This resource is hosted by the Nelson Mandela Foundation, but was compiled and authored by Padraig O’Malley. It is the product of almost two decades of research and includes analyses, chronologies, historical documents, and interviews from the apartheid and post-apartheid eras.
3. Analysis
3.1 Overview of inequality
Two spatial units are employed in this empirical analysis. The first is nine development regions with the boundaries as defined by the Development Bank of South Africa.1 Development regions, as noted above, included both former homelands and provincial (non-homeland) areas. (See Map 1.) The second spatial unit considered is the 23 subregions, which distinguish between former homeland and non-homeland areas within each development region. Grouping these subregions into homeland and non-homeland areas serves as a rough proxy for black and white areas respectively. This disaggregation also allows measurement of differences among non-homeland regions, namely the Cape Province, Transvaal, Natal and the Orange Free State, which is important because perceived inequalities across these former provinces may pose serious political limitations for national redistributive economic policy. Finally, significant differences in inequality emerge among former homelands when they are grouped into 'self-governing' and 'independent' subregions.
As an introduction to presentation of the study results, Figures 1-4 give an overview of regional economic performance and inequality for three groups of regions, former provinces, homelands and, as a subset of homelands, the TBVC regions.
Figure 1 shows the performance of the former provinces compared to that of all former homelands in GGP per capita, personal income per capita, urbanisation rates and education, all measured as a percentage of the national average. The most unequal performance appears for personal income per capita, in which the former homelands achieved only some 20 per cent of the national average, compared to the almost 140 per cent of the former provinces. GGP per capita is almost as unequal, with the former homelands having only slightly more than 20 per cent of the national average compared to the former provinces with 150 per cent. Urbanisation rates are most similar, the former homelands exhibiting a rate almost 80 per cent that of the national average.
Figure 2 shows the inequality within each group, former provinces and homelands, measured by the coefficient of variation. Homelands are much more unequal than provinces, revealing the differential impact of population relocations and migrations as well as resource distribution across the homelands. Personal income per capita is less unequal than GGP per capita for both groups of regions and considerably less unequal for the homelands. The reduction in inequality across former homelands moving from GGP per capita to personal income indicates that there has been an equalising effect of transfers and remittances to the former homelands. Urbanisation rates are more than twice as unequal in the former homelands than in the former provinces but inequality in education is very similar in the two groups of regions.
The results shown in Figures 1 and 2 demonstrate the seriousness of both intergroup and intragroup disparities in GGP, personal income, urbanisation and education. The redistributive tasks facing the new government are not limited to redressing white/black differentials but also must include reducing black/black inequality across geographic areas. Further, the figures suggest that education and urbanisation are not sufficient to achieve equality. Figure 1 shows relatively small gaps between urbanisation and education levels in homelands and provinces compared to the degree of inequality in GGP and personal income per capita.
Figures 3 and 4 compare economic performance and inequality between two groups of homelands – all homelands and those which became nominally independent of South Africa. This comparison reveals that the 'independence' of the TBVC states had little effect on their status. The indicators shown in Figure 3 are very similar for the two groups. Figure 4, which presents intragroup inequality for the two groups, does highlight some important differences. The TBVC states were more equal in GGP per capita but much less equal in personal income per capita.
1 Regional boundaries have changed since the election in April 1994. Comparing Map 1 with Map 2, changes are most pronounced for the former Cape province and there is as yet no firm determination of location of the Transkei in the new regions. Overall, however, current boundaries differ only slightly from the development region demarcation.
3.2 Structure and performance by development region
Table 1 provides an overview of the nine former development regions based on several indicators of economic performance and structure. Three regions, namely G, D and E, are in the worst positions for most of the indicators. Region G is clearly the poorest in both GGP and personal income per capita. In 1989, for example, the region's per capita GGP was only 26,2 per cent of the national average for all development regions. Moreover, the relative position of the region had deteriorated from 1970, when – although it was still the poorest development region – GGP per capita was 42,4 per cent of the national average.
Regions D and E were also well below the national averages for income. In 1989, the per capita GGP of Regions D and E was only 55,6 per cent and 63 per cent respectively of the national average. However, unlike Region G, the relative positions of Regions D and E improved over time, up from 45,4 per cent and 52,5 per cent of the national average in 1970.
Moving along the columns of Table 1, it becomes clear that Region G is significantly different from the other regions in levels of urbanisation and functional urbanisation. Regions D and E are also below the national average for these two characteristics, but the gap is much smaller. While urbanisation in 1989 in Region G was only 18 per cent of the national average, the rates were 83,7 per cent and 71,4 per cent of the national average for Regions D and E respectively.
