Dualism or underdevelopment in South Africa: what does a quantitative assessment of poverty, inequality and employment reveal?

Abstract

Debate over the nature of South African society has continued through the first decade of democratic government, including the question of a ‘First’ and a ‘Second’ economy. This paper focuses on the numbers and characteristics of those who might fit into the ‘Second’. The data examined suggest that both poverty and inequality have increased in South Africa since 1993, while also suggesting the presence of poverty traps that may hinder mobility and prevent individuals getting ahead despite the dramatic economic and political reforms since 1994. We estimate that some 6.2 million could be described as belonging to the ‘Second’ economy. However, we argue that there is more to the relationship between those who are marginalised and those who are benefiting from South Africa's achievement than a dualistic and disarticulated ‘First’ and ‘Second’ economy. Linkages of inclusion and exclusion determine the size, characteristics and future of the two economies.

1. INTRODUCTION

The success of South Africa's transition to democracy is linked to the extent to which the political, social and economic reforms introduced since 1994 can address the legacy of purposeful and institutionalised underdevelopment resulting from apartheid policies. It is therefore a matter of some concern that recent research has shown that the intransigence of social and economic forces set in motion by apartheid policies may lead to the persistence of poverty even though many other aspects of the South African political economy are being transformed (Moser, 1998; Carter & May, 1999). The result can be seen in the peculiar poverty profile of South Africa that some may claim continues to represent the dualistic development problematique of plenty amidst poverty identified by economist Jill Nattrass more than two decades ago (Nattrass, 1983). Others may argue that this represents the continuance of more active processes of underdevelopment suggested by Harold Wolpe a decade earlier (Wolpe, 1972), requiring nothing short of a complete transformation of the foundations of the South African economy.

While debate over the nature of South African society has continued through the first decade of democratic government, this question of dualism was revisited during 2004, and has been referred to as depicting the characteristics of a ‘First’ and a ‘Second’ economy plagued by the dynamics of uneven development. The first is described as ‘the modern industrial, mining, agricultural, financial, and services sector of the economy that, every day, becomes ever more integrated in the global economy. It is this sector of our economy that produces the wealth’ (Mbeki, 2003). The second ‘constitutes the structural manifestation of poverty, underdevelopment and marginalisation in the country’ (Mbeki, 2004).

In this paper we focus on the numbers and characteristics of those for whom this characterisation of a ‘Second’ economy might apply. We comment on the data problems involved in arriving at an accurate estimate of the scale and trends of poverty and inequality in South Africa, before ending with an assessment of how these might be analysed for future policy and the potential usefulness of a dualistic mode of analysis.

2. POVERTY AND INEQUALITY

2.1 1990–1995

The experience of dualism mentioned by Nattrass (1983) can readily by seen in South Africa's social indicators of the early 1990s with the Human Development Index (HDI) developed by the United Nations Development Programme (UNDP) revealing significant spatial and racial differences. In 1991, while white South Africans had an HDI similar to that of Canada or Israel, the HDI score for Africans was lower than that of Egypt and Swaziland, and a provincial comparison shows that the score for the Limpopo Province was lower than that of neighbouring Zimbabwe (May et al., 2000: 23). In terms of indicators of poverty and inequality that rely on money-metric measurement, at the time of the transition in 1993 South Africa was described as among the world's most unequal economies, with a Gini coefficient measuring 0.58 (May et al., 2000: 25). More recent analysis using the 2000 Population Census data put the Gini coefficient for income as high as 0.73, worse than that of Brazil and of 33 other developing countries (Leibbrandt et al., 2005: 7).

Although income or consumption based poverty is not the only component of deprivation, most analysts would agree with Lipton (1997: 1003) as to the usefulness of this approach when identifying those who may be poor and for evaluating progress made by policy interventions. At the start of South Africa's transition in 1993 almost half of the population, some 19 million people, were categorised as poor using a national consumption-based poverty line, with over 60 per cent of Africans being poor compared to just 1 per cent of the white population (Klasen, 1997). The World Development Report of 2000 uses the same data to show that 11.5 per cent of the South African population lived on less than Purchasing Power Parity (PPP) $1 per day, some 4.3 million people, while 35.8 per cent of the population lived on less than PPP$2 per day, equal to 13.5 million people (World Bank, 2000: 64). At that time, poverty rates in South Africa could thus be compared to countries such as Bolivia (11.3 per cent), Colombia (11.0 per cent) or Cote d'Ivoire (12.3 per cent) in terms of the PPP$1 per day measure of poverty.

