Parliament may correct the voting system of its own accord
(Sponsored content)
Shortly after South Africa’s 2024 elections a multi-seat party indicated that they requested clarification on a matter. Time lends perspective. It may be worthwhile revisiting the broader topic, as not all is well with the voting system, which is but one piece of legislation having passed through our law-making processes. With Parliament there may be this predicament in that whatever the objectionable, it may still be the prerogative. Whatever the wrong, though, it is to be righted, and that will be the focus here: on the culminations of things undesirable and on improved redress – ideally set in motion from within.
Analysis of 2024 election data
For the 2024 elections the Independent Electoral Commission reports 31,917,889 valid votes cast in aggregate across two ballots – after excluding 19,304 votes for independent candidates, as prescribed for the next step. It puts the Droop quota at 79,596 and slightly down from 79,644 had votes for independents been included. Both the inclusive and exclusive figures fall within the range of 79,368 to 79,866 that all produce the same interim allocation-by-quota outcome for 2024. Incidentally, towards its upper end, the range also includes 79,795 (rounded) calculated as the Hare/natural quota, again slightly down from 79,843 if all-inclusive. All four quotas (Hare or Droop; inclusive or exclusive) would have left sixteen residual seats with which to attempt/achieve/perfect proportionality. A choice was made and the lowest quota is prescribed.
A contestant may, or may not, qualify for a residual allocation. Most fell short of a quota seat, failed to qualify for a residual seat and exited the contest, with 581,716 (party only) votes cast in their favour contributing almost half of the sixteen residual seats. The three one-seat parties in Parliament also fell short of the quota, but received the first three allocations of the ‘largest remainder’ method, as did one two-seat party that claimed the fifth/final allocation of this method on top of the quota seat that it started out with. Two other two-seat parties (numbers 7 and 8 in line if not for the cut-off at five) claimed their second seats via the subsequent method, as did two more two-seat parties who benefitted from the change, for they would have been back in line (behind sixteen and eighteen) at 31 and 49.
At the high end, the largest remainder method ranks the biggest party at 51 for the Hare-inclusive quota, or 36th for Hare-exclusive, or 16th for Droop-inclusive, or, as the quota shrinks, at 11 for the prescribed Droop-exclusive. The method that it gives way to is not similarly affected by quota changes, cuts all that short, and ranks the party 18th overall.
The dualistic method is tailored by the stipulations of the Act. The proportional part tempers the effect of a change of method (suggesting purpose). The non-proportional part introduces a threshold (possibly its purpose). For 2024 it reduced the number of qualifying parties by five (representing votes worth 2.55 quotas). The result is more seats for fewer parties; exclusivity over inclusivity; deviation from proportionality. The number of residual seats will have to be prevented from outnumbering parties vying for them, though, as this may cross a red line. A valid quota takes every party to within one seat (if still short) of proportionality. Adding a second deviates from proportionality. Slack of two seats (16 seats contested for by 18 parties) seems to protect the 2024 results from a possible challenge of this kind.
A: It may be best to discard the troublesome method or to substitute it for a method befitting a proportional system. If it stays, then the dualistic method should disallow a second residual seat from being allocated to any contestant and introduce a remedial clause (e.g. to run the largest remainder method beyond its cut-off point, when needed).
Court case CCT 144/23 does dip into proportionality, but a detour via deliberations on another matter offers a helpful analogy and context: Overhang is recorded [e.g. para 69 of the judgement] as occurring ‘when a party wins more regional seats than (the number that) it is overall entitled to’. The Court concluded that, should overhang arise, it does so for two reasons – the second being the ‘not insurmountable’ absence of the (fixing) formula [para 91] that avoids breaching the limit on the number of seats set by the Constitution (and given as the first reason).
The absence of a formula may follow from an absence of the risk of occurrence [para 92], of which the Court is convinced [para 83]. Overhang will have implications: No enacted overhang-fixing formula can remedy the fundamental issue associated with it, which is the bypassing of the proportionality correction; overhang breaches two limitations if ever it is to occur, one fixable by a formula on occurrence, the other avoidable by addressing its risk. Risks revealed to the Court invariably involves an allocation that exceeds its entitlement.