Sources of the relative poverty of Region G are suggested by the data in the next set of columns of Table 1. The population growth rate has been highest in this region, as have the percentage of females and the dependency ratio (which measures the number of dependents per income-earning family member). Population growth between 1985 and 1990 in Region G was 32 per cent higher than the national rate; the share of females in the population in 1989 was 22 per cent higher and the dependency ratio was 240 per cent of the national average. At the same time, Region G in 1989 had the lowest levels of economically active population (51,6 per cent of the national average) and participation rates (64,5 per cent of the average), but the highest informal sector participation rate (206 per cent of the average). Regions D and E are again similar to Region G in that, with the exception of population growth, both are well away from the national averages in all of these indicators of regional performance.
3.3 Structure and performance by subregional group
From Table 1 an identifiable bloc of poor regions within South Africa is evident even at the level of aggregation of the development region. However, the nature and source of the relative poverty of these areas are as much obscured as revealed by a development region comparison. The lower part of Table 1 presents a disaggregation into the 23 subregions, arranged in three groups. The first group, labelled for brevity 'Republic of South Africa' in Table 1, is the Republic of South Africa with the former homelands subtracted. This group comprises the four former provinces of South Africa, namely Natal, the Orange Free State, the Cape and Transvaal. In constructing development regions provincial boundaries were replaced except for those of Natal, which remained largely intact in the new Region E. As a result each of the former provinces besides Natal extended into more than one development region, as indicated by the letters preceding the province names in Table 1 (for example, Transvaal was divided among development regions F, G, H and J). The second group of subregions is the former so-called 'self-governing' homelands, black areas officially part of South Africa, and the third group is the former 'independent' homelands.
For the provincial subregions, the second block of results in Table 1, there is noticeable variation in economic performance and status. For income measures, the clear leader among these subregions is the Region H portion of Transvaal, which includes the former PWV (now Gauteng province), the most developed part of the country. This subregion was also the most urbanised, the most educated (averaging education levels for 1980 and 1989) and the highest in population growth rate and economically active population ratio. At the same time, it was the lowest in dependency ratio and below average in the share of females in the working-age population.
These indicators describe an area in which the conditions for growth were favourable, but the PWV still fell behind other successful regions during the 1980s, particularly the Region F portion of Transvaal. Thus apparently favourable indicators for the PWV were not sufficient to sustain growth during the national economic decline during the 1980s, whereas Eastern Transvaal conditions did prove able to support growth.
In Table 1 Eastern Transvaal (Region F Transvaal) shows very low urbanisation, education and population growth compared to both the PWV subregion and the national average. On the other hand, this subregion had shares of females and dependency ratios similar to those of the PWV subregion. These differences further complicate an assessment of the sources of regional growth, since low urbanisation and education levels typically are associated with lower growth rates.
The other well-off provincial subregion is region A, which includes most of the former Western Cape province. This ubregion has high urbanisation and education, while both the share of females and the dependency ratio are also high. The labour force in this subregion is therefore in between that of the PWV and Eastern Transvaal – neither extremely biased toward migrant workers who leave dependents behind nor excessively reliant upon jobs requiring very low education levels.
The variability in economic and social indicators among these successful subregions implies that growth can be accomplished by many economic structures and strategies, ranging from the mixed economy in the Cape to the industrial concentration of the PWV, to the extraction and beneficiation of Eastern Transvaal. The poorer subregions, on the other hand, have a number of features in common.
Poor provincial subregions include parts of Regions B, C, D and G. The provincial subregions of B and D are part of the former Cape Province, C was in the Orange Free State and G in Northern Transvaal. These areas have in common lack of proximity to urban or industrial centres, as evidenced by low urbanisation rates (with the exception of Region D which includes Port Elizabeth and East London), low education levels and high dependency ratios. The poverty of the Region G subregion of Transvaal is such that the population growth rates have been negative during the 1980s. The only development regions with subregional data, the Cape and Transvaal, thus both appear as internally highly unequal provinces including both very rich and very poor subregions.
Variation is also pronounced across the homeland subregions. Among the 'self-governing' homelands, for example, two are in Region G. The 1989 GGP per capita in Gazankulu was 3,7 times higher than that of Lebowa, despite Gazankulu's lower urbanisation rates, participation rates and share of females in the population, which typically would be associated with a lower level of development. That the GGP per capita is in fact higher in Gazankulu suggests again that sources of poverty are more complex or localised than data on a development region level can capture. A similar situation is evident for the 'independent' homelands. Within Bophuthatswana, for example, the part located in development region J is considerably richer even though it has a relatively low urbanisation rate, high population growth and an almost average share of females. Location of mineral resources is therefore the primary determinant of the variation in income within Bophuthatswana.
