Jump to ContentJump to Main Navigation
Electoral Systems and Democratization in Southern Africa$

Andrew Reynolds

Print publication date: 1999

Print ISBN-13: 9780198295105

Published to Oxford Scholarship Online: November 2003

DOI: 10.1093/0198295103.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 24 February 2017

Crafting Districts for Re‐runs

Crafting Districts for Re‐runs

Electoral Systems and Democratization in Southern Africa
Oxford University Press

Re‐running the Malawian, Zambian, and Zimbabwean plurality elections under alternative electoral system formulas is more straightforward than in the South African and Namibian cases because the single‐member constituency results give an authoritative data basis for assessing the alternative vote in single‐member districts and different forms of list PR.

Malawi. The 177 constituencies used to elect members of Malawi's parliament were clustered into 25 administrative districts, which provide excellent multi‐member districts for the alternative vote MMD and provincial list PR re‐runs. These MMDs range in size from four members to eleven. I have split the capital of Lilongwe into three constituencies of six, six, and five members. For the purposes of the alternative vote preference methods I have assumed that supporters of the UDF would be more likely to transfer their votes to AFORD rather than the MCP and likewise AFORD voters would choose a UDF candidate as their second preference. As noted in Chapter 3, it is quite clear that Malawians were split into two camps in the May elections: either for Banda and the traditional ruling party, or against it and for the opposition and multi‐party democracy. This second group was then subdivided (along regional lines) between the UDF and AFORD. Indeed, in the referendum on one‐party rule in June of 1993 the UDF and AFORD joined together (although not always harmoniously) to campaign for a ‘multiparty democracy’ vote. The comparison between the results of the 1993 referendum and 1994 general election clearly illustrates how Malawian politics evolved from ‘the opposition versus the MCP’ in 1993, into ‘the UDF and AFORD versus the MCP’ in 1994.

For the national and provincial list PR re‐runs, results have been calculated using the same Droop quota as used in South Africa. The national list PR result takes the entire country (all 177 constituencies) as one constituency (à la Israel, Namibia, and the Netherlands) and the provincial results are based on the three regions acting as multi‐member constituencies.

Zambia. The design of Zambia's 150 single‐member districts may have been flawed in a number of respects—as noted in Chapter 3 the largest constituencies had nine times as many voters as the smallest—but they were also drawn from nine provinces in which the seat apportionment was more faithful to population and voter distribution. These nine provinces (ranging in size from the Copperbelt with twenty‐two seats to the North‐Western Province with twelve seats) therefore provide an adequate basis for designing provincial list PR districts and smaller multi‐member districts for the alternative vote. In the the first elections of 1991 huge majorities built up by the United National (p.277)

Table A.1. A Comparison of the 1993 Referendum and 1994 General Election in Malawi


National (%)


One‐party (MCP)



Multiparty (UDF and AFORD)






General election—1994










Independence Party (UNIP) in the Eastern Province and the Movement for Multi‐Party Democracy (MMD) in the rest of the country meant that voters' second preferences were irrelevant under either system. Not one of the 150 constituencies was won by a plurality rather than a majority and if the alternative vote had been used in larger multi‐member districts the absolute majorities would have been inflated even further. The only provinces not swept by a single party were in the Lusaka, Northern, and North‐Western Provinces, but here UNIP's 14 to 30 per cent of the vote was not enough to take any reasonably sized multi‐member district. In 1996 twenty‐eight seats were won on a plurality of the vote and in these cases I have made the assumption that on balance opposition party (ZDC, NP, NLP, AZ) votes would be more likely to transfer to the candidate of another opposition party rather than the candidate of the MMD. Where the transfers of independent voters could have altered the final result I have left the SMD plurality winner intact as it is too speculative to guess where such transfers might have gone. While many of these independent candidates were hostile to the government, in most cases they had represented the MMD until very recently. As in 1991, if the electoral system had been the alternative vote in multi‐member districts the MMD would have swept the board, the opposition NP being the only other successful party, picking up a six‐member seat in the North‐West Province. As in Malawi the national list PR results have been calculated under the Droop largest remainder method from the aggregated votes for all 150 constituencies, while the provincial results are based on the nine provinces acting as multi‐member constituencies.

