Bronze-level article

Instant Runoff Voting

From RationalWiki
Jump to navigation Jump to search
UK general election results (2015). You wouldn't think from this picture that only 37% of voters voted Blue, would you? Such is the folly of first past the post — which IRV aims to fix.
Oh no, they're talking about
Icon politics.svg
As usual
Country sections
United States politics British politics Canadian politics Chinese politics French politics Indian politics Iranian politics Israeli politics Japanese politics South Korean politics

Instant Runoff Voting ("IRV"), also known as the Alternative Vote ("AV")[1] or ranked choice voting ("RCV") is a system of voting designed to allow voter preferences below their first choice to be considered.

It is a popular reform among people whose knowledge of voting theory is limited to having watched a few CGP Grey videos, but experts consider it mediocre at best.[2][3][4][5][6][7] In fact, its use in national-level elections is limited to four countries in the anglosphere and that's it, compared to the sixty-six countries (from Argentina to Zambia) that use a Two-round system for presidential elections. Several of the positive claims made about it are actually myths.[8]


In IRV, a voter picks not only their first choice for an office, but gives an order of preference for the remaining candidates. When the polls close, if a candidate has 50% "+" of first-choice votes, they win. If no candidate has reached this threshold, the candidate with the lowest number of votes is "removed", or excluded, and their votes are redistributed according to the ballots' second choice. The votes are re-tallied to see if a candidate has attained 50% "+" threshold. This process continues until one candidate reaches the threshold number of the votes, whereupon a winner is declared.

The Australian Electoral Commission gives this example[9], from the 1972 federal electionWikipedia to the seat of McMillanWikipedia, illustrating this point.

Round Armitage, Barrie

(Liberal Party)

Buchanan, Alex


Hewson, Arthur

(Country Party)

Houlihan, Michael

(Democratic Labour Party)

Mountford, Frank

(Australian Labor Party)


Redistributed votes












Redistributed votes



0 10262







Redistributed votes



0 13406


0 23173



Redistributed votes

0 0 26096


0 23709

As can be seen, even though Hewson initially did not obtain the most votes (he was third), he won on the back of the preferences of the Country Party's ally, the Liberal Party, and those splitters the People's Popular Front of Judea Democratic Labour Party. In a voting system where votes aren't redistributed, Labor's Mountford would have won in the first instance with 45% of the vote—IRV/AV advocates would insist on expressing this as 55% voted against Mountford, as evidenced by preferences of the eliminated candidates skewing heavily conservative. By the end of this IRV example, Hewson won with 52% of the backing of the voters. In practice, these situations are rare, and in most cases the round-one winner will also end up being the final winner under AV; the difference being that they have a confirmed majority support. Situations where eventual winners receive relatively few first-preference votes in round one are rarer still. IRV is aimed at these situations, however, and attempts to reconcile the broadest possible support for a candidate, by eliminating the problems associated with a split vote across similar parties such as tactical voting.

Purported benefits[edit]

IRV is marketed as a boon to alternate or third-party candidates in strong two-party systems, since a voter can pick their "real" choice without "throwing their vote away": It purportedly allows voters to "vote their conscience" rather than hold their nose and pick the lesser of two evils, while still indicating which of those lessers they could be more comfortable with.

This is true, but only so long as the third parties have no chance of winning anyway.[10] If a third party becomes more popular, IRV still suffers from vote-splitting, and voting honestly for one's true favorite can still spoil an election and help "the greater of two evils" win.[11] This means that IRV tends to perpetuate a two-party system, rather than fix it.[12]

IRV really only works as designed when there are two strong parties and several weaker ones, and most of its proponents don't think beyond this scenario. It does not cope well with three strong candidates or more, and can elect a candidate even though a (super)majority of voters preferred someone else (though it sweeps this under the rug by eliminating the preferred candidate, so that most voters are none the wiser).

IRV is also claimed to promote consensus candidates and reduce polarization, because candidates can get second- and third-preference votes by appealing to a broader set of voters. This doesn't take into account IRV's elimination method, however, which ignores those same second- and third-preference votes! In fact, IRV can eliminate multiple consensus candidates (for not having enough first-preference votes) and winnow down the field until there are only extremists left, whereupon it claims one of them has "majority support" (though this majority was really contrived by eliminating all the better candidates).[13]

Another benefit to small parties is that they can determine their "real" polling strength without damaging a larger-party candidate that their voters would prefer over another large-party candidate.