Despite this variation within subregional groups, Table 1 clearly shows that regional poverty runs along racial lines. For all indicators, the former homelands (which are virtually all black) as a group fall substantially below the group of subregions excluding homelands. The lesson to be learned from interhomeland inequality, nonetheless, is that national policy must target the poorest of these areas specifically and not rely on general policy aimed at the former homeland populations.
3.4 Regional inequality
The relative importance of variation within and across the several groups of regions and subregions is described in Tables 2 and 3. The inequality measures in Table 2 are calculated for two levels of aggregation and six groups. At the top of the table is the development region group, followed by five groups at the subregional level.
. The first of the subregional groups includes all 23 subregions of the nine development regions.
. The second group, with nine subregions, comprises the four former provinces.
. The third, with 15 subregions, is the former provinces plus the former 'self-governing' homelands.
. The fourth is the former TBVC states, or the 'independent' homelands, with eight subregions.
. The fifth is the TBVC plus the 'self-governing' homelands, comprising 14 subregions.
A proxy for racial differences is comparison of the group comprised of the Republic of South Africa less homelands (the former provinces) with the group including all former homelands. It must be emphasised that these two groups are only a rough approximation for measuring racial inequality because the majority population across all subregions is black.
Table 2 presents two measures of inequality (the coefficient of variation and the ratio of maximum to minimum) for all the groups as well as the group averages for the indicators shown in Table 1. As expected from the relative levels of aggregation, variation is less across development regions than within the groups of subregions. For example, the coefficient of variation for the nine development regions in personal income per capita for 1985 is 0,4421, while the coefficient of variation for the 23 subregions is 0,5936. The same is true for the second measure of variation, the ratio of maximum to minimum personal income per capita, which is 6,5 for the development regions but 58,9 for all subregions.
Comparison of inequality measures for the nine regions with those for the group of 23 subregions (in the second block of Table 2) illustrates an additional consequence of the level of aggregation. While the coefficients of variation for personal income and GGP per capita for the development regions dropped between 1970 and 1989, the same measures of inequality for all subregions increased. Thus, at the development region level inequality seems to have improved overall, while for the subregional data inequality has grown. In other cases, the reverse inconsistency applies. For example, inequality in the economically active population declined across the 23 subregions but increased slightly for development regions. These inconsistencies in measured inequality by level of aggregation stem from the inclusion in development regions of both provinces and homelands, with the former provincial areas raising the regional average relative to the homeland average.
Where results for different degrees of aggregation conflict with each other, the disaggregate measure gives more information and should therefore be taken to be a better description of inequality. The conclusion follows that inequality has increased in the summary indicators of economic performance, GGP and personal income per capita, despite small improvements in inequality of other variables such as urbanisation rates, population growth, dependency ratios, economically active population, participation rates and education. A tentative hypothesis then emerges from these findings: much greater equalisation of the improved variables must take place before they will have a significant effect on the overall economies of the poorer regions.
Table 3, an ANOVA table for homeland versus non-homeland subregions, quantifies and evaluates the statistical significance of the inequality measures in Table 2. For all but the 1980 functional urbanisation rate, the table shows a statistically significant variation between former homeland and provincial groups relative to within-group variation. The absolute difference in the averages for former homeland and non-homeland subregions is shown in the last column as 'White region premium', labelled such with the qualification noted above that the former provinces are also predominantly non-white. The figures in this column reflect the size of the gap between the homeland and the provincial averages for each of the variables listed. For example, the per capita personal income in former provinces in 1985 was R2948 higher than the per capita income in the former homelands. Similarly, the GGP per capita was higher in 1989 by R4036 in the provinces. Also greater in the 'white' areas were urbanisation and functional urbanisation rates, participation and economically active rates, education and provision of health services. Lower in the provincial regions (entries in the column with negative signs) were population growth rate, the percentage of females, dependency ratio and informal sector participation rate.
The picture painted by these data is not unexpected, but the size of the gaps revealed is quite dramatic. Moreover, for many key variables the gap has widened. For example, the personal income and GGP per capita have become more unequal over time as shown by the increasing white premium, as have the majority of the other variables. Widening gaps in the percentage of females and the economically active population premiums point to growing inequality in the future, although this possibility may be offset by the declining gaps in population growth rates and in the dependency ratio. Note also that the informal sector participation rate has been falling in former provinces relative to homelands, suggesting that the structure of employment is becoming more dissimilar, with homeland employment becoming relatively more concentrated in the low pay and highly uncertain informal sector jobs.
