The proportion of income derived from wages increases linearly by income quintile. If a person is a member of a household situated in the poorest five deciles, the person is likely to receive relatively little wage income and to depend quite heavily on government grants and subsidies. Figure 3: Share of household income from various sources, 2008 100% 90% 80% 70% Investment 60% Remittances 50% Wages 40% Old age pension 30% 20% Disability 10% Child grants 0% 1 2 3 4 5 Quintile Source: Leibbrandt et al. (2010). 2.3 Trends in non-money-metric poverty Most studies in the post-apartheid era have focused on the assessment of trends in money-metric poverty and inequality, using income and/or expenditure data. Money-metric measures of welfare are extremely important in terms of our understanding of poverty and inequality in South Africa, and have the distinct advantage of being measured in consistent and easily comparable (currency) units. However, the concept of welfare extends beyond the simple flow of income or expenditure into and out of a household and includes, among other things, the various assets accumulated by households over time. Further, in the context of public policy, many government interventions involve the transfer of assets and provision of services to households that are not picked up in income measures and are not necessarily easily valued in currency terms. Such assets include, for example, the provision of sanitation services or housing. The assessment of trends in non-money-metric welfare is less commonly attempted in South Africa when compared with the volume of publications dedicated to the assessment of income/expenditure poverty and inequality. One reason for this is that the aggregation of various disparate assets and services into a single measure of welfare for comparison is a complex task. Fortunately, although various methods have been devised to allow such comparisons, very few published analyses that cover the period after 2005 have been located. The non-money-metric welfare story of the post-apartheid era is considerably more straightforward than the money-metric story. Not least among the reasons for this is the fact that measures of non-money-metric welfare include a large number of services and assets that are directly impacted by the state’s roll-out of services as it addresses some of the infrastructural and other inequalities inherited from apartheid. Thus, the provision of low-cost housing, the provision 8