Alternatives to the Present System

A “better” electoral system for South Africa is one that incorporates elements of both the present PR system and systems based on some form of district (i.e. constituency) representation. It must also be a system that is acceptable to the leaders of all major political parties and one that sustains high voter turnout at future elections.

Presented below are simulations of five alternatives to the present electoral system. These alternatives were computed by the application of the SDSS, and demonstrate the relevance of location-allocation models for electoral system design.

The various districts (i.e. constituencies) presented in these simulations were computed by the SDSS using population and electoral data for Gauteng Province for the 1999 election. All simulations assume that the province will continue to elect 92 members of the National Assembly.

Gauteng Province currently elects 46 members to the National Assembly via PR. Voters in the province also cast a second ballot for half the members selected on the basis of the national vote. This effectively means that residents of the province elect 92 members (two times 46) to the National Assembly. It is therefore possible to estimate the total number of seats won by each party in the province in the 1999 election as follows.

As indicated by the legend to the right of the map, the ANC won 63 seats in the province (32 on the basis of its provincial vote, and 31 on the basis of its national vote within the province), the DP won 17 seats, and so on. The Index of Disproportionality which measures the gap between the proportion of votes and the proportion of seats received by all parties is very low at 0.8, a typical outcome of PR systems.

To simulate a variety of district based electoral systems, we applied the SDSS to analyze population and electoral data for Gauteng Province as reported for 1,837 voting districts in the province for the 1999 election. The following is a map of these voting districts.

For the simulations that follow, the SDSS was applied to construct separate systems of constituencies consisting of 92, 46, 12, 18 and 23 districts for which the model was to optimize for the criteria of an equal number of voters per district while minimizing the loss of geographic compactness. The votes were then “counted” in each of these simulated districts, and seats allocated to each party on the basis of the formula used for each alternative.

The greatest departure from the present system would be a return to the single member district (SMD) system or “first past the post” used for most South African elections prior to 1994. With the exception of Namibia, all Anglophone countries in Africa and most Francophone countries elect their national legislatures from single member districts.

If Gauteng were divided into 92 single member districts of nearly equal population the province would elect its representatives from a series of relatively small and homogeneous geographic areas. Members of the National Assembly would presumably visit their constituencies on a regular basis or risk defeat at the next election. Communication between citizens and their representatives would be high or have the potential for such. As can be seen from the map, the varying size of constituencies reflect the population density across the province with the largest number of small districts located in the townships. One can easily identify the districts for SOWETO that are located slightly southwest of the center of the map.

In “first past the post” versions of SMD, the party or candidate that gains a plurality of the vote wins the seat. This tends to reward large parties and disadvantage small ones. Thus, while the ANC and DP respectively won 68.2 and 17.7 percent of the vote in Gauteng, they captured 81.5 and 18.5 percent of the seats. No other party obtains a seat under SMD and the Index of Disproportionality is in the upper-middle range of 10.4.

SMD systems, particularly those consisting of a large number of small homogeneous districts, also reflect the demographic base of each party that wins a seat. The simulation confirms the well-known fact that the ANC draws the bulk of its votes from African voters who reside in areas of high population density, while the DP draws the bulk of its votes from white voters who reside in suburban areas in and around Johannesburg and Pretoria.

Although this system would no doubt raise the level of accountability by individual members to the electorate, it violates the constitutional requirement that the electoral system results “in general, in proportional representation.”

In a parallel system half or some portion of the seats are elected from single member districts while the remaining half or portion of the seats are elected by PR. To simulate a parallel system, the SDSS was used to create 46 single member districts while the remaining 46 seats were allocated by PR.

A parallel system of 46 districts would create a series of larger constituencies than a system with only single member districts. One result is that whomever represents these areas will have more territory to cover to maintain contact with the electorate. Note also how the DP wins only 4 of the district seats though this is matched by 8 obtained via PR for a total of 12. The larger and more socially diverse the constituency, the less likely small parties will win seats in single member districts. Finally, the Index of Disproportionality is reduced only slightly to 9.4.