B: The entitlement to seats is at the heart of matters. It is readily calculable upfront. It may be regarded as allotted by an unofficial election (across two ballots conducted in one all-encompassing voting district), which allocates quota seats and residual seats into a tally, as a provisional result i.e. the entitlement per contestant. It then relinquishes a number of regional seats while compensatory seats remain (in deficit when overhang occurs). Positioning its calculation between the regional and compensatory processes [para 32] goes with the omission of data. Non-proportionality has no role here as compensatory seats correct towards the entitlement, which is where proportionality resides if it is achieved.
Searching for proportionality
While important for overhang, the entitlement to seats functions primarily as a provisional allocation [para 32]. Parliament observes that parties can hold numbers of seats proportional to numbers of votes [para 64]. Yet the provisional allocation – the entitlement per contestant – is determined in part by a non-proportional method. It brings into question the stated purpose of compensatory seats [para 63]. The non-proportional method is brought into the fold with the infusion of independent candidates and it adds to differentiation professed as legitimate ‘for the express purpose of achieving proportional representation’ [para 62]: to then have the method feature where independent candidates do not, seems disillusioning.
That what happens to be ‘put before the court’ matters. With respect to proportionality we have the applicant and a report in support of the applicants accepting that the split they oppose is proportional, and we have the respondents accepting that the split they defend against is proportional [para 66]; there is acknowledgement by the applicant of the implications for achieving proportionality, as put forward by respondents. This seemingly all-round agreement is also apparent when arguing direct access to the Court [paras 45 to 48]. A report filed on behalf of a respondent is interpreted to conclude that the contested split achieves proportionality – which may not hold true for 2024 results data.
C: Proportionality is an objective inquiry. For each contestant a pair of ratios exists: votes cast in its favour to total votes cast, and number of seats secured to total seats available. It is the only data needed for the exercise. It includes independent candidates. After correctly combined where parties field candidates (not applicable for votes cast for a party directly), the difference between each seat-ratio and vote-ratio is a surplus/deficit that may be squared into a positive value measuring deviance per contestant, to be added up across all contestants (parties and independents alike) into a value for variance, with which to assess the ‘best fit’ i.e. the valid allocation with the lowest possible variance.
Valid allocations may be randomly generated, repeatedly, while calculating variance at each iteration and if lower, to unseat the allocation still standing on having less variance than predecessors – up to the proportional one.
The sequence of diminishing contribution to variance reduction (SDCVR) starts with any seed value: blank allocations are fine, as is any possible interim quota allocation, or an election result. Rank contestants on their (squared) deviations: start with the highest and verify its origin as a surplus/deficit to appropriately subtract or add one seat for the contestant, subject to validity (e.g. independents may be allocated at most one seat – skip/avoid adding more and move along). If the new variance value is lower, execute the adjustment, update the ranking, and repeat. SDCVR characteristically stops when proportionality is reached (which is controllable to let it stop where required to).
With a favoured seed, SDCVR reaches proportionality at its eleventh allocation still short five, which are obliged in sequence of least disruptive to base variance to be comparable with the 2024 results: While the allocation line-ups differ, the pure largest remainder method (with a quota too low) allocates a single mismatch (11th instead of 20th in line) compared to the extended SDCVR sixteen. The dualistic method’s first seven allocations and three others are agreeable, but there are six mismatches (#8 #9 and #12 #13 #14 #15) not found in SDCVR as extended. On its own the non-proportional method rehashes these six and adds three more mismatches, one of which as a second residual seat (which is troublesome).
Correcting system deficiencies
D: A number of possible electoral systems (might) comply with (constitutional) requirements [para 68] – the choice of words matter as proof of proportionality lies in the result. Section 46(1)(d) reads: ‘system that… results… in’.
D1: A quota that is too low violates proportionality. Droop is the lower limit for an election quota and any recalculation thereof that results in its value reducing, sends it into forbidden territory. Deciding the entitlement only after the regional allocation process, effectively recalculates the (Droop-exclusive) quota for ‘inter-party purposes’ relative to an (absent Droop-inclusive) upfront calculation (point B above), which is the appropriate positioning – and reach – for achieving proportionality.
