There are a number of complications and issues surrounding the application and use of the Single Transferable Vote that form the basis of discussions between its advocates and detractors.
Contents |
A frequent concern with STV among electorates considering its adoption is its relative complexity compared with plurality voting methods. For example, when the Canadian province of British Columbia held a referendum on adopting the BC-STV single transferable vote in 2005, according to polls most of the "no" voters gave their reason as "wasn't knowledgeable" when they were asked why, specifically, they voted against STV.[1]
However, as with all voting systems, once STV is understood there remain a number of areas of controversy surrounding its use. In particular, arguments for and against proportional representation in general are frequently referenced in debates among electorates considering STV, however the specific implications of a particular STV system can be examined as well. Most of the arguments for and against STV and proportional representation in general are based around the expected outcomes of the alternative proposed system and not on the system itself.
STV differs from all other proportional-representation systems in actual use in that candidates of one party can be elected on transfers from voters for other parties. Hence, the use of STV may reduce the role of political parties in the electoral process and corresponding partisanship in the resulting government. Unlike proportional representation systems employing party lists, voters in STV are not explicitly constrained by parties even when they do exist; voters may ignore candidate party labels and mix their preferred candidate rankings between parties. Similarly, candidates may achieve electoral success by obtaining a quota of voters not generally within their own party, perhaps by winning transfers from moderates or by championing a specific issue contrary to party doctrine. STV advocates argue that, by requiring a candidate to appeal to the supporters of other candidates for their second and further preferences, it reduces adversarial confrontation, and indeed, gives a substantial advantage to candidates who broaden their appeal by being not only collegial but as open-minded and flexible in their principles as they can manage. STV detractors see this as a flaw, arguing that political parties should be able to structure public debate, mobilize and engage the electorate, and develop policy alternatives.
Unlike List PR, STV can be used in elections in organizations without any political parties at all, such as in municipal non-partisan elections, trade unions, clubs, and schools.
Some STV variations, however, may encourage the role of political parties and actually strengthen them. In Australian Senate elections, where a combination of large districts, mandatory complete ballots, and compulsory voting results in the near 95% usage of partisan group voting tickets, political parties gain significant power in determining election results by adjusting the relative ordering of their tickets, both in terms of their own candidates and of transfers to other parties.
Successful campaign strategy in STV elections may differ significantly from other voting systems. In particular, individual candidates in STV have little incentive for negative campaign advertising, as reducing a particular opponent's ranking among voters does not necessarily elevate one's own; if negative campaigning is seen as distasteful by the voters, the practice may even prove harmful to the attacking candidate. Conversely, in order to avoid elimination in early counting rounds by having too few first place votes, candidates have a significant incentive to convince voters to rank them explicitly first as their top preference, rather than merely higher. This incentive to attain top preferences, in turn, may lead to a strategy of candidates placing greater importance on a core group of supporters. Avoiding early elimination, however, is usually not enough to win election, as a candidate must still subsequently win enough votes on transfers to meet the quota; consequently, strategies which sacrifice wide secondary support in favor of primary support amidst a core group may ultimately fail unless the group is particularly large.
There are also tactical considerations for political parties in the number of candidates they stand in an election where full ballots are not required. Standing too few candidates may result in all of them being elected in the early stages, and votes being transferred to candidates of other parties. Standing too many candidates might result in first-preference votes being spread too thinly amongst them, and consequently several potential winners with broad second-preference appeal may be eliminated before others are elected and their second-preference votes distributed. This effect is amplified when voters do not stick tightly to their preferred party's candidates; however, if voters vote for all candidates from a particular party before any other candidates and before stopping expressing preferences, then too many candidates is not an issue. In Malta, where voters tend to stick tightly to party preferences, parties frequently stand more candidates than there are seats to be elected. Similarly, in Australian Senate elections, voters also tend to vote along party lines due to the relative ease of selecting a party's declared preferences rather than individually casting their own complete list. In the Republic of Ireland, the main political parties usually give careful consideration as to how many candidates to put forward in various Dáil (parliamentary) constituencies. Transfers are often not along party lines, but rather go towards more prominent local personalities. Election posters for the most prominent candidate of a political party usually list preferred 2nd (and possibly 3rd) preferences for that party.