Zimbabwe. In 1985 ballot results for the eighty common voters' roll single‐member district seats were reported by eight districts (although these districts were slightly different from the previous electoral districts used for the PR election of 1980). Only one seat, Chipinge in Manicaland, was not won with an absolute majority but here Ndabaningi Sithole of ZANU‐N would have needed only two transfers out of the PF‐ZAPU candidate's 838 votes to transform that plurality victory into an alternative vote absolute majority victory—I assume (p.278) Sithole would have achieved this. Similarly, there is little need to speculate about second preferences for the AV‐MMD re‐running. Both the AV‐MMD and provincial list PR re‐runnings have used the real‐life eight reporting districts as a basis of analysis. For the elections of 1990 and 1995, the number of elected members of parliament increased from 80 to 120, but the boundaries of the eight districts remained largely the same. In 1990 there were three seats won with a plurality which could have changed hands under the AV‐SMD re‐run. Within these three seats I have made the assumption that voters of the opposition ZUM, ZANU‐N, or UANC parties would be more likely to transfer their votes to another ‘opposition’ party than to the candidate of the governing ZANU (PF). In 1995 ZANU PF dominance meant that not a single seat was won with less than an absolute majority. As in previous re‐runs, AV‐MMD and provincial list PR results have been calculated on the basis of the eight regions outlined in the Appendix.

In the 1980 ‘white roll’ elections for twenty single‐member seats the Rhodesian Front (RF) of Ian Smith won every seat, with fourteen of them unopposed. This renders a re‐running of the results unnecessary because no alternative political party was available to take seats under a PR system, and none of the six independent candidates gained enough votes to stop the RF candidate gaining an absolute majority (and thus allowing for the possibility that AV‐SMD might have made a difference). But the 1985 entry of the Independent Zimbabwe Group (IZG) into competitive white politics (see Chapter 5) means that re‐running this election under AV‐SMD and national list PR does prove informative. Three seats were won with a plurality (all won by the renamed Rhodesian Front, now the Conservative Alliance of Zimbabwe (CAZ)) and here I have assessed where the third placed independent candidates' votes were more likely to go. The PR results are calculated on the basis of one twenty‐member national constituency.

For the 1980 common roll election held under PR rules in eight multimember districts, I have relied upon the district‐level breakdown of the results and the evidence of the geographic concentration of party support. The hegemonic dominance of ZANU (PF) in Shona‐speaking areas and PF‐ZAPU in Matabeleland meant that there were no plurality SMD seats which would have been affected by AV‐SMD. For the plurality SMD re‐run in the Midlands district (the only somewhat competitive region of the country) I have assumed that ZANU (PF)'s 60 per cent of the vote, and PF‐ZAPU's 30 per cent, was again concentrated in the distinct Shona‐ and Ndebele‐speaking areas of the Midlands. This would have led to eight seats being won by Mugabe's ZANU and four seats by Nkomo's ZAPU.

South Africa. For the PR election case studies of South Africa and Namibia crafting districts is a little more complicated as now large areas need to be reduced down to smaller geographically contiguous single‐member constituencies. However, it is possible to use previous administrative boundaries and detailed counting district data to define reasonable SMDs which are of roughly equal size and have geographic integrity. In South Africa, by taking the district‐level voting results supplied by the IEC, and applying them to the 361 magisterial districts illustrated in Voting for a New South Africa,1 (p.279) I created 400 single‐member constituencies which adhere to the following guidelines: Constituencies are:

  1. (i) contiguous and based on current administrative boundaries;

  2. (ii) of roughly equal size, where the average number of voters is 48,834;

  3. (iii) based within the nine provinces, each province having as many single‐member districts as was apportioned under the real electoral system, April 1994.

