Abstract
Why did whites in South Africa come to support the dismantling of the apartheid system that gave them a monopoly of political power? We use a reformulated version of symbolic politics to address this puzzle, showing that white attitudes toward political change were primarily driven by symbolic predispositions regarding race, ideology, party, and specific leaders, as well as various sorts of threat perceptions. Strong attachments to the National Party and de Klerk, low perceptions of threat, more tolerant racial attitudes, and more socially and politically liberal values increased the likelihood of whites supporting policies consistent with the ending of apartheid. We also find that assessments of the economy, both personal and national, have no influence on this attitude. We use South Africa’s Human Sciences Research Council data collected during the crucial 1991–1992 period.
South Africa’s first truly democratic elections, held in April 1994, resulted in an overwhelming victory for the African National Congress (ANC) led by Nelson Mandela. This constitutional transition from apartheid to majority rule had required radical change in both elite policy and in public opinion over the previous decade. In his 1985 “Rubicon speech” President P.W. Botha had defiantly proclaimed that whites would not give up power: there would be no “abdication and suicide,” he declared (quoted in Welsh, 2009: 231). White public opinion supported his position, with 62% of whites rejecting the idea of negotiations with the ANC even as late as 1988 (Welsh, 2009: 347). Less than a decade later, however, Botha’s successor F.W. de Klerk had agreed to an arrangement denounced by a prominent white historian (Giliomee, 1997) as “surrender without defeat,” and de Klerk carried public opinion with him: more than two-thirds of white voters supported his efforts when asked in a March 1992 referendum.
While the elite politics and the negotiation process of South Africa’s transition are well-studied affairs, the public opinion dynamics are not. Why did whites’ views on negotiations with the ANC change so dramatically in just a few years? Whites understood that talks with the ANC would result in empowering the country’s black majority; why did they support a course of action that would result in their giving up most of their political power?
This is a question of great importance and political urgency. Big political issues typically involve questions of redistribution of power, money, or other tangible benefits—often from the powerful to those with less power. For example, addressing income inequality means redistributing wealth from the rich to the working class; immigration redistributes opportunity from the native-born to the immigrants; and addressing climate change redistributes wealth away from the coal and oil industries and current consumers in favor of future generations. In all of these cases resistance to change is reinforced by hostile attitudes toward the beneficiaries or proponents of change; a prominent example is the anti-immigrant sentiment currently so widespread in the USA, Europe, and South Africa. The South African transition is a rare case in which a visionary leader succeeded in overcoming both short-term material incentives and deeply-rooted prejudice to build a winning political coalition in favor of redistributive political change. It is worth understanding how this was done and the dynamics behind it.
Traditional theories associated with individual demands for political change have stressed either relative deprivation explanations or resource mobilization capabilities. But these theories help in explaining why people decide to join movements that demand change and how they are successful in doing so. By contrast, almost no attention has been given to the reasons why people would accept the kind of redistributive change that leads them to give up political power, which is the situation whites confronted in South Africa in the early 1990s.
Kaufman (2015) has argued that his reformulated version of symbolic politics theory can account for these dynamics. According to this model, political support for any sort of political mobilization is a function primarily of individuals’ symbolic predispositions (especially prejudice and party identification), their threat perceptions, and the way leaders frame the political alternatives. From this perspective, once President de Klerk framed the idea of a negotiated settlement as necessary for South Africa’s whites, support for his policy should have been a function of white voters’ symbolic predispositions and threat perceptions. In this paper, we test this argument. To determine to what extent strictly economic-interest explanations add to our symbolic politics model, we control for the role of assessments of the economy in support for change. Given the fluidity of political conditions and the rapidly changing nature of institutional arrangements, we examine polls conducted at different points of time during this transitional period. What we lose in parsimony by relying on polls with different dependent variables and predictors we gain in accuracy by subjecting our hypotheses to repeated testing. Although our specific dependent variables differ from survey to survey, they are good proxies of what we are trying to explain: the willingness of South African whites to peacefully give up political power. Our goal is not to explain how white public opinion changed during this volatile period but to solve the puzzle of why white voters supported change, using polls conducted at different points in time. Our data come from a series of surveys conducted from 1991 to 1992. The samples were not nationally representative, as they oversampled whites and other racial minorities. This could be a significant problem if we were trying to understand the dynamics of national public opinion. But we are not, and in fact, this overrepresentation of whites is a good thing for our analysis, as we therefore have enough white respondents in the samples to analyze white public opinion. Since blacks and whites in the early 1990s were effectively in different political systems—democratic for whites, authoritarian for blacks—it makes sense to analyze the two groups’ attitudes separately. We focus on the attitudes of whites towards a political process that was designed to end their monopoly of power.
