Re-Running the 1994 Elections with Different Systems         

Andrew Reynolds

        The 1994 national parliamentary elections held in South Africa marked the high point of a period of tumultuous change from authoritarian rule to multi-party democracy in the region. There is strong evidence to suggest that institutional factors have an important role to play in the consolidation of a new democracy. Nowhere is this more apparent than in an ethnically divided or plural society like South Africa, with a history all too often filled with conflict over fundamental preferences and beliefs.
        Constitutional design carries most weight when politics is at its most dangerous; i.e., when it is seen as, or in practicality is, a "zero-sum game." The recent experience in Angola, Algeria and elsewhere in Africa illustrates that the concept of a "loyal opposition" is a difficult one to entrench when one segment of society sees that losing an election is equivalent to being completely shut out of governmental power.
        It is for these reasons that the issues of: whether to adopt majoritarian or consensual democracy; whether to choose a presidential or parliamentary executive system; and whether to constitute parliament by a plurality, majoritarian, or proportional representation type electoral system, become critically important to the prospects for democratic consolidation in a divided society.

Assessing Different Electoral Systems

        This article will concentrate on the third of those constitutional engineering choices -- the type of electoral system to be used. By re-running the national parliamentary elections in South Africa we can assess how the use of alternative electoral system formulae would have changed the composition of government and parliament. The resulting empirical evidence can then be used to assess how stable the system would have been likely to be considering the pre-existing divisions within each society.
        For the purposes of this exercise I have re-run the election under: (1) single member district (SMD)
plurality; (2) a majoritarian system (the Alternative Vote, or AV) in the same single member districts; (3) regionally based constituency list proportional representation (PR). To these main electoral systems I have added: (4) the Alternative Vote in multi-member districts (MMD), which was the important and widely discussed, proposal made for South Africa by Donald Horowitz in his award-winning book A Democratic South Africa: Constitutional Engineering in a Divided Society.

Methodology and Assumptions

        The data used in this paper comes from the full district level voting figures released by the Independent Electoral Commission of South Africa which gives voting figures, by party, for nine provinces and 343 counting districts. The underlying assumption of the simulation exercise is that voter preferences would have, by and large, remained the same regardless of the electoral system used in South Africa and therefore electoral system incentives would not have significantly affected the votes cast.
        But indeed the predominant objection to re-running elections under alternative systems is that a voter's party preference may not be constant under varying electoral rules as some systems, chiefly plurality type ones, provide powerful incentives for minority party supporters to vote tactically where they believe their first choice party has no feasible chance of winning in a single member district.
        Preference voting systems, such as the Alternative Vote, help in part to negate such tactical voting by allowing supporters of minority parties to express their second preference on the ballot. At the other extreme, list PR, held in districts of large magnitude, free the voter to choose their first preference party almost regardless of that party's perceived support because the low threshold for representation makes "wasted" votes far less likely.
        However, I would argue that intuitively this objection is far less troublesome in the case of democratic elections in South Africa as party identification and voting preferences were very strongly held, reflecting polities highly polarized along ethnic, linguistic, cultural, ideological and regional lines. Polling evidence from South Africa supports this intuition that there was little scope for voting "defection" among party supporters.


Re-Running the 1994 South African National Elections Under Alternative Electoral Systems

Key: ANC= African National Congress, NP= National Party, IFP= Inkatha Freedom Party, FF= Freedom Front, DP= Democratic Party, PAC= Pan-Africanist Congress, ACDP= African Christian Democratic Party, I.D.= Index of Disproportionality

Overview of Current System

        South Africa's general election of April 1994 was held under a form of national list PR with half the National Assembly (200) being constituted from nine provincial lists and the other half being constituted from a single national list. In effect the country used one nation-wide constituency (of 400 members) for the conversion of votes into seats and no threshold for representation was imposed. The Droop quota (votes/ [seats + 1]) was used to apportion seats and surplus seats were awarded by an adaptation of the largest remainder method.
        Complete results for each of the four electoral system simulations, and the actual national list PR results, are given in table 1.