Potential problems[edit]

One difficulty arising from IRV is the loose-knit coalitions that form, where parties agree to assist one another against one singled out party by printing "how to vote cards" with that party in last place on all cards. This is really only viewed as a problem by people who support the party being singled out. A typical occurrence is the Australian party "One Nation", which has appeared last on most parties' how to vote cards, due to its extreme anti-immigration position and poor economic platform (they once proposed printing more money as a solution to budget deficits, known as "Qualitative Easing" elsewhere). As One Nation has never won a House seat, this is probably a good thing.

Another problem that is less to do with the system but more how it is implemented is that on occasions, during the recount, the winner changes. Generally, the electoral commission will attempt to get a result on the night of the election and will start redistributing votes before they are all counted. If it is a close contest on who comes last, the order of redistribution can affect the final outcome of the election. As a result, there is usually a mandatory recount where the redistribution is checked in full to ensure that the result is correct.

A related problem that exists in smaller elections (like more than 10 candidates and fewer than 100 voters) is elimination order can be ambiguous if there is a tie for last place. The tie can be broken by various methods, including by lot, or which candidate was behind in the first round, or which candidate was behind in the previous round. It is mathematically possible, though unlikely, that any candidate tied for elimination could win later on transfer votes, if they avoid elimination. One approach to reduce these cases is to set a vote threshold as a fixed minimum number of votes or percent of vote in the first round, like eliminating all candidates below 5% in one step. Of course, rules like this can't be used if there aren't at least two candidates above the threshold.

Another danger of having many choices is that voters may not rank enough candidates to prefer one of the final two candidates, or if the ballots themselves have limited the number of preferences allowed. Ballots that have no active choices left are considered exhausted and the majority threshold for victory is reduced with these ballots are removed from the count. This can cause the winner to have less than 50% of the original votes cast. This problem doesn't exist in sequential voting rounds of an actual runoff process. One solution is to trigger a new ballot vote among the final two candidates if neither has 50% of the original ballots. However, a new vote may not help a democratic outcome and the winner of the runoff round may end up with fewer total votes than the winner in the first round, due to a drop in voter turnout.

Donkey vote[edit]

In systems where Instant Runoff Voting exists, there is the potential for a donkey vote, where the voter simply numbers all of the boxes in order they appear. This is not so much a problem with the system as a problem with the voter, but it does mean that the votes are skewed towards the order of the names on the ballot paper. This problem is greatest in countries such as Australia that combine IRV with mandatory voting.[note 1] As a result a lottery is usually held to determine the order of the candidates on the voting sheet. On voting machines, the impact of this behavior could be reduced by randomizing the order in which the candidates are presented to each individual voter. Ballot papers in the ACT[14] are subject to "Robson rotation", wherein the order in which the candidates appear on the ballot is evenly distributed.[15]


In some elections, such as the 2009 election for mayor of Burlington, Vermont[16], raising the ranking of the winning candidate on some ballots can paradoxically cause that candidate to lose the election (and vice versa). For example: in the following 25 voter election between candidates A, B, and C with these ballots:

Vote Quantity
A>B>C 8
B>A>C 7
C>B>A 10

B is eliminated in the first round, and A defeats C 15–10. But if 4 C>B>A voters promote A on their ballots and vote A>C>B, the election becomes:

Vote Quantity
A>B>C 8
A>C>B 4
B>A>C 7
C>B>A 6

Now, B wins, defeating A 13–12 after C is eliminated in the first round. Four voters ranking A higher paradoxically causes him to lose.

In this election, the consensus candidate, Andy Montroll was eliminated, and Bob Kiss was selected as the winner, even though a 54% majority of voters preferred Montroll over Kiss.[17][18] (The voters expressed an unambiguous overall preference ranking of Montroll > Kiss > Kurt Wright > Dan Smith > James Simpson.)

Spoiler Problem[edit]

Much like in Plurality Voting, in IRV it is possible for a candidate's decision to run to change the outcome of an election in a way that is contrary to the interests of their voters. For example, consider the following 99-voter election with the ballots:

Vote Quantity
Left>Center>Right 35
Center>Left>Right 12
Center>Right>Left 12
Right>Center>Left 40

Center is eliminated, and Right defeats Left 52–47. However, if Left decides not to run, Center defeats Right 59–40, an outcome the Left voters would prefer.

