Consolidating South Africa's New Democracy: Geographical Dimensions of Party Support in the 1999 Election¶
Roddy Fox; Anthony Lemon (2000). Consolidating South Africa's New Democracy: Geographical Dimensions of Party Support in the 1999 Election. Tijdschrift voor Economische en Sociale Geografie 91(4). Blackwell Publishers / Royal Dutch Geographical Society KNAG.
| Type | journal article |
| Scope | Wider apartheid literature |
| Method | descriptive quantitative |
| Period studied | 1994–1999 |
| Regions | South Africa, Western Cape, Northern Cape, Eastern Cape, KwaZulu-Natal, Gauteng, North West, former homelands, Coloured Labour Preference Area |
| Themes | politics ideology elections, transition post apartheid, bantustans removals influx, demography urbanisation, origins segregation |
| Citation key | fox2000 |
Argument¶
The paper analyses the spatial pattern of party support in South Africa's 1999 election using polling-district data matched to magisterial districts. It argues that voting still follows the racial divide, with the ANC and IFP (and the UDM) tracking the black share of the electorate and the DP and NNP as essentially 'non-black' parties. The regional parties (IFP, UDM, UCDP) rest on factors other than ethnicity alone, and only the IFP depends on ethnic mobilisation, and it declined. Former apartheid boundaries (homelands, the CLPA) remain visible in the electoral geography, mainly because they reflect race and the urban/rural divide. The authors conclude that South Africa is a one-party dominant democracy, and that fragmented, racially based opposition poses problems for consolidating liberal democracy.
Evidence and method¶
IEC spreadsheeted results for 14,535 voting districts (supplied confidentially by a government official), classified as rural/urban/metropolitan by local government structure and by apartheid region (CLPA, homelands, rest) via magisterial districts; 1996 census maps of race and language; secondary election analyses (Reynolds, Lodge, Taylor & Hoeane, Habib & Naidu, Mattes et al.).
Place in the debates¶
- Race versus ethnicity as the basis of voting in South Africa's elections (the 'racial census' thesis) (see White electoral politics, delimitation and party systems). Supports the racial-census view for 1999: the black/non-black divide is clearer than in 1994. Rejects reading regional-party support as primarily ethnic, though it concedes race does not fully determine identification.
- One-party dominance versus consolidation of liberal democracy (see White electoral politics, delimitation and party systems). Describes a one-party dominant democracy that has dangers (weak, racially bounded, fragmented opposition) but may also bring consistent reconstruction policies.
- Persistence of apartheid spatial legacy (see White electoral politics, delimitation and party systems). The homeland and CLPA boundaries are still visible in voting, but this reflects race and the urban/rural divide more than ethnicity.
Findings¶
-
ANC and IFP vote shares together match the black share of the electorate nationally. (p. 4)
In 1999 the ANC (66.36%) and IFP (8.59%) together accounted for three-quarters of all votes cast, which again corresponds closely to the black share of the electorate.
-
In KwaZulu-Natal the ANC and IFP combined matched the black electorate almost exactly. (p. 5)
In KwaZulu-Natal, the ANC came a close second to the IFP, but together they accounted for 80.22% of the poll, matching almost exactly the black proportion of the electorate.
-
The DP and NNP drew very little black support; the NP's black vote fell sharply. (p. 6)
only 200,000 blacks supported the NP and 180,000 the DP, out of over 12 million voters. For the DP this is more than in 1994, but the NP's black vote has declined by 60%
-
The IFP's ethnic mobilisation weakened even in its KwaZulu-Natal heartland. (p. 8)
The IFP vote declined between 1994 and 1999 in all four provinces, most steeply (by one-third) in its KwaZulu-Natal heartland.
-
UDM support rests on Transkei patronage and traditional leaders rather than ethnicity as such. (p. 9)
suggests that it won the votes of those who benefited from the trappings of Transkeian `indepen-dence' between 1976 and 1994 and some rural support flowing from its cultivation of tra-ditional leaders
-
The DP and NNP are overwhelmingly urban parties. (p. 9)
both the DP and NNP are overwhelmingly urban parties: 94% of the total DP vote came from urban and metro-politan areas.
-
The CLPA boundary still shows in voting: the NNP does far better and the ANC worse inside it. (p. 11)
in the Northern Cape the NNP doubles its share of the poll in the CLPA, at the expense of the ANC. In the Eastern Cape the NP share is more than four times higher
-
Former homelands give almost no support to the NNP and DP, and the ANC does 12-13% better in ex-homeland areas of three provinces. (p. 11)
The ANC performed markedly better (12%±13%) in ex-homeland areas of the Free State, Mpumalanga and the Northern Province than in the rest of these provinces.
-
The black-white divide is more marked than in 1994, and the DP's success has increased racial polarisation. (p. 12)
The black±white divide is even more marked than in 1994, as the NNP has lost most of its limited black support.