More recent South African measures of poverty based on a minimum acceptable standard of living suggest that poverty was more severe during the mid-1990s than the rather arbitrary international rules of thumb implied. Woolard and Leibbrandt (2001) use a range of thresholds to provide a rigorous analysis of poverty in South Africa in the period up to 1995. ¹ Using data from the 1995 Income and Expenditure Survey (IES), they conclude that some 40–50 per cent of South Africans could still be categorised as poor in 1995 (15.7 m to 19.7 m), while 25 per cent (9.8 m) could be categorised as ultra- or chronically poor. They also found that the poverty rate was far higher in rural areas than in urban (65 per cent of individuals compared to 22 per cent) and 27 per cent of rural dwellers were below half the poverty line, and thus were likely also to be chronically poor in the sense of being unable to escape poverty over time (Woolard & Leibbrandt, 2001: 59–60). In line with other studies, the Eastern Cape emerged as the poorest province in South Africa, containing 27 per cent of those likely to be chronically poor, while KwaZulu-Natal and Limpopo Province accounted for 19 and 17 per cent of the chronically poor respectively. Also in 1995, the poverty gap, a measure of the depth of poverty calculated to show the amount that is needed annually to wipe out poverty through a perfectly targeted transfer to the poor, was about R15 billion, or about 4 per cent of GDP. In the case of the two poorest provinces, the Eastern Cape and Limpopo, the poverty gap amounted to 11 per cent and 21 per cent of the provincial Gross Geographic Product (GGP) respectively (May et al., 2000).

2.2 1995–present

Despite some instability during the late 1990s, the South African economy has grown progressively in the post-apartheid period while social spending has increased alongside steady and substantial investment in infrastructure and basic services. As a result, an improvement in the social indicators just discussed might have been expected. Indeed, at PPP$11 240 per annum in 2001, South Africa's per capita GDP corrected for purchasing power parity now places it as one of the 50 wealthiest nations in the world. However, the strikingly poor social indicators of the country have persisted, with the result that South Africa was ranked 111th of 175 countries in terms of its HDI in 2001, down on its ranking of 93rd in 1992 (UNDP, 2002). ² In fact, South Africa is one of a handful of countries that has experienced a decline in the HDI since 1995. Despite being among the 35 largest economies in the world, the country now has life expectancies among the 30 worst, and projections of mortality suggest that these will deteriorate further as deaths from the AIDS epidemic increase (UNDP, 2003; Dorrington et al., 2004: 25).

A comparison of human development indicators in South Africa during this period with countries with similar income levels, provided in Table 1, shows that South Africa has continued to fare poorly when compared with other countries that are also ranked as middle-income in terms of their per capita Gross National Product (GNP).

Table 1: Comparison of social indicators from selected middle-income countries

Index	Turkey	Algeria	South Africa	Chile	Mexico	Malaysia
HDI rank	85	106	107	38	54	59
Life expectancy (years)	69.8	69.6	52.1	75.3	72.6	72.5
Adult literacy	85.1	66.7	85.3	95.8	91.4	87.5
Combined enrolment	62	72	93	78	71	66
GDP per capita ($)	6974	5308	9401	9417	9023	9068
HDI value	0.742	0.697	0.695	0.76	0.796	0.782
Population (2000) (m)	66.7	30.3	43.3	15.2	98.9	22.2
< $2 PPP poverty line	18	15.1	35.8	8.7	37.7	na
% pop using improved water	83	94	86	94	86	na
% 1yrs immunised for measles	80	83	82	96	95	88
Infant mortality	38	50	55	10	25	8
Under 5 mortality	45	65	70	12	30	9
Gini coefficient	41.5	35.3	59.3	56.6	53.1	49.2
Source: UNDP, 2002; World Bank, 2002. Most data refer to 2000.

Table 1 shows the inadequacy of using per capita GDP as the sole indicator of development in a dualistic economy such as South Africa's. All the countries to the right of South Africa in the table have a similar per capita GDP to South Africa, yet generally they performed far better on indicators such as life expectancy, infant mortality and adult illiteracy. Algeria has a similar HDI rank, but a per capita GDP that is almost half that of South Africa, while Turkey has a similar percentage below the PPP$2 per day poverty line and ranks 20 countries ahead of South Africa in terms of HDI rank, although with a per capita GDP that is three-quarters that of South Africa.

Changes in the incidence and severity of money-metric poverty since 1995 are less clear and have been a source of debate. Using large sample surveys collected by Statistics South Africa between 1994 and 1998, Budlender (1999: 93) suggests that both poverty and inequality may have increased, a finding supported by an official publication using the results of the 2000 IES (Stats SA, 2002). The average annual per capita income in 1995 was reported to be R12 135 adjusted to 2000 prices, higher than the per capita income of R11 755 per annum reported in 2000. However, many analysts have raised serious concerns about the quality of the data collected by this survey, pointing to methodological and weighting problems and evidence of sloppy fieldwork and data processing (van der Berg & Louw, 2003: 2; Meth & Dias, 2004: 61; Simkins, 2004).