In some cases the voting data were given for areas which were entitled to more than one single‐member district. In the majority of these cases the counting district covered only two or three SMDs and here party vote totals were merely divided by the number of districts, assuming that party voting within a geographical area of 80,000–150,000 electors would be more consistently dispersed than not. However, in the cases of larger districts, e.g. Johannesburg (25 seats), Pretoria (13), Randburg (5), Roodepoort (4), Germiston (3) in the PWV area; Durban (11), Pietermaritzburg (7), Pinetown (5) in KwaZulu‐Natal; Wynberg (7), Mitchell's Plain (6) in the Western Cape; and Bloemfontein (5) in the Orange Free State, seats have been awarded in proportion to party vote across the larger counting district. This method was used because these urban conurbations are composed of homogeneous pockets, constituted along the lines of ethnicity, wealth, and party identification, within the larger heterogeneous administrative area. For example, Pretoria, which reported 655,385 valid votes, has ‘white areas’ interspersed with ‘black townships’, and very few mixed areas. Forty‐eight per cent of the total poll went for the ANC with the NP gaining 37 per cent. We know that nationally the ANC won very few white votes and the NP very few black ones,2 and so we can then extrapolate that the NP won the ‘white districts’ of Pretoria (three‐sevenths of the total) with the ANC taking the other four‐sevenths which were predominantly ‘black districts’. The reasoning used to decide how constituencies would have been won within larger counting districts reflects the fact that the SMD plurality system provided relatively proportional results in the National Assembly overall. In this sense Pretoria was merely a microcosm of the whole country, which demonstrated high degrees of geographically concentrated party support. The detailed constituency breakdown cited here can indeed be disputed and alternative constituency boundaries could be drawn. But it is my contention that due to the patterns of party support any redrawing of these constituencies, which remained true to the logical guidelines I have outlined previously, would produce highly similar overall results.

For the majoritarian AV‐SMD simulation I have used the same plurality SMDs as previously outlined and then assessed whether the seat would have changed hands if the alternative vote had been in operation. Three hundred and eighty‐three (96 per cent) of the seats were won with an absolute majority (over 50 per cent plus one) and therefore the plurality winner would have also been the AV winner. In total only 17 seats (4 per cent) were won with a plurality (less than 50 per cent of the vote), leaving open the possibility of a second placed party overcoming the plurality leader with the vote transfers of other unsuccessful parties. In assessing the likely second preferences of (p.280) losing party supporters I have used polling evidence from the eight Launching Democracy Institute for Multi‐Party Democracy reports from November 1993–February 1994. But in most cases the likely AV outcome in these 17 constituencies was easy to predict (i.e. the ANC had 49 per cent of the vote with the NP on 35 per cent and the PAC 4 per cent, or the NP led with 45 per cent and could rely upon receiving a substantial share of a FF/DP transfer totalling up to 10 per cent).

The South African counting district data enable us to run an even more accurate AV‐MMD simulation than for plurality because the larger districts do not need to be broken down into smaller SMDs. I have maintained the constituency guidelines outlined earlier and created what I believe to be logical, contiguous districts that remain within provincial boundaries. Similarly, where a MMD was won with a plurality rather than a majority I have used the same data and rationale as cited for AV‐SMD to assess if the second placed party could have leapfrogged the plurality winner and thereby won the constituency. One further assumption is needed here, as AV‐MMD requires the voter to vote for a number of candidates. Voters have the choice of voting for all the candidates of a single party or they may split their votes between the candidates of two or more parties. My assumption is that, on balance, South African voters would have been more likely to vote ‘straight party ticket’ rather than splitting their votes between a number of different party candidates. This again goes back to my premiss that there was particularly high voter party identification in these first democratic elections.

