A Systematic View on the Electoral Reform Debate in South Africa

A wide variety of electoral systems is in operation throughout the world, and in principle an infinite number of electoral systems can be designed. According to Dieter Nohlen 1 , all electoral systems can nevertheless be traced back to either one of the two basic principles of representation: the principle of majority representation or the principle of proportional representation (PR). The distinction between these two basic principles is mainly based on their respective political objectives regarding the nationwide electoral outcome. The major political aim of the princip1e of majority representation is to produce a parliamentary majority for one party or for a coalition of parties. The major political aim of the principle of proportional representation is an accurate reflection of social and political groups in Parliament. Empirically, the difference between the principles of majority and proportional representation can be seen best in their effects on the relation between votes and seats. In majority representation systems there is usually a significantly higher degree of dispropor­ tionality between votes and seats than in proportional representation systems.


Introduction
A wide variety of electoral systems is in operation throughout the world, and in principle an infinite number of electoral systems can be designed. According to Dieter Nohlen 1 , all electoral systems can nevertheless be traced back to either one of the two basic principles of representation: the principle of majority representation or the principle of proportional representation (PR). The distinction between these two basic principles is mainly based on their respective political objectives regarding the nationwide electoral outcome. The major political aim of the princip1e of majority representation is to produce a parliamentary majority for one party or for a coalition of parties. The major political aim of the principle of proportional representation is an accurate reflection of social and political groups in Parliament. Empirically, the difference between the principles of majority and proportional representation can be seen best in their effects on the relation between votes and seats. In majority representation systems there is usually a significantly higher degree of dispropor tionality between votes and seats than in proportional representation systems .
In the past few decades, the variety of both majority representation systems and propor tional representation systems has been widened considerably without sufficient theoretical discussion accompanying this trend. In the continuous debate on majority representation versus proportional representation, advocates and critics of either princip1e are still inc1ined not to adequately accept the variety of subtypes of electoral systems . They still try to estab lish the pros and cons of either basic princip1e of representation from the extreme ends of the scale, the plurality system (first-past-the-post system) on the one hand 2 and the pure PR 2 26 Nohlen, D. , Wahlsysteme der Welt, Piper, München! Zürich 1978; ders. , Wahlrecht und Parteien system, Leske & Budrich, Opladen 1990; ders. , Elections and Electoral Systems, Macmillian India, New Delhi 1996. At present the plurality system in single-member districts is applied in Great Britain, Canada, the United States, in many former British territories in Africa (for example Botswana, Kenya, Malawi, Tanzania, Zambia, Zimbabwe), in Asia (for example India, Malaysia, Singapore, Pakistan, Bangladesh) as weil as in all anglophone Carribean countries except Guyana. A few countries apply the plurality system in multi-member districts, for example Mauritus with its three-past-the post system. system (PR list system at national level) on the other 3 . Other subtypes of majority repre sentation and proportional representation are hardly dealt with in such theoretical discus sions. This is also the case with other "classic " variations such as absolute majority systems 4 and PR list systems in multi-member constituencies 5 , and even more so with electoral systems where the seats are allocated on different levels (e.g. constitueneies, constituency assoeiations, national constituency) and in multi-phase and separate electoral procedures. Such systems are for example a) PR list systems in multi-member constituen eies with additional seats at national level, which are applied, for example, in Poland and Guatemala; b) PR list systems in multi-member constitueneies with compensatory seats at the national level, which are applied, for example, in Denmark 6 ; c) the mixed-member proportional system (personalised proportional representation), a system which has proved worthwhile in Germany and which was introduced a few years ago in New Zealand and, slightly modified, also in Venezuela. Under the mixed member proportional system, a part of the parliamentary seats is distributed in single-member constitueneies, whereas the votes obtained by the parties at the national level constitute the only basis for determining, proportionately, the seat shares of the parties 7 ; d) the compensatory mixed-member propor tional system, which was only recently introduced in Italy and Hungary. Under this system one share of the seats is assigned under the plurality fo rmula in single-member constituen-4 5 6 7 The classic example of the pure PR system is the electoral system of the "Weimar Republic" in Germany. Today the pure PR system is in operation in a few countries only, for example in Israel (with a threshold of 1.5% at national level), Guyana, Namibia and the Netherlands. France, in particular, is the force behind this system. Different forms of absolute majority systems are applied, among other countries, in Gabon, Haiti, the Comoro Islands, in Congo, Mali, Macedonia, the Ukraine, White