Public opinion and symbolic predispositions
At the risk of oversimplification, theories about public opinion choices can be divided into two general paradigms: the rational-choice paradigm, which includes economic assessment variables; and the symbolic politics paradigm, which includes fear but adds values-driven variables. While we argue that symbolic politics offers a better approach to understand people’s willingness to accept unfavorable political change, we acknowledge that a case can be made for examining the role of material interests in this calculation. Springing from Anthony Downs’ (1957) and Morris Fiorina’s influential works (1981), rational-choice approaches argue that political preferences or choices are a function of voters’ cost-benefit calculations. When assessing the choice to support or reject a political leader, for instance, scholars look at voters’ evaluations of economic conditions (Duch and Stevenson, 2008). Within this paradigm, there is a vibrant debate about how citizens take into account economic information. Some assert that voters focus on retrospective performance, so they shift away from supporting incumbent parties during or after economic declines (Fiorina, 1981); others argue that voters look prospectively to economic conditions (Erickson et al., 2000; Kiewit and Rivers, 1985; Lewis-Beck, 1988; MacKuen et al., 1992). Originally designed to account for political behavior in the United States, these models were soon applied to multi-party systems in Western Europe and Latin America (e.g. Arce and Carrión, 2010; Benton, 2005; Brug et al., 2007; Duch and Stevenson, 2008; Dorussen and Taylor, 2002; Echegaray, 2005; Lewis-Beck, 1988; Norpoth et al., 1991; Remmer, 1991; Weyland, 2003).
Early 1990s South Africa is not the most likely case for the relevance of this rational-choice approach because it was a period of political crisis, dominated by issues of race relations, political violence, and constitutional change. On the other hand, the period was also one of economic difficulty, so economic voting of some kind could have been a factor. For this reason, we control for both prospective and retrospective evaluations of the economy in our analysis, to discern the extent to which these factors might have displaced symbolic emotional ones in influencing public opinion at the time.
Reformulating the symbolic politics approach: the influence of values and threat on political support
The work of David Sears (Sears et al., 1980; Sears and Citrin, 1982) was seminal in articulating a symbolic politics theory of public opinion. According to this theory, voters’ political evaluations are based not on economic assessments or cost-benefit calculations but on “symbolic predispositions,” defined as “stable affective responses to particular symbols” (Sears, 2001). Sears found that some of the most powerful and enduring symbolic predispositions are ideology, party identification, racial prejudice and religiosity. One of his early studies showed that symbolic predispositions accounted for 27% to 37% of vote choice in the 1976 American election, “while self-interest had almost no [effect]” (Sears et al., 1980).
The importance of Sears’ symbolic predispositions is widely accepted in the social-psychological literature on public opinion and voting behavior, even if some of his terminology is not. For example, a recent reformulation of the “Michigan” model of voting focuses exclusively on symbolic predispositions: feelings about the candidates, about the parties, and about specific issue clusters (Lewis-Beck et al., 2008). Marcus (1988) reached a similar conclusion about the 1984 American presidential election, attributing most explanatory power to symbolic predispositions.
This emphasis on symbolic predispositions ties in with recent research in neuroscience demonstrating the degree to which many social evaluations are not rational but intuitive—prerational and heavily biased by stereotypes and symbolic predispositions (e.g. Bartholow and Dickter, 2007). Even calculations of self-interest, from this point of view, are heavily influenced by ideology and other predispositions, becoming so dominant as to often cause distortion of perceived facts to fit the preexisting predispositions (Bartels, 2002). Symbols and stereotypes frequently elicit emotional responses, with important effects on political learning, opinion formation and decision making (Brader, 2012).
In psychology, these insights have been built into the increasingly widely-accepted “dual-systems model” of human decision-making (Evans and Frankish, 2009). According to the model, most decision-making occurs in the “intuitive” system located in evolutionary older portions of the brain. Symbolic predispositions reside here, built out of emotionally-laden associative memories. This intuitive system is extremely fast and powerful; it is the source of “knee-jerk” or “instinctive” behaviors. In contrast, the conscious “reflective” system is much slower and has much less processing capacity. It is also weaker: psychologist Jonathan Haidt (2012) refers to its role as that of a “public relations firm,” generating rationalizations for decisions that were actually made intuitively.
