Overhang becomes apparent where proportional representation (PR) is expected but not reliably delivered. A supposedly proportional system fails when it allocates more seats to a contestant than the proportional entitlement permits. The seat count in excess of the entitlement is called overhang. It poses a particular risk for two-tier systems that allow for non-proportionality in tier-one on the supposition that any deviation could be recovered within tier-two – which raises the reliability issue.
Rewind the clock a few years back, and South Africa (ZA) was formally discussing overhang in case CCT 144/23. The respondents were better positioned (relative to the applicants) on the risk-of-overhang axis. This was made use of – framed as a (second) stated objective [Para 69], which (amongst separate findings against the applicants) weighed with the Court [Paras 82, 83]: The (national) election may not be able to be declared should overhang occur absent a remedy for it. Why not have a remedy at the ready? It ‘may well be a result of the fact that there is no risk of overhang on the 200/200 split’ [Para 92].
Germany (DE) was cited as an example [Paras 79, 80]. Voter support determines the proportional entitlement to seats (as per their rules, which may encourage a form of tactical voting). In their tier-one election, representatives for constituencies are directly elected and allocated on the first-past-the-post (FPTP) method, covering half of the statutory seats. These allocations stand. In the trouble-free scenario, all tier-one allocations remain within the proportional entitlement, in which case tier-two fills up each entitlement with the other half of statutory seats. However, excess allocations above parties’ entitlement do occur in tier-one (where proportionality is not prioritised). Overhang takes up seats meant for others, who (are entitled to it but) remain under allocated by the time tier-two’s allocations conclude. Therefore, additional PR seats are created and allocated so as not to deny any contestant their entitlement.
New Zealand (NZ), that is also cited as an example [Para 89], stops at this point. The DE Bundestag used to stop at this point too, with 34 seats added in 2021 for overhang. But since overhang contestants remain allocated at above their entitlement, additional corrections are necessary if proportionality is to be restored. Post a ruling in 2017, the Bundestag also started creating levelling PR seats, to scale up everyone’s ratio to match the largest overhang, amounting to 104 seats for 2021. Together, overhang and levelling added 138 seats to restore Bundestag proportionality for it to reach the unprecedented 736 total seats, whereas the statutory seats at the time was 598 seats, split 299/299.
A: Where overhang occurs, the proportionality stipulation requires that it is cured or corrected for, which is evidenced when the ratio of tier-one seats decreases (e.g. from half to less than half) while that of PR-type seats increases correspondingly (through seat creation and addition). The know-how for curing exists. Yet, prevention may still better cure: the only way to achieve proportionality without having to adjust the seat-type ratios, is by avoiding overhang altogether.
Not deeming the challenges insurmountable, the Court, by way of illustration [Para 90], indicates that room to manoeuvre may be created: if an additional 2 allocations are expected to cure overhang, then set the target for total ZA seats at 390 (down from 400) – presumably again split equally at 195 seats each (for tier-one and tier-two, down from 200 each). The increased total of 392 seats then falls within the allowed 350 to 400 range, as expected. The tier-one seat ratio that starts out with 50% (195/390), reduces to less than 50% (195/392), precisely what curing is expected to do (point A above), while tier-two increases to more than 50% (197/392).
The context within which the illustration of the Court is made, is that of a reprimand [Para 84] for a submission about a method utilised at the local municipal level (supposedly) to prevent overhang. It is unfortunate that this submission is not properly before the Court [Para 86] as the formula/rules at the heart thereof [Para 91] certainly recognises overhang, but prescribes allocations that do not produce the evidence associated with a cure for overhang (point A above) with the tier-one ratio seemingly unaffected. It may never have been a potential solution to the risk of overhang, also not at local municipal level. It may allow the election to be declared, but remains to be scrutinized.
While South Africa deliberated on overhang and related matters, in the same year of 2023, Germany underwent (contested) electoral reforms to mend its legislature’s tendency to inflate. Tier-one for the Bundestag was kept at the statutory 299. Tier-two increased to 331 PR seats with the total number of seats being set/fixed at 630. The rules governing seat allocations were substantially changed (approved by their Federal Constitutional Court in 2024). And for their following election in 2025, all 630 Bundestag seats were allocated proportionally (i.e. ignoring the deviation caused by the threshold that they apply) and without an incidence of overhang.
Regarding the risk of overhang, the Court asserts [Para 83] that Parliament’s second stated objective of the 200/200 spilt (to avoid the risk of overhang) is achieved. This may be true for the design before the Court, but may be conditional on some feature within that design not changing. Note that the same equal split is in force at local municipal level, which is a (comparable but) different design, where incidences of overhang may have occurred already (which is very different from virtually no risk). More may need to be made of the risk of overhang.
B: One notable characteristic stands out when contrasting the arrangements at the national level and at the local municipal level: Light partitioning (of electoral districts) is applied at national level, stopping at broad multiple-seat regions. Heavy partitioning exists at local municipal level that prescribes single seat (ward) constituencies. Partitioning is acknowledged for affecting system performance. Should heavy partitioning be considered at national level, e.g. to mimic the Bundestag, the risk of overhang for an unchanged 200/200 equal split is best reassessed.