Russia and in Australia (in conjunction with the alternative vote). This type of PR system is used in many Western European and some Eastern European countries (for example Austria, Finland, Czechia, Lichtenstein, Luxembourg, Portugal, Spain, Siovakia and Switzerland), in most Latin American states (for example Argentina, Brazil, Colombia, Costa Rica, EI Salvador, Paraguay, the Dominican Republic) as weil as in some African and Asian countries (for example Benin, Cape Verde, Indonesia). For a detailIed description of the Danish electoral system see: Parliamentary Election Act of Denmark (Act NO.27 1 of May 13th 1987, latest amendments added on April 10th, 1991) as weil as Elklit, J. , Simpler than its Reputation: The Electoral System in Denmark since 1920, in: Electoral Studies, Vo1.l2, 1993. It is noteworthy, that the effects of the Danish electoral system -namely, high proportional outcome (which results from calculating the proportional seat share of each party at the national level) and personal vote (by open lists) -are very similar to the mixed-member proportional system, discussed below; see Elklit, J./ Roherts, N., A Category of their Own? Four PR Two-tier Compensatory Member Electoral Systems in 1994, in: European Journal of Political Research, Vo1.30, 1996 For a descripction of the electoral systems of Germany, New Zealand and Venezuela see, among others, Nohlen, D. 1990Nohlen, D. , 1996 eies, another share is assigned according to the proportional formula in anational constitu ency (or in large multi-member constitueneies at provineial level). Contrary to the parallel system (see below) the assignment of seats is not carried out in completly separate procedures. The allocation of the list seats aims at compensating to a great extent the disproportionality effect of the plurality formula in single-member constitueneies 8 ; e) mixed-member majority systems (e.g. majority systems with additional proportional lists/parallel systems), on which elections in Mexico, Japan, Russia, Croatia and Lithuania are based. Under this system, in a separate procedure, one share of the seats is allocated according to the majority fo rmula in single-member constitueneies and one share according to the proportional formula in anational constituency (or in large multi-member constitu eneies at the provineial level). The total electoral outcome is, as it were, composed of two independently ca1culated results.
It is essential that theorists of electoral systems give their increased attention to such complex electoral systems -and not only to the effects of these systems but also to their origins.

Constituency elements within Proportional Representation systems
In the light of these more complex systems the constituency elements within Proportional Representation (PR) systems will be discussed next. In the c1assical debate on electoral systems, constituency elements have been traditionally connected with majority representa tion systems, not with PR systems. The reason for this is that the debate usually started out from the extreme ends of the scale of subtypes -that is the plurality system on the one hand and the pure PR system at national level on the other. The pure PR system at national level operates without any constitueneies or rather with one national constituency.
The situation became somewhat more complicated after PR systems began to operate in smalI, medium and/or large multi-member constitueneies in many European and Latin American states today during this century. This was when the distinction between single member and multi-member constitueneies became more important. Usually, only majority representation systems used single-member districts, where individual party candidates were elected. Under proportional representation systems, members of Parliament generally were elected in multi-member constituencies. With the combined electoral systems coming into existenee, or rather after taking note of these, it beeame eompletely questionable to dassify eleetoral systems where single-member eonstitueneies are used exdusively as plurality or majority systems. The reason for this lies in the fact that some of these more eomplex eleetoral systems eombine single-member eonstituencies with the prineiple of proportional representation. We are referring here to the mixed member proportional system applied in Germany, New Zealand and Venezuela, and to the eompensatory mixed member proportional system used in Italy and Hungary.
The question arises what the funetion is of eonstitueney elements within PR systems. Single-member distriets -as weil as very small multi-member distriets -have two impor tant effeets: Firstly, they produee a high degree of disproportionality between votes and seats. As is well-lmown, the vote-seat-proportionality or rather disproportionality depends to a high degree on the magnitude (size) of the eonstitueney (defined as the number of representatives eleeted in a eonstitueney). Nohlen puts it as fo lIows: "The smaller the size of the eleetoral destriet, the less the degree of proportionalit� of the eleetoral system. And generally, smaller parties have less chances to obtain seats" . Mathematical and empirieal data, among others, from Douglas Rae, TaageperalShugart and Arend Lijphart support this statement 10 . Seeondly, small eonstituencies, and especially single-member eonstituencies, are eonsidered to be a key element to ensure a dose relationship between voters and repre sentatives. They allow the voter to ehoose between individual party eandidates and tend to favour a relationship based on knowledge of the candidate, on trust and on his or her aeeountability. Candidates who represent single-member eonstitueneies or at least very small multi-member eonstituencies are usually doser to the eleetorate they serve. For this reason eonstitueney-based systems are regarded as high on aeeountability.