Electoral stasis is a condition which arises when an electorate cannot realistically change its political composition, regardless of the swing occurring in a general election.[2] A safe seat in a single member electoral system is one that is in electoral stasis. Proportional representation systems, such as STV ballots, can also be in electoral stasis. These “safe seats” can be ignored by political party strategists who will allocate resources to other electorates.
In an electorate returning two members, the quota, (33.34%) is too large to be affected by swings in an election. For example, the Territories in the Australian Senate have always returned one member from each the two major parties. In an electorate returning three members electoral stasis occurs when the weaker party is able to gain a quota (25.01%) The more popular party will then win two seats and the weaker party one seat.
The smaller the number of members to be elected, the more likely that the electorate will be in electoral stasis. The larger the number of members returned by an electorate the greater the chance that a voter who changes his or her vote can influence the outcome of the election.
As seen above, voters in an STV election rank candidates on a preferential ballot. STV systems in use in different countries vary both as to ballot design and to whether or not voters are obliged to provide a full list of preferences. In jurisdictions such as the Republic of Ireland and Northern Ireland voters are permitted to rank as many or as few candidates as they wish. Consequently voters sometimes, for example, rank only the candidates of a single party, or of their most preferred parties. A minority of voters, especially if they do not fully understand the system, may even 'bullet vote', only expressing a first preference. Allowing voters to rank only as many candidates as they wish grants them greater freedom but can also lead to some voters ranking so few candidates that their vote eventually becomes 'exhausted'–that is, at a certain point during the count it can no longer be transferred and therefore loses an opportunity to influence the result.
To prevent exhausted ballots, some STV systems instead oblige voters to give a complete ordering of all of the candidates in an election (if a voter does not rank all candidates their ballot may be considered spoilt). However, when there is a large set of candidates this requirement may prove burdensome and can lead to random voting, or "donkey voting" in which a voter who has no strong opinions about their lower preferences simply chooses them in the given order. Some jurisdictions compromise by setting a minimum number of preferences that must be filled for a ballot paper to be valid (for example Tasmania, which requires five preferences).
To facilitate a complete ballot, some STV systems provide the voter with the option of using group voting tickets rather than having to identify manually a complete list of individual preferences. For example, in elections to the Australian Senate a voter can either rank the candidates individually or place the number one in a box "above the line" to vote for a predetermined ordering of candidates drawn up by one of the political parties. This system diminishes the emphasis on individual candidates and increases the power of party leaders who submit the predetermined rankings; in practice it may even lead to a system resembling party-list proportional representation. However it is still possible for independent candidates to be elected as has been shown in the 2007 Federal Senate elections where Nick Xenophon was 3rd elected before preferences.[3]
The simplest way to list candidates on a ballot paper is alphabetically, though they may also be grouped by party. However, any fixed ordering will give some candidates an unfair advantage, because some voters, consciously or otherwise, are influenced in their ordering of candidates by the order found on the ballot paper. For example, studies conducted in the Republic of Ireland, where candidates are listed alphabetically, have shown that candidates whose surnames begin with an early letter in the alphabet enjoy a small electoral advantage over candidates with later letters. To solve this problem some systems involve a random ordering of candidates, or an ordering that changes from one ballot paper to another (the latter is often called the Robson rotation, after Neil Robson, a Tasmanian MHA who championed such a system).
There is evidence that the STV system has given a small advantage to candidates whose surnames are near the top of the alphabet. In the Irish Dáil (lower house of parliament) elected in 2002, about 4.5% of members had surnames beginning with A, 8% beginning with B, and 12% beginning with C.[4] From listings in the Eircom telephone directories for the 01, 06 and 07/09 areas combined, the expected percentages would be about 1.5%, 7% and 9.5% respectively. Similar deviations from the norm have been noted after previous elections. The effect seems to be minimal, as the Dáil median surname currently falls within the letter K, which accurately reflects the distribution of Irish names. An analysis of likely voting patterns seems to predict a small bias of this kind. A voter votes 1 for the candidate they most favour. If they wish to support other candidates of the same party, but have no strong preference between them, the voter is likely to number them downwards from the top of the ballot sheet, in normal reading fashion. Later preferences to candidates of another party are likely to be numbered in the same way. If candidates' names are listed alphabetically, this pattern will mean that earlier preferences will go to candidates earlier in the alphabet. Over a series of elections, such a minor bias would have a cumulative effect, as elected members stand for election again. In Ireland, it has been suggested that names on ballot papers should be printed in random order to prevent this.