3.5 Sectoral structure and inequality
In explaining inequality, sectoral structure across regions is a key variable. Table 4 describes the sectoral composition of both GGP and the economically active population for five regional groupings. For the nine development regions, manufacturing is the largest sectoral contributor to GGP, followed by community and personal services. However, large differences among regions are obvious from the table.
One of the most notable differences between the former provincial and homeland groups arises in variation of the GGP share accounted for by community, social and personal services. The average for all nine development regions in Table 4 is a share of 17,8 per cent of GGP in this sector. For the former 'self-governing' homelands, the range of share in GGP is from 40,2 to 59,8 per cent, and for the former TBVC regions from 14,5 to 52,8 per cent. These levels reflect the dependence of the homelands on transfers from the South African government. Formal government welfare expenditure in the homelands was 56 per cent for the 'self-governing' areas and 42 per cent for the TBVC territories, compared to 12 per cent for South Africa as a whole (Mbongwa & Muller, 1992). For the TBVC, an average of only 15 to 20 per cent of the budget came from their own revenues at the end of the 1980s (Nattrass & Nattrass, 1990).
Variation in sectoral shares of GGP within groups of regions is shown in Table 5. The most interesting feature of this table is that variation between the former provinces group ('RSA less homelands') and the former homelands is not necessarily greater than differences within either group. In agriculture, for example, the ratio of maximum to minimum share in GGP for the former homelands group of subregions is only 7,14, compared to a ratio of 12,73 for the former provincial group. At the same time, inequality as measured by the coefficient of variation for the percentage of GGP derived from agriculture is 0,55 in the homelands and 0,49 in the former provinces. Moreover, the ratio of the average in the provincial group to the average in the homelands (shown in the last row of the table) is low at 1,09.
A similar pattern emerges for the other major sectors (manufacturing, wholesale and retail trade, and finance and insurance), but the sector including community and social services is an important exception. Community, social and personal services is the largest sector in all regional groupings and in this sector variation within the two homeland groups is larger than variation within the group of former provinces. The coefficient of variation is 0,28 for the latter group, versus 0,46 for the TBVC. The homelands have had a very unequal distribution of services, implying inequality in the location of government expenditure in these regions. The policy implication is to disperse government activities within the homelands to equalise expenditure. Note, however, that the former homelands already have a considerably higher share of GGP accounted for by services than the rest of the economy. Intrahomeland redistribution rather than expansion of existing expenditure would therefore be the more attractive alternative from the perspective of restraining government expenditure. Unfortunately, this would redistribute resources away from other homeland regions that are only marginally better off, exacerbating the already high tensions among former homelands.
In any case, accelerating growth of the less-developed regions of South Africa will require expansion of production activities as well as government services. Identification of sectors which should expand requires choice among sometimes competing goals. For example, increasing labour productivity may conflict with growth in employment. Table 6 presents the GGP per employed (a proxy for labour productivity) for the major sectors in 1990. These data are available only for the nine development regions, but even at this level of aggregation large differences among sectors and regions are evident.
Of the dominant sectors, manufacturing exhibits the largest interregional variation in GGP per employed with a coefficient of variation of 0,6660. The GGP per employed is highest in Region F (R93,6) and lowest in Region G (R20,6), compared to the average of R34,5 for all development regions. The result for Region G is consistent with the indicators of economic development shown in Table 1, which identify this region as the poorest. Two other poor regions, namely B and C, are close to Region G's low level of productivity.
These figures imply that expansion of manufacturing in Region F would yield better results as measured by increasing national average productivity. The participation rate in Region F, however, is 66 per cent for 1989 compared to only 38 per cent for Region G (from Table 1). Creation of manufacturing jobs concentrated in Region F will only exacerbate the interregional maldistribution of job opportunities.
South Africa thus faces the usual trade-off: based on the existing structure and location of manufacturing, a choice must be made between increasing productivity and improving access to employment in poor areas. However, the terms of the trade-off are unusually unfavourable for this country, since the GGP per employed in manufacturing in Region G is only about one fifth of that in Region F. Reorientation to Region G (and to Regions B and C) would require a very large sacrifice in productivity (assuming, of course, ceteris paribus).
Moving from manufacturing to agriculture, Tables 7 and 8 reveal the extent of interregional differences in agricultural potential. Arable land per capita in Region C is, at 1,59 hectares, more than 12 times greater than the 0,13 hectares per capita available in Region D. (Region H, the former PWV, is not considered because of its unique urbanisation.) However, the value of agricultural production is not determined solely by the availability of land, as evidenced by the fact that Region A's agriculture achieves three times the value per hectare of Region C, with less than a third as much arable land per capita. A number of additional factors intervene, such as the mix of farm and grazing land, rainfall and market conditions for the various crops.