In a rough attempt to manage some of these data quality problems, van der Ruit and May eliminate extreme outliers and use a PPP adjusted $1 a day poverty threshold. They conclude that poverty levels appear to have increased from the 11.5 per cent (4.3 m) calculated by the World Bank for 1993 to 19.8 per cent in 2000 (8.6 m), a doubling in the absolute number of people (World Bank, 2000: 64; van der Ruit & May, 2003: 23). ³ Using a different methodology both to clean the data and to measure poverty, and using the IES data from 1995 and 2000, van der Berg and Louw (2003: 18) conclude that the numbers in poverty reached 17 m people in 2000, an increase of 1.2 m people from 1995. Carefully managing the data quality problems and using different methodologies to derive a cost-of-calories poverty line, Hoogeveen and Özler (2005: 38) calculate that 12.6 m South Africans were living on less than PPP$1 per day in 1995 compared to 14.4 m in 2000, an increase of 1.8 m people, and that 22.9 m South Africans were living on less than PPP$2 per day in 1995 rising to 25.2 m in 2000, an increase of 2.3 m. They also show an increase in both the poverty gap index and the severity measure of poverty.

Irrespective of the poverty line chosen, inspection of the entire distribution of income suggests that there has indeed been a decline in well-being in the post-apartheid period when measured in terms of private consumption. Figure 1 depicts cumulative distribution functions from the 1995 and 2001 IES. ⁴ Following May et al. (2000), the data use household expenditure calculated per adult equivalent and include a modest economies-of-scale parameter. Figure 1 shows that for all incomes higher than R150 per adult equivalent per month, the distribution function for 1995 lies slightly above that for 2000. In other words, the percentage of individuals at successive income levels was consistently greater in 2000 than in 1995.

Figure 1: Cumulative distribution function

In an attempt to interpret these and other official statistics, Meth (2004) employs a mix of procedures to adjust for the data and measurement problems and concludes that the case for an increase in the numbers of people in poverty seems convincing. At the $2/day level (R214 per capita in 2002) the numbers of people in the lower of two expenditure groups identified as encompassing the potentially poor increased by about 3.5 m people between 1997 and 2002. Using the General Household Survey (GHS), this analysis provides an estimate that there were 17.4 m people in poverty in 2002. Attempting to take account of cash and non-cash transfers to the poor from government in the form of pensions and the CSG, water, electricity, health care, housing, sanitation, education and transport (perfect targeting to every eligible beneficiary being assumed), Meth amends this estimate, but still finds an increase of about 800 000 people (Meth, 2004: Table 6a). ⁵

Turning to the distribution of income, Hunter et al. (2003) use the data from the 2000 IES to show the continued high levels of income inequality in South Africa. The top decile in the distribution accounts for 49 per cent of total expenditure compared to just 8 per cent accounted for by the bottom four deciles, suggesting little change in income inequality since 1993. Hoogeveen and Özler (2005: 15) use the same data to show that the Gini index went up slightly from 0.56 in 1995 to 0.58 in 2000, while Fields et al. (2003) use household panel data to show that the Gini coefficient increased from 0.515 to 0.543 in KwaZulu-Natal. Van der Berg and Burger (2002: 10–11) try to take account of shifts in social spending in terms of their impact on the distribution of income adjusted for non-cash transfers. Noting that per capita incomes of the African elite have almost caught up with those of the white population, they calculate a Gini coefficient for South Africa of 0.66, similar to most other studies. Turning their attention to the impact of social spending, they show that social spending to the African population has increased from 51 per cent in the immediate transition period to 80 per cent in 1997. When income is adjusted for South Africa's relatively progressive taxes and for the strongly progressive non-cash transfers that are made, they conclude that the Gini coefficient can be re-estimated to 0.44. In particular, they conclude that fairly good targeting of Old Age Pensions and Disability Grants has resulted in rural areas receiving an unusually fair share of social spending compared to the situation in many developing countries. However, they do not attempt to measure the impact that this social spending might have had on the well-being of the poor, noting that the quality of the service delivered substantially determines this. As an example, while inequalities in terms of inputs to education have narrowed, the outcomes in terms of educational attainment have not (van der Berg & Burger, 2002: 18).

