Instant-runoff voting
|
|
When the single transferable vote voting system is applied to a single-winner election it is sometimes called instant-runoff voting (IRV), as it is much like holding a series of runoff elections in which the lowest polling candidate is eliminated in each round until someone receives majority vote. IRV is often considered independently of multi-winner Single transferable vote (STV) because it is simpler and is a widely advocated electoral reform in the USA.
Instant-Runoff Voting was invented around 1870 by American architect William Robert Ware. Ware was not a mathematician, thus never subjected his election method to any rigorous analysis. He evidently based IRV on the single winner outcome of the Single Transferable Vote or STV developed in 1855 originally by Carl Andrae in Denmark. It was introduced into England in 1857 by the barrister Thomas Hare, where it earned public praise from John Stuart Mill, an English philosopher, member of parliament, and employee of the East India Company.
IRV is a form of preferential voting, also known as the Alternative Vote (AV), or the Hare System. It is sometimes also known as Ranked Choice Voting (RCV), a term useful for describing the voter's experience as well as the appearance of the ballot, however, this term could also describe Borda count as well.
IRV was first used in Australia (where it is known as preferential voting) by the self-governing colony of Queensland, in 1893. The system gradually spread to other Australian colonies (states after 1901) and has been used to elect the Australian House of Representatives since 1919. It is also used for the lower houses of most of Australia's state and territorial parliaments. IRV is also used to elect the President of Ireland, the Papua New Guinea National Parliament, and the Fijian House of Representatives. (See below for a more detailed list.)
| Contents |
How IRV works
Voting
There are two forms of "IRV", the "additive" and the "eliminative". Each voter ranks at least one candidate in order of preference. In most Australian elections, voters are required to rank all candidates. In other elections, votes may be "truncated", for example if the voter only ranks his first five choices.
Counting the votes
In the "additive" form, if no candidate gets 50% of the first choices the next choices are ADDED IN, & so on, until someone finally does. This is the less commonly mentioned, but also more exactly top dead center form.
In the "eliminative" form, first choices are tallied. If no candidate has the support of a majority of voters, the candidate with the least support is eliminated. A second round of counting takes place, with the votes of supporters of the eliminated candidate now counting for their second choice candidate. After a candidate is eliminated, he or she may not receive any more votes.
This process of counting and eliminating is repeated until one candidate has over half the votes. This is equivalent to continuing until there is only one candidate left. However it is possible, with voter truncation, for the process to continue until there is only one candidate left, who does not end up with more than half the votes.
An example
Imagine an election for the capital of Tennessee, a state in the United States that is over 500 miles (800 km) east-to-west, and only 110 miles (180 km) north-to-south. In this vote, the candidates for the capital are Memphis, Nashville, Chattanooga, and Knoxville. The population breakdown by metro area is as follows:
- Memphis: 826,330
- Nashville: 510,784
- Chattanooga: 285,536
- Knoxville: 335,749
If the voters cast their ballot based strictly on geographic proximity, the voters' sincere preferences might be as follows:
42% of voters (close to Memphis)
|
26% of voters (close to Nashville)
|
15% of voters (close to Chattanooga)
| 17% of voters (close to Knoxville)
|
| City | Round 1 | Round 2 | Round 3 |
|---|---|---|---|
| Memphis | 42 | 42 | 42 |
| Nashville | 26 | 26 | |
| Chattanooga | 15 | 0 | |
| Knoxville | 17 |
Chattanooga, having the smallest vote, is eliminated in the first round. All of the votes for Chattanooga have Knoxville as a second choice, so they are transferred to Knoxville. Nashville now has the smallest vote, so it is eliminated. The votes for Nashville have Chattanooga as a second choice, but as Chattanooga has been eliminated, they instead transfer to their third choice, Knoxville. Knoxville now has 58% of the vote, and it is the winner.
In a real election, of course, voters would show greater variation in the rankings they cast, which could influence the result.
Special cases of IRV eliminations
Instant Runoff Voting as an ideal does not explicitly define how to handle special cases such as ties and different rules can be considered. A good IRV election must define rules to handle these cases before the votes are cast. The reason why is that there are cases where one set of rules will select a winner different from another set of rules and the set of rules used may affect how the voters cast their ballots.
Especially when performing IRV counts on smaller elections, there can be frequent last-place ties that prevent clear bottom elimination.
Here are some approaches to consider, individually and combined. The first class of rules allows many candidates to be eliminated at the first count regardless of actual ties. These are practical rules before the first round that reward stronger candidates among the full set of competition. Such rules won't likely affect the winner but they will reduce the number of elimination rounds and thus the number of opportunities for ties to develop. A second class of rules consider actual ties that can't be avoided.