A Re-Run With Plurality: One Party Domination

        The first, and probably most salient, issue to note concerning plurality's effect on the South African National Assembly election is that the ANC would have won, under this system, more parliamentary seats than under any alternative system and furthermore that they would have received well over two-thirds of the Constitutional Assembly giving them enough power to write the new constitution without formal reference to any other party in the Assembly.
        The 283 ANC seats represent a "seat bonus" of 31 members, over and above what the party would have won under a purely proportional electoral system. Interestingly the vast majority of ANC seats turned out to be "safe" ones, in the sense that there was no competitive challenge in the district from another party. Only eight (3%) of the ANC's seats were not won with an absolute majority and the average ANC seat was taken with over 80% of the popular vote (in seats the ANC lost, it averaged 19% of the popular vote).
        These results illustrate, that under plurality at least, the likelihood of the ANC losing their grip on parliamentary power in the next, or subsequent, elections would be slim. If the Nationalists were to make gains from the ANC (or if Inkatha, or the PAC, were to win converts from the ANC) those defectors would more likely be from already safe ANC constituencies and thus the ANC could easily hold constituencies despite a moderate drop in their national popular vote.
        In Britain marginal seats are categorized as those seats won with a majority of less than 10 percent of the recorded vote. These then are considered the "battle-ground" seats -- those that determine which party forms a government. In the 1992 British parliamentary elections there were 171 such marginal seats, or 26 percent of the total.
        However, under the plurality simulation for South Africa only 13 (3 percent) seats were won with majorities of less than 10 percent, and thus would be considered "marginals." In effect the system would be frozen in terms of seats won and requiring a political earthquake to jar the patterns of party voting concentration.
        This finding is a serious blow to the case for plurality in South Africa, for such majoritarian prescriptions rely on the perception that executive power can change hands from election to election. If plurality leads to a de facto "one-party state" then there are no incentives for losing parties (in this case the NP, IFP, FF, and PAC) to remain "loyal parliamentary opposition" parties. A deeply divided society like South Africa is most threatened in its fledgling democratic times if incentives for pro-system loyalty are lacking.

Plurality and Incentives for Violence

        The second most serious threat to stability posed by the plurality results is the total exclusion from parliament of the smaller minority parties (the Freedom Front [FF], the Pan Africanist Congress [PAC], the Democratic Party [DP] and the African Christian Democratic Party [ACDP]).
        The exclusion of the liberal Democratic Party might be a sad loss to parliamentary debate in the new South Africa and clearly Helen Suzman's 13 years as a lone voice in the previous "whites-only" parliament illustrated that the level of parliamentary effectiveness does not always correlate with a party's number of seats.
        Similarly the loss of the ACDP's two seats might eliminate another small but growing and dynamic section of opinion. But the threat really would emanate from the exclusion of representatives of the Afrikaner Freedom Front led by General Constand Viljoen and the Pan Africanist Congress of Clarence Makwetu. Both parties are currently undergoing an internal "battle for the soul" waged by proponents of moderation versus those who have in the past advocated and practiced anti-system violence.
        The pressures on the Freedom Front to withdraw from the democratic system come from Afrikaners within the Conservative Party (CP) and Afrikaner Weerstandsbeweging (AWB), and PAC withdrawal is argued for by similarly vociferous racists in the Azanian People's Liberation Army (APLA) the armed wing of the PAC nationalist movement.
        I believe that the 14 combined seats won by these diametrically opposed parties under the list PR system invests each group with a "pro-system" democratic mentality which helps to tip the balance in favor of the moderate wings of each respective party. If, as under the plurality electoral system, the FF and PAC had won not a single National Assembly seat, then their incentives for anti-system violence might have been greatly increased.