Put in rather simplified terms, in proportional representation systems the seeond effeet (that is the dose relationship between voters and representatives) is a desired one, the first effeet, however, (that is the high degree of disproportionality between votes and seats) is an undesirable one. Certainly, a PR system eannot be eomposed only of single-member or very small multi-member eonstituencies; otherwise it would not eorrespond with the principle of proportional representation, but with the principle of majority representation. Therefore, it is neeessary to eompensate the disproportional effeet of these eonstituencies (as is the ease within eompensatory mixed-member PR systems) or even to prevent them from influencing the vote-seat-relationship (as is the ease within the mixed member proportional systems). To summarise: From the functional point of view, the "strange" combination of constitu ency elements within PR systems serves to secure a close voter-representative relationship and, at the same time, to guarantee a fair (proportional) representation of parties according to the votes they receive.
It has to be pointed out, however, that from a systematic point of view the voter-represen tative relation can be influenced not only via the magnitude of the constituencies, but also via the form of the party list. "Contrary to many assumptions, the personality of the candi date can play a larger role in party lists than in individual candidacies, given that there are different forms of lists and some are open to personal votes,, 11 . In the case of single member districts -as weil as in muIti-member districts in conjunction with closed party lists -provision is made only for inter-party competition (that is, between candidates of different parties). In the case of open or free lists, however, intra-party competition is also possible (that is, among candidates of the same party). Open lists allow the voters to decide which candidate should represent the party in Parliament. The ranking order of candidates on the party list could be altered by the voters. The free party list even allows the voter to cross party lines. These non-closed lists aim at greater participation of voters, giving them the opportunity to elect not only the parties, but also the candidates of their choice.
However, non-closed lists may have harmful side-effects as can be seen from the experi ence in Italy and Japan, prior to recent electoral reforms, and also in Panama and Peru. In these countries, open party lists do not only increase significantly intra-party competition, but also favour political clientelism, corruption and even electoral manipulation. In coun tries like Panama and Peru, where democracy, political parties and the party system are still in a stage of development, it is a problematic phenomenon that open lists weaken the parties by favouring individual candidates and intra-party factionalism l 2 . Furthermore, it should be noted that non-closed lists are not easily handled by the electorate and the elec tion administration. In Peru as much as 41 % of the votes in the parliamentary election in 1995 were ruled invalid compared to 9% to the presidential election. This is mainly due to the voters' lack of understandi�f and the electoral staff s maladministration of the voting ballots and electoral documents . Aiming at an improvement of the voter-representative relation, both options, that is intro dueing constituency elements on the one hand and non-closed party lists on the other, may be considered in electoral reform debates. A good example of this is the reform debate in Venezuela. There, a PR list system in multi-member constitueneies with additional seats at the national level had been applied since 1946. The electoral system was criticized particu larly because of the closed party lists wh ich fo stered the widely lamented anonymity between voters and representatives l4 . In order to solve this problem two basic alternatives were considered: introdueing open party lists or creating a certain number of single member constitueneies, without affecting the prineiple of proportional representation. Finally the second alternative was agreed on, not least because single-member-constituen eies have less of a weakening effect on the party organisation as open party lists. Today Venezuela applies a personalised PR system (mixed-member proportional system) in a form similar to the German and New Zealand electoral system.
Also in other countries where the prineiple of proportional representation is widely accepted, the introduction of non-closed lists or a certain number of single-member constituencies or at least of very small multi-member constituencies has been discussed out of functional considerations, without electoral reform being envisaged at this stage. This applies, for example, to South Africa: The electoral reform debate in South Africa Traditionally, the plurality system in single-member constitueneies had been used in South Africa since its establishment in 1910 for the election of members of the House of Assem bly and later also the Houses of Representatives and Delegates in the Tricameral Parliament under the 1983 Constitution l 5 . Under the interim Constitution of 1993 1 6 a pure PR system was adopted after a short, but intensive electoral reform debate 1 7 . The adoption of this system was motivated by various factors, inter alia the fact that there was not time enough The Constitution-writing process The 490 member Constitutional Assembly (consisting of all the members of the National Assembly and Senate) was assigned the task of drafting and adopting with a two-thirds majority a "final" Constitution within two years of the first sitting of the National Assembly representation". Constitutional principle XIV in turn provided that provision should be made for participation of the minority political parties in the legislative process in a manner consistent with democracy while principle XVII stated that at each level of government there must be democratic representation.