D2: The dualistic method violates proportionality to the extent that the known non-proportional method comes into play. The recorded conclusion [para 78] of deviation not exceeding one seat (ideally cut back to tolerance short of a seat), may stem from the model used, or assumptions made, or lengths of residual seat sequences it generated. SDCVR may corroborate that the dualistic method (or amended electoral formula) is restrictive of proportionality and by several mismatches more restrictive than the largest remainder method (former electoral formula) – as dependent on residual seat numbers (sixteen for 2024, eleven beyond the first five).
‘At the heart of this case lies the question of how Parliament chooses to give effect to (constitutional) requirements’ [para 5]. The ‘give effect to’ part may not have been scrutinized sufficiently: Neither the (all) inclusive determination of the entitlement (point D1) nor the replacement of the troublesome method (point D2) should have to wait for court orders to be enacted.
The number of seats available is set at an upper limit, but [contrary to affidavit_70.3(c)] having them all filled is not a priority stipulation. Within the scope that the lower limit allows, their number may be decided by proportionality, for which the judgement [para 90] offers another analogy.
Unlike Hare, the Droop quota disregards a quota worth of votes at every voting district it is applied to because of the ‘+1’ in the denominator. Expressed as a percentage, the relative reduction (1–Droop/Hare), is small at 0.25% for the entitlement involving 400+1 seats, higher at 2.08% for the largest voting district (GP; 47+1) with 81,984 as quota for 2024, and 16.7% for the smallest (NC; 5+1) with its 65,880 quota for 2024. At the low end, single seat districts (1+1) can expect a 50% reduction. Choose Hare, the quota in use for local elections, with the largest remainder method: combined they are efficient at proportional allocation, especially where universally determined and applied to all regions.
Droop’s original premise is that counting may stop once a target tally lower than Hare is reached, which may have been practical in times gone by, or appreciable in settings taken a dim view of today [para 156]. Parliament may accept guidance from a Panel that they trust not to cast them in a bad light, and review an argument for Droop as submitted. Close the chapter on quotas being high/low or reducing (or moderating/countering) that may have been pursued for the wrong reasons.
Towards the core of the issue
Parliament is constrained by safeguards aimed at ensuring appropriate protection for citizens who desire foundational rights [para 53 extract]: Safeguard One is compliance with the threshold of rationality [para 52]; it asks whether enough (if not 100%) is achieved of what is targeted. The section 36 limitation of rights is Safeguard Two, which may just as well assume that the rationality test is passed, for it asks (amongst others) whether too much is achieved, infringing elsewhere.
Of the contention that six sections are violated [para 94], some challenges are framed unexpectedly. One claims that the contended split is arbitrary/irrational therefore it violates section 1(c): it seems self-defeating when the split passes the rationality test with Safeguard One nullifying the call on Safeguard Two. The challenge fails abruptly [para 97], ruled to be ‘without merit’. Another claims that the split differentiates arbitrarily between contestants [para 100], seemingly pushing Safeguard One to cripple Safeguard Two.
E: According to the Democratic Party (DP) judgement, its parallel New National Party (NNP) judgement deals ‘with the constitutional framework relevant for the consideration of the issues’ [DP_9] and ‘concludes that the impugned provisions… do not constitute limitations on the rights relied upon by applicants in both cases’: Under the heading ‘Rationality of the statutory provisions’ the NNP rationality test covers twelve paragraphs. Under ‘Denial of the right to vote’, its section 36 inquiry covers eleven more. Findings are penned as the investigations progress, before concluding as if in summary on both [NNP_48]; regrettably susceptible of being read out of context, as discernible from an extract [para 110] where added emphasis covers findings on rationality, followed by findings on the section 36 inquiry (that avoid emphasis).
E1: Consider reading the essence of the summary as follows: (a) section 9 rights are allegedly breached, (b) however, from the foregoing, and (c) despite differentiation between categories, (d) the rationality test is passed for Safeguard One, (e) while no limitation is established for Safeguard Two. Consider interpreting the text receiving added emphasis as follows: differentiation is no proof of irrationality.