Plurality and Regional Fragmentation

        Another telling implication that plurality would have had for the evolution of politics in South Africa is the way in which the system would have created regional/provincial fiefdoms of homogeneous party power. Under the actual results each province sent representatives from at least two parties -- five parties in the PWV case -- to sit in the National Assembly.
        But under the plurality simulation all the seats were won by a single party (the ANC) in four provinces (North-West, Northern Transvaal, Eastern Cape, Eastern Transvaal) and all but one seat was won by the ANC in the Free State. The NP would have been far more dominant in the Western Cape (winning 34/42 seats instead of 24/42) and the IFP more dominant in KwaZulu-Natal (taking 49/80 seats as opposed to 39/80).
        Such concentrations of representation would first, increase pressures for the politics of clientelism -- the exclusion of some regions from government resources and the inclusion of others -- and second, eliminate the ability of minority party supporters in most regions to approach a representative from their region who sympathized with their views.

Re-Running the Election Under the Alternative Vote

        The first thing to note about the impact of the alternative vote on South Africa's election is that, for all intents and purposes, the results would have been the same as under a SMD plurality system. This is due to the fact that most seats (94 percent) were won with absolute majorities (making the AV redundant) and of those 17 seats won by a plurality only six would have been likely to have been won by the second placed party in the plurality vote.
        The ANC would still have had their two-thirds parliamentary majority although with six seats less than their plurality victory. The NP and IFP's handful of extra seats would not have been enough to alter the dynamics of the new National Assembly. The complaints about both the exclusion of the FF, DP, PAC and ACDP and also the troubling occurrence of regional fiefdoms would have remained.

The Alternative Vote in Multi-Member Districts

        Before explaining the results of the alternative vote in multi-member (MMD AV) simulation in South Africa, it is useful to give some background to the rationale Donald Horowitz cites for his proposal and the ensuing pre-election debate which then took place within scholarly publications.
        In his 1991 book A Democratic South Africa? Horowitz argued that "The incentive to compromise, and not merely the incentive to coalesce, is the key to accommodation" in an ethnically divided society. This rested, on the empirically well-founded premise, that parliamentary coalitions between differing ethnic parties often fall apart because there is no substantive common interest glue to keep them together.
        Horowitz therefore argued that incentives for "pre-election" compromise that encouraged "vote-pooling" or party appeals across ethnic boundaries, were key to crafting a stable and less ethnically divisive constitutional order. In practical terms Horowitz argued that these incentives would be engendered by the Alternative Vote, a majoritarian type electoral system also called Majority Preference Voting, which forces the winning candidate of a SMD to win an absolute majority of the district's vote (i.e., 50 percent plus one).
        If no one candidate wins over 50 percent of the first preferences then the lowest polling candidate is eliminated and his or her second preferences are redistributed to candidates remaining in the race. This process continues until a single candidate surmounts the 50 percent threshold. However, these incentives only "kick in" if no one party has an absolute majority of (or perceives it has an absolute majority of) the districts votes. As Horowitz himself noted, "If a party can win on first preferences, second preferences are irrelevant".
        Due to the correlation of ethnicity, geographical concentration and party preference in South Africa, Horowitz recognized that constituencies would have to be heterogeneous to allow for the possibility of no single party winning an absolute majority. "To achieve this, the constituencies may have to be large, and they may therefore need to be multi-member constituencies."
        The two most detailed analyses of the "Horowitz proposal" are Arend Lijphart's 1991 article in the South African Journal of Political Science, Politikon, and chapter five in my own 1993 book Voting for a New South Africa. Both of these pieces questioned the applicability of MMD AV for South Africa and prophesied that the system would have disastrous implications for stability and democratization in South Africa if used in practice.
        Lijphart's critique rested on three main points: firstly, that party coalitions within parliament carry similar incentives to compromise as "vote-pooling" incentives on the electoral level. Secondly, that AV, in theory and practice, resembles the "majority-runoff" method which was found to be highly unsatisfactory when used in divided societies in 19th century Europe. Thirdly, Lijphart refuted Horowitz's claim that AV would "mitigate the winner-take-all aspects of plurality systems"...and that it would by necessity..."achieve better proportionality of seats to votes than plurality systems do".
        As Douglas Rae stated in his 1967 comparison of electoral system consequences, "the Australian system (of AV) behaves in all its particulars, including its degree of disproportionality, as if it were a single member district plurality (FPTP) formula." Furthermore, Lijphart notes that AV operated in multi-member constituencies makes the system even less proportional and more majoritarian. "In PR systems proportionality increases as district magnitude increases, but the relationship is just the other way round for majoritarian systems. AV's disproportionality will rise sharply when it is applied in multi-member districts."
        Indeed the only empirical case study we have of MMD AV -- the elections for the Australian Federal Senate between 1919 and 1946 -- proves this to be the case. J.F.H. Wright has argued that in all ten elections with this system the results were unsatisfactory. "On two occasions parties with less than half the votes won majorities of seats, and on eight occasions, parties or coalitions with more than 40 percent of the votes won three seats or less (15 percent)." The system displayed its worst anomalies in 1925 when Labour, with 45 percent of votes, failed to win a single seat, but then in 1943 Labour won all 19 seats up for election with 55 percent of the popular vote.