Putting aside concerns about the quality of official statistics, the problems of poverty measurement and the differences in the methodologies used to overcome these, analysis of other survey data supports evidence that the incidence and severity of income poverty continued to increase in the immediate post-apartheid period. Using panel data, Roberts (2001) and Carter and May (2001) adopt different methodologies to distinguish those who move in and out of poverty from those who are structurally poor. Both analyses suggest that while some 40–50 per cent of South Africa's population can be described as poor, around 20–25 per cent of the sampled African and Indian population in KwaZulu-Natal can be thought of as being chronically poor in terms of either a ‘time-spell’ or ‘trajectories’ definition. Furthermore, despite a 50 per cent chance of upward mobility from the poorest to the next expenditure class, Carter and May (2001) calculate that some 75 per cent of those measured as being ultra-poor in 1993 would still be below the poverty line in 1998 and that there was little chance that this group would ever escape poverty in the absence of fundamental change. They ascribe this to the persistence of multiple market failures that prejudice the chances of the poor making use of the assets that they do have, or accumulating additional assets. Using the same data, Woolard and Klasen (2005) identify three poverty traps that hindered the upward mobility of the poor between 1993 and 1998: large initial household size, low initial levels of education and low initial participation in the labour market. As noted in the introduction, the origins of each of these traps could be found in past apartheid policies.

3. WORK AND ECONOMIC INCLUSION

The statistics above give an approximate idea of the changes and scale of poverty and inequality in South Africa and suggest quite conclusively that both poverty and inequality have increased in South Africa since 1993. While some of the methodologies employed might be disputed, and much must be done to improve data quality, these data show that at least in money terms poverty and inequality remain rife in South Africa. They also suggest the presence of poverty traps that may hinder mobility and prevent individuals from getting ahead even in the face of the dramatic economic and political reforms that took place after 1994. But is this enough to assess the extent of dualism and to identify a Second economy that is somehow disarticulated from the mainstream of the South African economy? Recalling the roots of theories that sought to explain underdevelopment as an uneven but articulating relationship between the modern capitalist economy and earlier modes of production, it would seem appropriate to also consider the activities that are undertaken by those who might find themselves in this Second economy. ⁶ The data required to do this accurately are hard to come by and the remainder of this paper draws on two sources of information: official labour force and quality of life studies undertaken by Statistics South Africa between 1997 and 2001 and the KwaZulu-Natal Income Dynamics Study (KIDS), a panel survey of household collected between 1993 and 1998. ⁷

Table 2 shows employment data for the national economy over the period 1997–2004 using the various October Household Surveys (OHS) and the more recent Labour Force Surveys (LFS).

Table 2: Formal and informal labour market trends, 1997–2004

(1000s)	October Household Survey			Labour Force Survey
	1997	1998	1999	2000-1	2000-2	2001-1	2001-2	2002-1	2002-2	2003-1	2003-2	2004-1
	Formal	6 406	6 527	6 813	6 678	6 842	6 678	6 873	7 037	7 034	7 358	7 461	7 827
Commercial agriculture	496	726	804	757	667	699	666	734	811	865	832	930
Subsistence agriculture	163	202	287	1 508	965	653	359	792	520	420	350	337
Informal	966	1 077	1 574	1 820	1 934	2 665	1 873	1 767	1 703	1 845	1 899	1 834
Domestic work	992	749	799	1 001	999	914	916	972	875	1 005	1 022	1 013
Unspecified	71	108	93	115	306	227	146	57	65	73	17	44
Total employed	9 094	9 390	10 369	11 880	11 713	11 837	10 833	11 359	11 008	11 566	11 581	11 985
Unemployed (official)	2 451	3 163	3 158	4 333	4 082	4 240	4 525	4 738	4 837	5 250	4 627	4 611
Participation rate (%)	46.6	48.8	51.5	61.3	58.7	59.3	56.1	58.3	56.7	56.9	54.4	54.5
Discouraged	2 751	2 471	2 724	2 220	2 477	2 721	3 173	3 138	3 088	3 171	3 819	3 801
Not eco active	13 961	13 157	12 753	10 242	11 100	11 044	12 006	11 543	12 118	12 740	13 638	13 833
Total not employed	16 412	16 320	15 911	14 575	15 182	15 284	16 532	16 281	16 955	17 990	18 265	18 444
Total pop, age 15–65	25 505	25 710	26 280	26 454	26 895	27 121	27 365	27 673	27 984	29 555	29 917	30 429
Note: There are two discontinuities in the series, one between 1999 and February 2001, the other between September 2002 and March 2003. The first is caused by the changeover from the OHS to the LFS (note the large jump in the total number employed, and the large fall in the number ‘not economically active’). The second is caused by the weighting of the survey results to the 2001 population census instead of the 1996 census. Note the substantial increase in the size of the working age population. Re-weighted figures for earlier years than 2003 are not yet available.
Source: Devey et al. (2004: 4) (1997–2001); Labour Force Surveys, 2002–2004.

After making allowances for the two discontinuities in the results, at 1999 and at September 2002, it would appear that:

• Formal employment may have grown by about 150 000 to 180 000 jobs per annum over the period, moving from 6.4 m people in 1997 to 7.8 m people in 2004.