E2: Part E1(b) of the reworded summary above, suggests that anything read into the wording should be able to find support in earlier deliberations. The Court, ‘not relieved of the obligation to test the rationality of (some) provisions’ challenged on other grounds [NNP_31], goes ahead and conduct the test, which may add to the perceived connection between parts E1(c) and E1(d) of the reworded summary. The NNP judgement confirms both safeguards [NNP_19; 20]; it engages Safeguard Two when investigating alleged violations of the section 9 right to equality – the challenge fails for reasons summarised in part E1(e) based on earlier findings [NNP_41; 44; 47].
E3: Parliament submits [para 101] that the Court made it clear that – for (certain sections, including) section 9’s right to equality – passing the Safeguard One rationality test serves to end proceedings: somehow nullifying the ensuing Safeguard Two inquiries that (presumably) are to be blocked, ignored, or ruled in the negative. In reality, part E1(d) above may not have nullified the ensuing part E1(e) back in 1999.
E4: In 2023 the Court seems to embrace what it supposedly clarified when it states: ‘in keeping with what (is) said in (NNP), as there is a proven rational connection… it cannot be maintained that the split violates section 9(1)’. By way of NNP precedent, this assertion is bearable subject to the section 36 inquiry and conditional on its negative outcome. The section 9(1) inquiry ends abruptly [para 116] like section 1(c)’s did, ruled to be ‘without merit’ – possibly from self-defeating framing – then still seemingly reliant solely on the rationality test. The Court may be revered for jealously guarding its obligations. After an ensuing inquiry, the section 3(2)(a) challenge addressed concurrently [para 115], is ruled upon separately [para 127] and in the negative.
The rational basis is the threshold safeguard. Like the section 36 inquiries were formalised into the Walters test [para 139] a few years after the NNP judgement, it may have become pressing to develop the scrutiny granted and required by section 36, lest inadequate legislation prone to litigation may be encouraged.
The New Nation II (NNII) judgement ‘dispels the notion that proportional representation is consonant only with representation through political parties’ [NNII_80]. Respondents hold the opinion [affidavit_70.11] that no system (supposedly also not SDCVR) can perfectly align with the imperatives. They avoid arguing how close a system may come to alignment (decidedly closer absent seat reservation). They do argue: firstly, that proportionality tends to be distorted by independent candidates [para 64]; secondly, that proportionality in a two-tier system (in effect) requires seat reservation for parties [para 65]; and thirdly, that the concept of inter-party proportionality (the subject of a report, and nuance hopefully not lost on the Court [para 66]) may stand-in for what is required [affidavit_70.3(b)]. The onus may rest on respondents to proffer justification [NNII_119] once a limitation of a right might be established.
Shortcomings in the applicant’s case regarding limitations of political rights had the Court consider how it has dealt with section 19 challenges in five other cases [paras 138 to 152], to address the content and scope of the rights alleged to be infringed and the effect of the impugned provisions on them.
An inter-relatedness central to the NNII case is not mentioned – it was not raised/challenged and respondents would not have addressed it in their arguments. ‘It can never be that the envisaged legislation could not be subject to constitutional curbs’ [NNII_75]: The potentially unjustified limitation of section 18 freedom of association in the voting process remains not scrutinized and awaits undergoing the Walters test.
Questioning constitutionality
Voters know that their votes translate into seats: The vote is to elect representatives. Free choice ranks higher than system peculiarities. Compounded information improves the outcome: when sourced from an additional vote (that translates into seats), it must occur via an all-inclusive ballot paper.
F: An issue remains. The Court states [in the context of para 125] that ‘the Constitution guarantees citizens a right to make free political choice’, which may hold true generally: If both votes translate into seats and a voter finds a favourite on one of the two ballot papers, but not on the other, may the voter opt for ‘free ballot paper choice’ and peacefully exchange it for the all-inclusive one where the candidate-of-choice appears, especially with it not undermining the system – but regardless if it does? Does the flexibility exist? Is there a better way of ensuring free political choice?