MMD AV In Practice

        The results in table 1 illustrate that MMD AV in South Africa would not only have been far less proportional than PR but even less proportional than results simulated for a straight SMD plurality system. This is primarily due to the occurrence of minority parties winning plurality SMDs, but being submerged, and defeated, in larger multi-member AV majoritarian constituencies.
        The practical implications for South Africa were; firstly that the ANC would have gained a "seat bonus" of 30 and, as under plurality and SMD AV, would have had enough parliamentary seats to write the new South African constitution unfettered; secondly that, similar to the other majoritarian systems, all other minority parties (bar the NP and IFP) were excluded from representation; and finally, that the creation of homogeneous provincial party fiefdoms became even more pronounced.
        The NP lost five seats from their plurality total -- they picked up two seats in the Western Cape and one in the Northern Cape, but these gains were outweighed by losing eight seats in KwaZulu-Natal. The IFP were the main beneficiaries of that loss, picking up six seats, over and above their plurality total, from KwaZulu-Natal.
        However, the most important implication for the Horowitz proposal is that the geographical concentration of each parties vote meant that only three multi-member constituencies (totaling 19 seats) were not won with an absolute majority and therefore in only 5 percent of the National Assembly seats was there any incentive for a party to appeal outside its particular support base.
        Horowitz could well argue that I simply have not tried hard enough to draw heterogeneous constituencies but the reality is that drawing substantial numbers of plurality-won MMDs in South Africa, under current electoral realities, would be a task making "affirmative districting" in the U.S. for compliance with the Voting Rights Act look like a picnic.
        There is no conceivable way of drawing such districts in the Orange Free State, North-West, Eastern Cape, Northern Transvaal or Eastern Transvaal provinces, and in vast swathes of the PWV area and the Western Cape. Only in the Northern Cape and KwaZulu-Natal could such districting be countenanced and even there it would require massive gerrymandering of boundaries losing any semblance of geographical compactness.
        This is precisely the sort of electoral manipulation that Horowitz has railed against in black majority districts in North Carolina and elsewhere in the United States. Increasing the magnitude of the districts does not help and would add the disadvantage of a highly complicated ballot in a multi-candidate, multi-member, preference voting constituency.
        From the evidence it appears as though the Horowitz proposal might only have merit when either; a) there is a massive realignment of voting behavior, or b) when there is substantial dwelling mobility among different ethnic groups and districts become far more integrated and diverse than they are at the present -- neither occurrence seems likely in the foreseeable future.

Re-Running the Election Under Provincial List PR

        Re-running the election on a provincial list PR basis with no national or compensatory list seats would have largely left the national results unaltered as is illustrated in column 4 of table 1. The ANC would have gained three seats, one from the Freedom Front and both the parliamentary seats of the ACDP, but this would not have altered the dynamics of the new parliament as the ANC's absolute majority still would have fallen short of a two-thirds super-majority.

