Explaining FairVote’s position on STAR Voting

Posted by Rob Richie, Drew Penrose on December 04, 2017

This year has underscored the value of ranked choice voting (RCV). Nationally, our broken politics demands changes to accommodate greater voter choice and give voters more voice. Each of the four cities using RCV experienced remarkable surges in voter participation, and voters handled the system well. State legislators in 20 states, reflecting a balance of partisan views, have introduced RCV legislation and/or held hearings on it, and voters responded with energy and determination when incumbents tried to slow down its expansion. The Ranked Choice Voting Resource Center is working with those involved in election administrator for efficient, straightforward ways to run RCV elections, with one entire state (New Mexico) now having a “turnkey” solution for running RCV elections with what essentially is the flip of a switch.

FairVote is all the more focused on supporting state allies interested in expanding use of RCV. One of those states is Oregon, where voters in Benton County in 2016 approved enactment of RCV in its general elections for the county commission and where the League of Women Voters of Oregon this year formally endorsed RCV after an exhaustive two year study. RCV makes great sense to use in more cities and counties and as a statewide reform.

Some Oregon reformers are proposing an alternative reform called STAR voting, however. While we don’t want to discourage those seeking to reform our current rules, we won’t support their efforts. Here’s an explanation of the system and the reasons for our skepticism and decision to maintain a neutral position on the idea:

How STAR voting works

With STAR voting, the winner is determined in two steps. First, voters provide a score (that is, a rating), anywhere from 0 to 5 to each candidate. If there are four candidates, for example, one voter might score each candidate with a 5 and award a total of 20 points in scores. Another voter might score each candidate with a zero, and award a total of 0 points in scores. More realistically, voters would indicate a mix of scores with a mix of total points.

Each candidate’s total scores from each voter are totalled. Those totals determine which two candidates advance to an automatic runoff. Each voter’s ballots is then examined to see how the runoff candidates were scored. If one candidate was scored higher than the other, that ballot would count for the higher-rated candidate in the runoff. If the two candidates were scored the same, then the ballot would be set aside and not count in the runoff. The winner is the candidate who receives more votes in the automatic runoff.

Why FairVote does not support STAR voting in governmental elections

We appreciate electoral innovation. Rating systems like STAR voting make sense in some contexts. It is similar to how some Olympic events are scored, for example, and “cardinal” systems more generally can work well in informal settings. But we’re highly skeptical that STAR voting is ready for elections to political office where voters have a real stake in the outcome and usually have a clear preference for one candidate after a competitive campaign. Here are a few concerns we have about the system.

 

No history of use: The only voting methods that are used to elect a single winner office for any governmental position at any level anywhere in the world:

  1. plurality voting: vote for one, and the top finisher wins;

  2. forms of runoff elections: after a plurality vote in the first round, hold a second round between the top two finishers under certain conditions like no candidate earning more than half the first round vote); and

  3. forms of RCV -- with variations involving what share of the vote is necessary to win without further RCV tallies and how many candidates advance to the next tally after the first count (and a single case of the points-based Borda count, with two seats reserved for ethnic minorities in Slovenia that were uncontested in 2014).

That means STAR voting isn’t used for governmental elections anywhere in the world. Furthermore, unlike RCV -- which is widely used in non-governmental elections as well, and is recommended by Robert’s Rules of Order when repeated voting is impractical --  STAR voting has never been used in a contested election for private organizations. Its trials have involved snapshot polls where there was no campaign.

As a result, there is no data or research as to STAR voting’s effects in practice for competitive elections. But we know its theoretical properties and can make informed projections about how it would likely work in real elections, which raise serious concerns.

Fails the majority criterion and mutual majority criterion: Unlike ranked choice voting, STAR voting doesn't guarantee a candidate with the first choice support of 51% of voters will win. As an example, see this spreadsheet with a not unreasonable example. Failure to elect a candidate who is backed by an absolute majority of voters as a first choice violates fundamental democratic principles.

Failure of the majority criterion implies that STAR voting also fails the mutual majority criterion. That is, it is possible to have two or more candidates in a group be preferred by a majority of voters to all other candidates, yet for none of those candidates to win. Traditional runoff elections can fail this criterion as well, but RCV upholds it.

Easy understanding of how to vote tactically: Unlike RCV, STAR voting doesn't uphold the later-no-harm criterion. Under systems that violate later-no-harm, voters feel pressured to bullet vote, so as to not dilute their vote for their favorite. High-levels of bullet voting would mean the results differ insignificantly from a plurality election.