This can result in a two-party–dominated political system, since it is often not in voters' interests to rank minor candidate(s) above their favorite major candidate, even when they sincerely prefer the minor candidate(s).[19]

Counting efficiencies[edit]

Counting many rounds of elimination can be time-consuming. There are two mathematically rigorous ways to speed up elimination:

  1. If a set of candidates at the bottom combined all have fewer votes than the next lowest candidate, they can all be eliminated in one step. For example if there are 8 candidates with vote 1%, 2%, 3%, 5%, 8%, 20%, 30%, 31%, the bottom 5 candidates have a combined 19% which is below 20%, so they can be eliminated together.
  2. An initial tally can count how many ballots each candidate is ranked on. Any candidate who is ranked on fewer ballots than the votes of the top candidate in a given round mathematically can't win. For example if the top candidate has 20% of the first-rank votes, and 10 candidates each are ranked on under 20% of the ballots, they can all be eliminated at that step. This rule can also apply to write-in votes combined. If the total number of ballots with write-in preferences are fewer than the top candidate, all write-in votes can be eliminated together.

These efficiencies may not be necessary for computer counts, but they can also simplify election reports, allowing fewer elimination rounds to be presented.

Examples where it might have been illuminating[edit]

Recent examples of how this might have made U.S. presidential politics clearer include Nader voters in 2000 being able to support Gore as a second choice, or Perot voters in 1992 being able to show their preference between Bush and Clinton at some point on their list.[more detail please]

In U.S. Presidential elections, every state except Nebraska and Maine holds a separate FPTP election, then casts its electoral votes for whichever candidate receives the most popular votes. In 2000, though, Florida's vote was too close to call:[20]

Candidate Votes in FL
George W. Bush 2,912,790
Al Gore 2,912,253
Ralph Nader 97,488
Pat Buchanan 17,484
Harry Browne 16,415
everyone else 6,680

If Florida had used IRV instead of FPTP, the election might not have been nearly so close. Let's assume that everyone who voted for Ralph Nader picked Al Gore as their second choice. Even if the folks who voted for any of the other low-ranking candidates had picked George W. Bush as their second choice, Gore would have gained 97,488 votes while Bush would have gained only 40,579, giving Gore an incontestable 50,000+ vote lead.

Real-world implementations[edit]

Some jurisdictions have implemented Instant Runoff Voting. The only countries to have set it up on a national level are Australia (in the House of Representatives, at both the state (except Tasmania, those wily devils) and federal level), Ireland (straight IRV for President, and a multi-member variant for Dáil [lower house of parliament]) and Papua New Guinea (for Parliament). Several major U.S. cities use it for municipal elections, including Minneapolis, Oakland, and Santa Fe, and it was adopted for most elections in the state of Maine in 2016 as well as Alaska in 2020. Several jurisdictions that adopted in the past have since repealed it.[21]

From 1998 onwards, Conservative Party leaders have been elected via a form of IRV involving rounds. Each round eliminates the least popular candidate and the ballot is cast again until two candidates are left and the election proceeds to the full membership vote—assuming no one changes their mind unless their favoured candidate is eliminated, the differences between this and IRV/AV are negligible. In 2005, David Cameron was elected Tory Party leader by this method.[22] This is particularly interesting as Cameron spearheaded the "No 2 AV" campaign against electoral reform in the UK. So, while the system is good enough for MPs, the hoi polloi aren't deemed smart enough to deal with preferential voting.

UK referendum[edit]

I voted Yes and all I got was this lousy JPEG.

On 5 May 2011, the UK held a referendum on switching to instant runoff voting (billed exclusively as Alternative Vote, AV) for its parliamentary elections.

This was primarily driven as a deal by the coalition between the Conservative Party and the Liberal Democrats as a condition of having a coalition government. The Liberal Democrats favoured full proportional representation (PR), not least because such a thing would reduce a massive disadvantage to the party inherent in the current system. The Lib Dems won 23% of the popular vote in the 2010 election, but got only 8.7% of the parliamentary seats—an actual decrease in power despite an unprecedented increase in popular support. Thus the Lib Dem leader, Nick Clegg, once referred to AV as a "miserable little compromise" next to PR. On the other hand, small parties with strong local support can easily gain seats. That's why more than once the government only stayed in power because some Ulster Unionist Party or some other trifle you've never heard of lent them their support in exchange for expensive pork-barrel spending and/or politically disastrous concessions to their pet peeves.

AV, rather than proportional representation, was the most they could get out of the Conservative Party's majority control of the coalition government. The referendum was held, and thanks to voting for or against the "miserable little compromise" as well as the following lies as misrepresentations, the UK was saddled with the old FPTP system instead. Joy.


The disastrously misspelled No2AV campaign worked against AV during the referendum, favouring the current first-past-the-post (FPTP) system over AV. The Tory-led group ran a fairly dirty campaign throughout that was, frankly, riddled with lies and misrepresentations of AV. The following are claims mentioned on the main No2AV leaflet that was sent out in the weeks running up to the 5 May referendum.