In a mixed member proportional system, half or some portion of the seats are elected from single member districts while the remaining half or portion are allocated to achieve overall proportionality for the system as a whole. In this simulation, the SDSS has created the same 46 single member districts as for the parallel system. However, the remaining 46 seats are allocated to achieve results that are the same as would have been obtained by PR alone.

The total number of seats for each party are thus the same as for the current PR system, and the Index of Disproportionality is also the same at 0.8 The mixed member proportional option is a viable alternative to the present system for two reasons. First, it would provide a sufficient number of districts to establish a relationship of accountability between members of the National Assembly and the electorate. Second, it would fulfill the constitutional requirement of allocating seats in a manner that results in proportional representation. In short, a mixed member proportional system incorporates the best of constituency based systems and PR.

A mixed member proportional system of the type simulated here has proved to be a viable system in the Federal Republic of Germany. Lesotho has recently adopted a mixed member proportional format for its next parliamentary election scheduled for 2002. Two-thirds or 80 seats of its National Assembly will be elected from single member districts while an additional 40 seats will be distributed across all parties to achieve proportional representation overall.

We now turn to three examples of multi-member district systems, because they provide another approach to increasing the accountability of legislators to the residents of specific geographic areas while achieving results that are consistent with proportional representation.

Our first example consists of 12 districts, each of which elects 7 or 8 members of the National Assembly. The seats within each district were allocated by using the same PR formula currently utilized in South Africa at the provincial level and for the country as a whole (see next section on methodology for details).`

Although the map indicates that the ANC would be the dominant party in all 12 districts, the DP, NP and Freedom Front would all have gained representation under this system, and the Index of Disproportionality would be reduced to 7.3. While these results are not as proportional as a system based solely on PR, they are much better than any of the previous alternatives with the exception of the mixed member proportional system.

If the number of districts were raised to 18, the number of representatives per district would be lowered to 5 or 6. As with the previous option, the ANC is the dominant party in all districts. The advantage of this option is that the average size of each district is somewhat smaller and the number of representatives the electorate must follow is less. The Index of Disportionality is relatively modest at 7.5, but is not reduced.

By increasing the number of multi-member districts to 23 with 3 to 5 members each, the DP becomes the dominant party in two areas, and all parties except the Freedom Front and PAC obtain seats. The Index of Disproportionality is lowered still further to 6.4, but whether this would qualify as a system that “results, in general, in proportional representation” would probably be a matter for the South Africa Constitutional Court.

Like the mixed member proportional system, a multi-member system based on a large number of districts creates a large number of relatively small districts to facilitate accountability of legislators to the electorate while achieving a low to modest score on the Index of Disproportionality. However, the mixed member proportional system is arguably “better” in respect to reducing the geographical area to be covered by each member as well as achieving near perfect proportionality in the final allocation of seats.

We conclude with a simulation of a mixed multi-member proportional system. In this alternative, the number of multi-member districts remains at 23, but only 2 to 4 seats are allocated to each district for a total of 69. The remaining 23 seats are then allocated to achieve proportional representation overall.

The 23 districts for a mixed multi-member proportional system are the same as for the multi-member system with 23 districts presented in the previous two maps. However, the results are the same as for a system based solely on PR or the mixed member proportional system with 46 single member districts presented above. Thus, the ANC would win 63 seats (56 district seats plus 7 of the 23 seats to be allocated to achieve proportional representation), the NNP 4 seats (none from districts, but 4 of the 23 PR seats), the DP 17 seats (13 from districts plus 4), the IFP 1, the Freedom Front 3 and the PAC 1. The three smallest parties obtain all of their seats from the 23 seats allocated by PR.

We present a final table to quickly compare the results of all the alternatives presented in this section. South Africans will have to decide which, if any of these options, is preferred over the present system. Other options for which we did not create simulations might also be considered. The main point of this presentation is that the SDSS is a useful analytic tool that can simulate any district based electoral system that one can imagine with the exception of those that employ transferable votes. By comparing the results of different systems and the trade-offs between them, political leaders in South Africa and elsewhere can use the SDSS to design a better electoral system or arrive at a renewed consensus to maintain the present one.