Ethnic Voting and Political Change in South Africa

Literature › Works

Stanton Peele; Stanley J. Morse (1974). Ethnic Voting and Political Change in South Africa. The American Political Science Review, Vol. 68, No. 4. American Political Science Association.

Type	journal article
Scope	Wider apartheid literature
Method	survey fieldwork
Period studied	1966–1974
Regions	Cape Town, Cape Province, South Africa (white electorate)
Themes	politics ideology elections, comparative theory, methods data sources, business and state
Citation key	peele1974

Argument

Using a survey of 462 white voters in three Cape Town constituencies taken just before the April 1970 election, the authors argue that ethnicity (Afrikaans vs English) is the dominant determinant of white party vote. Class, education, religion and rural origin predict vote only because they covary with ethnicity, and the American Lazarsfeld-style multi-factor model and the cross-pressures hypothesis find little support. Parties act as ethnic and racial symbols rather than ideological choices, and the Dutch Reformed Church and Nationalist-controlled channels reinforce loyalty. Young voters and upwardly mobile Afrikaners show no significant shift. The exception is a group of 'ambiguous Afrikaners' who identify as English. Meaningful change is therefore expected to come from within the Afrikaner Nationalist camp (verligte vs verkrampte), not through the opposition parties.

Evidence and method

Original attitude survey of 462 white voters, interviewed in the two weeks before the 22 April 1970 election. The sample was stratified by income across Bishopscourt (upper class), Gardens (middle class) and Epping (lower-class Afrikaans), with a 4.7 per cent refusal rate. It is compared with actual 1970 returns, 1960 census sampling frames, Horrell's race-relations survey, and Lever's South African Voter study. Instruments include a 15-item Likert liberalism scale, Cantril ladder closeness questions, an information index, and party-seats-preferred questions.

Place in the debates

• Applicability of American voting-behaviour models (Lazarsfeld/Michigan) to other societies (see White electoral politics, delimitation and party systems). Partly applies: party labels are symbolic and sociologically conditioned, as in the US, but ethnicity is a single dominant predictor rather than an additive class-religion-residence index.
• Cross-pressures and social mobility (Barber) as sources of electoral change (see White electoral politics, delimitation and party systems). Little support for cross-pressures; Barber's thesis holds only for 'ethnic mobility' (ambiguous Afrikaners), not socioeconomic mobility.
• Why the 1970 election went against the Nationalists (Lever's thesis on HNP-induced Afrikaner abstention) (see White electoral politics, delimitation and party systems). Disagrees with Lever: Nat-preferring abstainers in Gardens were mainly English and more liberal than Nat voters, so they were not drawn to the HNP; their abstention reflects disillusionment or reversion to the UP.
• Prospects for political change in white South Africa (see White electoral politics, delimitation and party systems). Electoral change within the existing party system is unlikely; change must come from the Afrikaner Nationalist majority or non-partisan labour and business organisations.

Findings

• Ethnicity is the primary determinant of white party vote. (p. 19)

  This study shows that ethnicity is the major determinant of party vote in South Africa.

• Other demographic correlates of vote are largely spurious, since they covary with ethnicity. (p. 19)

  Party vote also varies with other demographic categories (SES, religion, place of origin), but this is largely because these factors covary with ethnicity.

• Origin and ethnic background predict vote better than current status; childhood language beats current language. (p. 8)

  then it is his ethnic background, rather than his current status in society, that most strongly determines his party preference.

• UP voters are weak supporters, often voting UP without wanting UP control or as a substitute for the Progressives. (p. 13)

  It seems that many people vote for the UP without any real desire that it overthrow the Nationalist regime, while others fall back on the UP as a substitute for their real choice, the Progressive Party.

• The Cape white electorate is more liberal in attitude than the party it keeps in power, but attitudes do not drive votes. (p. 15)

  the South African electorate, at least in the Cape, is more liberal than the party it helps maintain in power.

• Afrikaners who self-identify as English ('ambiguous Afrikaners') are far more likely to vote UP. (p. 11)

  Afrikaans respondents who identify themselves as English are far more likely to vote for the UP than those with a consistent (Afrikaans) objective and self-identification.

• Feelings of solidarity with whites generally best distinguish Afrikaner Nat voters from UP voters. (p. 18)

  Among Afrikaners (although not at all among the English), this is the social distance measure which best differentiates Nat and UP voters (8.5 vs. 7.0, p <.005).

• White respondents grossly underestimated the African share of the population, Nat supporters most of all. (p. 18)

  Our respondents estimated that Africans comprise 44 per cent of the country's population (the actual figure is 69 per cent).

• Abstainers in Gardens were liberal rather than HNP-leaning, contradicting Lever's hypothesis. (p. 16)

  This decisively disconfirms Lever's hypothesis that abstainers in 1970 were primarily Afrikaners wavering between the Nats and the more conservative HNP.

• Change is expected to originate within Afrikaner nationalism rather than through opposition parties. (p. 21)

  Rather, political change in South Africa must come from movements arising within the Afrikaner Nat majority, possibly within the NP itself.

• The authors note that labour unions, once co-opted by the Nats, were becoming more independent, including accepting unionisation of black workers. (p. 21)

  Labor unions, once co-opted by the Nats, have been taking a more independent line, in some cases accepting the unionization of black workers.