• From February 2000, employment in the informal economy has been roughly static at about 1.8 m (there are disagreements about whether or not the large jump in February 2001, detected by a special survey, actually occurred). Despite the attention the relatively small size of South Africa's informal economy has attracted, the phenomenon is not well understood. ⁸

• Unemployment (official) may have stabilised at about 4.6 m (the drop between March and September of 2003 coincided with an increase in the number of discouraged work seekers).

• The non-employment rate (not economically active plus unemployed, as a proportion of the working age population) fell from a high of about 64 per cent in 1997 to about 55 per cent in February 2001. It seems to have hovered in the region of 60–61 per cent since September 2001.

• The partial explanation favoured by the state for the increase in unemployment – rapidly rising participation rates – is not correct. The rate jumped between 1999 and 2000 because of definitional changes and the leap in subsistence agricultural employment. Since then it has fallen steadily.

• Employment in formal agriculture may be rising slowly. In subsistence farming, the numbers are erratic. It is not known how much of this is an artefact of the statistics, and how much a reflection of reality.

Behind the changes in the numbers of employed and unemployed is a not very well understood process of social and economic transformation. On the demand side, there has been a strong movement away from low-skilled employment. On the supply side, some of the features of the transformation include a rapid increase in the number of households, accompanied by a decrease in average household size and increased migration between South Africa's provinces, as well as transnational migration. A distressing characteristic of this transformation is the emergence of increasing numbers of households whose members are all unemployed. According to Pirouz (2004: 8) the proportion of households containing unemployed persons rose from 13.4 per cent in 1995 to 27 per cent in 2002. Over the same period, the proportion of workless households containing unemployed people grew from 4.7 to 11.6 per cent. Even more distressing is an apparent increase in the numbers of workless households containing unemployed who do not receive any remittances. Households of this type seem to have increased in number from 1.3 m in 1997 to about 2 m in 2002. They were home to about 5.8 m people in 1997, and roughly 7.4 m in 2002. Of the 7.4 m, 4.4 m were adults, 1.3 m of whom were classified as officially unemployed. More than half the 7.4 m people concerned lived in households where the total monthly expenditure was less than R400 per month (Meth, 2004: 30). In this respect, the increase in the discouraged unemployed from 2.7 m people in 1997 to 3.8 m in 2004 is of concern as these are people who truly are cut off from independent income earning opportunities, whether in a modern (First) or subsistence (Second) economy.

Casale et al. (2004: 9) report further disturbing trends from these data using the earnings of those who are employed. Using 2000 as the base year, they note that in real terms average monthly earnings declined over the period 1995 to 2003, falling by more than 20 per cent from R3014 in 1995 to R2360 in 2003. Moreover, they go on to show that the largest relative fall in average real earnings occurred in informal sector self-employment. In 2003, they show that average real earnings among this group were less than a third of their 1995 value. In order to emphasise the impact of this decline in earnings, Casale et al. (2004: 11) follow the World Bank convention discussed above and adopt a PPP$2 per day poverty line. This shows that

the number of the employed living in poverty more than doubled: from just over 900 000 to approximately 2 m individuals based on the conservative poverty line … In 2003, almost 20 per cent of those with employment were earning less than the equivalent of PPP$2 a day, and almost 30 per cent reported earnings that were lower than the minimum wage for a domestic worker. (Casale et al., 2004: 11)

Particularly noteworthy are the 1 536 000 people that Casale et al. (2004: 12) calculate are employed (First?), but still earn less than PPP$2 per day, a class of working poor (Second?), and the 486 000 informally self-employed who are in the same position (Second!).

Taken together, these features of the workings of the labour market tend to confirm a worrying set of trends identified by Bhorat and Hodge (1999). Also providing some support, Rama (2001) cited in Hoogeveen and Özler (2005), argues that there has been a tendency towards outsourcing in the manufacturing sector that accounts for the increase in the number of workers in the informal sector between 1995 and 2001.

The overall implications of these findings would be that in 2003 those who might meet the criteria suggested by SARPN (South African Regional Poverty Network) for inclusion in a ‘Second’ economy might amount to some 6.2 m people, as is shown in Table 3. This 6.2 m amounts to some 20 per cent of the 29.5 m people in the working age population in 2003. The largest group of working age people who might meet the criteria to be included in the ‘Second’ economy are those categorised as the discouraged unemployed; in other words, those who have given up looking for work or who are unable to afford the costs of looking for work. Interestingly, the working poor are the second largest group, comprising almost one-quarter of the people eligible for inclusion in the Second economy. This suggests that the relationship between the institutions that might be expected of the dualistic ‘First’ and ‘Second’ economies is rather more complex than has been thought and that individuals may be engaged in both segments of the national economy at the same time and may also move between these segments over time. The situation will be even more complex at the level of households in which different individuals will be in different positions vis-à-vis the economy, and in which gender is certain to play an important part in determining these positions.