Voters would also be incentivized to tactically "bury" the strongest runoff candidates -- that is, try to keep the strongest opponent out of the runoff. This gives an edge to those voters who know that fact compared to voters "who vote like the ballot suggests they vote." An example would be an election where there's a clear Condorcet winner, such as the 2017 French presidential election won by Emmanuel Macron.

In that election, five candidates earned between 19% and 25% of the vote in the first round, yet all polls indicated that Macron clearly was the strongest head-to-head. The other candidates' campaigns all knew about Macron's strength against them and would have had every incentive to keep him out of a runoff by elevating a weaker candidate into the runoff instead. With STAR, any backer of the other French presidential candidate (and 76% of voters fit that category), would have been making a mistake to score Macron with any points if they wanted their favorite candidate to have a chance to win, no matter what they actually thought of Macron.

A backer of the leftist Jean-Luc Melenchon, for example, would score Melenchon with a 5 and  score Macron a zero, then score a 4 for the three other candidates who might finish second ahead of Macron. If backers of each of Macron’s four main challengers pursued this tactic, he wouldn’t have come close to making the runoff.

Imagine any election in which a relatively strong incumbent was facing more than one challenger, and suppose the incumbent was likely to be the first choice of 40% of voters, but short of a majority. Backers of the challengers could keep the incumbent out of the runoff by burying him or her with ratings of zero, and giving all other candidates 5’s and 4’s.

Advantage to voters who understand the system: Because tactical voting can work, STAR gives those voters who understand tactical voting an advantage over voters who simply vote honestly. This is highly problematic. All voters who vote as the STAR ballot suggests are likely voting differently than voters who understand the system and, as a result, may hurt the interests of their favorite candidates.

In contrast, all voters who vote as the RCV ballots suggests are almost certainly voting just the same as fully informed voters, and won’t hurt the interests of their favorite candidate. While it is true that that there post-election analysts could suggest different tactics, those are almost never going to affect what voters do in elections -- and that’s what matters for avoiding tactical voting.

Tactics exaggerate undemocratic outcomes: Unlike RCV, the tactical incentives associated with STAR Voting would directly lead to undemocratic outcomes. Nullifying runoff ballots would be common due to voters scoring the two runoff candidates the same for tactical reasons. The Melenchon voter above has essentially nullified his/her ballot if neither Melenchon nor Macron make the runoff. While it was perfectly logical to cast a ballot with 4’s for those other candidates to try to keep Macron out of the runoff, but now they aren’t differentiated. Scores of zero for the two runoff candidates would also nullify their runoff ballot -- even if those scores would makes sense if trying to keep both those candidates out of the runoff election..  

Inconsistent translation of preferences into scores: Unlike RCV, STAR Voting doesn't factor in the fact that ratings are subjective in how they translate into points. Any score system will results in different voters with the exact same views giving differing numbers of points to candidates. It will relates to whether a voter is a "hard grader" or an "easy grader" and based on awareness of tactical incentives. So two voters might feel exactly the same way about Macron, but one voter might score him a 3 on a scale of 0 to 5 when giving an "honest grade" due to being a hard grader and the other might score him a 5 due to being an easy grader. This is related to why Youtube abandoned its 5-star scoring system, followed later by Netflix replacing its similar scoring system with a  “thumbs up” or “thumbs down” approach: that is, the ratings were resulting in inconsistent outcomes that could be confusing to potential viewers of a video or movie.

Less likely to elect “beats all” Condorcet winner: Just like RCV and runoff systems, STAR voting doesn't guarantee the election of the “Condorcet winner” - the candidate who would defeat all other candidates if paired head-to-head. But due to the incentives to "bury" strong candidates, it’s much more likely that Condorcet winners would lose. Yet pairwise comparisons are built into the reporting of the system, meaning it will showcase immediately when the Condorcet candidate has lost in actual scores.

Conclusion

FairVote is not sectarian or inflexible in only supporting one model of reform. That’s why we are willing in context to support compromise forms of RCV, such as limiting the tally to a second round where only the top two finishers in first choices advance. That’s why we are open to combine RCV with other approaches -- like adapting the “Top Two primary” to one where four candidates advance from a primary held with RCV to a second round where RCV is used to pick the winner

But STAR voting is a distraction that we don’t see as politically viable nor likely to work like its advocates believe. Rather than explore a weaker system with questionable viability, we will keep working with Oregon allies who support using and expanding use of RCV.

 


 
Show Comments
comments powered by Disqus

Join Us Today to Help Create a More Perfect Union