Traditional Instant Runoff or Version 2.0?
Traditional Instant Runoff Voting (IRV) is the Ranked Choice O.G. - it's been used around the world and the states for more than a century. STAR Voting (Score-Then-Automatic-Runoff) is a new kid on the block that also uses an instant runoff to determine the final winner.
Metrics For Comparison
Equal.Vote evaluates voting systems according to five primary criteria: equality, honesty, accuracy, simplicity, and expressiveness. The narrative below describes and then compares IRV and STAR Voting according to these five metrics.
The Rules and the Ballots
IRV uses a rank-order ballot
IRV is one of many forms of Ranked Choice Voting that use a preference order ballot: first choice, second choice, third choice and so on. Because full rank ordering makes for a complex ballot (basically an N x N choice matrix for each office), IRV proponents typically advocate a limit of three or four choices on the ballot.
STAR Voting uses a score ballot
STAR is a form of Ranked Choice that uses a score ballot: the voter can assign a level of support for each candidate, from zero to the maximum score.
Who gets to the final runoff?
Both IRV and STAR Voting have a final instant runoff round to determine the winner. What differentiates the two system is how they determine which candidates are present in the final runoff.
In IRV, the final contestants are determined by successively eliminating the candidate with the fewest first-choice votes amongst the voters. With each elimination, the votes of those who chose the eliminated candidate in top position transfer to the next choice on the ballot, until the final instant runoff that gives one candidate a majority of first choice votes amongst the ballots that remain.
In STAR, the final contestants are the two overall highest scoring candidates.
Equality: Does the voting system provide an equal weight vote and thereby comply with the fundamental principle of 'One Person, One Vote'?
The U.S. Supreme Court has found unequivocally that 'One Person, One Vote' requires that "each vote be given as much weight as any other vote." Put simply, our very notion of democracy requires that the voters and the candidates are on an even playing field in every election, and every proposed reform should be scrutinized through this lens.
STAR Voting provides equality to voters
STAR Voting passes the test of balance that tells us definitively that a voting system provides an equal weight vote to all the voters. With STAR, every possible vote expression in both the scoring and automatic runoff stages has an equal, balancing expression, and all the voters' expressions about all the candidates are counted.
IRV fails 'One Person, One Vote'
IRV, by its very design, gives some voters more than one bite at the apple while giving others just one. In IRV, second choice votes only ever count for those voters who happen to see their first choice eliminated before their second choice. Whether you pick a strong consensus candidate or a weak extremist second, if your first choice hangs around in the count longer, your second choice is never considered by IRV. The 2009 Mayoral IRV election in Burlington, Vermont is a real-world example of how IRV's failure to provide an equal vote can result in an obviously non-representative outcome and subsequent repeal of the system.
Honesty: Can the voter safely express her honest opinion on the ballot, and likewise, to what level does the system disincentivize voters from strategically voting insincerely in order to produce a better outcome?
Voters in IRV can't always safely vote 1st choice for their honest favorites
Instant Runoff Voting proponents often make the inaccurate claim that with IRV, "you can honestly rank candidates in order of choice without having to worry about how others will vote and who is more or less likely to win." In fact, you can only safely rank candidates honestly in IRV if your favorite either has no chance at all or is a very strong candidate. This video shows clearly how IRV suffers from a more general form of the vote-splitting spoiler effect:
Because IRV doesn't eliminate the spoiler effect, major parties can honestly propagandize against supporting up-and-coming third party candidates, as they do in Australia: "This confusion is often encouraged by the major parties who do not want people to give a first preference to a minor party. In left-wing inner-city seats around Sydney and Melbourne, where the Greens are now challenging the hold of the Labor Party, Labor campaigners often will claim that a vote for the Greens would help the Liberal Party, sowing confusion about our electoral system, in order to bring progressive voters back to Labor."
Plain Score Voting has potential strategic issues
IRV advocates argue that rating systems are vulnerable to tactical 'bullet voting' - that is, support of just one candidate on the ballot in order to maximize the chance of one's favorite choice winning. Rating advocates have demonstrated that in a significant number of IRV elections, giving full support to your favorite can actually help elect your least favorite candidate.
STAR Voting breaks the tradeoff
STAR's scoring phase ensures that the two strongest candidates overall advance to the final runoff, and the runoff phase reduces the incentive to score second choices tactically. With STAR Voting, you can honestly support your true favorite and second choice without worrying you'll be promoting a losing candidate over a stronger consensus choice or unduly harming your favorite's chance of winning.
Accuracy: How accurately does the voting system reflect the will of the people?
Voter Satisfaction Efficiency models representation accuracy
Voting method simulation is used primarily to determine how accurately a voting method will produce an outcome that represents the will of the electorate. In a real human election it's hard to actually know what the voters really wanted - they could have been voting strategically, exit polling is imprecise and so on. By running simulated voters through thousands or millions of simulated elections, voting systems can be concretely evaluated and compared.
STAR (aka SRV) tops the list
STAR Voting simulates best-in-class in the several election method simulators that have included it. Across the board, with both honest voters and mixtures of strategic voters, STAR is "unquestionably a top-shelf method."