The symbolic politics model is distinct from Marcus’s prominent “affective intelligence” model. Space limits preclude a detailed comparison here; the key point for our purposes is that the two approaches, while differing in focus, are largely compatible given their common stress on party identification and candidate attachments. In particular, Marcus finds that 72% of voters’ behavior is based on party identification and attitudes on specific issues (that is, symbolic predispositions) (Marcus et al., 2000: 116). Kaufman (2015: 36) further argues that Marcus’s third set of major determinants of voting behavior—attitudes toward the candidates themselves—are also symbolic predispositions. Thus, while Marcus does not use the terminology of symbolic politics theory, his findings are compatible with its logic.
Kaufman’s (2015) reformulation of symbolic politics theory builds on these insights, using dual systems theory to reinforce the argument for focusing on symbolic predispositions to explain public opinion. It also adds a new element, based on a different school of psychological thought: “terror management theory”. Studies of the “terror management effect” have shown that stimulating people to think about their mortality has powerful effects at the individual level. When primed to think about their own mortality, participants become more ethnocentric and aggressive; more favorable to those who praise group values; more unfavorable, punitive and aggressive toward members of outgroups; and less concerned with incidental harm to innocent members of outgroups (Greenberg et al., 1990; Pyszczynski et al., 1997; Cuillier et al., 2010). One study found a similar effect of mortality salience on support for George W. Bush in the 2004 U.S. presidential election (Cohen et al., 2005). In a similar vein, Stenner (2005: 8) argues that antidemocratic predispositions can be triggered by “changing conditions of threat.”
Kaufman’s version of symbolic politics theory thus focuses on the interaction between symbolic predispositions and threat perceptions. It also recognizes that physical threats that generate mortality salience are not the only threats that matter. Studies of prejudice have shown that when people perceive threats to their group identity, status or values, they are also inclined to react aggressively (Riek et al., 2006; Kaufman, 2015). Sears’ work includes a similar logic in analyzing the phenomenon of “symbolic racism” in the USA, which includes feelings that social welfare programs that benefit blacks threaten whites’ values of hard work and self-reliance (Sears and Citrin, 1982). To sum up, the revised version of symbolic politics theory is a “values plus threat” theory of public opinion. The structure of our analysis will focus on determining how well this symbolic politics model does in explaining white South Africans’ willingness to support the transition to black majority rule. We control for a number of socio-demographic variables and those related to perceptions of the economy. We analyze three surveys conducted at different points in the transition. Given that these surveys were designed for different purposes than testing theories of support for political change, we are forced to rely on different dependent and independent variables to test our hypotheses. We acknowledge that this is not an ideal situation but believe that this strategy diminishes the changes of “overfitting” theory to one particular data set. Multiple surveys offer the opportunity for repeated tests of our theory, thus enhancing the validity of our findings.
Research design and main hypothesis
Our data come from surveys conducted by South Africa’s Human Sciences Research Council (HSRC). Each survey included a nationwide sample of more than 2000 subjects, with all three official racial minorities (whites, “coloreds” and Indians) oversampled relative to the black majority. This sample structure makes it possible for us to measure the attitudes of the dominant white minority separately from the black majority. Since we are interested in examining white public opinion and how it reacted to De Klerk’s political reform, we only analyze whites. 1
The diversity of the surveys conducted by HSRC similarly provides opportunities but also limitations to our analysis, which we need to acknowledge upfront. The opportunity is provided by the availability of data from numerous surveys conducted by HSRC in the early 1990s. Our first survey (September 1991) was conducted two years into President de Klerk’s reform effort. The second survey (February 1992) was conducted on the eve of the pivotal whites-only referendum that provided crucial public support for that effort. Our final survey (July 1992), happened when negotiations had ceased and the ANC had shifted to a mass action strategy. As we mention at the outset, the surveys oversampled whites, which gives us the opportunity to have enough cases in each data set to conduct regression analyses.
In the tumult of the 1990s, the HSRC’s questionnaire constantly changed. The effect is that there is little consistency in the measures either of our dependent variables or of many of our independent variables. Substantively, it is right that this should be so: the question that white voters were asked in the March 1992 referendum (and which was previewed in the February 1992 survey) was utterly irrelevant to the transformed situation even at the end of that year. We have no choice, then, but to embrace the diversity of information. Thus, our analyses are not strictly comparable from one survey to the next. This is not the problem it might appear to be, however, because we are not attempting to track changes in one specific dependent variable over time; we are examining the influences on support for constitutional change—the general concept each of these variables measures—within each sample. Thus, the variables available in the surveys do allow us to test whether symbolic politics helps us understand why white South Africans were willing to give up power. Indeed, our findings are more robust because we can show that similar patterns hold up across what can be seen as different specifications of the same attitude.