As noted earlier, the actual results of the South African National Assembly election were calculated on the basis of a 400‐seat constituency by the Droop quota and largest remainder method. The one slight list PR adaptation I have shown results for in Table 7.10 is for constituency‐based list PR without any provision for national or compensatory seats. This I have titled ‘provincial PR’ because party seats won are calculated by province and then combined to give each party's national total. I assume each province receives an equivalent proportion of the seats to its actual entitlement in the 1994 elections.3

Namibia. The same base methodology previously outlined for re‐running the South African elections was utilized for Namibia. In 1989 the country was split into 23 electoral districts of varying size: these areas then formed the basis of drawing 72 single‐member districts which averaged 9,317 votes cast. Through a process of splitting some EDs and combining others, constituencies were crafted which retained compactness and contiguity and did not deviate too greatly from the mean size (in the smallest district (Hereroland North) 7,802 votes were cast, in the largest (Kaokoland) 12,794 were cast). As in the South African case some EDs reported data covering many hypothesized constituencies. For the largest ED, Ovamboland, containing 26 districts, this was not a problem as SWAPO's 92 per cent of the vote made it clear that they would sweep every SMD in the region, but for some of the smaller multi‐member EDs a more nuanced analysis and projection was needed. For example, in the Kavango ED which consisted of seven single‐member districts SWAPO won 50 per cent of the popular vote and the DTA won 40 per cent. The clear evidence outlined in Chapter 4 suggests that SWAPO won urban (p.281) areas within this, and other regions while the DTA won the more rural outlying areas. Therefore, and in line with evidence from actual plurality SMD elections in Africa, I have awarded the Kavango seats proportionally: four went to SWAPO while three went to the DTA.

Thirteen of the 72 seats were won by pluralities and thus could have switched hands under AV‐MMD. For this simulation I assumed that the ACN and FCN supporters would have been more likely to transfer their votes to the DTA than to SWAPO, while the UDF votes would have been roughly split between the two main contenders. The AV‐MMD re‐runnings were based on the twenty‐three electoral districts being consolidated into ten multi‐member districts ranging from three members in size (Swakopmund and Caprivi) to twenty‐six members (Ovamboland).4 As in the South African AV‐MMD simulation the assumption of ‘straight party ticket’ voting was made. The ‘provincial list PR’ calculations were also based on these ten districts.

The re‐running of the 1994 elections was made considerably easier by the 23 old South‐West African electoral districts being replaced by 13 new regional districts, in turn broken down into a total of 95 subdistricts (councilmanic districts). These were the constituencies used for the ‘first‐past the‐post’ local government Namibian elections of 1992. Of particular use to this exercise was the breaking down of the huge Ovamboland district into four smaller areas (Ohangwena, Omusati, Oshana, and Oshikoto). Polling data reported for the 95 subdistricts then gives us an excellent basis for crafting 72 parliamentary SMDs. The average simulated district reported 6,800 votes with the largest being Sesfontein, Khorixas, Kamanjab, and Outjo with 9,708 votes and the smallest Soweto, with 4,200 votes. Six constituencies were won with a plurality (three SWAPO, two DTA, one UDF) and in these second preferences were calculated along the same lines as the 1989 exercise although in this election the ACN had been renamed the Monitor Action Group (MAG). AV‐MMD and provincial list PR results were calculated around the new thirteen multi‐member districts and as in 1989 the national list PR results were re‐run using a Droop largest remainder quota as opposed to the Hare method actually used.


(1.) Reynolds 1993a: 79–103.

(2.) See Reynolds 1994 and Mattes and Gouws 1998.

(3.) e.g. each province would receive the following number of seats: Western Cape (42), Eastern Cape (56), Northern Cape (8), KwaZulu‐Natal (80), Orange (p.308) Free State (30), North West (30), Northern Transvaal (40), Eastern Transvaal (28), Pretoria‐Witwatersrand‐Vereeniging (86).

(4.) In practice Ovamboland would have been split up into smaller MMDs for the alternative vote, but SWAPO's electoral dominance in that region makes such a detailed analysis irrelevant. With 92% of the vote they would have taken every seat no matter how the boundaries were delimited.