IRV offers middle-of-the-road representation accuracy
Traditional IRV, by contrast, doesn't come close to the top of the list. Recent work by Robert Norman, a mathematician at Dartmouth, suggests that IRV's topsy-turvy math issues would create non-representative outcomes in one in five close contests among three candidates and that with larger numbers of candidates, it would happen even more often.
Burlington demonstrates IRV's counting failure with crystal clarity: Three strong candidates faced off - we'll call them A, B and C. A majority of voters expressed a preference for A over B, and a plurality of voters expressed a preference for A over C, yet the IRV count eliminated A before either B or C, and elected a candidate not supported by the majority. Not surprisingly, Burlington repealed IRV the next year.
Simplicity: How easy is the system for voters to understand and cast ballots, and how easy is it for elections officials to tabulate and hand-recount?
For a small number of candidate choices, it's arguable whether scoring or rank ordering is simpler: if you only want to support one, bubbling in "first choice" or "top score" are essentially equivalent.
As the number of candidates and choices increase, however, scoring clearly becomes by the simpler method, because each candidate can be considered individually, where ranking systems essentially require the voter to sort the candidates before filling out the ballot. As a Wikimedia board member stated:
"I actually voted in an election conducted under a Condorcet method (specifically Schulze's beatpath method): the Wikimedia board election. There were more than a dozen candidates, and about the only way I could vote intelligently was to first rate them, then turn that into a ranked ballot. So obviously a rating-type ballot (where the second step was not needed) would be easier and quicker to cast."
Multi-round elimination is complex
The mechanism for tabulating IRV can remain opaque even for sophisticated voters. Steve Pond wrote the following about IRV after its adoption as the voting system for choosing Best Picture by the Motion Picture Academy: "A year and a half after the Academy went to a different system for counting Best Picture ballots, nominees and voters and campaigners still don't understand how it works. And it's driving me crazy."
Because IRV's counting algorithm goes through one round for each losing candidate until a winner is decided, elections can take many rounds to compute. The image below represents a sample election vote tally in the multi-round IRV process. It's hard for observers and lay voters to draw meaningful conclusions about voter preferences from these intermediate results:
STAR Voting always completes in two rounds
STAR is simple to tabulate, with transparent results that show overall popular support for every candidate. You can see the election results of the sample election run with STAR below:
IRV requires centralized counting
The winner of an IRV election cannot be tabulated until all ballot information is reported to a central location. Maine, which just passed IRV, is dealing with the headaches of this system now, expecting that "ballots would have to be shipped by town clerks to a central location in Augusta for additional voting rounds. Thus the outcome of an election in a multiple-candidate race might not be known for several weeks."
Oregon would likely use the internet to transfer election data from the counties to the state HQ, so we expect it wouldn't take weeks, but regardless, the lack of precinct summability is a real weakness of IRV. Each locale has to come up with all of the many possible combinations of rank orderings and report the total for each, in order that the central counting authority can begin the process to determine the winner. These partial results give little meaningful information to the voters. A small subset of one election's partial results in IRV might look like this:
STAR Voting results can be summed by precinct
STAR gives information that is useful in understanding how the local electorate felt about the election and makes the results easier to verify by hand-recount. Each precinct sums the scores for each candidate and creates a preference matrix tallying how many ballots prefer each candidate to each other candidate in the race. The central election authority adds up all the precinct sums and all of the precinct preference matrices. The final winner is determined by looking up the two top scoring candidates in the final preference matrix to see which one more voters preferred.
Reading the preference matrix: each element in the matrix represents the voter preferences of one candidate over another. So the number "15" in the first row, second column means that 15 voters gave A ("A>" row label) a higher score than B (column label). The 10 in the second row, first column means that 10 voters gave B a higher score than A. The head-to-head total is A:15 to B:10.
Expressiveness: Can the voter express a nuanced opinion on the outcome?
Ranking is more expressive than plurality
Our current voting system, where we are limited to picking a single favorite in each election, is the least expressive voting system humans have ever constructed. Instant Runoff Voting allows the voter to express an opinion on multiple candidates by placing them in preference order.
Preference order alone leaves out information
The same ordering A > B > C could mean that the voter thinks any of the following:
- A is an awesome candidate, B is mediocre and C is the devil incarnate
- A is awesome, B is almost as awesome, and C is just a hair less awesome than A and B
- A is above average, B is mediocre, and C is mediocre and dishonest
Furthermore while IRV lets you express support for more than one candidate, it doesn't actually count that secondary support at all until your first choice candidate is eliminated.
Scoring is more expressive than ranking
Instead of only counting the voter's support for one candidate at a time, SRV allows the voter to express, and counts, a nuanced level of support for any number of candidates on the ballot.
It's Time For 2.0
The need for true election reform is more apparent now than ever before, and the recent adoption of IRV in Benton County, OR and the state of Maine show that electorates on both coasts are ready for more expressive voting systems. What we need now is the upgraded version, that actually gets rid of the spoiler effect, once and for all, and isn't vulnerable to repeal after adoption due to complexity of implementation and non-representative outcomes.
The American electorate is hungry for a real solution to our broken political system. We clearly need an election system that gives us all an equal say, accurately reflects our collective will in the outcome, is simple for us to ballot and for election officials to tabulate, and that allows us to expressively share our honest opinions on the outcome. By all these measures, the new IRV 2.0 - STAR Voting - is the clear winner.