The outcome of voting under STV is proportional within a single election to the collective preference of voters, assuming voters have ranked their real preferences and vote along strict party lines. However, due to other voting mechanisms usually used in conjunction with STV, such as a district or constituency system, an election using STV may not guarantee proportionality across all districts put together. Differential turnout across districts, for example, may alter the impact of individual votes in different constituencies, and when combined with rounding error associated with a finite number of winners in each constituency the election as a whole may throw up anomalous results.
For example, the 1981 election in Malta resulted in the Labour Party winning a majority of seats despite the Nationalist Party winning 51% of the first-preference vote.[5] Controversy over the election ultimately resulted in a constitutional crisis, leading to an amendment adjusting the voting system to allow for the possibility of bonus seats and making the Maltese voting system more similar to an open-list PR system; STV alone would also have given the second most popular party a parliamentary majority in the 1987, 1996 and 2008 Maltese elections. This kind of difference due to rounding error can occur with any PR system used at a district level, although greater rounding error occurs with smaller districts and there is a tendency for STV elections to use smaller districts when compared with PR elections employing party lists.
Similarly to differences in voter turnout, instances of malapportionment across districts can also cause disproportionate results for the legislature as a whole. In STV elections to the Australian Senate, states with vastly different populations have the same number of seats, and so while the results for individual states are proportional, the nationwide result is not, giving greater voting power to individual voters in less populated states. This is because the senate's role is at act as representatives of their states rather than representatives of the people. Thus the number of Senators does not take into account the population of the state or territory and thus by nature it is not intended to be proportional across the nation. However the system is intended to be proportional at a state level as the make up of that states senators will be comparable to the percentage that they gain of the state vote. This system is intended to remove potential geographical discimination that can occur due to Australia's large size. If the system was proportional across the nation then a party could focus on the eastern states that have 77% of the population. Thus if allocated proportionately the eastern states would gain a disproportionate amount of political attention and thus government funding due to the concentration of voting power.
By contrast, the New South Wales Legislative Council avoids the use of districts entirely, electing all 21 members using a single, statewide constituency and guaranteeing results that are proportional to the final allocation of preferences. In Victoria, Australia, legislation is in place to ensure that each electorate contains the same number of constituents. Problems occur when the number of candidates to be elected is different in each electorate, as the percentage of voters required to elect a candidate varies, whilst the number of voters remain the same. The percentage breakdown varies considerably, adding to the argument that the number of voters and the percentage required for election should be the same across all electorates. Under Victorian legislation an electoral review is required should the number of electors in each electorate fall outside the nominated tolerance level. Municipal Electoral reviews are undertaken every second term of office.
STV differs from other PR systems in that it allows the voter to decide the issues that should matter for the proportionality. In most PR systems the voter can only influence a single aspect - party representation. STV allows the voter to choose other criteria that can be used to create the proportionality: gender, ethnicity, age, place of residence or character as examples. A perceived problem arises from the fact that voters and the political elite are often not in accord with what sort of proportionality should be achieved. Increased voter choice and control reduces the power and influence of the political elite to determine the political debate.
STV provides proportionality by transferring votes to minimise waste, and therefore also minimises the number of unrepresented/ disenfranchised voters. In this way STV provides Droop proportionality - an example STV election using the Droop quota method for 9 seats and with no exhausted preferences would guarantee representation to every distinct group of 10% of the voters, with at most only 10% of the vote being wasted as unneeded excess (in most cases it would be much less). Unlike other proportional representation methods employing party lists, voters in STV do not explicitly state their preferred political party (with the exception where above-the-line voting systems are in place); this in turn can create some difficulty when attempting to analyze how an STV election's results compare with the nationwide partisan makeup. One common method of estimating the party identification of voters is to assume their top-preference on their ballot represents a candidate from their preferred party, however this method of estimation is made more complicated by the possibility of independent candidates and of cross-party voting. However valid comparisons have and can still be made if sufficient data and information is available. In Victoria, Australia it is possible to make a direct comparison between the Australian Senate election and the Victorian Upper House elections although individual circumstance will always exist voting patterns have shown that most voters stick to their chosen party within a limited percentage range based on local issues and circumstances. The main advantage in Victoria's case is that both systems are similar in design with one being a subset of the other. Victoria held its first multi-member proportional representation election in November 2006 for the Legislative Council
Another issue commonly considered with STV elections is the size of the voting districts in terms of the number of candidates elected and, to a lesser extent, the total size of the body being elected.
