User:DemocracyProfessor/Comparison of instant runoff voting to other voting systems

This article is a comparison of various voting systems with "Instant-runoff voting" (IRV), also called the "Alternative Vote", "preferential voting" and "ranked choice voting."

Categories
Voting systems fall into three broad types.
 * Plurality/majority systems, sometimes called majoritarian.
 * Proportional representation systems, sometimes called proportional.
 * Semiproportional systems.

The types of various legislatures and their voting systems are given below.

Voting system criteria
Political scientists rate voting systems using voting system criteria. Underlying these criteria are Arrow's Theorem and the Gibbard-Satterthwaite theorem, which presume that voters rank all candidates in a strict preference order, among other assumptions. No ranked preference method can satisfy all of the criteria, because some are mutually exclusive. Some of those criteria are given below, Which in turn references: along with a list of voting systems. Voting systems that pass the criterion are labelled with a "Y", those that fail with an "N" and those that suffer an especially intolerable failure with an "IN".

Only three of these voting systems are used to elect candidates in one-winner elections held for national, regional and local government offices: First-past-the-post voting, FPTP; Instant-runoff voting, IRV; and Two-round systems of runoffs, RO. Because both systems involve automatic elimination of trailing candidates, every criteria failed by IRV is also failed by traditional runoffs, although IRV meets certain criteria that runoffs fail to meet.

Legend/Notes: Voting systems that pass the criterion are labelled with a "Y", those that fail with an "N" and those that suffer an especially intolerable failure with an "IN". Data held in the "#" column denotes the same sequence of each tabulated criterion as listed by the references' authors. Criteria in need of references is tabulated last. Additional references are needed to verify claims and to help resolve (3) disputes between claims from existing references.

Tabulated references:

Are the results proportional?
The intention of IRV is to find one candidate acceptable to a majority of voters. It is intended as an improvement on the 'first-past-the-post' (plurality) voting system. Under 'first-past-the-post' the candidate with the most votes wins, even if that candidate has less than a majority of votes and even if considered the worst candidate by a majority of voters (the "Condorcet loser"). IRV will always result in the defeat of the Condorcet loser. When used to elect legislative bodies, however, IRV can produce results that can be unrepresentative of voter preferences across the entire jurisdiction. Like all winner-take-all voting systems, IRV tends to exaggerate the number of seats won by the larger parties. For that reason some backers of proportional representation oppose IRV for legislative elections.

How unproportional are the results?
Significance 868543 deviation from proportionality 1945-2010 average.PNG for sources).

]] The "deviation from proportionality" for four systems since 1945 are as follows:

The "deviation from proportionality" is a linear scale that measures the degree of unfairness of a given voting system from 0 to a higher number. In the case above, 0 = perfectly proportional (seats assigned match the votes cast), and 200 = perfectly unproportional (a party with no votes gets all the seats).

Deviation from proportionality adjusted scale.PNG for sources).

]] Another study, (this time of 24 councils in the United Kingdom) came up with the result below:

That study adjusted the score so that 0% was perfectly proportional and 100% was the most unproportional score possible in practice.

What are the results under different systems?
How well do different political parties do under different systems? Some results are given below:

A different study focusing just on FPTP and AV gave these results:

Are the results decisive?
How decisive are elections under different systems? Some results are given below:

No overall majority
No overall majority 1945-2010.PNG for sources).

]] "No overall majority" is achieved when no single party has more seats than the other parties combined. How often does this happen? Some results are below:

Number of ballots per voter
The "number of ballots per voter" is the number of times a voter has to complete a single ballot.

Voter choice
"Voter choice" is the number of candidates on the ballot.

Like the two-round system, IRV tends to give voters a wider choice of candidates than plurality. More independent and third party candidates are likely to run because the spoiler problems are less severe. However, most jurisdictions (regardless of voting system) limit the total number of candidates by requiring deposits, a large number of nominators, or other measures.

Voter exhaustion
"Incomplete ranking" occurs when a voter fails to list a preference for all the candidates. The ballots on which incomplete ranking has occurred are called "exhausted ballots". The result is referred to as "voter exhaustion".