To determine whether distribution of agricultural resources is more unequal in the provinces or in the homelands, Table 8 provides measures of inequality by regional grouping for the variables given in Table 7. For the nine provincial subregions comprising the group 'RSA less homelands', the coefficient of variation for hectares per capita and the percentage of arable land is larger than for the group comprising all the homelands ('TBVC & homelands'). However, the coefficient is smaller for arable land per capita and value per hectare.
Although in the share of arable land the homelands are more similar than the former provinces, they are less similar in arable land per capita. This difference is due to the large population shifts into the various homelands resulting from relocation of black populations and migrancy. Thus, although the homelands were relatively equal in the amount of land available, varying population densities resulting partly from apartheid have stretched resources unequally in the different areas.
From the bottom block of Table 8, it is also clear that the former provincial areas of South Africa have almost six times as much arable land per capita as the former homelands, but only 1,3 times as great a share of arable land. The provinces also have almost seven times the amount of total land per capita, which is a result of the small area ceded to homelands when they were created. These ratios reinforce the argument that scarcity of resources in the homelands was created by the policies of first limiting homeland territory and then removing populations to the homelands.
Agriculture in the former provincial areas is more similar because of the size of government support for large-scale commercial farming. With the high levels of subsidy, any natural regional differences which normally would show up in profitability and viability of farming are obscured. The similarities within the group 'RSA less homelands' thus should not be taken to mean that conditions of production are similar, or, that in a market environment the agricultural sector could be relied upon as a source of equalisation across regions.
The variation in agricultural resources and uses shown in Tables 7 to 9 results in wide variation across regions in the structure of commercial agriculture. From Table 10 it is clear that the average size of a field crop farm varies from 261 hectares in Region D to 1,075 for Region A, while still wider ranges obtain for other agricultural sectors. For agriculture as a whole, the average farm in the region with the largest farms (Region B, concentrated in animal products) is over nine times the size of the smallest average regional farm (as found in Region H).
Regional variation in gross income per farm is also high, as is disparity in gross income per hectare. For the latter, which is an approximate measure of farm productivity, the ratio of highest to lowest (Region E to Region B) average is 4,1 for field crops and 10,4 (Region H to Region B) for horticulture.
Farm debt is highly variable as well, indicated by 'Debt to income ratio' in Table 10. By crop type, for field crops the ratio of highest to lowest (Region D to Region E) average farm debt is only 5,5. For horticulture the ratio is 3,1 (Region F to Region G) and for animal products 4,8 (Region B to Region H).
Finally, labour use is very different by region as well as by crop type. In field crops, the range is from 0,379 employees per hectare (Region A) to 0,018 (Region C), giving a ratio for maximum to minimum of 21,1. A similar range appears for animal products, but for horticulture the regions are more similar, with a ratio for maximum to minimum of 5,2 (Region H to Region C).
Several patterns can be observed in this table, although given the serious data limitations and the level of aggregation, they must be regarded as tentative. First, regions with the lowest utilisation of labour (B, C and D) are also the highest debt regions. These regions seem to be paying the price for capital-intensive agriculture in an era of high interest rates. Similarly, Region C has the second largest average farm size in commercial agriculture but only average gross income per hectare as well as high debt, suggesting that farm size may not be positively correlated with performance. This suggestion is strengthened by the data in Table 7, which shows that Region C has a relatively high percentage of arable land and the highest arable land per capita.
For the former development regions ranked lowest in per capita GGP and highest in unemployment (Regions D, E and G), farm size, farm income and income per hectare are all at a medium level. However, in Regions E and G, the debt/income ratio is only average or slightly below average. These two regions also have relatively low labour use per hectare, although this may be explained for Region E by the very large share (92,64 per cent) of land devoted to animal products. However, the same explanation does not hold for the other two regions, which have close to average allocation of land by type of production. A provisional result then is that the poorest regions do not use more of their abundant resources and cheap labour, than regions with higher incomes and lower unemployment. These regions therefore form a group which is distinguished by low success in agriculture as measured by indebtedness relative to income and by a structure of production that appears to be inconsistent with their resource endowments.
Even at this level of aggregation, more trade-offs in policy choices are apparent. For example, if employment growth is the goal, expansion of field crops would be the appropriate strategy for region B, but from Table 10 it is clear that in this region field crops earn only half the income per hectare than horticulture. On the other hand, horticulture – the highest income generator – employs fewer than half as many workers per hectare as field crops. Fortunately, such trade-offs do not apply everywhere: in South Africa the highest level of employees per hectare is in horticulture, which is also the sector with the highest gross income per hectare.