Because STV is proportional, larger districts reduce the support a candidate requires to become elected as a percentage of the district. With 9 to be elected, for example, any who reach (with transfers) 10% electoral support will win a seat, whereas with 5 to be elected 16.7% is required. Some STV elections make use of districts with the number of seats available as small as three. A larger number of candidates elected also results in a smaller number of wasted votes on the final count. However, larger districts and the implicit larger number of candidates also increase the difficulty of giving meaningful rankings to all candidates from the perspective of the individual voter, and may result in increased numbers of exhausted ballots and reliance on party labels or group voting tickets.
While Ireland originally had a median district magnitude of five (range three to nine) in 1923 [6], successive governments lowered this. A parliamentary committee in 2010 discussed the "increasing trend towards the creation of three-seat constituencies in Ireland." They found "a larger district magnitude is shown to have positive effects in relation to the representation of women and minorities, as well as allowing for a plurality of ideas in terms of policy." They recommended not less than four-seaters, except where the geographic size of such a constituency would be disproportionately large.[7] When Northern Ireland adopted STV, they found five-seaters not sufficiently proportional and chose six-seaters. Tasmania had seven-seaters until the Green Party won the balance of power in 1996, after which the two large parties shrank the district magnitude to five and called an early election which cut the Green representation from four seats down to one.
Larger electoral districts can also significantly reduce the effects of gerrymandering; because gerrymandering relies on wasted votes to award the "last seat" in each district, proportional representation systems such as STV with larger multimember districts are intrinsically more difficult to gerrymander.[8] Larger districts can also make for significantly harder tactical voting: since the problem of making correct assumptions about other voter's behavior and rearranging one's tactical ballot is NP-hard, the difficulty of tactical voting increases sharply as the number of candidates grows.
There is no theoretical upper limit to the size of districts in STV, and they may not even be needed at all: Thomas Hare's original proposal was for a single, nationwide district. In theory, STV could allow for the election of particularly small minorities provided they secure a quota's worth of votes, if very large districts were used. No jurisdiction currently uses districts larger than nine for election to its main legislature, although the 1925 Irish Senate election used one district to elect nineteen places.[9]
A related question is the number of people in the district, and the possibility of an effective democratic campaign by a candidate in a large district where campaign costs are high. In Northern Ireland districts range from 96,000 people in six-seaters for the Northern Ireland Assembly, and even smaller districts municipally, to a single three-seater district of all 1,737,000 people for election to the European Parliament.
According to the Gibbard-Satterthwaite theorem tactical voting is possible in all non-dictatorial deterministic voting systems. A number of methods of tactical or strategic voting exist that can be used in elections conducted using STV. In general these methods are only effective in marginal districts and only affect the allocation of a single seat per district.
One potential strategy in STV involves casting a first preference vote for a candidate that has no chance of winning. This vote will then be transferred at full value and carry more weight in determining the winners at the later stages of the count. This strategy can be effective due to two features of some STV counting systems.
For example in an election in which 5 candidates are contesting 3 seats the following 400 votes are cast:
The quota is 100 and A is elected at stage 1. E is the first candidate to be eliminated. Under SGT the 50 E>A>C>D votes ‘skip’ candidate A and transfer to candidate C; who is then elected. These votes then transfer again to candidate D who wins the final seat. The SGT winners are A, C and D.
Under WIGT the 50 E>A>C>D votes again 'skip' candidate A and transfer to C who wins the second seat. All C's votes (both the E>A>C>D votes and the C>B votes) then transfer at reduced value. As a result of this transfer candidate B wins the third seat. The WIGT winners are A, C and B.
The Meek system, which is immune to this strategy, elects A, B and D.
Under SGT the E>A>C>D voters effectively determine the winners of the two final seats. Under WIGT as a result of the E>A>C>D votes failing to contribute to the election of candidate A and transferring at an increased value, C wins a seat as opposed to D. SGT systems are more vulnerable to this form of tactical voting than WIGT systems.