Table 3: Estimate of working age people meeting ‘Second’ economy criteria

	‘000	%
Subsistence farmers:	420	6.7
Informal self-employed earning less than $2 per day	486	7.8
Discouraged unemployed	3819	61.0
Employed earning less than $2 per day	1536	24.5
Total	6261

The KIDS study sheds some light on the fluidity of these shifts for households, and Table 4 shows the changing incidence of livelihood activities within households between 1993 and 1998. ⁹

Table 4: Incidence of livelihood activity, 1993 and 1998

Activity	KIDS 1993	KIDS 1998
Wage labour
Primary labour market	35.8	35.1
Secondary labour market	21.3	20.4
Casual labour	11.1	15.1
Agricultural
Crop production	23.6	38.8
Livestock production	22.5	24.2
Entrepreneurial
Non-farm entrepreneurial	18.5	14.5
Claiming
Old age pensions	29.0	35.1
Other welfare payments	6.7	8.7
Contributory social protection	5.7	7.8
Regular remittances	21.1	24.2
Irregular remittances	17.8	18.4
n =	1031	1031

Although the KIDS data presently available cover only 1993 and 1998, these trends have some resonance with the national data from 1997 to 2004. Formal employment in better paying and secure jobs has shown comparatively little change, while employment as casual labour has increased, as has employment in agriculture and reliance on social grants. Between 1993 and 1998, the KIDS data shows an increase in the number of households surveyed in 1993 participating in agricultural production, increasing the proportion of African and Indian households involved in this work to 49 per cent. Thus, just over half (51 per cent) of the households that had been involved in agricultural production in either 1993 or 1998 performed this activity in both years, 12 per cent had been involved in 1993 only, and 37 per cent were involved in 1998 only. This result suggests a comparatively high level of stability over the five-year period, with 81 per cent of households still producing in 1998 also having been involved in agricultural production in 1993. This stands in dramatic contrast to those working in the non-farm informal sector. In this case, although the total proportion of households in the non-farm informal sector was similar for both years, just 17 per cent of those in the non-farm informal sector in 1993 were still working in this sector in 1998, while 18 per cent involved in informal sector activities in 1998 were also involved in such activities in 1993. Indeed, agricultural production emerges as the destination of almost half of these ‘dissatisfied’ informal sector producers in 1998.

There are other indications that the dependency on wage labour has increased over a longer time frame. Comparing income and expenditure studies undertaken in 1985 and 1992 in the former KwaZulu homeland, the proportional breakdown of the source of income does not appear to have changed radically between 1985 and 1992, although there are a number of important shifts (Nattrass & May, 1986: 592; own calculations derived from DRA, 1992). In 1985, wages and remittances accounted for 76 per cent of household income compared to 64 per cent in 1992. Pensions and transfers were virtually unchanged as a proportion although the amount received had substantially increased. Income from agriculture had declined from 8 per cent in 1985 to 6 per cent in 1992. Finally, income from entrepreneurial activities had increased from 1.5 per cent to 15 per cent.

4. STRUCTURAL POVERTY

Trying to understand the economic causes behind these dynamic patterns of economic inclusion and poverty, Carter and May (1999) use an estimated joint distribution of well-being from the KIDS sample to show that a large clump of initially poor households have either held steady or fallen behind. In the language of conventional empirical analysis of poverty dynamics, these households are chronically poor. In contrast, upward mobility is concentrated among households that were initially better off. This suggests a more nuanced notion of poverty based not just on observed levels of poverty, but on the likelihood of being poor. A re-categorisation of poverty based on analysis of household assets, and the returns that were achieved as these are used, allows us to distinguish between changes that appear to be stochastic and those grounded on changes in the household's ability to generate an income and thus the result of successful accumulation. The categories of the poor can now be further broken down into those who clearly fell behind during the five-year period between the two waves of the survey and those who could improve their position. A significant proportion of the sample (coincidently also some 20 per cent of households) now emerge as not only poor but also falling behind, that is to say, their ability to generate an income declined between 1993 and 1998, perhaps moving them from a ‘First’ to a ‘Second’ economy? While still poor, just 4 per cent of the sample were able to improve their situation. This novel approach breaks the KIDS sample into the seven groups shown in Table 5.

Table 5: Structural poverty classes

Structural poverty classes	HHS
Dual failures	1.5
Structurally poor	16.2
Stochastically upward	5.3
Structurally upward	3.9
Stochastically downward	3.8
Structurally downward	21.1
Never poor	48.2
n =	1031

The categories may be explained as follows:

• Dual failures are households whose assets are such that they are expected to be above the poverty line in both periods but that have experienced entitlement failures in periods during which data were collected and, subject to measurement error, would be observed to be chronically poor.

