This is the first of three Briefs and sets out two key aspects of the rules of the game. The second brief will discuss geography and a baseline estimate of outcomes, based on the pattern of party support in the 2014 national elections. The third will project, under specific assumptions, how outcomes will change if party support changes.
The geographical basis of the system
The 2014 national elections demarcated 22 263 voting districts, each with a polling station. The 2016 local government election will use this demarcation and will aggregate voting districts into local government wards. The number of voting districts making up a ward will depend on population density. In densely populated areas, wards will contain few voting districts. In sparsely populated areas, wards will contain a greater number of voting districts. The rule is that the population in any particular ward should not vary by more than 15% either way from the average population per ward in the relevant municipality.
In addition, Schedule 1 of the Local Government Municipal Structures Act, 1998 (the Act) sets out additional criteria for ward delimitation. Two important requirements are that ward demarcation should not fragment communities and that ward boundaries are clearly identifiable.
Ward demarcation is the responsibility of the Municipal Demarcation Board (MDB). The MDB has endeavoured to maintain stable ward boundaries as far as possible, but the number of wards in a municipality usually changes between elections, partly as a result of changes in the number of registered voters. Indeed, municipal boundaries themselves change between elections.
As an input to this process, the Minister responsible for local government determines a formula for the number of councilors for each municipality, in terms of Section 20(3) of the Act. The MECs responsible for local government in the nine provinces use this formula to determine the number of councilors for each municipality in terms of Section 18 of the Act. Under certain conditions, the MECs may deviate from the formula. The Minister’s determination is published in the Government Gazette and the Section 18 notices are published in the Provincial Gazettes.
The MDB follows a lengthy consultative process before ward boundaries are finalized. It provides draft maps to stakeholders and communities, and convenes meetings with them with a view to reaching agreement about the boundaries. According to the MDB’s published schedule, ward information was published in Provincial Gazettes in April. The MDB is now considering objections and finalizing ward boundaries which are to be handed over to the Independent Electoral Commission (IEC) in August.
The allocation of seats following an election
If the total number of seats determined for a municipality is an even number, the ward councillors have half the seats and the PR (proportional representation) councillors have the other half. If the number is odd, then there is one more ward councillor than PR councilors. The term ‘PR councillor’ is slightly misleading, because it is not the PR councilors who reflect proportionally the party support in the election. It is all councillors.
An example will make this clearer, and the IEC’s own sample calculation (available on its website) can be used. Suppose a municipality has 50 000 registered voters and 21 seats in total, implying that there are eleven wards. Suppose that 42 000 valid votes are cast and, of these, 35 000 indicate a party affiliation, the remainder going to independents. Suppose there are five parties contesting the elections. The election outcomes are as follows:
Wards won  Votes for parties  
Party A  1 
5 300 
Party B  2 
7 700 
Party C  2 
8 900 
Party D  1 
8 100 
Party E  1 
5 000 
Independents  4 

Total  11 
35 000 
Then a quota is defined by taking the number of votes cast for parties and dividing it by the number of seats (21) minus the number of independents (4). In this case the quota is 35 000 divided by 17 or 2 059. In turn, the votes for each party are divided by the quota. This in general will give a whole number and a fraction. For instance, Party A’s quotient is 2.57. What is then done is a first round allocation of seats disregarding the information following the decimal point. This first round allocation will not exhaust the seats available, so a second round allocation of a seat apiece as given to the parties who have the highest numbers after the decimal point, working from the top until all the available seats have been filled. The point is that this process determines the total number of seats available to parties. The number of ward seats is subtracted from the total number of seats to determine the seats allocated to PR councilors. Once the independents have won wards seats, all remaining seats, whether ward or PR, are distributed proportional to the number of votes.
Two things follow. First, provided that the number of independents are known, the allocation of all the remaining seats will depend only on the party votes cast. This makes projection of party representation in metros easy to calculate.
Secondly, the conundrum in national elections has been regarded as the tradeoff between proportional representation and constituency accountability. The present national system comes down on the side of proportional representation and it follows that the loyalty of MPs is entirely to parties and not to constituencies. The artificial ‘constituency system’ that we currently have does not change this fact. But the local government system indicates that one can have both. Suppose the country was divided into 200 constituencies. Then you could have 200 constituency MPs and 200 PR MPs chosen in the same way as in local government elections, to obtain proportional representation as a whole. One would then have 200 MPs whose fate depended on keeping their constituencies happy, creating a stronger bond between parliament and the electorate.
Charles Simkins
Senior Researcher
charles@hsf.org.za