There are two sources of incomplete ranking:
 * Some implementations don't allow complete ranking, either due to voting machine limitations or other reasons; for example, San Francisco allows only three ranks, even with over twenty candidates.
 * Some voters may simply choose not to rank all the candidates.

To avoid exhaustion, Australia generally does not count exhausted ballots. But New South Wales and Queensland do count such ballots (a variant known as optional preferential voting, or OPV). Antony Green notes that "The exhaustion rate has approached 80% in some seats....optional preferential voting almost always assists the party with the highest primary vote."

Voter confusion
"Voter confusion" takes place when a voter cannot discern who to vote for.

To cast a valid ballot, Australian voters must rank every candidate. Australian political parties use "how to vote" cards to encourage votes for their candidates. Without compulsory rankings, voters would have the option to stop ranking when indifferent to their remaining choices.

Invalid ballots
Invalid ballots.PNG for sources).

]] "Invalid ballots" are ballots that cannot be counted. The most recent rate of invalid ballots for two systems is as follows:

"Invalid ballots" are called "informal ballots" in Australia (where voting is compulsory) and "spoilt ballots" in the United Kingdom (where voting is optional).

Voter participation
Voter participation.PNG for sources).

]] "Turnout" is the number of people who do vote compared with the number of people who can. The most recent turnouts for two systems is as follows:

Voting is compulsory in Australia and optional in the United Kingdom.

Risk of fraud in the United States
Fraudulent counting in the United States takes place when the count is interfered with.

Most counties in the United States use voting machines to register and count their votes at the polling place. Although first choices can be counted at the polling place, AV ballots usually are counted in a central location in races without a first round majority winner. Changing the voting system by counting the vote centrally may increase the risk of fraud or may not increase the probability of fraud associated with such machine counts. FPTP need not be counted centrally.

Risk of fraud in Australia
Fraudulent counting in Australia takes place when the count is interfered with.

Australia (except for ACT ) uses paper and pencils to vote and counts the votes by hand in a central location after unofficial tallies at the polling place. The fear of fraud is less.

Tactical voting
"Vote-splitting" occurs when voters split their votes between similar candidates, allowing a dissimilar candidate to win. One or more of the similar candidates may be characterised as a "spoiler" and the result as the "spoiler effect". "Tactical voting" occurs when a voter lists an "insincere preference" for a candidate due to the "spoiler effect". An "insincere preference" is when a voter deliberately gives a candidate a dishonest preference to prevent somebody winning (or another undesirable outcome).

Is tactical voting possible in theory?
The Gibbard–Satterthwaite theorem demonstrates that no voting system can be entirely immune from tactical voting unless it is dictatorial (there is only one person who is able to choose the winner) or incorporates an element of chance.

IRV permits tactical voting if voters have complete and reliable information about the other voters' full preferences. The 2009 mayoral election in Burlington, Vermont provides an example where tactical voting could have happened: most supporters of the candidate who came in second (and led the first-round) preferred the Condorcet winner (who IRV gave third place) over the IRV winner. If some of these voters had known that beforehand and insincerely raised their second-choice to first-choice, the Condorcet winner would have won the instant runoff, which these voters would have preferred. In a three-party election where voters for both the left and right prefer the centrist candidate to stop the "enemy" candidate winning, those voters on the left and right who care more about defeating the "enemy" than electing their own candidate may cast a tactical first preference vote for the centrist candidate.

Does tactical voting occur in practice?
Under FPTP, voters have an incentive to vote insincerely for one of the two major candidates instead of their true favorite, because a vote for their true favorite is likely to be "wasted."

In his book Collective Decisions and Voting Nicolaus Tideman uses real-world voting data to analyze all proposed election methods in terms of resistance to tactical voting, and states on page 194 that "alternative vote [IRV] is quite resistant to strategy."