Keep One Person, One Vote

This is based on the misrepresentation that AV allows people to vote more than once. This is an odd claim: if someone got multiple votes, they'd probably put them all behind one party/candidate. In fact, AV is also one person, one vote—it's just that votes get counted in rounds. Each vote is counted once in each round, and so no single person can vote more times than anyone else.

None of your taxes have been used to print this leaflet

Good to mention it, as campaigns like this do not use taxpayer money. This is a clear use of poisoning the well to imply that the Yes to Fairer Votes campaign was using tax money—which wasn't true, and nobody would be able to get away with that sort of lie, but is a great example of how sneaky rhetoric can be.

The cost of AV is £250 million

This £250-million cost of AV, effectively a PIDOOMA figure, was present in much of the No2AV campaign from the beginning. By the day of the referendum, however, it had become clear that the figure was simply made up from nowhere, and proponents of the No2AV campaign admitted as much the second the polls closed.[23]

The figure breaks down thus: £91 million for the referendum (which included local electionsWikipedia that were happening anyway) that would occur regardless of the ballot's outcome; £130 million on electronic voting machines that nobody with authority has ever been implied as even being necessary; £26 million on explaining it to voters, which is usually spent explaining each election anyway. The remaining £3 million is presumably a rounding error as the No campaign have frequently been shown to be very bad at maths. As an extension to this point, the No campaign have frequently said, using an appeal to emotion, that the money would be better spent elsewhere. But, since such money clearly wouldn't be spent on AV, nor would it be spent on the 2,503 doctors that they claim it would pay for instead. Indeed, AV was rejected at the referendum, but this money didn't magically make its way into taxpayers' hands.

The second or third best can win under AV
With a little context added, it can be shown that winners in FPTP elections seem far from the "best".

The thing is that "best" is a subjective concept—and "best" in the sense of an election is measured in a different way. If two different systems selected the same "best", they effectively would be the same and a switch would be pointless. So of course two electoral systems would produce different results—that's the point! AV supporters deem the candidate with the largest, broadest support "best", whereas FPTP suggests only the largest minority is "best". Impressively, the diagram that comes with this part of the leaflet highlights pretty well what is wrong with the FPTP system when multiple opposition parties exist: that someone can win with 60 to 70 per cent of people voting against them.

AV is unpopular

AV is famously used in Australia (where they don't use electronic voting machines). Many other countries use proportional representation systems or alternatives like Single Transferable Vote (STV). But given that the "miserable little compromise" was forced upon voters by the people who want to vote No, is this a surprise? Regardless, the popularity of the system is not an argument against (or for) it.

How the systems compare

The No2AV campaign frequently overcomplicated AV to make it seem impossible to understand. In reality, if you don't understand how preferential voting works (it requires complex things such as "counting numbers"), you may find yourself qualified for employment at the Creation Museum.

The BNP card

Godwin's Law applies here as both campaigns have deployed the BNP card to say that the other system would favour fringe groups such as far-right and neo-Nazi groups. Specifically, the No2AV leaflet says that BNP votes will be counted "again and again". It is most likely, though, that the BNP wouldn't gain enough first preference to survive the eliminations long enough to be counted "again and again" (not that votes are counted "again and again" under AV anyway, see above). Also, if the people want a BNP candidate, they should get a BNP candidate. It's called democracy.

It is also worth noting that the BNP itself opposed AV in favor of First Past the Post which they felt would give them a better chance of winning (FactCheck later verified that this was true). In fact, BNP deputy chairman Simon Darby is quoted as saying, "AV is a retrograde step—it's worse than what we've got now. We are never going to get our feet under the table under the AV system."[24]

In the end, however, the UK voted overwhelmingly to keep FPTP and not switch voting systems, proving that if you attack a straw man often enough with a well-funded campaign of propaganda, you too can win any vote you choose.

See also[edit]

External links[edit]


  1. One RationalWikian, acting as scrutineer at a by-election with 17 candidates, saw about 50% donkey or reverse-donkey votes.


  1. UK Electoral Reform Society
  8. Luckily, there are dozens of other voting methods that don't have these problems. Have fun learning about all of them!
  9. Australian Electoral Commission - How the House of Representatives votes are counted
  14. Australian Capital Territory, the area in which the federal capital Canberra is located (think of an Aussie version of Washington, D.C.).
  16. Equal Vote Coalition
  22. - Conservative Leaders
  23. New Statesman: “No campaign used made-up figures”, says David Blunkett