While this appears an effective strategy in theory there are a number of problems with it in practice. Firstly accurate information is needed about how other voters are going to vote, in practice this information is difficult to obtain. Secondly if all voters vote for candidates they believe will lose, these candidates will win. In general, this strategy will not work if everyone uses it. Little evidence exists of the use of this strategy in the real world.[10]
Two vote management systems exist in STV. Vote equalisation works for all systems except those that allow group voting (above the line voting). The other method, party designated transfer, only works for systems that allow group voting. A party employing a vote equalisation strategy would attempt to ensure that all its candidates obtain an equal number of first preference votes, in the hope that transfers from candidates of other parties eliminated early in the count will lead to the election of an increased number of its candidates.
For example in an election for 3 seats, 2 parties (A and C) present two candidates and a third party B presents a single candidate. All party A voters prefer party B to party C. All party C voters prefer party B to party A. All party B voters prefer party C to party A. The following 1000 votes are cast:
The quota is 250. C2 is eliminated first and candidates A1, B and C1 are elected.
If party C distributes its vote equally between its two candidates for example:
B is eliminated first and candidates A1, C1 and C2 are elected. By equalizing the distribution of its vote party C wins an additional seat.
Evidence from Ireland[11] indicates that this form of vote management has been attempted on a number of occasions. To work, vote equalization requires accurate information about voting intentions and party strengths and also the active cooperation of the voters themselves. If a party misjudges the situation a vote equalization strategy can go badly wrong. For example in the 2003 elections to the Northern Ireland Assembly, a vote management strategy by Sinn Féin in West Belfast intended to replace an Social Democratic and Labour Party member with a Sinn Féin member led instead to the election of a member of the Democratic Unionist Party, a party extremely hostile to Sinn Féin.[12]
The party designated transfer system of vote management is applicable only to systems that allow group voting. Under group voting a voter places a '1' next to his or her first choice of party list. His or her vote then transfers down the party list in an order designated by the party and then to the lists of other parties in an order predetermined by the voter's first choice party. The published lists that show the designated transfers can be long and complicated and can effectively disguise who an ‘above the line’ vote is likely to end electing. An example of the use of this strategy is provided by the 2004 Australian Senate election in Tasmania. The Liberal and Labor parties and an independent Labor candidate designated that their above the line vote should transfer to the Family First party in preference to the Green party. Had all voters in Tasmania cast ‘above the line' votes the Green party candidate, who obtained almost a quota of first preference votes, would have lost to a candidate of the little supported Family First party.[13]
Academic analysis of voting systems such as STV generally centers on the voting system criteria that they pass. No preference voting system satisfies all the criteria described in Arrow's impossibility theorem: in particular, STV fails to achieve independence of irrelevant alternatives (like most other vote-based ordering systems) as well as monotonicity. Failure to satisfy independence of irrelevant alternatives makes STV slightly prone to strategic nomination, albeit less so than with plurality methods where the spoiler effect is more pronounced and predictable. Non-monotonicity, in turn, makes it possible under some circumstances to elect a preferred candidate by reducing his position on some of the ballots; by helping elect a candidate who displaces the preferred candidate's main rival, a voter may cause the preferred candidate to profit from transfers resulting from the rival's defeat. STV fails the participation criterion which can result in a more favorable outcome to an STV voter by not voting at all. However, a voter who truncates a candidate off the ballot does not harm a ranked candidate, nor is another truncated candidate helped on the ballot.
STV is also susceptible to the Alabama paradox: a candidate elected in an n seat constituency may or may not be elected in the same constituency with n + 1 seats even when voters express exactly the same preferences. This is due to the use of quotas; list PR by a largest remainder method is similarly affected, though a highest averages method is not. Intuitively, a candidate who was elected largely because of transfers from two similar groups (neither obtaining a quota) may not be elected when the number of winning candidates increases, as both groups would instead get their preferred candidates elected (with the new, smaller quota) rather than automatically compromising on their mutual second choice as their votes transfer.
Some modifications to STV have been proposed in order to pass monotonicity and other criteria. The most common method of proposed modification to STV is to alter the order in which candidates are eliminated: theoretically, a candidate who ranked second on every ballot could be the first candidate eliminated even if he is a Condorcet winner. Meek noted this problem in proposing his variation of transferring votes to nearly eliminate tactical voting in STV, however Meek himself did not propose a method for satisfying the Condorcet criterion. Other theorists have proposed further refinements of STV, such as using a Condorcet method to rank candidates for elimination order. Some of these modifications alter STV in a way such that it no longer reduces to instant-runoff voting when applied to a single seat but instead reduces to some other single winner system, such as a Condorcet method.