An initial design required votes to be cast on one all-inclusive ballot paper, customised per voting district. It offered voters free choice. Respondents contemplated the prospects of a case against it (on other grounds) and made late adjustments to the Bill that changed the design [para 26]. Two ballots are prescribed, but only one vote is cast on an all-inclusive ballot paper – not the other. For all the balancing going into the design [affidavit_70.11], free choice apparently had to give. Voters’ free choice would still be intact had all-inclusive ballot papers been carried forward into the current design – or when reintroduced as a remedy.
For a different type of remedy, hone in on the ‘influencing’ of inter-party proportionality [affidavit PA7_38] achievable by inter-party seat allocation with no net effect on their collective sum total. All-inclusive regional ballots may decide all 400 available seats, as final allocations for independent candidates, and as a provisional allocation for parties combined – a collective party entitlement, per region, towards which party support from the regional ballots count – before being allocated inter-party proportionally utilising aggregate data across two ballots: fill all seats from regional party lists for widespread geographic representation; or alternatively, split each regional section of the provisional allocation into two portions for the party allocations. Duly influenced ‘inter-party proportionality’ is achievable without the (hard-coded) reservation of seats. ‘Any scheme (that) is not sufficiently flexible… has the potential of infringing the right’ [NNP_23].
G: Parliament may know where reservation of seats originates from, if not from their initial design. Reservation of seats for political parties characterises the current system. It evaded challenge [para 34].
G1: Whether an alleged limitation of voters’ free choice has purpose intrinsically is to be verified (if only in anticipation of the second stage Walters test inquiry). Would Parliament knowingly infringe on a right? If deemed justifiable it is perceivable that they might: achieving proportionality in the current design [para 68] may be argued as justification for limiting freedom of choice (available through the all-inclusive ballot paper of the initial design). One constitutional requirement is then pitted against the other [affidavit_70.11], with proportionality arguably at risk – supposedly supported by assessments – should both ballot paper types become all-inclusive.
G2: Parliament may choose to argue that seat reservation [para 65] is the means to the ‘end’ of achieving proportionality – while pointing at the proven rational relation [para 83]. Section 36(1) lists another item for consideration: ‘less restrictive means to achieve the purpose’ that (Parliament may agree) is by definition also rationally related. Where such means may exist (as explored above), they constitute an improvement in that the two constitutional requirements are both met (simultaneously and harmoniously), absent the wedge seemingly driven by seat reservation (that is not a requirement).
A borrowed analogy may apply: The tools exist already. Broken down, all the components are at hand. The voting systems only have to be assembled correctly as per the guidelines laid down. It may be advisable to remove from the build anything destructive/corrupting/devious – however well intended (and without a doubt when).
Members of Parliament in support of electoral reforms may be ideally positioned, also to inform voters on whether their right to free political choice extends to all-inclusive ballot papers. If the impetus has to come via the courts, then it is crucial to (assume rationality and) prove the limitation of a right: everything hinges on success at this first stage Walters test inquiry. Whoever contemplates a legal challenge has to independently assess its prospects of success – and then secure the result.
Upcoming 2026 local elections
H: The local elections share in the principles even with their requirements customised to a greater extent. One of the featured ballots is being used to accommodate independent candidates [affidavit_70.10] – but when a contestant is registered, the system likely comply with requirements only if the approved voter choice is accessible on both ballot paper types (which may have been overlooked). The entitlement-to-seats calculation must extend to cover all contestants: support for independent candidates must be proportionally corrected for, before their seats are restricted to (at most) one. Issues may become magnified as the demarcated voting districts for local elections may bring about fewer quota seat allocations and more residual ones.
It may be prudent to activate the Panel. Our voting systems should achieve proportionality also with all votes cast for independent candidates – and should not anticipate voting behaviour [para 125].
Many things may go wrong in ensuring compliance with the Constitution. Know that this may be the norm. Monitor the roles of advisors. Learn to appraise people and political options alike. Come to grips (once more, if you would) with why it is that we each do what we do. Do qualify for what lies ahead. Differentiate on what matters. Recover something astray.
Thomas R Labuschagne, 18/03/2025