Because the structure of the questions measuring our dependent variable changes, we must also adjust the statistical techniques used to analyze the data. When the survey yields interval data, we use ordinary least squares (OLS) regression; when the dependent variable is binary, we use logistic regression. Ultimately, we believe the gain from repeated analysis outweighs the limitations on the degree to which findings from different surveys are strictly comparable.
Our main predictors are feelings of threat (mortality salience) and values (symbolic predispositions). Since we utilize surveys conducted at different times, we have no option but to operationalize our main predictors in different ways. We assume that questions such as, “How safe/unsafe do you feel in South Africa today?” and degree of agreement with statements such as “Violence is a threat to a genuine democratic order,” “My life in the new South Africa will be worse,” “Inter-racial relations will be worse with a democratically elected government,” and “Security is the most important thing in life,” can serve as proxies for mortality salience, thus enabling us to test the terror management effect by which feelings of threat encourage ethnocentric and aggressive attitudes. Questions such as, “The group’s language and culture will be threatened with a democratically elected government,” measure social threats which tend to prime prejudice, thus encouraging ethnocentric and aggressive attitudes though a different mechanism.
The key question regarding values is which symbolic predispositions should be expected to be powerful. After consideration, we conclude that the symbolic predispositions Sears emphasized—racial attitudes, party identification, and an appropriate measure of ideology—are also relevant for the South African case. Additionally, the other key variable emphasized by the other social-psychological approaches—attitudes toward particular leaders—also seems relevant. Therefore, the key values we examine are: prejudice, predispositions towards political leaders, party identification, and ideology. We operationalize prejudice by examining attitudes towards affirmative action, school integration, apartheid, and whether South Africans share a common sense of fellowship, taking attitudes toward these issues as proxies for prejudice. To measure attitudes about affirmative action we use the following questions: degree of agreement with the policy of “Affirmative action for black citizens,” and degree of agreement with the statements, “Business should be forced to hire more blacks,” and “Fewer whites in civil service to allow more non-whites.” Support for school integration is measured by the degree of agreement with the statement “Students from different background should attend the same schools.” Endorsement of apartheid is measured by a specific question that probes support for it. Feeling of shared fellowship is measured by the degree of agreement with the statement, “South Africans share a sufficient degree of mutual goodwill to ensure a happy future.” Our expectation is that white South Africans who show evidence of prejudice (did not favor affirmative action-type policies, reject school integration, favor apartheid, and reject the idea that South Africans share mutual good will) would oppose the transition to a democratic political system for all.
Predispositions towards political leaders are measured by two straightforward questions that probe whether the respondent believes that de Klerk and Mandela are sincere in wanting a non-racial democracy. As in the case of the American presidency that Sears and colleagues studied, our hypothesis is that the stronger people’s symbolic predisposition toward any political leader, the more likely they are to support the policies advocated by that leader. Party identification is measured by questions about how close the respondent feels to de Klerk’s National Party and the opposition Conservative Party. Our expectation, as in the case of predispositions towards leaders, is that the stronger people’s symbolic predisposition toward their political party, the more likely they are to support the policies advocated by that party.
The final symbolic predisposition emphasized by Sears is political ideology. While a simple liberal-conservative ideological spectrum in the American sense does not directly apply to South Africa, there are other ideological underpinnings that can be examined. For instance, citizens might assign different importance to constitutional guarantees to express political dissent. Also, citizens may hold different views on the proper role of the state in providing for a degree of income redistribution versus the unfettered operation of the free market. Likewise, respondents may have different attitudes about accepting unconventional forms of political activism. Finally, some citizens may embrace more socially liberal values or be in general more satisfied with their lives. Political liberalism, in the sense of being accepting of constitutional provisos for political rights, is operationalized here by the question “Agree with Bill of Rights that limits government powers.” Attitudes towards income redistribution are measured by the degree of agreement with the following statements: “More housing is needed even if that means more taxes,” “Whites must be taxed more to provide for the poor.”
Analysis
September 1991
The release of Nelson Mandela from prison on February 11, 1990 began a period of preliminary “talks about talks” regarding South Africa’s future. Horrific violence also took place, with one tally finding 3699 deaths in political violence in 1990 alone (Guelke, 2010: 241). This violence undermined the credibility of all the key actors. Revelations about ANC efforts to subvert the government and kill its opponents alternated with charges of government aid for ANC rivals. Revelations of past outrages by de Klerk’s predecessor, P.W. Botha, further undermined the government’s credibility in spite of de Klerk’s assertions that he had been unaware of many of them. This was the context in which South Africans were polled in September 1991, with no negotiating progress yet visible.