- Consider multicandidate elimination of weak candidates as the first step:
- CANDIDATE COUNT: Define a maximum number of candidates that can survive the first round.
- Example top-two
- VOTE MINIMUM: Define a minimum vote threshold (5 vote for example) and eliminate all weaker candidates together.
- Requires limitations for rule to apply
- PERCENT MINIMUM: Define a minimum percent vote threshold (5% for example) and eliminate all weaker candidates together.
- Again, requires limitations for application
- PERCENT RETENTION: Define a minimum percent of votes by top candidates to be retained.
- Example - retain the top set of candidates who combined control 50% of the vote
- CANDIDATE COUNT: Define a maximum number of candidates that can survive the first round.
- Tie-breaking rules:
- LOGIC: If the tied candidates combined have fewer votes than the next highest candidate, the entire tied set can be eliminated at once.
- Logically deterministic, but may not apply
- FIRST ROUND: Eliminate the candidate in the tie with the lowest number of votes at the earliest stage in the count at which the candidates in question had an unequal number of votes (in effect this means the candidate with the lowest number of first-preference votes - except in the unlikely circumstance that the same candidates are tied both in the round when one has to be eliminated and in the first round).
- Traditional rule; violates purity of one person, one vote principle
- ALL: Eliminate all tied candidates at once.
- Good for weak candidates (with less than 5% of votes), but can lead to strategic nominations, which cause IRV implementations using this method to not be spoiler proof
- RANDOM: Eliminate one randomly to break the tie.
- ORDER: If the order of the candidates on the ballot paper has been determined by lot, then ties can be eliminated by choosing say the top candidate.
- Random Voter Hierarchy (RVH) (http://alumnus.caltech.edu/~seppley/MAM%20procedure%20definition.htm): Randomly determine a strict ordering of the candidates and when selecting a candidate to eliminate, pick one based on this strict ordering.
- Similar to random elimination, but with many nice properties not found with random elimination
- LOGIC: If the tied candidates combined have fewer votes than the next highest candidate, the entire tied set can be eliminated at once.
Where IRV is used
The single-winner variant of STV is used in Australia for elections to the Federal House of Representatives, for the Legislative Assemblies ("lower houses") of all states and territories except Tasmania and the Australian Capital Territory, which use regional multi-member constituencies. It is also used for the Legislative Councils ("upper houses") of Tasmania and Victoria, although the latter will switch to the multi-member variant from 2006.
In the Pacific, IRV is used for the Fijian House of Representatives. Papua New Guinea has also decided to adopt it for future elections, starting in 2007. The Fijian system has been modified to allow for both "default preferences", specified by the political party or candidate, and "custom preferences", specified by the voter. Each political party or candidate ranks all other candidates according to its own preference; voters who are happy with that need only to vote for their own preferred candidate, whose preferences will automatically be transferred according to the ranking specified by the candidate. Voters who disagree with the ranking, however, may opt to rank the candidates according to their own preferences. In the last election, however, only about a tenth of all voters did so. The ballot paper is divided by a thick black line, with boxes above (for the default options) and below (for customized preferences).
The countries mentioned above all use STV for some or all of their municipal elections. Starting in 2004, some municipal areas in New Zealand also adopted STV to elect mayors (by the single-member variant) and councilors (by the multimember variant). Political parties, cooperatives and other private groups also use STV and/or IRV.
A voting method similar to IRV, known as [1] (http://mathforum.org/dmpow/solutions/solution.ehtml?puzzle=46)plurality vote with elimination, is used to select the winning bid of both the Summer and Winter Olympics in the International Olympic Committee.
See Table of voting systems by nation
Adoption in the United States
Suggested by a recent version of Robert's Rules of Order, instant-runoff voting is used in the United States for some non-governmental elections, including student elections at some major universities.
Notable supporters include Republican U.S. Senator John McCain, 2004 Democratic presidential primary election candidates Howard Dean and Dennis Kucinich, and consumer advocate Ralph Nader. The system is favored by many third parties, most notably the Green Party and the Libertarian Party, as a solution to the "spoiler" effect third-party sympathizers suffer from under plurality voting (i.e., voters are forced to vote tactically to defeat the candidate they most dislike, rather than for their own preferred candidate). In order to increase awareness of the voting method and to demonstrate it in a real-world situation, the Independence Party of Minnesota tested IRV by using it in a straw poll during the 2004 Minnesota caucuses (results favored John Edwards).