B1: We used to have one dominant party and a pack of smaller ones in our not too distant past. In a heavily partitioned setup, a party that achieves 50%+ of the vote may easily be leading the pack in each of the single seat constituencies. The FPTP allocation method then grants it the (bulk of) tier-one seats (assuming fair homogeneity across constituencies): It gets its 50% mostly through tier-one, but may need PR seats for excess support above 50% (or for some heterogeneity causing it to lose certain wards). Should its support drop to 40% though, while the pack further fragments to remain lagging it, it may still be allocated about 50% by taking most if not all of tier-one. Overhang becomes inevitable, unless the size of tier-one is shrunk accordingly to about 40% to maintain the symmetry between leading-contestant share-of-vote and tier split.
B2: Instead of lowering tier-one in advance, it may remain at its equal 200/200 split with tier-two, but the rules may be amended: when overhang occurs, then in all affected constituencies the seats are not allocated (and not directly elected), but are instead converted into PR seats, allowing the election itself to change the ratio between tier-one and tier-two. This is an outcome recognised as a cure for overhang (point A above). It may resemble what the recent 2023 DE reforms led to. It draws attention to perceived differences between tier-one and tier-two seat types, if any. The method is effective and may be applied as a catch-all feature backing other attempts at curbing overhang.
B3: Overhang may (differently) be countered by lightening the partitioning a notch, e.g. by merging single seat constituencies pairwise into dual seat ones. The logic behind this is to count the expected support of the two biggest contestants. If their sum is at about 50% or above, it is likely that tier-one will allocate one seat each to the top two contestants in the merged wards. This spreading (is the objective and) splits tier-one allocations between two parties and reduces the risk of overhang, but adds a secondary component in that either of the two may still get over-allocated, exceeding its entitlement, and requiring an intervention like converting the affected seat into a PR seat.
B4: Another distinguishable approach is to venture into multiple votes. Several run-off choices may be needed for multiple eliminations to happen. Contestants are eliminated one by one from the least popular, with the secondary votes redistributed to those remaining in the race as the run-off votes dictate. For a single seat constituency, the bar should be set at 50% i.e. for achieving a majority (to restore the symmetry mentioned above). For a dual seat constituency, the combined tally of the two leading candidates should achieve that majority. The run-off method may only have to be activated in (otherwise) overhang constituencies. For the Bundestag, eliminations in tier-one should not affect the PR entitlement calculation, but in ZA it may.
B5: Nothing caps the bar at 50% though: it could theoretically be raised up to 100% with sufficient run-off options (that do not get exhausted). Raising the bar counters deviation from proportionality. As it is raised, above 50% and going, the symmetry starts favouring tier-one that may increase in size and become dominant over tier-two. The current 200/200 split at national level carries excess numbers of (dormant for their purpose) tier-two PR seats. The applicants in the case may have tried to accentuate this reality.
B6: Should raising the bar be pursued aggressively then tier-one may approach proportionality, and the need to keep tier-two (to correct for deviation) diminishes and disappears. Single Transferable Vote (STV) type systems are setting the bar at 100%. They are designed around direct election of individuals (for all seats), transferrable (run-off) voter instructions, and achieving proportionality off a single ballot in moderately partitioned multi-seat voting districts. The Lekota Bill was ZA’s attempt at it, likely ahead of its time.
B7: There may be good purpose in allowing run-off optionality for the PR seats of tier-two as well: It may be used there to determine the election’s proportional entitlement, right at the start, to ensure that every voter’s input is taken into account. a): In the absence of an existing threshold, the eliminations that start with the overall lowest ranked contestant, should stop short of the lowest ranked contestant that gained entitlement (either outright or) from previous eliminations. b): Where a traditional threshold is replaced, rules may stipulate that eliminations proceed further than that, with run-off votes appropriately re-allocated, up to the elimination of all parties that would have been eliminated by the traditional threshold. The improvement is that the voters whose favourite choices were eliminated, were allowed to decide where their votes go next. The effect is similar to the threshold that it replaces, but votes are not discarded by the system.
The respondents did not oppose the application for direct access [Para 46]. In considering the application, the Court mentions [Para 44] that granting direct access deprives parties of the right of appeal and deprives the Court of the benefit of other courts’ insights. The Court’s perspective [Para 89] that the NZ format resolves the intricacies with less flourish, may have been informed differently absent the urgency.
If the trend of fragmentation persists, then the 2026 local municipal elections are likely to reveal a growing tendency in overhang allocations, despite its built-in method/formula for recalculating the entitlement for the rest [Schedule 1 Item 16(4)(a)]. Such allocations may be contested, especially if it favours some more than others, or losses are not offset by gains. Whether contested or not, growing shortcomings in a system can’t be ignored. We now have the example of DE reforms that managed to resolve their seat overhang problem. The first port of call may be to have Parliament consider incorporating it.
Thomas R Labuschagne, 2026 11 03