Given our interest in political support for the transition process, we chose as our dependent variable respondents’ agreement, neutrality or disagreement with the proposition, “The policies of President F.W. de Klerk will bring peace to South Africa.” Answers were measured on a five-point scale from agree fully to disagree completely. We inverted the scale for ease of interpretation and used OLS regression. The main predictors of interests are feelings of threat (agreement that “political violence is a threat to a genuine democratic order”; assessment of “prospects of a better life in the new South Africa”; and the question, “How safe do you feel in South Africa today?”), racial animosity or prejudice (support for racial segregation in classrooms; and agreement that “South Africans share a sufficient degree of mutual goodwill to ensure a happy future”), party identification (closeness to the ruling National Party), and a set of baseline variables that include economic assessments (retrospective and prospective economic performance and perception of income in relation to the inflation rate) and sociodemographic factors (age, gender, education, political interests, ownership of a telephone, and whether Afrikaans is spoken at home).
The results of our analysis of whites’ views are reported in Table 1. There is some evidence that supports the expectation that feelings of threat affect support for the political transition, but it is mixed. On one hand, people who believe that they would have better lives in the new South Africa are more likely to agree that de Klerk’s policies would bring lasting peace. On the other hand, the questions about how the person currently feels in South Africa and whether violence is a threat to a democratic order do not have a significant effect on this attitude. In contrast, values-driven factors, especially proxies for racial prejudice, emerge as consistent influences on the decision to support political change that would diminish the power of white South Africans. Those who favor the desegregation of schools and believe that South Africans share mutual goodwill are significantly more inclined to agree with the statement that de Klerk’s policies would bring lasting peace than those who hold opposite views.
De Klerk’s policies “will bring lasting peace,” white respondents only, September 1991.
SA: South Africa; SAs: South Africans; NP: No prediction; AIC: Akaike Information Criterion; BICR: Bayesian Information Criterion Robust.
≤0.05; **≤0.01; ***≤0.001. Standard errors in parentheses.
Party identification was a core factor in Sears’s original specification of symbolic politics theory, and as predicted by it, party identification among white respondents also influences views about de Klerk’s policies. White respondents closer to the National Party are also more willing to trust the policies of its major leader, de Klerk. In fact, among all the values variables, party identification has by far the most powerful effect, with a coefficient triple the size of the second-strongest values variable (“goodwill”). 2 At this stage of the political transition, assessments of the personal or national economic situation do not play a significant role in driving opinions towards the end of apartheid, as none of the “economic voting” variables emerge as statistically significant. Among the socio-demographic variables, only age turns out to be a significant predictor, with older people more trusting of de Klerk’s policies than younger respondents.
The weak effect of threat perceptions on attitudes towards the political transition is somewhat surprising. We believe what the data reflects is merely the dilemma outlined at the beginning of this section: the alternatives were few. Though political violence was intense and left many South Africans feeling unsafe, the conclusions they drew were ideologically driven. Moderate South African whites felt that in spite of the threat of violence—perhaps even because of it—it was necessary to press on with negotiating a solution. Conversely, extremists had their hostility and distrust toward blacks confirmed by the violence, so they showed no confidence in the negotiation process.
February 1992
Formal negotiations, the Convention for a Democratic South Africa (CODESA) talks, opened in December 1991, so there was room for optimism at the time this poll was taken. However, in November 1991 and February 1992, the National Party lost two by-elections to the more hardline Conservative Party, leaving de Klerk open to charges that he had lost the support of (white) voters. In response to these charges, de Klerk suddenly announced that there would be a nationwide referendum for white voters to be held on March 17, 1992, asking whether they supported continuing negotiations for a new constitution. The proposition was approved, with 68.7% of voters supporting the talks after a brief campaign in which even the ANC campaigned in favor of de Klerk (Welsh, 2009: 439-41).
The February 1992 survey was conducted around the time of this referendum campaign. While apparently designed before the referendum was announced, it asked a related question: whether respondents accepted or rejected de Klerk’s proposal in CODESA to establish an interim government that “must be representative of the total population.” Though imperfect, we use responses to this question as a proxy for support for the negotiation process and for the redistribution of power in favor of South Africa’s black majority. In proposing this measure, de Klerk took a giant step toward meeting ANC demands, and the responses to the question reflected this fact: only 47% of whites supported de Klerk’s CODESA proposal, while 34% opposed it with the rest uncertain. From the white point of view, a “yes” answer to this question reflects a more reformist attitude than a “yes” vote on the referendum would, as the referendum asked only about support for the reform process.