This issue rose to attention in the United States in the 2000 election. Supporters of Ralph Nader who nevertheless preferred Democrat Al Gore to Republican George W. Bush found themselves caught in a dilemma. They could vote for Nader, and risk Gore losing to Bush, or, they could vote for Gore, just to make sure that Bush is defeated. It has been argued that Bush won largely due to the "spoiler effect" of Nader supporters in Florida.
In March 2002, an initiative backed by the Center for Voting and Democracy passed by referendum making instant runoff voting the means of electing local candidates in San Francisco. It was first used in that city in Fall of 2004. (Note: The San Francisco Department of Elections (http://www.ci.sf.ca.us/site/election_page.asp?id=24269) prefers the term "Ranked Choice Voting" because "the word 'instant' might create an expectation that final results will be available immediately after the polls close on election night.") The new system did not work as well as was hoped due to software and logistical difficulties; the results took several days to produce definitive results.
Voters in Ferndale, Michigan amended the city charter in 2004 to allow for election of the mayor and city council by instant-runoff voting. On March 1, 2005, voters in Burlington, Vermont voted to amend their city charter to use instant-runoff voting.
In September 2003, an amendment to the California State Constitution was proposed (SCA 14 (http://info.sen.ca.gov/pub/bill/sen/sb_0001-0050/sca_14_bill_20030912_introduced.html)) with wide-ranging goals of election reform, including ranked-choice voting for statewide offices.
Activists in the state of Washington have been urging adoption of instant-runoff voting there for several years. An initiative seeking ballot access in 2005 failed to garner enough signatures. Vancouver, Washington has voted to adopted instant-runoff voting, but the state legislature has yet to enact enabling legislation.
Adoption in Canada
The British Columbia Citizens Assembly on Electoral Reform (http://www.citizensassembly.bc.ca/) met during 2004 and selected STV as the preferred method of voting in British Columbia.
Assessing IRV
Comparison of IRV to normal runoff voting
Advantages to instant runoff ballot: (IRV)
- FEWER GAMES: Voters and parties have less opportunity for playing games in early round(s) to influence the elimination order in favor of easier competition. (Runoffs allow more flexibility in tactical votes, influencing elimination, and still having a chance to move back to a favorite in the final round)
- MORE POSITIVE: Candidates are discouraged from negative campaigning. (A winning candidate will usually need first, second and lower ranked preferences to win, and can't safely afford to make enemies with no second chance vote)
- AVOIDS SOME PERVERSE RACES: In a runoff, a major political interest may fracture into a variety of parties, and its vote so splits in the first ballot that all those parties candidates are eliminated in the first ballot. (Compare the 2002 French Presidential election, when the minority Front National candidate won through to the final ballot when the more numerous Socialist vote split.)
Advantages to sequential balloting: (runoff voting)
- EASIER TO VOTE: A runoff allows voters and factions to refocus their attention on remaining candidates in each round. (In IRV, voters must make careful choices among a large set of candidates in one ballot and may not have enough information to make informed rankings among the competitive candidates.)
- CHANCE FOR APPEAL: Candidates that were eliminated are given another chance to endorse and remaining candidates have another chance to court voters supporting the eliminated candidates.
- PRESENTATION OF MINOR PARTIES: In a runoff, minor parties stand on their own merits in the first round of voting. Under the single event IRV, interest in the minor parties only focuses on how they recommend their supporters cast their preferences between the major parties: the electoral system configures minor parties as preference 'cows'. In a runoff minor parties have the power to recommend second preferences in the final round.
Effect on parties and candidates
Unlike runoff voting, however, there are no chances to deal in between rounds, change voters' minds, or gain support of the other candidates.
Giving them only one chance to do so, instant runoff preference voting encourages candidates to balance earning core support through winning first choice support and earning broad support through winning the second and third preferences of other candidates' core supporters. As with any winner-take-all voting system, however, any bloc of more than half the voters can elect a candidate regardless of the opinion of the rest of the voters.
This is considered a weakness by the advocates of a more deliberative democracy, who point to the French system of presidential election where such between-round dealings are heavily exploited and useful (they say) to draw together a very factionalized electorate. However, critics of the French runoff system point to the dreaded "votez escroc, pas facho" (vote for the crook, not the fascist) phenomenon, which awarded Chirac an undeserved landslide victory in 2002.
The Australian system also allows minority parties to have key planks of their platforms included in those of the major parties by means of so-called "preference deals". This is seen as legitimate political activity. If enough people care about (for instance) Green Party issues that that party's second preference can swing the vote, then it is fair enough that it have some limited say in policy.