Does the spoiler effect occur?
IRV removes the spoiler effect when there are two major candidates and one or more minor candidates. In Australia, (a nation that uses IRV for its House of Representatives election) a smaller third party co-exists with its coalition partner, and functions without losing seats to other parties due to vote splitting. In Australia's national elections in November 2007, at least four candidates ran in every constituency, with an average of seven, and every constituency was won with an absolute majority of votes including several where results would have been different under plurality voting.

In the United States of America (a nation where states use FPTP for allocating Electoral College votes in Presidential election), third party and independent candidates are often characterised as "spoilers" - such as in the 2000 Bush/Gore Presidential race where Ralph Nader was described as drawing support away from Al Gore, allowing George W. Bush to win.

One person, one vote
"One person, one vote" (sometimes called "One man, one vote") is the principle that no voter be given greater weight in his or her vote over the vote of another voter.

In Ann Arbor, Michigan arguments over IRV in letters to newspapers included the belief that IRV "gives minority candidate voters two votes," because some voters' ballots may count for their first choice in the first round and a lesser choice in a later round. This argument was addressed and rejected by a Michigan court in 1975. In Stephenson v. the Ann Arbor Board of City Canvassers, the court held that "majority preferential voting" (as IRV was then known) was not a violation of one-man, one-vote.

Under the "MPV System", however, no one person or voter has more than one effective vote for one office. No voter's vote can be counted more than once for the same candidate. In the final analysis, no voter is given greater weight in his or her vote over the vote of another voter, although to understand this does require a conceptual understanding of how the effect of a "MPV System", is like that of a run-off election. The form of majority preferential voting employed in the City of Ann Arbor's election of its Mayor does not violate the one-man, one-vote mandate nor does it deprive anyone of equal protection rights under the Michigan or United States Constitutions.

Power of an individual vote
Voter Power Index.PNG for sources).

]] The "Voter Power Index" (VPI) for two systems is as follows:

The "Voter Power Index" (VPI) is a linear scale that measures how powerful an individual vote is. In the case above, 0 = perfectly powerless (a vote in a very safe constituency of very large size), and 1 = perfectly powerful (a vote in a very marginal constituency of average size).

Voter equality
Voter equality.PNG for sources).

]] "Voter equality" compares the least powerful fifth to the most powerful fifth of voters. Some results are below:

The "Voter equality" is the VPI for the least powerful fifth of voters divided by the VPI for the most powerful fifth of voters. In the case above, 0% = perfectly unequal (the least powerful fifth have no power) and 100% = perfectly equal (the least powerful fifth have the same power as the most powerful fifth).

Marginality
A constituency is "marginal" if there is a 1 in 5 chance or more that it will change hands in an election. A constituency that is not "marginal" is called "safe". How many constituencies would be safe in the United Kingdom (excluding Northern Ireland) in the 2010 election under different systems? Some results are below:

The same study covered the United Kingdom (including Northern Ireland). The results are below:

"Kicking the rascals out"
Significance 868543 churn 1945-2010 average.PNG for sources).

]] The "average churn rate" (characterised in the UK as "kicking the rascals out" ) for four systems since 1945 are as follows:

The "average churn rate" is a percentage that measures how easy it is to remove an elected member - the higher the number, the easier to "kick the rascals out". In the case above, 0% = impossible to remove any member in any election, 100% = every member loses every seat in every election.

Operating cost
On 2010-08-21 the 2010 Australian federal elections were held. The cost has not yet been released. On 2007-11-24 the 2007 Australian federal elections were held. The cost have been released. It does not give a quote for the cost of the House of Representatives election (held under AV) separately from the cost of the Senate election (held under STV) held on the same day. The costs are below.

On 2010-05-06 the 2010 UK general election was held. The cost has not yet been released but an estimate of the cost has. It does give a quote for the cost of the House of Commons election (held under FPTP) separately from the cost of the local elections (also held under FPTP) held on the same day. The costs are below.

Because it does not require two separate votes, AV costs less than two-round primary/general or general/runoff election systems.

Conversion cost in the USA
The "conversion cost in the USA" is the cost to convert from one voting system to another in the USA.