In explaining white attitudes, as shown in Table 2, we find that two out of the four indicators we use to measure perceptions of threat emerge as statistically significant predictors of support for an all-inclusive government. Those who feel safer and those who have less fear about future race relations in a democratically elected government (which clearly meant a black-majority government) were more supportive of de Klerk’s negotiating policy. Although the evidence is only partial, there seems to be an influence of feelings of threat in the determination of the attitudes towards the political transition. From the symbolic perspective, the question about prospective race relations taps values more than fear, probably reflecting relatively positive symbolic predispositions toward blacks. People who expect better race relations in the future would tend to believe that racial tensions of the time were primarily the result of the injustice of apartheid; those who did not probably believed that racial tensions also reflected hostile racial feelings (including, quite likely, the respondent’s own).
White support for de Klerk’s proposal for all-inclusive government, February 1992.
SA: South Africa; NP: No prediction; AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion.
≤0.05; **≤0.01; ***≤0.001. Entries are logistic regression coefficients (robust standard errors in parentheses).
Values variables, as was the case in the September 1991 survey, remain a consistent predictor of endorsement or rejection of political change. People who hold more politically tolerant and liberal (i.e. constitutionalist) views tend to support de Klerk’s proposal for an all-inclusive government in greater proportion than those who hold opposite views. Thus, agreement with the idea that there should be a bill of rights to limit governmental powers and rejection of the proposition that laws are needed to stop strong protests in a future government are associated with greater support for an all-inclusive government. In terms of racial animosity, we could identify only one question that would measure this attitude (agreement with affirmative action policies for black citizens), and it did not turn out to be a significant influence. By contrast, party identification, as was the case with the September 1991 survey, is a strong factor in determining support for the transition, with people who declare feeling close to the National Party more supportive of the political transition. Assessments of the economy, national or personal, once again fail to obtain statistical significance. Similarly, none of the sociodemographic variables turned out to be significant predictors. The non-significance of demographic variables reflects the strength of the symbolic politics model, indicating that it captures the specific attitudes for which demographic variables often work as proxies.
De Klerk and the National Party were keenly aware of their constituents’ fears and symbolic concerns; their campaign for a “yes” vote in the referendum illustrates how they shrewdly used framing effects to gain support for their policy of negotiations. First, de Klerk worked to shift the influence of fear by bluntly arguing, “a ‘No’ vote would be suicidal;” National Party advertisements emphasized that it would lead to escalating racial strife. They also addressed economic fears, arguing that a return to apartheid would spur intensified international economic sanctions; the relevant slogan was, “You can vote yourself out of a job.” Additionally, National Party ads framed support for negotiations as a necessity for South African whites to regain international status, emphasizing that a “no” vote in the referendum would mean that South Africans would again be banned from international athletic competitions, and that “nobody wants the state we seek to make” (Kaufman, 2015: 194–195). Their overall message, in paraphrase, was: don’t fear our policies; fear our opponents.
July 1992
The middle of 1992 was a winter of violent discontent for South Africa. The CODESA talks deadlocked, prompting the Congress of South African Trade Unions, a key ANC ally, to prepare a series of mass protests to demand quicker action. In mid-May, the ANC walked out of the talks. ANC radicals were now in the ascendancy, pursuing a “Leipzig Way” strategy—that is, a strategy of using mass protests to force the South African government to collapse the way the Communist governments of Eastern Europe had folded in late 1989. Meanwhile, a June 17 massacre in the black township of Boipatong served to symbolize an atmosphere of continuing lethal violence, with the ANC and the government each blaming the other for it.
This was the context in which HSRC carried out a poll in July 1992. A key question in that survey got to the core of the political problem, asking participants:
There are different views about how political change should take place in South Africa: which of the following would you support? 1. The ANC and other parties should organize very large protest meetings, marches, stay away strikes and boycotts to force the government to resign from power. 2. The ANC and other parties can organize some protest meetings and marches, but it should keep on talking to the government about change. 3. The ANC, the government and other parties should keep on talking and negotiating, even if it takes many months before agreement[s] are reached. 4. The present talks will not work. The government should run the country and look for other solutions.
The first option represented support for the “Leipzig Way” strategy of revolutionary change; the second, “talks plus protest” option was the position associated with Mandela; and the last represents the preference of unyielding white hard-liners. In our analysis we use answers to this question as the dependent variable (although we reversed the orientation to model support for the radical alternative). We use ordinal logistic regression to estimate the models. Among whites, the majority (52%) favored continued negotiations (option 3 in the original distribution), though fully 42% preferred no change (option 4); 6% tilted toward the second, “talks plus protest” option. If the variable is, essentially, support for more or less radical modes of change, what explains variation in it?