Another advantage of runoff voting is that it allows a "protest vote" to be made without penalty. A person voting for a minority party doers not "throw their vote away", as with first-past-the-post systems, so allowing the electorate to send clear signals to the major parties.
Criteria passed
There are several voting system criteria that have been defined by political scientists for assessment of voting systems.
IRV meets the majority criterion, the mutual majority criterion, the Condorcet loser criterion, and the independence of clones criterion.
IRV is invulnerable to the burying strategy.
Criteria failed
IRV does not pass the monotonicity criterion, the consistency criterion, the Condorcet criterion, the summability criterion, the participation criterion, or the Independence of irrelevant alternatives criterion.
These theoretical objections may correspond with practical "failures" of IRV, discussed below. The first two, compromise and push-over, are types of tactical voting, where voters vote insincerely to increase the likelihood of a favored outcome. Some argue that Condorcet methods and Approval voting are better at selecting compromise candidates and at reducing the spoiler effect.
Compromise
Assume the earlier Tennessee example. The voters from Memphis can get a better result by "compromising": They can rank Nashville over Memphis, and thus ensure that Nashville, their second choice, will win, rather than Knoxville, their last choice.
Alternatively, if the voters from Memphis do not vote tactically (perhaps because they think they have a chance of winning outright, perhaps because they dislike insincerity, etc.), voters from Nashville can improve their result by "compromising" and ranking Chattanooga over Nashville. This would allow Chattanooga to defeat Knoxville in the first round and go on to become eventual winner, a better result for Nashville voters than a Knoxville win.
Nearly all voting methods produce an incentive to use the compromising strategy in some scenarios, i.e. in scenarios where there is no Condorcet winner. However, IRV clearly has more frequent compromising incentive than Condorcet methods, in that it sometimes produces a compromising incentive even when there is a Condorcet winner.
Also, IRV typically does not allow equal rankings, whereas Condorcet and approval do. If voters do choose to compromise in a version of IRV that does not allow equal rankings, they will have to so by ranking a more-preferred candidate below a less-preferred candidate, which is a more severe distortion of their sincere preferences than ranking them equally.
Push-over
IRV is unusual in that it fails the monotonicity criterion. The implication of this is that it is vulnerable to the "push-over" strategy. Tactical voters can benefit by raising "push-overs" (candidates unlikely to win) above their real favorite, causing a stronger opponent to their favorite to be eliminated early, leaving only the push-over to contend with their favorite in the last round.
This strategy typically requires a great deal of coordination, and typically entails a substantial risk of backfire. Thus, some argue that IRV's vulnerability to the strategy will rarely or never manifest in practice.
"Return of the 3rd-party spoiler effect"
Although third parties without much support are unlikely to "spoil" the race between the major parties, a third party with a substantial share of the vote can potentially cause another type of "spoiler effect" by causing the early elimination of the more similar major party, thus causing the "opposite" major party to win instead. Thus, some argue that IRV is only reliable at stopping this "spoiler effect" as long as the 3rd party clearly does not have a chance to win.
For example, imagine that the following votes are cast in an IRV election:
| 45% of voters | 13% of voters | 13% of voters | 25% of voters | 4% of voters |
|---|---|---|---|---|
| 1. Republican | 1. Democrat | 1. Democrat | 1. Green | 1. Green |
| 2. Green | 2. Democrat |
The IRV winner is the Republican candidate. However, if the Green party candidate withdrew before the race, the Democratic candidate would win instead, a result that is preferable for most Green party voters. Thus, it is difficult to argue that IRV entirely eliminates the spoiler effect.
Failure to pick a compromise candidate
Imagine that candidates are located along a one dimensional ideological spectrum, and that the center of the spectrum is defined by the median voter. Unlike Condorcet methods, IRV does not reliably choose the option closest to the center of the spectrum. Thus it can be argued that IRV is less apt at choosing compromise candidates, and more likely to choose an ideologically polar candidate.
For example, this failure can occur in a 3-choice election where parties A and C are bitterly opposed, and party B is first choice for a minority but tolerable for a large majority. For a real-life example, consider the 17th-century Europe struggle over "government-enforced Catholicism" versus "government-enforced Protestantism", with "freedom of private worship" as the compromise B.
Imagine that votes are cast as follows:
| 38% of voters | 11% of voters | 13% of voters | 38% of voters |
|---|---|---|---|
| 1. A | 1. B | 1. B | 1. C |
| 2. B | 2. A | 2. C | 2. B |
| 3. C | 3. C | 3. A | 3. A |
In IRV, the compromise (choice B) is eliminated immediately. Choice C is elected, arguably giving severely lower total satisfaction amongst voters than choice B, who is preferred by a large majority to A, and who is also preferred by a large majority to C.