Evidence ledger

Sources used

Source	Kind	Period	How used
South African Census 1960 (Cape Town income-earning white male population)	population census	1960	Sampling frame and quotas by income class; sample checked against census demographics and language shares.
Official 1970 election returns for Constantia, Cape Town Gardens, Tygervallei and Germiston	election returns	1970	Actual constituency results are compared with respondents' intended votes to validate the survey.
Muriel Horrell, A Survey of Race Relations in South Africa: 1968 (SAIRR, 1969)	official statistical series	1968	Source of population group sizes and Afrikaans-speaking share, and of white university enrolment figures.
Henry Lever, The South African Voter (Juta, 1972): 1966 and 1970 election data from Johannesburg marginal constituencies	secondary literature compilation	1966 and 1970 elections	Prior survey findings serve as comparison, and Lever's hypothesis about abstentions is tested and rejected with the authors' data.
Open-ended survey responses on likes and dislikes of parties	author survey or fieldwork	1970	Used qualitatively to suggest deviant and changing voters acted on feelings about the Nats, though too few responses for statistics.
Peele doctoral dissertation, 'The Ethnic Basis of Voting Behavior in South Africa' (University of Michigan, 1973)	author survey or fieldwork	1973	Fuller report of the same survey, cited for further detail.
Orpen doctoral dissertation and Journal of Psychology article on authoritarianism and prejudice in South Africa	secondary literature compilation	1970-1971	Prior survey findings that prejudice is culturally normative are said to be supported by the authors' own data.

New data

Stratified area survey of 462 white voters in three Cape Town constituencies (Bishopscourt, Gardens, Epping) in April 1970, with closed-ended items on voting intention, party preferred, 15 Likert attitude items, social-distance ladders, political information, and demographics; about 4.7% refusal rate.

Inference

Quantification: Statistical analysis.

Cross-sectional survey analysis: chi-square, contingency coefficients and lambda for vote against demographic predictors, ethnicity controlled by sub-sample tabulation, ANOVA and Pearson correlations for attitudes and party-preferred data, validated against actual constituency returns. No multivariate regression, and no temporal or panel evidence on change.

Key claims

Claim	Basis	Strength
Ethnicity (childhood language) is the dominant predictor of party vote; SES, religion and birthplace matter mainly because they covary with ethnicity. (p. 1537)	Author survey: chi-square, contingency coefficient and lambda for each predictor (Tables 3-8), with ethnicity controlled.	Partial data
Young voters, and Afrikaners rising into business and higher SES, show no significant shift from the NP to the UP. (p. 1538)	Survey cross-tabulations of vote by age, SES and education within Afrikaners; Table 6 has very small cells (32 NP, 11 UP).	Partial data
The cross-pressures hypothesis finds little support, as information avoidance and abstention patterns are weak. (p. 1530)	Survey ANOVAs on information sources and concern about the election by ethnic group; N's for vote changers are tiny (21 English, 9 Afrikaners).	Partial data
Intended vote is a valid indicator, as survey intentions match actual constituency returns and are not an interviewer effect, except in uncontested Epping. (p. 1529)	Comparison of survey intentions with official returns in Constantia, Gardens and Germiston; checks on interview language.	Partial data
A group of 'ambiguous Afrikaners' (43 of 210 Afrikaans respondents) who identify as English splits roughly evenly between NP and UP and may be a source of political change. (p. 1529)	Survey social-distance questions and Table 9 (n=27 self-identified English among Afrikaans respondents).	Partial data
UP voters are less firm supporters than Nat voters, and Nat-preferring abstainers in Gardens are mostly English and more liberal, contradicting Lever's HNP-conflict explanation of abstention. (p. 1531)	Survey party-preferred seat allocations (Tables 10-12), attitude means and t-test for Gardens abstainers; small cells.	Partial data
Feeling of closeness to South African whites and to one's own ethnic group best differentiates Nat from UP voters, whereas racial attitudes toward nonwhites have little bearing on vote. (p. 1536)	Survey closeness-ladder ratings by ethnicity and vote (Table 17 and text).	Partial data
The Cape Town sample is liberal relative to the parties; political change must come from within the Afrikaner Nat majority rather than via the existing party structure. (p. 1539)	Survey attitude scale means (self 3.5 vs NP perceived 2.8), plus post-1970 developments referred to without sources.	Asserted

The author's caveats

The authors acknowledge small N's in controlled tables, the small numbers of Prog and HNP voters, the unusual Epping nonvote and a sampling error in Bishopscourt. They note the sample over-represents Afrikaans speakers (39% vs 33%), that analysis of variance is used heuristically, and that the Cape may not be typical. They also acknowledge too few vote changers to test Lever's thesis and that the attitude scale explains little variance.

Gaps in the evidence

The argument rests on a single-city, single-wave survey of whites in Cape Town and cannot test change over time. A panel, or a companion 1966 study, or national or multi-region samples would test electoral change and generalizability; Lever's Johannesburg data existed but are only compared loosely. Constituency-level election returns across censuses, and the 1970 census once released, could show urbanization and ethnic voting at the aggregate level. Small cell sizes make claims about young Afrikaners and ambiguous Afrikaners fragile.

Replicability: Data poorly documented.