To assess the impact of threat perceptions on attitudes toward change, we take as a proxy of threat the question about what is “most important in your life?” The resulting dummy variable distinguishes those who choose “security” as their response from those who selected another option. This variable is the only indicator we have for perceptions of threat, and it does emerge as a significant predictor of support for radical change, with people who value security being less likely to endorse change (Table 3). In terms of the impact of values, we find some support for the expectation that social conservatism and racial animosity would lessen support for political change. Those with negative views of homosexuality and those less willing to endorse affirmative action for blacks are also less likely to embrace political change to end apartheid. As in previous analyses, we find again that trust in one’s own political leader is an important factor in influencing support for the policies advocated by that leader. Among whites, trust in de Klerk’s sincerity is a statistically significant influence in whites’ attitudes towards radical political change, whereas trust in Mandela’s sincerity has no impact. Clearly, Mandela’s charm and his later-legendary moderate image had no effect in assuaging the concerns of white South Africans at this point in the process. This confirms an important insight of the reformulated version of symbolic politics, namely that trust in one’s own leaders is a powerful influence of political attitudes.
White support for radical political change in South Africa, July 1992.
NP: No prediction; AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion.
≤0.05; **≤0.01; ***≤0.001. Entries are ordinal logistic regression coefficients (robust standard errors in parentheses).
In a manner consistent with previous analyses, none of the variables measuring assessments of the national or the respondent’s economic situation turn out to be significant predictors of support for change. Among the sociodemographic factors, only the dummy variable for Afrikaans spoken at home—effectively a proxy for conservative leanings—emerges as statistically significant.
The denouement
The ANC’s “Leipzig Way” strategy failed. Its major initiative was a September 1992 march on Bisho, the capital of the Ciskei homeland. Refusing to be intimidated, however, Ciskei dictator Oupa Gqozo deployed his army. The result was a massacre: the troops fired indiscriminately on the crowd of marchers, killing 29 and injuring hundreds, many of them shot in the back as they tried to flee (Welsh, 2009: 453–454). The apartheid government would not be intimidated into collapse; the only option was a return to negotiations.
Negotiations had, in fact, been going on through quiet back channels ever since the collapse of CODESA. At the same time, major shifts in thinking in both the government and the ANC impelled both parties to change their position in those talks. The result, just weeks after Bisho, was the signing of the Record of Understanding between the government and the ANC on September 26, 1992. This document laid out the basic outlines that would be followed in the transition process, including the provisions for a transitional government and principles of constitution-building (Welsh, 2009: 456–464). Two years later, with all parties finally on board, South Africa held its first democratic election and chose Nelson Mandela as its first black president.
Assessment and conclusion
These findings help us understand the puzzle we posed at the outset, i.e. why did whites endorse political change that would end their political hegemony? We show here that our model, based on a reformulated version of symbolic politics, provides a powerful theoretical framework to understand this puzzle, and accounts, on average, for more than half of the variance in white attitudes toward change, even though the questionnaires were not ideally suited to testing it.
One of the most potent factors we identified was the respondent’s symbolic predisposition toward political parties: in every case in which feelings of closeness to political parties was measured, white voters’ feelings about the National Party strongly predicted white support for de Klerk’s policy of negotiations, even when it was clear to everyone that the ultimate consequence of these negotiations was to end white supremacy in South Africa.
Symbolic predispositions toward political leaders also proved important where we had clear measures. The July 1992 data showed that white attitudes toward Mandela were irrelevant at the time, but trust for de Klerk was still associated with support for the negotiations process. We found the same pattern in the 1991 survey data—trust in de Klerk strongly predicted support for his negotiations policy—though we omitted this variable from our model because the wording of the questions was too similar. In short, charisma—first de Klerk’s, and later (after the transition) Mandela’s—seems to have mattered greatly.
Also very important were feelings of threat, consistent with the “mortality salience” literature that argues that individuals who exhibit high perceptions of threat are more ethnocentric and aggressive (Greenberg et al., 1990; Pyszczynski et al., 1997; Cuillier et al., 2010). White South Africans with higher threat perceptions fairly consistently opposed political change. We had especially good measures of this influence in the two earlier surveys. In the 1991 survey, the more comprehensive fear, “My life in the new South Africa will be worse,” was the more powerful influence. Whites’ responses in the February 1992 survey showed that two different kinds of fear were influential, with concerns for physical safety (“how unsafe do you feel”), and concerns for inter-racial relations in a future government significant when other values variables were not accounted for.