Logistical issues
IRV fails the Summability criterion, which means that the results for a particular precinct cannot be conveniently summarized for transfer to the central tally location. Instead, the precincts must send a total record of every ordering of the candidates made by a voter, times the frequency of that ordering. The unwieldiness of this data may prolong the counting procedure, and provide more opportunities for undetected tampering than in summable methods.
If counting takes place in several places for a single IRV election (as in Australia), these counting centers must be connected by a securely authenticated channel (historically the telegraph was used) to inform them which candidate has come last and should be dropped.
Logistical issues in Australia
House of Representatives
Initially, in Australia, ballots are counted at the booth level, with first preference results reported to the Divisional Returning officer and then to the National Tally Room. If it is clear who the two leading candidates will be, a notional distribution of the preferences of the minor candidates may be made. Postal and absentee ballots are of course yet to be processed - that takes another week or two.
Over the next few weeks, ballots and matching documentation are concentrated in the offices of the Divisional Returning Officer, where a actual distribution of preferences is made. This may be done by physically moving the ballots around, or by entering ballot data into a suitable computer.
If a candidate wins 51% of first preferences, a distribution of minor party preferences is strictly speaking not necessary, however the law now allows that such preferences be distributed to see what the "two-party preferred vote" actually is.
Federal elections are conducted by the Australian Electoral Commission, who employ all the workers at all the booths, to a common standard of neutrality and efficiency. Candidates may appoint scrutineers to watch (but not touch) what is going on.
Other single-winner methods
IRV is not the only alternative to the plurality system. Other possible reforms include several different Condorcet methods (e.g. ranked pairs and beatpath), approval voting, range voting, the Borda count, Bucklin, and many others. (See the voting systems article.)
See also
- List of democracy and elections-related topics
- Table of voting systems by nation
- Voting system - many other ways of voting
- First Past the Post electoral system
- Australian electoral system
- Electoral systems of the Australian states and territories
- Runoff voting
- Single Transferable Vote
- Approval voting
- Condorcet method
- Borda count
- Majority Choice Approval
- Papua New Guinea uses an IRV variant known as Limited Preferential Vote
- IRV - Sample election - favorite season
External links
- Advocacy organizations
- Instant Runoff Voting (http://www.fairvote.org/irv) at the Center for Voting and Democracy
- instantrunoff.com, by the Midwest Democracy Center [2] (http://www.midwestdemocracy.org/)
- FIRV (http://www.firv.org) (Ferndale, Michigan for Instant Runoff Voting)
- California IRV Coalition (http://www.calirv.org/)
- Advocacy positions
- "Single Transferable Vote Resists Strategic Voting" (http://www.isye.gatech.edu/~jjb/papers/stv.pdf) (PDF file) by John J. Bartholdi, III and James B. Orlin.
- "What is Instant Runoff Voting?" (http://thedemocraticagenda.blogspot.com/2005/01/what-is-instant-runoff-voting.html) a weblog The Democratic Agenda, January 1, 2005
- Con positions
- "The Problem with Instant Runoff Voting" (http://electionmethods.org/IRVproblems.htm) at electionmethods.org
- Analysis positions
- Nonmonotonicity in AV (http://www.mcdougall.org.uk/VM/ISSUE15/P2.HTM) Article by Eivind Stensholt.
- Comparison (http://www.condorcet.org/rp/IRV.shtml) with Condorcet Voting by Blake Cretney
- Voting methods page (http://fc.antioch.edu/~james_green-armytage/voting.htm) by James Green-Armytage (e.g. 1 (http://fc.antioch.edu/~james_green-armytage/vm/survey.htm#irv) 2 (http://fc.antioch.edu/~james_green-armytage/vm/define.htm) 3 (http://fc.antioch.edu/~james_green-armytage/vm/introduction.htm#irv) 4 (http://fc.antioch.edu/~james_green-armytage/vm/cvdletter.htm) 5 (http://fc.antioch.edu/~james_green-armytage/vm/value_of_first_choice.htm))
- Examples
- IRV poll for U.S. President, 2004 (http://www.mnip.org/caucusresults.htm) by the Independence Party of Minnesota
- pSTV -- Open source software for computing IRV and STV (http://stv.sourceforge.net/)
- Australian Electoral Commission [3] (http://www.aec.gov.au/)de:Instant-runoff voting