-
The authors characterise South Africa as a one-party dominant democracy with fragmented opposition. (p. 13)
At the national level, South Africa thus remains a one-party dominant democracy, a situation that has obvious dangers but may also offer the advantages of firm, consistent
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| IEC spreadsheet of 1999 national election results by voting district (14,535 voting districts) | election returns | 1999 (national and provincial elections) | Core dataset: voting-district results aggregated by province, rural/urban/metropolitan category, apartheid region (CLPA, homelands) and language area to map party support. |
| IEC voting district maps and GIS covers (approx. 14,000 voting districts) and IEC website | maps spatial data | 1999 | Context for how voting districts were spatially defined; the authors note that the GIS spatial information itself was not released, so they had to match districts to magisterial boundaries. |
| IEC (1999), Atlas of Results, National Election 1999, South Africa | commission or official report | 1999 | Official published election results atlas; mentioned as the published form of results but not used because of its prohibitive cost. |
| Republic of South Africa (1997), Green Paper on Local Government | commission or official report | 1997 | Basis for classifying voting districts as rural, urban or metropolitan by type of local government structure (Transitional Rural/Local/Metropolitan Councils). |
| Magisterial district classification into CLPA, homelands and rest of South Africa (apartheid regional divisions, Figure 3) | maps spatial data | apartheid-era boundaries applied to 1999 data | Voting districts were matched to magisterial districts and assigned to CLPA, ten homelands or remainder to test the legacy of apartheid boundaries in voting. |
| 1996 population census data on racial composition and dominant languages by province/district (Figures 1 and 2) | population census | 1996 | Provides the racial composition of provincial electorates and distribution of dominant languages used to relate party votes to race and ethnicity. |
| Reynolds (1994, 1999) estimates of racial breakdown of party support | secondary literature compilation | 1994, 1999 | Used for estimated shares of NP, DP and ANC support derived from each racial group in 1994 and 1999. |
| Mattes, Taylor & Africa (1999) voter opinion survey | secondary literature compilation | 1999 | Cited as survey evidence supporting a multi-causal interpretation of voting rather than race alone. |
| Habib & Naidu (1999) and Taylor & Hoeane (1999) polling district analyses of coloured and Indian voting | secondary literature compilation | 1999 | Support claims about income differences among coloured/Indian voters and ANC performance in former Indian group areas such as Lenasia. |
| Lodge (1995) analysis of 1994 election results | secondary literature compilation | 1994 | Source for 1994 comparison, including black NP support in Gauteng and the 'racial census' characterisation. |
| 1994 and earlier election results (party shares in 1994, 1996 local elections) | election returns | 1994, 1996 | Baseline for measuring change in party support, e.g. ANC 62.65%, NP 20.39%, IFP 10.54% in 1994, and IFP decline since 1996 local elections. |
| Party manifestos, campaign materials and press reports | newspapers periodicals | 1998-1999 | Provides the description of party platforms and campaign strategies in the section on the contending parties. |
New data¶
The authors built a classified dataset by matching the IEC's voting-district spreadsheet (14,535 districts) to magisterial districts and local government categories, aggregating 1999 votes by rural/urban/metropolitan type and by CLPA/homeland/rest-of-province, with language-area comparisons. The underlying IEC data were not publicly released.
Inference¶
Quantification: Descriptive tables.
Descriptive electoral geography: aggregation of 1999 voting-district results by province, rural/urban/metropolitan category and former apartheid region, compared with 1996 census racial and language distributions and 1994 results. Claims rest on visual and tabular correlation and comparison across areas and time, with no regression or formal statistical test.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| The 1999 vote again closely tracked race: ANC+IFP (+UDM in Eastern Cape) shares match the black share of the electorate, leaving DP and NNP as essentially non-black parties. (p. 4) | Provincial results from the IEC spreadsheet compared with 1996 racial composition (Figure 1, Table 1), plus Reynolds' estimates of black votes for the NP and DP. | Partial data |
| The regional parties (UDM, UCDP, IFP) are not primarily ethnic; the IFP's ethnic mobilisation declined, with its vote down by a third in KwaZulu-Natal between 1994 and 1999. (p. 8) | Polling district returns compared with the language distribution map (Figure 2) and 1994 results; turnout-based estimates of Zulu voting for the IFP. | Partial data |
| The DP and NNP are overwhelmingly urban parties; 94% of the DP vote came from urban and metropolitan areas, while the ANC did better in rural areas except in KwaZulu-Natal. (p. 9) | Rural/urban/metropolitan classification of voting districts by local government type (Table 1, Figure 4), acknowledged to be imprecise in KZN, Mpumalanga and Northern province. | Systematic data |
| Apartheid boundaries still structure voting: the NNP vote is much higher in CLPA districts of the Eastern and Northern Cape, and the ANC does markedly better in ex-homeland areas of Free State, Mpumalanga and Northern Province. (p. 11) | Voting-district data classified by magisterial district into CLPA/homeland/rest (Tables 2 and 3, Figure 5). | Systematic data |
| Coloured voters in the Northern and Western Cape swung toward the ANC, with ANC support among Western Cape coloureds probably around 35%, and white support for the ANC rose only slightly to around 100,000. (p. 5) | Provincial totals and polling district evidence, with racial group shares estimated rather than measured. | Illustrative or anecdotal |
| The NNP collapsed among whites (to the DP), and Afrikaner political fragmentation is unprecedented; less than one-third of NNP support came from whites. (p. 6) | Polling district evidence interpreted against 1994 results and Reynolds' estimates. | Partial data |
| South Africa remains a one-party dominant democracy with fragmented, regional opposition, raising problems for liberal democracy. (p. 13) | Interpretive synthesis of the party-share results across provinces and regions. | Asserted |
The author's caveats¶
The authors note that the IEC would not release voting-district data in machine-readable form or the GIS spatial covers, and that the rural/urban classification by local government type overstates rural and understates urban populations in KwaZulu-Natal, Mpumalanga and Northern Province. They also stress that numbers are very small in the semi-arid rural interior of the Western and Northern Cape.
Gaps in the evidence¶
Race of voters is not recorded at the ballot, so racial and ethnic voting shares rest on ecological inference from district-level aggregates and on estimates borrowed from Reynolds, without any formal ecological-inference method or individual-level survey data of their own. The census race and language data are only loosely documented and not matched to voting districts in a reported way. Linking voting districts to 1996 census enumerator-area data, and using exit polls or the Mattes et al. survey microdata, would test the race and ethnicity claims better. The IEC voting-district data and GIS boundaries now exist and could be used for such analysis.
Replicability: Data poorly documented.