• Structurally poor households are those that are observed to be poor in both periods and that do not have sufficient assets to generate incomes that would ensure they would not be poor. They too would be observed to be chronically poor.

• Stochastically upward households are those that are observed to be poor in the first period, and then non-poor in the second period due to entitlement windfalls, but that still lack the assets to generate sufficient income to be non-poor. These households would be observed to be transitory poor.

• Structurally upward households are those that are observed to be poor in the first period, and then non-poor in the second period, and that have accumulated the assets required to generate sufficient income to be non-poor.

• Stochastically downwards households are those that are observed to be non-poor in the first period, and then poor in the second period due to entitlement failure, but that have the assets to generate sufficient income to be non-poor in both periods. These households would also be observed to be transitory poor.

• Structurally downward households are those that are observed to be non-poor in the first period, and then poor in the second period, and that have experienced endowment shocks, thereby losing assets, and as a result can no longer generate sufficient income to be non-poor. Once again, these households would be observed to be transitory poor.

• And, finally, the never poor are those households that are both structurally non-poor and those that are observed to be non-poor in both periods.

In an effort to dig beneath these statistics, Carter and May (1999) employed flexible econometric methods to identify asset bundles that map onto livelihoods above the poverty line given the market structure in place at the end of apartheid. Analysis of the KIDS data supports the presence of dualism in some form, and reveals significant departures from the smooth asset additivity that would characterise the mapping in a world of full and complete markets and uncover three dimensions of the poverty problem:

1. The marginal returns on uneducated labour are positive, but so low that they are insufficient to generate an income that meets the poverty subsistence requirements of incremental units of labour. Claims on other economic or social assets are thus necessary to lift a family above the poverty line.

2. Financial constraints limit the ability of the poor to use effectively the productive assets and endowments (e.g. land) that they do have. Poverty is thus not only a matter of few assets, but also of constraints on the effective use of those assets.

3. The burden of meeting basic needs creates a ‘time poverty’ that further constrains households' ability to effectively employ those resources to which they do have access to generate livelihood.

It would seem then that the poor are poor not only because they have few assets, but also because they are constrained in their ability to use effectively the assets they do have. Thus poverty in South Africa is in part a measure of having few assets on which the extant economy pays significant returns. However, inefficiencies such as those described above can be thought of as multiple market failures in which wage opportunities are weak and ancillary factor markets are not working very well. As a result the poor remain poor because they cannot borrow against future earnings to invest in inputs for the production or accumulation of assets for future production, including education. In the absence of opportunities for engaging in paid employment they are unable or unwilling to enter into entrepreneurial activities because the costs of failure are too high, they are unable to insure themselves against risks, and they lack information about market opportunities. Finally they are deprived of many of the public goods necessary for such activities (such as property rights, public safety and infrastructure) and incur high costs in terms of time and expense when trying to obtain these goods. Trying to set a minimum value to the assets that would be sufficient to offer the opportunity for upward mobility, Carter and Zimmerman (2001) talk of a ‘Micawber Line’ referring to the asset bundle below which escape from poverty is impossible despite the best efforts of those who are in this position to work, scrimp, save and accumulate: a line that perhaps might represent the divide between ‘First’ and ‘Second’.

5. CONCLUSION: THE NATURE OF DUALISM AND UNDERDEVELOPMENT

Examining the impact of trade and technology on employment dynamics in South Africa since 1995, Bhorat and Poswell (2003: 51) conclude that recent economic growth has resulted in weak employment growth among highly skilled workers, while unskilled workers and those in poor households have carried the adjustment costs of greater openness. Moreover, using growth incidence curves that plot the annual growth rate in each percentile of the distribution of per capita expenditures, Hoogeveen and Özler (2005: 15) conclude that ‘Growth has not been pro-poor in South Africa as a whole, and in the instances when poverty declined for certain subgroups, the distributional shifts were still not pro-poor’. It is not surprising then that Nzimande (2003) makes a stronger case that both exclusion from the mainstream economy and underdevelopment have persisted in the post apartheid era:

But if we speak of an underdeveloped pole what exactly are we referring to? Are we referring to an apartheid ‘legacy’? Yes, obviously, our crisis of underdevelopment has much to do with apartheid … The trouble with thinking about underdevelopment simply as an apartheid legacy is that you may assume that as we move away from the apartheid years, so this reality will start, more or less spontaneously, to fade away. The idea of ‘legacy’ fails to capture the ongoing, the active, the daily re-production of underdevelopment and poverty.