Writing in the Canadian Journal of Political Science, Harold J. Jansen studied the Alternative Vote in Canada, concluding that "On balance, it differed little from the single member plurality system."

The United States uses machines to register their votes and machines to count their votes. When voting takes place on machines relying on software, changing the voting system can lead to new costs. Pierce County, Washington election officials outlined costs of $3,291,340 to implement IRV for its elections in 2008, covering software and equipment, voter education, testing, staff time, consultants and ballot printing and postage costs. In 2009 the auditor [chief elections director of Washington counties] reported ongoing costs that were not necessarily balanced by the costs of eliminating runoffs for most county offices, as those elections may be needed for other offices not elected by IRV. Other jurisdictions have reported immediate cost savings.

Conversion cost in Australia
The "conversion cost in Australia" is the cost to convert from one voting system to another in Australia.

Australia (except for ACT ) uses paper and pencils to vote and counts the votes by hand in a central location. The cost of running an election per capita is greater Australia, however, as AV involves more than one round.

Rules of Order
"Rules of Order" are rulebooks that recommend how parliamentary assemblies should be run. Their comparative assessment of voting systems are given below

The sequential elimination method used by IRV is described in Robert's Rules of Order Newly Revised, 10th edition. as an example of "preferential voting," a term covering "any of a number of voting methods by which, on a single ballot when there are more than two possible choices, the second or less-preferred choices of voters can be taken into account if no candidate or proposition attains a majority. While it is more complicated than other methods of voting in common use and is not a substitute for the normal procedure of repeated balloting until a majority is obtained, preferential voting is especially useful and fair in an election by mail if it is impractical to take more than one ballot. In such cases it makes possible a more representative result than under a rule that a plurality shall elect...."Preferential voting has many variations. One method is described ... by way of illustration." And then the instant runoff voting method is detailed.

Robert's Rules continues: "The system of preferential voting just described should not be used in cases where it is possible to follow the normal procedure of repeated balloting until one candidate or proposition attains a majority. Although this type of preferential ballot is preferable to an election by plurality, it affords less freedom of choice than repeated balloting, because it denies voters the opportunity of basing their second or lesser choices on the results of earlier ballots, and because the candidate or proposition in last place is automatically eliminated and may thus be prevented from becoming a compromise choice."

Two other books on parliamentary procedure take a similar stance, disapproving of plurality voting and describing preferential voting as an option, if authorized in the bylaws, when repeated balloting is impractical: The Standard Code of Parliamentary Procedure and Riddick's Rules of Procedure.

One-party dominance
"One-party dominance" is achieved when the opposition is so small that it cannot act as an opposition. How often does this happen? Some results are below:

If 90 per cent of council seats are held by a single party, opposition is effectively neutered. If 70 per cent of council seats are held by the majority party there will be some effective opposition but it can always be outvoted.

Conduct (ethnic campaigning)
"Ethnic campaigning" occurs when a candidate attacks another candidate's ethnicity to encourage the voter to vote against that candidate.

Benjamin Reilly suggests instant runoff voting eases ethnic conflict in divided societies. This was a leading reason why Papua New Guinea adopted instant runoff voting. The result in Fiji which adopted it for the same reason were not encouraging.

Conduct (negative campaigning)
"Negative campaigning" occurs when a candidate attacks another candidate to encourage the voter to vote against that candidate.

John Russo, Oakland City Attorney, argued in the Oakland Tribune on July 24, 2006 that "Instant runoff voting is an antidote to the disease of negative campaigning. IRV led to San Francisco candidates campaigning more cooperatively. Under the system, their candidates were less likely to engage in negative campaigning because such tactics would risk alienating the voters who support 'attacked' candidates", reducing the chance that they would support the attacker as a second or third choice. Others state that there is a lack of evidence that such an effect occurs as often as suggested. Indeed, Lord Alexander's objections to the conclusions of the British Independent Commission on the Voting System's report cites the example of Australia saying "their politicians tend to be, if anything, more blunt and outspoken than our own."