For purposes of deciding on the provisions to be included in the final Constitution regard ing an electoral system, an Inhouse Workshop was held by Theme Committee Two in Parliament on 8 May 1995. Papers at the workshop were presented by academics from the University of South Africa, Rand Afrikaans University, the University of Stellenbosch, the University of the Western Cape and the University of Heidelberg (Germany).

The eleetoral system applied in the 1994 eleetions
Before analyzing the different reform proposals presented at the workshop, we should have a look at the electoral system used in the 1994 elections. Among the main technical features of the South African electoral system, the following points should be noted: For purposes of the 1994 e1ection the country was divided into ni ne multi-member constituencies of variable sizes. They corresponded to the ni ne provinces established in terms of the interim Constitution. However, the primary basis for the allocation of the 400 seats was the votes cast for each party in the country as a whole. The 400 seats were filled from national and/or regional party lists. Parties could decide whether to submit only regional lists or to submit both regional lists and a national list, provided that lists of candidates could not in total contain the names of more than 400 candidates 23 . The list form was closed. The parties, therefore, decided on the ranking order of the candidates on their lists. Each voter had one vote for the e1ection of the National Assembly.
In order to allocate the seats, the procedure was as folIows: The National Party, for example, chose not to submit a national list of candidates, but only nine regional lists.
In the 1994 elections, these constituencies were identical to the nine provinces established in the Interim Constitution. The allocation was as folIows: to a given party at the regional level was subtracted from the total number of seats that the party obtained from the votes cast at the national level. The difference was filled by the national party list. If a particular party did not submit a national list, the difference was filled by the regional party lists (in the same proportions as set out in the first step).
Considering the proportionality between votes and seats, the South African electoral system secures a very fair representation of political parties in Parliament, given that the primary basis for the allocation of parliamentary seats are votes cast nationally for each party. The electoral outcome of the 1994 election was highly proportional by international standards.
And what is more, a broad consensus existed between almost all the political players that ("STV Droop"), and not the ordinary "LR Droop", being votes divided by (seats + 1). The difference between the two quotas, however, is of no practical consequence. 26 See fn. 24. We should be aware of the fact that the low fragmentation of the party system at the parliamentary level is not due to the electoral system. A comparison of the party systems' fractionalization at the level of votes and at the level of parliamentary seats indicates clearly the low extent to which the electoral system defractionalizes the party system by converting votes into seats. Moreover, the overwhelming parliamentary majority of the ANC is not produced by the electoral system. The seat share of the ANC corresponds almost exactly with its vote share. This leads one to the conclusion that the electoral system doesn't have a significant concentration effect. But, given the structure of the existing party system, this is not regarded as a problem at present. See Reynolds, A., The Case for Proportionality, in: Journal of Democracy, Vo1.6, 1995, No.4, 117-124. However, the lack of individual accountability of members of Parliament and the lack of "constituency representation" are widely regarded as defects of the electoral system. As   Apart from some technical differences in detail, these options are basically a variation of the personalised proportional representation or the mixed-member proportional system used in Germany and New Zealand, although without a legal threshold.

Proposals by academics
2) The second alternative provided for allocating one share of the seats in small multi member constituencies and the other share of the seats at the national level, either to national party lists (de Ville) or to the "best losers" in the constituencies (Kotze). Again, the national vote cast for each party should determine the final allocation of seats.
Jacques de Ville (1996) proposed a combination of small (2-5) multi-member constitu encies (using party lists) and a national party list for the election of members of the National Assembly. Under this system, voters would have two votes: one for a party list of constituency candidates and another for a national party list. The national party list vote would determine the final allocation of seats. 300 of the 400 members of the National Assembly would be elected in constituencies. The number of constituency candidates elected would be subtracted from the overall allocation of members of each party, wh ich would determine the number of national party list seats a party is entitled to. In the event of overhang seats a party would retain its excess seats but other parties would, where appropriate, be allocated additional seats to ensure full proportionality.
Hennie Kotze (1996) also proposed multi-member constituencies. He, however, suggested that constituencies should have either three or five members each and that voters should have only one vote which would be counted twice, once to determine the election of the 300 constituency representatives and a second time to establish the over all party representation in the National Assembly. The 100 "additional members" would not be allocated from party lists but from the "best losers" in the constituencies. He also favoured the introduction of the preferential vote in constituencies wh ich would enable voters to cast their vote for a particular constituency candidate. In the event of overhang seats a party would retain all its excess seats.