In an interesting micro-level study of these patterns, du Toit (2004) prefers to use the notion of adverse inclusion rather than social exclusion to draw attention to the fact that these dynamics do not simply result from being left out of the mainstream economy, but rather from the terms under which an individual is incorporated. May et al. (2004) reflect on these arguments, and suggest that progress along the current growth path will not be sufficient to reduce poverty. Indeed, as this quantitative assessment suggests, there is more to the relationship between those who are marginalised and those who are benefiting from South Africa's achievement than a dualistic and disarticulated ‘First’ and ‘Second’ economy. Linkages of inclusion and exclusion determine the size, characteristics and future of the two economies, a conceptualisation already reflected by Frank in 1967 in his well-known statement: ‘Economic development and underdevelopment are the opposite face of the same coin’ (Frank, 1967: 9). In this vision, underdevelopment is indeed an ongoing process that will continue to reproduce unacceptably high levels of hardship among a substantial proportion of South Africa's population.

To deal with these dynamics, rather than separating policies supporting the asset poor from those for the asset rich, May et al. (2004) call for stronger action to address the market ‘failures’ that shape the adverse relationships described by du Toit through further transfers and redistribution. They suggest that such redistributive policies should form part of the micro-economic strategy being sought by the government in various Budget and State of the Nation addresses (Manuel, 2001, 2002; Mbeki, 2001) and stress that such reforms can no longer be delayed for a more optimal moment as the economic costs of poverty and inequality continue to rise. Recent measures to increase the resources available for land reform, the access to finance for poorly resourced individuals and enterprises, and access to quality education are welcome steps in this regard. However, these need to be matched with improvements in the state's capacity both to deliver the relevant services and to monitor the progress of delivery. In addition to this, greater clarity must be achieved in the aims of crucial redistributive policies such as education and land reform, the former dogged by problems of quality and exclusion, the latter a distressing failure in terms of the pace of redistribution. The tendency to load policy goals with excessively ambitious targets needs to be curbed, as must the tendency to formulate potentially misleading conceptualisations of how the South African economy functions. Fewer measures, fewer ambitious theories and more precisely conceived and designed policies tailored to the country's capabilities ought to be the order of the day. Finally, the ability of the official statistics gatherers to collect and analyse data on development needs very urgently to be augmented. The data required include basic information on the measurement of money-metric poverty using instruments such as those reported earlier and the collection of social indicators such as employment in the informal economy and participation in small scale or subsistence agriculture.

Additional information

Notes on contributors

Charles Meth

Respectively, Associate Professor, and Senior Research Fellow, School of Development Studies, University of KwaZulu-Natal. An earlier version of this paper was presented to a conference on Overcoming Underdevelopment in South Africa's Second Economy, 28 and 29 October 2004, Sheraton Hotel, Pretoria. The financial support of the Development Bank of South Africa (DBSA) and the Department for International Development, South Africa (DFID-SA) is acknowledged, as are the comments of Ingrid Woolard and three anonymous referees. The views expressed reflect those of the authors only.

Notes

¹For most of their analysis Woolard and Leibbrandt (2001: 56) settle on the Household Subsistence Level and $1 a day ‘International’ line (R3509.00 and R2200 per annum per adult equivalence in 1995 rand). The latter may be thought of as the ultra-poverty line.

²The HDI calculated by the South African Human Development Report (UNDP, 2003) would place South Africa at 115th place for 2003.

³All these analyses calculate poverty and inequality in terms of ‘per capita’ or ‘per adult equivalent’ scales, and include an adjustment to take account of household economies of scale.

⁴The IES 2000 data have been cleaned and re-weighted for this analysis following Simkins (2004) and the 1996 data have been adjusted by the CPI for urban areas to 2000 prices. The data in the table depict incomes below R2200 per adult equivalent per month due to the long tail of the distribution arising from South Africa's extreme income inequality.

⁵Meth refers to this adjustment as a social wage as an attempt to take into account government reaction to the Statistics South Africa report.

⁶Noteworthy in this literature are Frank (1967), Laclau (1971), Taylor (1979) and Wallerstein (1976).

⁷KIDS was undertaken by a consortium of South African and international researchers in 1998 which re-interviewed 1100 households first surveyed in 1993 as a part of the national Project for Statistics on Living Standards and Development (PSLSD). KIDS has been extended by a further five years with a re-survey conducted in 2004 and the data are in the public domain (http://www.sds.ukzn.ac.za).

⁸Chen (2004) reports that informal employment comprises one-half to three-quarters of non-agricultural employment in developing countries. As examples, informal employment amounts to 48 per cent in North Africa, 51 per cent in Latin America, 65 per cent in Asia and 72 per cent in sub-Saharan Africa.

⁹The 1993 PSLSD questionnaire was not designed to provide accurate labour market information and as a result the KIDS panel is an imperfect instrument for analysing labour market segmentation. For the purposes of this paper, secondary labour has been defined as all labourers, agricultural workers and domestic workers. All other forms of regular employment have been defined as being in the primary labour market. Arguably, casual employment could also be included in the secondary labour market.