Eliminating losing candidates in reverse order of their Borda count, STV-B,[14] affects some issues of STV. Borda counts include deeper preferences, and so contain more information than is held in current first place votes. Borda counts would capture the popularity of that Condorcet winner described in the preceding paragraph, preventing immediate elimination. Elimination by Borda counts creates other features for STV including promotion of moderate candidates and increased stability in the face of small changes of preferences.
When compared with other voting methods, the question of how to fill vacancies which occur under STV can be difficult given the way that results depend upon transfers from multiple candidates. There are several possible ways of selecting a replacement:
The countback method is used in the Australian Capital Territory, Tasmania, Victoria, Malta, and Cambridge, Massachusetts. Casual vacancies are filled re-examining the ballot papers data from the previous election. The candidate who held the seat is eliminated, and a new election result is obtained by transferring votes that were originally allocated to the elected candidate at the time of his or her election to unsuccessful candidates. This process maintains the proportionality of the representation according to the voters' choice without the need and associated costs of holding fresh elections.
Although the countback method is designed to select a replacement representing the same group of voters who elected the original candidate, it remains possible that no similar candidates remain on the ballot. In 1985 the Tasmanian parliament amended the electoral act to allow true by-elections if no candidates of the same party as the outgoing representative remained on the ballot; in this circumstance the party may request that a by-election be held, however this has not yet happened as most political parties and community groups now nominate a surplus number of candidates in order to fill any casual vacancies.
Countback methods vary by whether or not wasted and exhausted ballots are additionally used during the countback. The effect this has on the result of the countback depends on the differences in the next preferences of voters. Moreover, in STV systems that use exhausted ballots during countbacks, it becomes theoretically possible that the order of multiple resignations will affect who the ultimate replacements are - this is a consequence of the order of election as votes allocated to elected candidates are locked that value of which are no longer used in the determination of the election of further vacancies. Additionally, if ballots are allowed to be exhausted in the election, then by either method it remains possible that the chosen replacement will only meet a fraction of a quota of voters; when this fraction is particularly small, and therefore no similar candidates remain on the ballot, election rules may call for a different method of filling the vacancy to be used. The method is the calculation of a candidates surplus also contributes with a paper based formula (see above) complicating the countback process. A value based system being more accurate and less complicated. With the aid of computer based technology the results of a count-back can be determined immediately.
Another option is to have a head official or remaining members of the elected body appoint a new member to fulfill the vacancy. In Australia, for example, the state legislatures appoint replacements members to the Australian Senate, now done at the suggestion of the party of the outgoing senator. Before this rule, disputes over Senate vacancies contributed to the Australian constitutional crisis of 1975, ultimately resulting in a 1977 amendment to the Constitution of Australia to provide that the legislature must elect a member of the same party as the outgoing senator. Vacancies in the New South Wales Legislative Council are filled in a similar way by a joint sitting of both the legislative council and assembly.
In the Republic of Ireland and Northern Ireland, vacancies on local authorities are filled by co-option of a candidate nominated by the departed councillor's party colleagues, although in Northern Ireland this requires unanimous approval from the Council.[15]
A third alternative to fulfill a vacancy is to hold a single-winner by-election (effectively instant-runoff unless there is more than one vacancy); this allows each party to choose a new candidate and all voters to participate. This often leads to a different party winning the seat (usually one of the largest parties, since the quota is large). In the Republic of Ireland, by-elections are used for the Dáil and for the Seanad where voters in the Dáil constituency or Seanad panel vote to fill the vacancy or vacancies.[16] This removes the balance of proportionality of the original election. The costs of holding a by-election in a multi-member constituency is considerable. In a recent review of the City of Melbourne, Australia, electoral requirements the cost of holding a City wide ballot to fill a casual vacancy was estimated to cost over 1 million Australian Dollars.
Another alternative is to have the candidates themselves create an ordered list of successors before leaving their seat. In the European Parliament, a departing Republic of Ireland or Northern Ireland member is replaced with the top eligible name from a replacement list submitted by the candidate at the time of the original election.[17]
This system is also used for the Northern Ireland Assembly.