What is new in this alternative is that it does not advocate the implementation of single member constituencies, but for small multi-member constituencies. Thus not only the strongest parties, but also other parties have an opportunity to obtain constituency repre sentation which is of vital importance considering the dominant role the ANC is playing in most South African provinces. Systematically, this alternative finds itself positioned between the German system, which combines highly proportional outcomes with the personalisation of the vote through single-member constituencies, and the Danish system, which combines highly proportional outcomes with the personalisation of the vote through open party lists in predominantly large constituencies. De Ville and Kotze are proposing a highly proportional system, which provides for a personalisation of the vote by introducing small multi-member constituencies with c10sed (de Ville) or non-c1osed (Kotze) party lists.
From the systematic point of view, the South African electoral debate is highly interesting.
Not only because the reform proposals present some original ideas, but also because the debate shows that the functioning of an existing electoral system can lead a reform debate into a certain direction and it can considerably influence the attractiveness or non-attrac tiveness of possible alternative electoral systems. The logical step from the South African electoral system to the alternatives mentioned above is easy to make: It simply results from the idea of reducing the size of the large constituencies in South Africa existing at the provincial level and thus to change them into single-member constituencies or small multi member constituencies. This would be a small technical alteration with major politicaI effects.
It is worth pointing out that within the constitutional writing process academics widely agreed in their opinions on the existing electoral system as weil as on the basic direction of the presented reform proposals. Various South African academics seek to improve the present electoral system in order to rule out its shortcomings (in particular, the lack of constituency representation and of individual accountability of representatives) while simultaneously retaining its strong points (especially, the high degree of proportionality of sions of the majority party, the ANC, however, was perhaps the most cautious, noting only that "the implementation of a system of both proportional and/or constituency-based repre sentation needs to be carefully considered".
The submission of the DP, however, was rather concrete. It provided for 80% of the members of the National Assembly being elected in multi-member constituencies and 20% from national party lists. Each voter should have two votes: one vote for a constituency party list and one for a national party list, the second vote determining the number of seats a party is entitled to in the National Assembly. In the event of overhang seats, the number of members of the National Assemby should be increased slightly. ANC to be able to run a disciplined parliamentary party was also critical in their decision 36 .
This latter consideration is supported by the theory of electoral systems which shows that there is a greater independence of the elected representative from his or her party in single member constituencies, given that he/she is elected as an individual candidate.
Perhaps the strongest argument for maintaining the existing system was, however, the time factor. Within the rather short constitutional writing process the decision on the electoral system was a political decision taken under time-pressure. The politicans had not much time to take into account the academic debate and to seriously consider the pros and cons of electoral reform. In order not to precipitate the decision they maintained the existing elec toral system at least for the 1999 election. Obviously, the two major political parties also did not wish to incur the risk of undesired side effects of electoral reform. The fact, how ever, that even the ANC and the NP consider it necessary to improve the voter-representa tive-relationship can be seen in the introduction of an intra-party system of informal constituency representation -which admittedly hasn't fu nctioned properly up to now.
One will have to wait and see whether the electoral reform debate will start again before or after the 1999 elections. It should be stressed that at least an astonishing consensus exists on the academic and political level with regard to the evaluation of the strong and the weak points of the electoral system (namely high representativeness versus lack of constituency representation), the objectives of possible electoral reform (improvement of the voter-repre sentative relationship within the principle of proportional representation) and the spectrum of technical means in order to reach these objectives (introduction of constituency elements one way or another).

A Systematic View on the Electoral Reform Debate in South Africa
By Michael Krennerich and Jacques de Ville In the past few decades, the variety of both majority representation and proportional representation systems has been widened considerably. Some of the more complex electoral systems combine highly proportional outcomes with the personalisation of the vote through single-member constituencies or through open party lists. The issues sUITounding the relationship between voters and representatives and constituency representation play aIso a prominent role in the South African electoral reform debate, rekindled within the Constitution-writing process (1994)(1995)(1996). In South Africa's first democratic elections (1994), a pure proportional system was applied. On the academic as weil as on the political level, an astonishing consensus exists with regard to the evaluation of the strong and weaks points of the existing electoral system (namely high representativeness versus lack of constituency representation), the objectives of possible electoral reform (improvement of the voter-representative relationship within the principle of proportional representation) and the spectrum of technical means in order to reach these objectives (introduction of constituency elements). However, reform proposals which aimed at supporting the indivi dual accountability of members of Parliament and constituency representation without giving up the widely accepted princip1e of proportional representation have failed. The strongest argument for maintaining the pure PR system was the time factor. Within the rather short Constitution-writing process the decision on the electoral system was a political one under time-pressure. In order not to precipitate the decision and not to incur the risk of undesired side effects of electoral reform, the two major political parties maintained the existing electoral system at least for the 1999 election.