Electoral Systems Make a Difference

        The results contained within this paper clearly illustrate that institutional design, and more specifically electoral system design, matter a great deal in the early stages of a multi-party democracy. In South Africa the choice of electoral system would have determined:
        • 1) the ANC's ability to write the new constitutional unfettered, or with reference to other parties;
        • 2) the ability of the FF, DP, PAC, ACDP, AMP and AMCP to gain representation;
        • 3) the nature of territorial representation by party.
        Furthermore, as the first Cabinet of President Nelson Mandela was apportioned along proportional lines, with each party winning over 5 percent of the Assembly seats being awarded a number of cabinet portfolios, the choice of system would have also effected the balance of power within the cabinet. If a plurality system had been used instead of a PR one the ANC would have taken one extra Ministerial portfolio and the NP one less.
        The evidence from this case study is not definitive but the results do highlight clear trends which have become apparent during the "third wave" of democratization. i.e., that plurality or majoritarian systems are detrimental to democratic consolidation in ethnically divided fledgling democracies and that proportional representational type systems are necessary to provide the inclusive ground rules to enable party compromise and ethnic accommodation to flourish.

The Impact of Plurality

        In South Africa plurality would have been detrimental in four important respects.
        • First, minorities would have been excluded, thus engendering a climate in which anti-system extremist violence would have been more likely.
        • Second, regional fiefdoms of concentrated party support would have been exaggerated and exemplified at the parliamentary level providing the recipe for government clientelism and the politics of inclusion of allies and exclusion of foes.
        • Third, the party system would have "frozen," with few seats being competitive and likely to change hands from election to election. An adequate level of seat competitiveness is a pre-requisite for a successful majoritarian system of governance.
        • Fourth, if plurality had been combined with majoritarian, "Westminster" system of government (as is usually the case), then the classic "zero-sum game" so detrimental to ethnic inclusiveness in a divided society would have been heightened.
        However, the simulation does illustrate one interesting (and as yet under recognized) aspect of plurality in that the system provides relatively proportional national results in countries which have highly concentrated ethnically based parties. Results from the 1994 Malawi elections further demonstrate this point.

The Use of Alternative Vote

        While the use of the Alternative Vote in single member districts would have had no great effect on plurality results in South Africa the Alternative Vote used in multi member constituencies (or the "Horowitz" proposal) would have had a much greater impact. The Horowitz proposal would have given rise to the most disproportional results, exaggerated regional fiefdoms of support, advantaged the largest and discriminated against the smallest parties.
        As I noted earlier, for "vote-pooling" incentives to work constituencies must be created that avoid single party majorities, but the realities of South Africa illustrate that crafting such constituencies would, at the very least, take massive degrees of gerrymandering and probably then would still be impossible to manage.

Choosing a PR System

        With a failure of plurality/majoritarian type electoral systems we are left to consider variations of PR systems. PR appears most attractive because it has in South Africa facilitated inclusive parliaments representing majorities and minorities, which form the bedrock of any viable power-sharing constitutional dispensation. PR helps to mitigate incentives to clientelism and the exclusion of those geographical areas which did not vote in large numbers for the government of the day.
        Added to this is the intuition that access to representation and advocacy in South Africa is better under PR systems because voters would be more likely to have a representative of their choice from their region, if not from their single member district. Finally, when deciding between "provincial" and "national list" PR it seems as though national list PR includes the wider array of opinion in parliament and so one would lean towards recommending that, even if the PR system was constituency based, there would be a national or compensatory list element to facilitate the inclusion of small parties whose support was widely geographically dispersed.

        This article is condensed from a chapter in Elections in Africa (Washington DC: USIP Press, 1995), edited by Andrew Reynolds and Timothy Sisk. Reynolds is a doctoral student at the University of Calif.-San Diego specializing in constitutional design and democratization in Southern Africa. He edited Election '94: South Africa: The Campaigns, Results and Future Prospects (St. Martin's Press, 1994) and wrote Voting for a New South Africa (Cape Town: Maskew Miller Longman, 1993).

Table of Contents
Spotlight Four