Unsurprisingly, racial attitudes—both positive and negative—also played an important role in influencing views about the negotiations process. Absolute levels of tolerance increased sharply around the time under study. Thus in 1988, 86% of Afrikaners stated opposition to “racially mixed schools” in a survey (Manzo and McGowan 1992); however, according to our data, only 52.6% of Afrikaners (and 41.5% of all whites) reported the same attitude in July 1992.
Furthermore, racial tolerance is associated with support for the negotiations to end apartheid. In the 1991 survey, both support for school integration and belief that there was goodwill among racial groups both predicted support for the negotiations process. Conversely, opposition of white racists to the process also remained robust. In the February 1992 survey a different question yielded a similar result, with expectations that future race relations would improve predicting support for the talks. Again, we see this as an indication of racial attitudes: those who felt hostile to the other race should have been less likely to expect improvement in race relations in the future. In July 1992, those whites still willing to defend apartheid practices tended to oppose the negotiations process.
Finally, other ideological predispositions proved the most difficult to test with the data available, but we did find some evidence of ideological influence. For example, in the February 1992 survey, support for a bill of rights limiting government power predicted support for the negotiations. Also, rejection of laws to ban strong protests against a future government was also associated with support for negotiations in this survey. Furthermore, in the July 1992 survey, we found that negative attitudes toward homosexuality (a proxy for general social conservatism) also predicted opposition to political change. In contrast to these insights, we find no support for the contention that whites’ preferences regarding political change were influenced by their views on their economic situation or the country’s.
What does all of this tell us about popular support for South Africa’s transition? If the central question is why white South Africans embraced a negotiation process that had to lead to their loss of power, we can offer some answers. First, declining levels of racial hostility meant that the ideological support for apartheid had eroded. While South Africa still contained many unreformed racists, it also included many whites—including many Afrikaners—who no longer supported such pillars of apartheid policy as segregated schools, and it was these people who tended to support de Klerk’s opening to the ANC. Indeed, by 1993, most white supporters of the process reported that they felt a moral obligation to do so.
Second, party loyalty and trust in de Klerk played important roles in maintaining white support for the transition process. National Party supporters showed deep loyalty to the party, including among people who did not agree ideologically with party policy. Relatedly, feelings of trust in de Klerk among white voters also made a very large difference in solidifying support for his policy until after it had become irreversible. While Mandela’s charisma did not play a role among white voters during the negotiation process, deep respect for de Klerk did. Some ideological concerns also played a role: those who valued the principle of a bill of rights were inclined to support a process of creating a new constitution built around a bill of rights.
Finally, though feelings of threat did play a measurable role in undercutting white support for the policy of negotiating with the ANC, those feelings did not overwhelm considerations in the policy’s favor. We can infer, then, that government policy—and, probably, ANC policy—was successful at managing and calming fears among most white South Africans. If white voters were disinclined to trust the ANC, at least they were not terrified of it.
The theoretical contribution of this article is twofold. First, it argues that the updated version of symbolic politics theory offers good theoretical reasons for including these variables—symbolic predispositions regarding race, ideology, party, and specific leaders, as well as various sorts of threat perceptions—into any model of public opinion. Second, it demonstrates that including such variables dramatically boosts the predictive power of statistical models of public opinion. In our particular case, it helps explain a fascinating puzzle in South Africa. These results are achieved using a theory originally designed for U.S. politics, but in the case of another country; and using survey data not especially well-suited for testing symbolic politics theories. We believe that the updated formulation of this theory contributes to our understanding of the cognitive and emotional dynamics that drive public opinion choices in contexts of conflict.
Applied to contemporary politics, symbolic politics theory offers a rigorous, intuitive and powerful framework for understanding contentious politics in the era of Trump and Brexit. It leads with the expectation, well-supported in the data, that support for such reactionary politics is driven by prejudice, ideology and partisanship—i.e. symbolic predispositions—as well as popular fears. It would hypothesize that resistance to the populist wave should prevail where prejudice and threat perceptions are lower, and/or where charismatic moderate politicians emerge (e.g. France’s Macron). Symbolic politics theory thus suggests replacing the slogan, “It’s the economy, stupid” with, “It’s bias and fear, stupid” to define the central determinants of political behavior. De Klerk won because one important bias (prejudice) was declining and two others (party loyalty and his own stature) favored him, and because he assuaged the voters’ fears. This is a political formula well worth learning from.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.